Processing math: 59%
Research article Special Issues

A-RetinaNet: A novel RetinaNet with an asymmetric attention fusion mechanism for dim and small drone detection in infrared images


  • Received: 17 December 2022 Revised: 06 January 2023 Accepted: 15 January 2023 Published: 02 February 2023
  • To solve the problems of texture lacking and resolution coarseness in the detection of dim and small drone targets in infrared images, we propose a novel RetinaNet with an asymmetric attention fusion mechanism for dim and small drone detection. First, we propose a super-resolution texture-enhancement network as an effective solution for the lack of texture-related information on small infrared targets. The network generates super-resolution images and enhances the texture features of the targets. Second, considering the inadequacy of feature pyramids in the feature fusion stage, we use an asymmetric attention fusion mechanism to constitute an asymmetric attention fusion pyramid network for cross-layer feature fusion in a bidirectional manner; it achieves high-quality semantic and location detail information interaction between scale features. Third, a global average pooling layer is employed to capture global spatial-sensitive information, thus effectively identifying features and achieving classification. Experiments were conducted by using a publicly available infrared image dim-small drone target detection dataset; the results show that the proposed method achieves an AP of 95.43% and a recall of 80.6%, which is a significant improvement over the current mainstream target detection algorithms.

    Citation: Zhijing Xu, Jingjing Su, Kan Huang. A-RetinaNet: A novel RetinaNet with an asymmetric attention fusion mechanism for dim and small drone detection in infrared images[J]. Mathematical Biosciences and Engineering, 2023, 20(4): 6630-6651. doi: 10.3934/mbe.2023285

    Related Papers:

    [1] Muhammad Akram, Saba Siddique, Majed G. Alharbi . Clustering algorithm with strength of connectedness for m-polar fuzzy network models. Mathematical Biosciences and Engineering, 2022, 19(1): 420-455. doi: 10.3934/mbe.2022021
    [2] Muhammad Akram, G. Muhiuddin, Gustavo Santos-García . An enhanced VIKOR method for multi-criteria group decision-making with complex Fermatean fuzzy sets. Mathematical Biosciences and Engineering, 2022, 19(7): 7201-7231. doi: 10.3934/mbe.2022340
    [3] Muhammad Akram, Maheen Sultan, José Carlos R. Alcantud, Mohammed M. Ali Al-Shamiri . Extended fuzzy N-Soft PROMETHEE method and its application in robot butler selection. Mathematical Biosciences and Engineering, 2023, 20(2): 1774-1800. doi: 10.3934/mbe.2023081
    [4] Muhammad Akram, Ayesha Khan, Uzma Ahmad, José Carlos R. Alcantud, Mohammed M. Ali Al-Shamiri . A new group decision-making framework based on 2-tuple linguistic complex q-rung picture fuzzy sets. Mathematical Biosciences and Engineering, 2022, 19(11): 11281-11323. doi: 10.3934/mbe.2022526
    [5] Muhammad Akram, Adeel Farooq, Maria Shabir, Mohammed M. Ali Al-Shamiri, Mohammed M. Khalaf . Group decision-making analysis with complex spherical fuzzy N-soft sets. Mathematical Biosciences and Engineering, 2022, 19(5): 4991-5030. doi: 10.3934/mbe.2022234
    [6] Xueyan Wang . A fuzzy neural network-based automatic fault diagnosis method for permanent magnet synchronous generators. Mathematical Biosciences and Engineering, 2023, 20(5): 8933-8953. doi: 10.3934/mbe.2023392
    [7] Yongfeng Yin, Routing Zhang, Qingran Su . Threat assessment of aerial targets based on improved GRA-TOPSIS method and three-way decisions. Mathematical Biosciences and Engineering, 2023, 20(7): 13250-13266. doi: 10.3934/mbe.2023591
    [8] Sumera Naz, Muhammad Akram, Mohammed M. Ali Al-Shamiri, Mohammed M. Khalaf, Gohar Yousaf . A new MAGDM method with 2-tuple linguistic bipolar fuzzy Heronian mean operators. Mathematical Biosciences and Engineering, 2022, 19(4): 3843-3878. doi: 10.3934/mbe.2022177
    [9] Bo Sun, Ming Wei, Wei Wu, Binbin Jing . A novel group decision making method for airport operational risk management. Mathematical Biosciences and Engineering, 2020, 17(3): 2402-2417. doi: 10.3934/mbe.2020130
    [10] Zhenzhen Zheng, Ching-Shan Chou, Tau-Mu Yi, Qing Nie . Mathematical analysis of steady-state solutions in compartment and continuum models of cell polarization. Mathematical Biosciences and Engineering, 2011, 8(4): 1135-1168. doi: 10.3934/mbe.2011.8.1135
  • To solve the problems of texture lacking and resolution coarseness in the detection of dim and small drone targets in infrared images, we propose a novel RetinaNet with an asymmetric attention fusion mechanism for dim and small drone detection. First, we propose a super-resolution texture-enhancement network as an effective solution for the lack of texture-related information on small infrared targets. The network generates super-resolution images and enhances the texture features of the targets. Second, considering the inadequacy of feature pyramids in the feature fusion stage, we use an asymmetric attention fusion mechanism to constitute an asymmetric attention fusion pyramid network for cross-layer feature fusion in a bidirectional manner; it achieves high-quality semantic and location detail information interaction between scale features. Third, a global average pooling layer is employed to capture global spatial-sensitive information, thus effectively identifying features and achieving classification. Experiments were conducted by using a publicly available infrared image dim-small drone target detection dataset; the results show that the proposed method achieves an AP of 95.43% and a recall of 80.6%, which is a significant improvement over the current mainstream target detection algorithms.



    Multi-criteria decision-making (MCDM) is an essential mathematical tool for solving various daily-life problems involving multiple parameters or attributes, and it is playing a vital role in several areas, including engineering, medical, economics, etc. Inspection of the past two decades show that the aggregation operator (AO) based MCDM methodologies are playing a significant role in solving several real-life problems by converting the raw data into a valuable piece of information. For the classification of alternatives in various daily-life scenarios, the experts or scientists used different types of traditional evaluation tools like crisp set theory. In different decision-making problems due to increasing uncertainties of datasets, it was difficult for the experts to tackle those situations with the help of exact numerical values. To remove this difficulty, Zadeh [1] originally launched the theory of fuzzy sets by proposing a membership function whose codomain is [0,1]. Thus, crisp set theory is a particular case of fuzzy set theory. After that, many experts from all over the globe have been attracted to the powerful idea of fuzzy sets and solved different decision-making problems comprising vagueness and imprecision in their data-sets more accurately than crisp sets, e.g., [2,3,4]. Several AOs in a fuzzy information environment have been explored to deal with different decision-making situations. For instance, Song et al. [5] proposed some parameterized AOs under fuzzy information and studied their basic properties. Merigo and Gil-Lafuente [6] investigated fuzzy induced generalized AOs and applied them to solve decision-making problems.

    As a direct extension of fuzzy sets, Atanassov [7] initiated the notion of intuitionistic fuzzy sets (IFSs) by adding a non-membership function with the membership function in the fuzzy set theory whose functional values sum should be bounded by 1. After the production of an IFS model, the experts moved their attraction to tackle decision-making situations using IFS theory. For example, Xu [8] investigated some IFS-based AOs, namely, IF weighted, ordered weighted, and hybrid weighted averaging AOs (see also [9,10] for IFS-based power AOs and IFS-based ordered weighted distance AOs). In addition, Xu and Yager [11] proposed some IFS-based geometric AOs. Wei [12] introduced different induced geometric and generalized IFS-based AOs with the solution of a decision-making application. Tan et al. [13] proposed IFS-based generalized geometric AOs and studied their applications to MCDM. After the invention of Pythagorean fuzzy sets (PFSs) by Yager [14], Peng and Yang [15] studied some fundamental notions of PFS-based AOs in an interval-valued environment. Garg and Kumar [16] proposed some power geometric AOs based on the connection number in intuitionistic fuzzy format. Shahzadi et al. [17] developed some novel AOs under Pythagorean fuzzy Yager operations. Ali et al. [18] introduced some novel arithmetic and geometric AOs by using complex T-spherical fuzzy sets and studied their application in an investment problem. Ashraf et al. [19] submitted certain spherical fuzzy Dombi AOs and explored their application to multiple attribute group decision-making. In recent years, several studies have been completed which directly involve the aggregation of bipolar data with the help of existing operations, that is, Dombi and Hamacher t-norms and t-conorms. For example, Wei et al. [20] presented bipolar data-based Hamacher AOs with their MCDM applications. Afterwards, Jana et al. [21] proposed bipolar data-based Dombi AOs and solved a daily-life problem. In addition, Jana et al. [22] introduced bipolar fuzzy Dombi prioritized AOs.

    Many daily life situations involve datasets from m different agents or sources (m2), which means the multipolar information emerges that cannot be portrayed mathematically through the traditional tools of crisp set theory, fuzzy set theory, IFS theory and PFS theory. The main goal of the work offered in this article is to tackle the shortcomings of mathematical tools considering multipolar, multi-attribute and multi-index information. These days, research scholars think that this world is nearing the concepts of multipolarity because multipolarity in information and data plays a substantial role in numerous disciplines ranging from arts to sciences. For example, a noisy communication channel may have different latency, bandwidth, radio frequency and network range. Concerning information technology, multipolar technology can be employed to analyze larger information systems. Concerning neurobiology, neurons in the brain collect data from other multiple neurons. Concerning a social network, the efficacy rate of distinct people may be distinct regarding trading relationships, proactiveness, and socialism. All of these multipolar scenarios contain fuzzy data. To deal with such multipolar situations, we need more innovative theoretical and mathematical models. In summary, the prevailing theories of fuzzy sets, IFSs and PFSs are very efficient mathematical tools to deal with vagueness and uncertainties; but they are inefficient in some scenarios, e.g., when the under-consideration datasets are multi-dimensional. To solve this difficulty in the implementation of fuzzy sets and their extensions, Chen et al. [23] generalized the theory of fuzzy sets and proposed the theory of m-polar fuzzy (mF) sets, which have the ability to deal with multipolarity in datasets of different domains of modern sciences. To date, some studies have focused on the aggregation of mF information by using different AOs. For example, Waseem et al. [24] launched mF Hamacher AOs and solved two MCDM problems. Khameneh and Kilicman [25] presented the ideas of mF soft weighted AOs and implemented them to solve MCDM problems. Additionally, Akram et al. [26] initiated the notions of mF Dombi AOs and explored some of their MCDM applications. Recently, Naz et al. [27] proposed some novel 2tuple linguistic bipolar fuzzy Heronian mean AOs for group decision-making.

    In the early 1980s, Yager proposed a t-norm (TN) and t-conorm (TCoN), which are more universal operators than the Lukasiewicz TN and TCoN, respectively. Recently, a number of researchers have been attracted toward these and introduced several new results in the area of MCDM. For example, Garg et al. [28] introduced Fermatean fuzzy Yager AOs and studied their application to COVID-19 testing facility. In addition, Liu et al. [29] presented some certain kinds of q-rung picture fuzzy Yager AOs for decision-making. Later, Akram et al. [30] launched the theory of complex Pythagorean fuzzy Yager AOs and illustrated their validity through an MCDM problem-solving method. All of these models do not consider the aggregation of mF information under Yager's TN and TCoN. We take Yager's operations due to their simple implementation compared to other TNs and TCoNs like Dombi, Hamacher and Frank. And, Yager's operations also consider a strong correlation between different estimated results compared to other operators. Therefore, in this article, we propose some other novel types of Yager AOs for the aggregation of mF information. For more related useful basic terminologies, the readers are referred to [31,32,33,34,35,36,37,38,39,40,41,42].

    The following reasons motivate us to develop the mF Yager AOs.

    1) The theory of mF sets being a generalized fruitful tool is playing a vital role in the execution procedure of uncertain decision-making problems involving multipolar information.

    2) The theory of fuzzy sets is only able to handle datasets in one dimension, and thus a loss of information may occur. This is because, in many practical situations, multiple attributes and all of their possible features can only be handled with mF set theory and its hybrid models.

    3) Until today, several results on the aggregation of complex real-world problems involving mF datasets have been presented under different MCDM-AOs (i.e., mF Hamacher and Dombi AOs), but the aggregation of mF information with the help of Yager's operations (that is, Yager's TN and TCoN) has not been elucidated.

    4) The mF Yager AOs provide an alternative approach for dealing with several MCDM problems like some existing mF AOs.

    To sum up, from the aforementioned discussion, we notice that the work on the aggregation of mF information under Yager's operations is not present in the existing literature. Due to these shortcomings, in this article, we have presented mF Yager AOs and operated them to solve a practical MCDM problem. This article mainly contributes the following:

    1) The concepts of some mF Yager arithmetic and geometric AOs are proposed along with their basic properties, including monotonicity, idempotency, boundedness and commutativity.

    2) An algorithm is designed step-by-step for dealing with daily-life MCDM problems in an mF information environment.

    3) A number of site selection problems have been explored in the literature via different fuzzy set-based hybrid models [43,44]. Thus, to verify the applicability of the initiated mF Yager AOs in practical scenarios, an application is presented which deals with the selection of an appropriate site for an oil refinery.

    4) To prove the feasibility and authenticity of the initiated mF Yager AOs, a comparison of these mF Yager AOs is investigated with existing mF Hamacher AOs [24], and mF Dombi AOs [26].

    This article is structured as follows: Section 2 first reviews basic definitions and properties associated with mF numbers and then introduces the mF Yager weighted averaging (mFYWA) operator, mF Yager ordered weighted averaging (mFYOWA) operator, mF Yager hybrid averaging (mFYHA) operator, mF Yager weighted geometric (mFYWG) operator, mF Yager ordered weighted geometric operator (mFYOWG) and mF Yager hybrid geometric (mFYHG) operator. Section 3 first develops a MCDM method under initiated mF Yager AOs to solve real-life problems containing complicated mF information and then explores an MCDM application in which the selection of a suitable site for an oil refinery is investigated. Section 4 provides a comparison of the developed methodology for mF Yager AOs with mF Hamacher [24] and Dombi [26] AOs. Section 5 concludes our work by providing advantages, disadvantages and some further future directions.

    The notations and abbreviations are provided in Table 1.

    Table 1.  Nomenclature of the research work.
    Acronyms and Notations Description
    mF m-polar fuzzy
    mFYOWG mF Yager ordered weighted geometric
    mFDWA mF Dombi weighted averaging
    mFHWA mF Hamacher weighted averaging
    mFHWG mF Hamacher weighted geometric
    COVID-19 Corona-virus disease 2019
    S(˜η) Accuracy function of mF number ˜η
    A(˜η) Score function of mF number ˜η
    ˜η=(p1η,,pmη) mF number
    Υ=(Υ1,Υ2,,Υn)T weight-vector
    S={S1,S2,,Sk} Universal set
    {E1,E2,,En} Universal set of parameters
    ˜M=(˜dit)k×n mF decision matrix
    ˜dr Preference values

     | Show Table
    DownLoad: CSV

    This section first reviews the definition of mF sets and some operations of mF numbers; it then presents some essential Yager operations for mF numbers by using Yager's TCoN and Yager's TN and establishes mF Yager arithmetic and geometric AOs together with illustrative numerical examples.

    Definition 2.1. [23] An mF set or mF set on a universal set S is a mapping η:S[0,1]m. The belongingness degree of each alternative is expressed as η(s)=(p1η(s),p2η(s),,pmη(s)) where sS, and for (j=1,2,,m), pjη:[0,1]m[0,1] is the j-th projection mapping.

    For an mF number ˜η=(p1η,,pmη), where pjη[0,1], for all j=1,2,,m, the score and accuracy functions of mF number ˜η are respectively given as follows:

    Definition 2.2. [24] For an mF number ˜η=(p1η,,pmη), its score S and accuracy A functions are provided by

    S(˜η)=1m(mt=1(ptη)),S(˜η)[0,1],A(˜η)=1m(mt=1(1)t+1(ptη1)),A(˜η)[1,1].

    Clearly, the above Definition 2.2 provides us an ordered relation criterion for mF numbers, which is given as follows:

    Definition 2.3. [24] For any two mF numbers ˜η1=(p1η1,,pmη1), and ˜η2=(p1η2,,pmη2), we have

    1) ˜η1<˜η2, if S(˜η1)<S(˜η2),

    2) ˜η1>˜η2, if S(˜η1)>S(˜η2),

    3) If S(˜η1)=S(˜η2) then

    ˜η1<˜η2, if A(˜η1)<A(˜η2),

    ˜η1>˜η2, if A(˜η1)>A(˜η2),

    ˜η1=˜η2, if A(˜η1)=A(˜η2).

    Some useful fundamental properties of mF numbers are given as below [24]:

    1) ˜η1˜η2=(p1η1+p1η2p1η1.p1η2,,pmη1+pmη2pmη1.pmη2),

    2) ˜η1˜η2=(p1η1.p1η2,,pmη1.pmη2),

    3) ς˜η=(1(1p1η)ς),,1(1pmη)ς),ς>0,

    4) (˜η)ς=((p1η)ς,,(pmη)ς),ς>0,

    5) ˜ηc=(1p1η,,1pmη),

    6) ˜η1˜η2, if and only if p1η1p1η2,,pmη1pmη2,

    7) ˜η1˜η2=(max(p1η1,p1η2),,max(pmη1,pmη2)),

    8) ˜η1˜η2=(min(p1η1,p1η2),,min(pmη1,pmη2)).

    Theorem 2.1. [24] Let ˜η1=(p1η1,,pmη1) and ˜η2=(p1η2,,pmη2) be mF numbers and ς,ς1,ς2>0, then, we have

    1) ˜η1˜η2=˜η2˜η1,

    2) ˜η1˜η2=˜η2˜η1,

    3) ς(˜η1˜η2)=ς(˜η1)ς(˜η2),

    4) (˜η1˜η2)ς=(˜η1)ς(˜η2)ς,

    5) ς1˜η1ς2˜η1=(ς1+ς2)˜η1,

    6) (˜η1)ς1(˜η2)ς2=(˜η1)ς1+ς2,

    7) ((˜η1)ς1)ς2=(˜η1)ς1ς2.

    Yager [41] initiated a useful TN (Yager product ) and TCoN (Yager sum ), which are respectively given by

    Y(s1,s2)=s1s2=1min(1,((1s1)σ+(1s2)σ)1σ), (2.1)
    Y(s1,s2)=s1s2=min(1,((s1)σ+(s2)σ)1σ), (2.2)

    where σ0 and s1,s2R (set of real numbers).

    We are now ready to present some essential Yager operations for mF numbers by using Yager's TCoN and Yager's TN. For two mF numbers ~η1=(p1η1,,pmη1)and~η2=(p1η2,,pmη2) and ς>0, we provide certain operations of mF numbers with Yager's TN and TCoN as below:

    ˜η1˜η2=(min(1,((p1η1)2σ+(p1η2)2σ)1σ),, min(1,((pmη1)2σ+(pmη2)2σ)1σ)),

    ˜η1˜η2=(1min(1,((1(p1η1)2)σ+(1(p1η2)2)σ)1σ),,1min(1,((1(pmη1)2)σ+(1(pmη2)2)σ)1σ)),

    ς˜η1=(min(1,(ς(p1η1)2σ)1σ),,min(1,(ς(pmη1)2σ)1σ)),

    (˜η1)ς=(1min(1,(ς(1(p1η1)2)σ)1σ),,1min(1,(ς(1(pmη1)2)σ)1σ)).

    In this subsection, we introduce some novel mF Yager arithmetic AOs with their useful properties:

    Definition 2.4. Let ˜ηt=(p1ηt,,pmηt) with t=1,2,,n be a finite set of mF numbers; then, a function mFYWAΥ:˜ηn˜η is called an mF Yager weighted average operator, which is given as

    mFYWAΥ(~η1,~η2,,~ηn)=nt=1(Υt˜ηt), (2.3)

    where Υ=(Υ1,Υ2,,Υn)T represents the weights for ˜ηt, t=1,,n and Υt>0 with nt=1Υt=1.

    Now we provide the main result to aggregate mF information with the proposed mF Yager operations.

    Theorem 2.2. Let ˜ηt=(p1ηt,,pmηt) be a finite collection of mF numbers, that is, t=1,2,,n; then, an aggregated value of these mF numbers using the mF Yager weighted average operator is given by

    mFYWAΥ(~η1,~η2,,~ηn)=nt=1(Υt˜ηt),=(min(1,(nt=1Υt(p1ηt)2σ)1σ),,min(1,(nt=1Υt(pmηt)2σ)1σ)). (2.4)

    The proof of this theorem is given in Appendix A.

    Example 2.1. Let ˜η1=(0.5,0.4,0.7),˜η2=(0.2,0.4,0.3),˜η3=(0.8,0.9,0.6) and ˜η4=(0.7,0.5,0.3) be 3-polar fuzzy (3F) numbers and Υ=(0.3,0.1,0.4,0.2)T be weights associated with these 3F numbers. Then, for σ=5,

    mFYWAΥ(~η1,~η2,~η3,~η4)=4t=1(Υt˜ηt)=(min(1,(4t=1Υt(p1ηt)2σ)1σ),,min(1,(4t=1Υt(p3ηt)2σ)1σ)),=(min(1,(0.3×(0.5)10+0.1×(0.2)10+0.4×(0.8)10+0.2×(0.7)10)1/5),min(1,(0.3×(0.4)10+0.1×(0.4)10+0.4×(0.9)10+0.2×(0.5)10)1/5),min(1,(0.3×(0.7)10+0.1×(0.3)10+0.4×(0.6)10+0.2×(0.3)10)1/5)),=(0.7395,0.8213,0.6364).

    In what follows, some essential properties of mFYWA operators are explored.

    Theorem 2.3. (Monotonicity) For two sets of mF numbers ˜ηt and ~ηt, with t{1,2,,n}, if each ˜ηt~ηt, then

    mFYWAΥ(~η1,~η2,,~ηn)mFYWAΥ(~η1,~η2,,~ηn). (2.5)

    Proof. It is straightforward by Definition 2.4 and Theorem 2.2.

    Theorem 2.4. (Idempotency) For a collection of mF numbers which are 'n' in number given as ˜ηt=(p1ηt,,pmηt) such that ˜ηt=˜η, we get

    mFYWAΥ(~η1,~η2,,~ηn)=˜η. (2.6)

    The proof of this theorem is provided in Appendix B.

    Theorem 2.5. (Boundedness) For a set of 'n' mF numbers ˜ηt=(p1ηt,,pmηt), if ˜ηl=nt=1(ηt) and ˜ηu=nt=1(ηt), then

    ˜ηlmFYWAΥ(~η1,~η2,,~ηn)˜ηu. (2.7)

    Proof. Its proof is easily followed by Definition 2.4 and Theorem 2.2.

    Now we discuss the notion of mFYOWA operators with some basic results.

    Definition 2.5. For a collection of mF numbers ˜ηt=(p1ηt,,pmηt), t=1,2,,n, an mFYOWA operator is a function mFYOWAΥ : ˜ηn˜η, which is given by

    mFYOWAΥ(~η1,~η2,,~ηn)=nt=1(Υt˜ης(t)), (2.8)

    where Υ=(Υ1,Υ2,,Υn)T is the weight-vector and Υt(0,1] with nt=1Υt=1. ς(t),(t=1,2,,n) represents the permutation, for which ˜ης(t1)˜ης(t).

    Theorem 2.6. For a set of mF numbers ˜ηt=(p1ηt,,pmηt) with t=1,2,,n, an accumulated value of these mF numbers by utilizing the mFYOWA operator is provided by

    mFYOWAΥ(~η1,~η2,,~ηn)=nt=1(Υt˜ης(t))=(min(1,(nt=1Υt(p1ης(t))2σ)1σ),,min(1,(nt=1Υt(pmης(t))2σ)1σ)). (2.9)

    Proof. It is similar to the proof of Theorem 2.2.

    Example 2.2. Let ˜η1=(0.5,0.4,0.7,0.6,0.2),˜η2=(0.1,0.5,0.4,0.3,0.6) and ˜η3=(0.6,0.2,0.4,0.3,0.7) be three 5-polar fuzzy numbers with weights Υ=(0.5,0.3,0.2)T. Then, for σ=4, by Definition 2.2, we calculate the scores as below:

    S(˜η1)=0.5+0.4+0.7+0.6+0.25=0.48,S(˜η2)=0.1+0.5+0.4+0.3+0.65=0.38,S(˜η3)=0.6+0.2+0.4+0.3+0.75=0.44.

    This implies S(˜η3)>S(˜η1)>S(˜η2); therefore,

    ˜ης(1)=˜η1=(0.5,0.4,0.7,0.6,0.2),˜ης(2)=˜η3=(0.6,0.2,0.4,0.3,0.7),˜ης(3)=˜η2=(0.1,0.5,0.4,0.3,0.6).

    Then, from Definition 2.5,

    mFYOWAΥ(~η1,~η2,~η3)=3t=1(Υt˜ης(t)),=(min(1,(3t=1Υt(p1ης(t))2σ)1σ),,min(1,(3t=1Υt(p5ης(t))2σ)1σ)),=(min(1,(0.5×(0.5)8+0.3×(0.6)8+0.2×(0.1)8),min(1,(0.5×(0.4)8+0.3×(0.2)8+0.2×(0.5)8),min(1,(0.5×(0.7)8+0.3×(0.4)8+0.2×(0.4)8),min(1,(0.5×(0.6)8+0.3×(0.3)8+0.2×(0.3)8),min(1,(0.5×(0.2)8+0.3×(0.7)8+0.2×(0.6)8)),=(0.5377,0.4272,0.6428,0.5505,0.6157).

    Remark 2.1. The mFYOWA operators verify different basic laws such as monotonicity, idempotency and boundedness as given by Theorems 2.3–2.5.

    Theorem 2.7. (Abelian Property) For every two sets of mF numbers ˜ηt and ~ηt with t{1,2,,n}, we have

    mFYOWAΥ(~η1,~η2,,~ηn)=mFYOWηΥ(~η1,~η2,,~ηn); (2.10)

    here ~ηt serves as an arbitrary permutation of ~ηt.

    Proof. Its proof is straightforward by Definition 2.5 and Theorem 2.6.

    From the above theory of arithmetic AOs (mFYWA and mFYOWA operators), we deduce that they efficiently aggregate mF numbers, but the first type of AOs do not consider ordering while the second type of AOs consider the ordering of mF numbers. In what follows, we provide a new kind of AOs, namely, the mFYHA operator, which keeps the characteristics of mFYWA and mFYOWA operators.

    Definition 2.6. For a set of mF numbers ˜ηt=(p1ηt,p2ηt,,pmηt) where t{1,2,,n}, an mFYHA operator is provided by

    mFYHAΥ,Ω(~η1,~η2,,~ηn)=nt=1(Υt˜˜ης(t)), (2.11)

    where Υ=(Υ1,Υ2,,Υn)T is the weight-vector corresponding to the mF numbers ~ηt with the following conditions: Υt(0,1],nt=1Υt=1 and ˜˜ης(t) represents the jth biggest mF numbers such that ˜˜ης(t)=(nΩt)~ηt,t{1,2,,n}, where Ω=(Ω1,Ω2,,Ωn)T is another weight-vector with Ωt(0,1],nt=1Ωt=1.

    Notice that, when Υ=(1n,1n,,1n), the mFYHA operator converts into the mFYWA operator. If Ω=(1n,1n,,1n), then the mFYHA operator becomes the mFYOWA operator. Thus, mFYHA operators investigate the mF degrees and ordering of mF numbers as an extension of both AOs, i.e., the mFYWA and mFYOWA operators.

    Theorem 2.8. For a set of mF numbers ˜ηt=(p1ηt,,pmηt) with t{1,2,,n}, an accumulated value of these mF numbers with the help of mFYHA operators is given by

    mFYHAΥ,Ω(~η1,~η2,,~ηn)=nt=1(Υt˜˜ης(t))=(min(1,(nt=1Υt(p1˜˜ης(t))2σ)1σ),,min(1,(nt=1Υt(pm˜˜ης(t))2σ)1σ)). (2.12)

    Proof. It is similar to the proof of Theorem 2.2.

    Example 2.3. Let ˜η1=(0.4,0.7,0.3,0.5),˜η2=(0.3,0.4,0.2,0.6), ˜η3=(0.7,0.3,0.4,0.1) and ˜η4=(0.5,0.6,0.8,0.7) be 4-polar fuzzy (4F) numbers with Υ=(0.4,0.1,0.2,0.3)T, a weight-vector corresponding to these available 4F numbers and another weight-vector Ω=(0.2,0.1,0.3,0.4)T. Then, using Definition 2.6, when σ=4,

    ˜˜η1=(min(1,(nΩ1(p1η1)2σ)1σ),,min(1,(nΩ1(p4η1)2σ)1σ)),=(min(1,(4×0.2×(0.4)8)1/4),min(1,(4×0.2×(0.7)8)1/4),min(1,(4×0.2×(0.3)8)1/4),min(1,(4×0.2×(0.5)8)1/4)),=(0.3890,0.6807,0.2917,0.4862).

    Similarly,

    ˜˜η2=(min(1,(4×0.1×(0.3)8)1/4),min(1,(4×0.1×(0.4)8)1/4),min(1,(4×0.1×(0.2)8)1/4),min(1,(4×0.1×(0.6)8)1/4)),=(0.2675,0.3567,0.1784,0.5351),
    ˜˜η3=(min(1,(4×0.3×(0.7)8)1/4),min(1,(4×0.3×(0.3)8)1/4),min(1,(4×0.3×(0.4)8)1/4),min(1,(4×0.3×(0.1)8)1/4)),=(0.7161,0.3069,0.4092,0.1023),

    and

    ˜˜η4=(min(1,(4×0.4×(0.5)8)1/4),min(1,(4×0.4×(0.6)8)1/4),min(1,(4×0.4×(0.8)8)1/4),min(1,(4×0.4×(0.7)8)1/4)),=(0.5303,0.6363,0.8484,0.7424).

    Now the scores of mF numbers for σ=4 are determined by

    S(˜˜η1)=0.3890+0.6807+0.2917+0.48624=0.4619,S(˜˜η2)=0.2675+0.3567+0.1784+0.53514=0.3344,S(˜˜η3)=0.7161+0.3069+0.4092+0.10234=0.3836,S(˜˜η4)=0.5303+0.6363+0.8484+0.74244=0.6893.

    Since, S(˜˜η4)>S(˜˜η1)>S(˜˜η3)>S(˜˜η2), thus

    ˜˜ης(1)=˜˜η4=(0.5303,0.6363,0.8484,0.7424),˜˜ης(2)=˜˜η1=(0.3890,0.6807,0.2917,0.4862),˜˜ης(3)=˜˜η3=(0.7161,0.3069,0.4092,0.1023),˜˜ης(4)=˜˜η2=(0.2675,0.3567,0.1784,0.5351).

    Then, from Theorem 2.8,

    mFYHAΥ,Ω(~η1,~η2,~η3,~η4)=4t=1(Υt˜˜ης(t))=(min(1,(4t=1Υt(p1˜˜ης(t))2σ)1σ),,min(1,(4t=1Υt(p4˜˜ης(t))2σ)1σ)),=(min(1,(0.4×(0.5303)8+0.1×(0.3890)8+0.2×(0.7161)8+0.3×(0.2675)8)1/4),min(1,(0.4×(0.6363)8+0.1×(0.6807)8+0.2×(0.3069)8+0.3×(0.3567)8)1/4),min(1,(0.4×(0.8484)8+0.1×(0.2917)8+0.2×(0.4092)8+0.3×(0.1784)8)1/4),min(1,(0.4×(0.7424)8+0.1×(0.4862)8+0.2×(0.1023)8+0.3×(0.5351)8)1/4)),=(0.5982,0.5938,0.7567,0.6671).

    In what follows, some other kinds of geometric AOs are presented under the conditions of Yager's operations on mF information, and they are mFYWG, mFYOWG and mFYHG operators.

    Definition 2.7. For a set of mF numbers ˜ηt=(p1ηt,p2ηt,,pmηt), t=1,2,,n, a mapping mFYWG : ˜ηn˜η is called the mFYWG operator, which is given by

    mFYWGΥ(~η1,~η2,,~ηn)=nt=1(˜ηj)Υt, (2.13)

    where Υ=(Υ1,Υ2,,Υn)T is the weight-vector, with nt=1Υt=1,Υt(0,1].

    Theorem 2.9. For a set of mF numbers ˜ηt=(p1ηt,,pmηt) with t{1,2,,n}, an accumulated value of the given mF numbers with the help of mFYWG operators is provided by

    mFYWGΥ(~η1,~η2,,~ηn)=nt=1(˜ηt)Υt,=(1min(1,(nt=1(Υt(1(p1ηt)2)σ))1σ),,1min(1,(nt=1(Υt(1(pmηt)2)σ))1σ)). (2.14)

    Proof. It is similar to the proof of Theorem 2.2.

    Example 2.4. Suppose that ˜η1=(0.3,0.7,0.5),˜η2=(0.8,0.9,0.6),˜η3=(0.4,0.3,0.1) and ˜η4=(0.5,0.4,0.8) be 3F numbers with the weight-vector Υ=(0.2,0.4,0.1,0.3)T. For σ=4, we get

    mFYWGΥ(~η1,~η2,~η3)=4t=1(˜ηt)Υt,=(1min(1,(4t=1(Υt(1(p1ηt)2)σ))1σ),,1min(1,(4t=1(Υt(1(p3ηt)2)σ))1σ)),=(1min(1,(0.2×(1(0.3)2)4+0.4×(1(0.8)2)4+0.1×(1(0.4)2)4+0.3×(1(0.5)2)4)1/4),1min(1,(0.2×(1(0.7)2)4+0.4×(1(0.9)2)4+0.1×(1(0.3)2)4+0.3×(1(0.4)2)4)1/4),1min(1,(0.2×(1(0.5)2)4+0.4×(1(0.6)2)4+0.1×(1(0.1)2)4+0.3×(1(0.8)2)4)1/4)),=(0.5168,0.5532,0.5535).

    One can easily prove from the above discussion that the mFYWG operators hold the following properties. So, we omit their proofs.

    Theorem 2.10. (Idempotent Property) Let ˜ηt=(p1ηt,,pmηt) be a set of 'n' equal mF numbers, that is, ˜ηt=˜η; then,

    mFYWGΥ(~η1,~η2,,~ηn)=˜η. (2.15)

    Theorem 2.11. (Bounded Property) Let ˜ηt=(p1ηt,,pmηt) be a set of 'n' mF numbers, ˜ηl=nt=1(ηt) and ˜ηu=nt=1(ηt); then,

    ˜ηlmFYWGΥ(~η1,~η2,,~ηn)˜ηu. (2.16)

    Theorem 2.12. (Monotone Property) For every two arbitrary sets of mF numbers ˜ηt and ~ηt with t{1,2,,n}, if ˜ηt~ηt, then

    mFYWGΥ(~η1,~η2,,~ηn)mFYWGΥ(~η1,~η2,,~ηn). (2.17)

    We now present some new mFYOWG operators as below:

    Definition 2.8. For a set of mF numbers ˜ηt=(p1ηt,p2ηt,,pmηt) with t{1,2,,n}, an mFYOWG operator is a mapping mFYOWG : ˜ηn˜η, which is given as:

    mFYOWGΥ(~η1,~η2,,~ηn)=nt=1(˜ης(t))Υt (2.18)

    where Υ=(Υ1,Υ2,,Υn)T is the weight-vector and Υt(0,1] with nt=1Υt=1. Here ς(t) with (t=1,2,,n) serves as an arbitrary permutation which satisfies ˜ης(t1)˜ης(t).

    Theorem 2.13. For a set of mF numbers ˜ηt=(p1ηt,,pmηt) with t{1,2,,n}, an accumulated value of the given mF numbers with the help of mFYOWG operators is computed by

    mFDOWGΥ(~η1,~η2,,~ηn)=nt=1(˜ης(t))Υt=(1min(1,(nt=1(Υt(1(p1ης(t))2)σ))1σ),,1min(1,(nt=1(Υt(1(pmης(t))2)σ))1σ)). (2.19)

    Example 2.5. Let ˜η1=(0.4,0.6,0.2,0.3),˜η2=(0.4,0.7,0.2,0.7),˜η3=(0.5,0.1,0.6,0.9) and ˜η4=(0.3,0.9,0.6,0.4) be 4F numbers and Υ=(0.3,0.4,0.2,0.1)T be a weight-vector. Then, the score values of these 4F numbers for σ=5 is calculated as:

    S(˜η1)=0.4+0.6+0.2+0.34=0.375,S(˜η2)=0.4+0.7+0.2+0.74=0.45,S(˜η3)=0.5+0.1+0.6,0.94=0.525,S(˜η4)=0.3+0.9+0.6+0.44=0.55.

    Since, S(˜η4)>S(˜η3)>S(˜η2)>S(˜η1), thus

    ˜ης(1)=˜η4=(0.3,0.9,0.6,0.4),˜ης(2)=˜η3=(0.5,0.1,0.6,0.9),˜ης(3)=˜η2=(0.4,0.7,0.2,0.7),˜ης(4)=˜η1=(0.4,0.6,0.2,0.3).

    Then, from Definition 2.8,

    mFYOWGΥ(~η1,~η2,~η3,~η4)=4t=1(˜ης(t))Υt,=(1min(1,(4t=1(Υt(1(p1ης(t))2)σ))1σ),,1min(1,(4t=1(Υt(1(pmης(t))2)σ))1σ)),=(1min(1,(0.3×(1(0.3)2)5+0.4×(1(0.5)2)5+0.2×(1(0.4)2)5+0.1×(1(0.4)2)5)1/5),1min(1,(0.3×(1(0.9)2)5+0.4×(1(0.1)2)5+0.2×(1(0.7)2)5+0.1×(1(0.6)2)5)1/5),1min(1,(0.3×(1(0.6)2)5+0.4×(1(0.6)2)5+0.2×(1(0.2)2)5+0.1×(1(0.2)2)5)1/5),1min(1,(0.3×(1(0.4)2)5+0.4×(1(0.9)2)5+0.2×(1(0.7)2)5+0.1×(1(0.3)2)5)1/5)),=(0.4054,0.4102,0.4516,0.5282).

    Remark 2.2. The mFYOWG operators verify different basic laws such as monotonicity, idempotency and boundedness as given by Theorems 2.10–2.12.

    Theorem 2.14. (Commutativity Property) For every two arbitrary sets of mF numbers ˜ηt and ~ηt with t{1,2,,n}, if ˜ηt~ηt, then

    mFYOWGΥ(~η1,~η2,,~ηn)=mFYOWGΥ(~η1,~η2,,~ηn), (2.20)

    where ~ηt is any permutation of ~ηt.

    Proof. Its proof is obvious by Definition 2.8 and Theorem 2.13.

    From Definitions 2.4 and 2.5, we conclude that mFYWG and mFYOWG operators are useful to efficiently aggregate mF numbers. The only difference is that mFYWG operators only aggregate mF information without considering the ordering of mF numbers while mFYOWG operators consider their ordering. We now present another general type of AOs called mFYHG operators, which keep the features of mFYWG and mFYOWG operators.

    Definition 2.9. For a set of mF numbers ˜ηt=(p1ηt,p2ηt,,pmηt) with t{1,2,,n}, an mFYHG operator is given as:

    mFYHGΥ,Ω(~η1,~η2,,~ηn)=nt=1(˜˜ης(t))Υt, (2.21)

    where Υ=(Υ1,Υ2,,Υn)T denotes the weights associated with the mF numbers ~ηt, t=1,2,,n, Υt(0,1],nt=1Υt=1 and ˜˜ης(t) represents the j-th largest mF numbers such that ˜˜ης(t)=(nΩt)~ηt,(t=1,2,,n), Ω=(Ω1,Ω2,,Ωn) is a weight-vector with Ωt(0,1],nt=1Ωt=1.

    Notice that, when Υ=(1n,1n,,1n)T, the mFYHG operator becomes the mFYWG operator. When Ω=(1n,1n,,1n)T, the mFYHG operators convert into mFYOWG operators. Thus, mFYHG operators are an extension of mFYWG and mFYOWG operators.

    Theorem 2.15. For a set of mF numbers ˜ηt=(p1ηt,,pmηt) with t{1,2,,n}, an accumulated value of the given mF numbers with the help of mFYHG operator is given by

    mFYHGΥ,Ω(~η1,~η2,,~ηn)=nt=1(˜˜ης(t))Υt=(1min(1,(nt=1(Υt(1(p1˜˜ης(t))2)σ))1σ),,1min(1,(nt=1(Υt(1(pm˜˜ης(t))2)σ))1σ)). (2.22)

    Proof. It is similar to the proof of Theorem 2.2 via a mathematical induction method.

    Example 2.6. Let ˜η1=(0.7,0.9,0.8),˜η2=(0.6,0.5,0.7), ˜η3=(0.9,0.8,0.4) and ˜η4=(0.5,0.4,0.5) be 3F numbers, and Υ=(0.1,0.2,0.4,0.3)T be an associated weight-vector and Ω=(0.2,0.3,0.4,0.1)T be another weight-vector. Then, using Definition 2.9, for σ=4

    ˜˜η1=(1min(1,(nΩ1(1(p1η1)2)σ)1σ),,1min(1,(nΩ1(1(p3η1)2)σ)1σ)),=(1min(1,(4×0.2×(1(0.7)2)4)1/4),1min(1,(4×0.2×(1(0.9)2)4)1/4),1min(1,(4×0.2×(1(0.8)2)4)1/4)),=(0.7195,0.9057,0.8121).

    Similarly,

    ˜˜η2=(1min(1,(4×0.3×(1(0.6)2)4)1/4),1min(1,(4×0.3×(1(0.5)2)4)1/4),1min(1,(4×0.3×(1(0.7)2)4)1/4)),=(0.5746,0.4637,0.6828),
    ˜˜η3=(1min(1,(4×0.4×(1(0.9)2)4)1/4),1min(1,(4×0.4×(1(0.8)2)4)1/4),1min(1,(4×0.4×(1(0.4)2)4)1/4)),=(0.8867,0.7714,0.2351),

    and

    ˜˜η4=(1min(1,(4×0.1×(1(0.5)2)4)1/4),1min(1,(4×0.1×(1(0.4)2)4)1/4),1min(1,(4×0.1×(1(0.5)2)4)1/4)),=(0.6353,0.5762,0.6353).

    Now the scores of mF numbers for σ=3 are computed as below:

    S(˜˜η1)=0.7195+0.9057+0.81213=0.8124,S(˜˜η2)=0.5746+0.4637+0.68283=0.5737,S(˜˜η3)=0.8867+0.7714+0.23513=0.6311,S(˜˜η4)=0.6353+0.5762+0.63533=0.6156.

    Clearly, S(˜˜η1)>S(˜˜η3)>S(˜˜η4)>S(˜˜η2); thus,

    ˜˜ης(1)=˜η1=(0.7195,0.9057,0.8121),˜˜ης(2)=˜η3=(0.8867,0.7714,0.2351),˜˜ης(3)=˜η4=(0.6353,0.5762,0.6353),˜˜ης(4)=˜η2=(0.5746,0.4637,0.6828).

    Now by Definition 2.8, we get

    mFYHGΥ,Ω(~η1,~η2,~η3,~η4)=4t=1(˜ης(t))Υt=(1min(1,(4t=1(Υt(1(p1˜˜ης(t))2)σ))1σ),,1min(1,(4t=1(Υt(1(p3˜˜ης(t))2)σ))1σ)),=(1min(1,(0.1×(1(0.7195)2)4+0.2×(1(0.8867)2)4+0.4×(1(0.6353)2)4+0.3×(1(0.5746)2)4)1/4),1min(1,(0.1×(1(0.9057)2)4+0.2×(1(0.7714)2)4+0.4×(1(0.5762)2)4+0.3×(1(0.4637)2)4)1/4),1min(1,(0.1×(1(0.8121)2)4+0.2×(1(0.2351)2)4+0.4×(1(0.6353)2)4+0.3×(1(0.6828)2)4)1/4)),=(0.6445,0.5763,0.5507).

    In this section, we present an MCDM methodology based on our initiated mF Yager AOs to tackle different real-world MCDM situations involving mF information. The terms used for this purpose are provided in the following subsection.

    Let {S1,S2,,Sk} be a universe and {E1,E2,,En} be universal set of parameters. Let Υ={Υ1,Υ2,,Υn} be a weight vector with nt=1Υt=1,Υt(0,1], t{1,2,,n}. Suppose that an mF decision-matrix ˜M=(˜dit)k×n=(p1ηit,p2ηit,,pmηit)k×n, which contains the experts' opinions in the form of membership degrees.

    An algorithm is developed to tackle MCDM situations using mFYWA (or mFYWG) operators.

    Algorithm: Selection of an appropriate alternative under mF Yager AOs

    Step Ⅰ: Input:

    ˜M, an mF decision matrix containing n attributes and k alternatives.

    Υ=(Υ1,Υ2,,Υn)T, the weight vector.

    Step Ⅱ: Utilize the mFYWA operators in the aggregation process of the given datasets in an mF decision matrix ˜M and determine the preference values ˜dr; here, the variation of 'r' is from 1 to k for given mF numbers ηt.

    ˜dr=mFYWAΥ(˜ηr1,˜ηr2,,˜ηrn)=nt=1(Υt˜ηrt)=(min(1,(nt=1Υt(p1˜ηrt)2σ)1σ),,min(1,(nt=1Υt(pm˜ηrt)2σ)1σ)).

    When we use mFYWG operators, then

    ˜dr=mFYWGΥ(˜ηr1,˜ηr2,,˜ηrn)=nt=1(˜ηrt)Υt,=(1min(1,(nt=1(Υt(1(p1˜ηrt)2)σ))1σ),,1min(1,(nt=1(Υt(1(pm˜ηrt)2)σ))1σ)).

    Step Ⅲ: Determine the scores S(˜dr), where the variation of 'r' is from 1 to k.

    Step Ⅳ: Write all of the alternatives Sr,(r=1,2,,k) in order in terms of their score values S(˜dr). In the case when the final score values of two alternatives are equal, one can use the accuracy function to find their exact ranking.

    Output: An alternative with the highest score in the last step is the decision.

    An oil refinery or petroleum refinery is an industrial process plant where crude oil is processed and refined into more beneficial commodities like liquefied petroleum gas, kerosene, heating oil, asphalt base, petroleum naphtha, diesel fuel and gasoline. A petroleum refinery contains very sensitive and important substances, which is why making a suitable site selection is not an easy task due to the effects of different factors (parameters), including the availability of land, availability of raw water, resources/labor, effluent disposal, natural and geographic conditions of the site, conditions of society (humanities) and economy, conditions of traffic and transportation and conditions of utilities.

    The government of Pakistan wants to build a new oil refinery and for this very important project, the first significant thing is site selection because the cost of this project is directly proportional to the site. Therefore, this crucial assignment is given to a team of experts of this domain from the eight areas proposed by the government officials. The proposed alternatives are S1,S2,,S8. After few meetings among the experts, they all agreed to evaluate the alternatives under their expertise with the following five common parameters:

    E1 denotes the "Natural Conditions",

    E2 denotes the "Traffic and Transportation Conditions",

    E3 denotes the "Conditions of Utilities",

    E4 denotes the "Cost",

    E5 denotes the "Geographical Conditions".

    Some other sub characteristics of these parameters are provided below to better understand the construction of 3F numbers.

    ● The parameter "Natural Conditions" in the site selection procedure includes temperature, humidity and wind.

    ● The parameter "Traffic and Transportation Conditions" in the site selection process includes by road, railway and sea.

    ● The parameter "Conditions of Utilities" includes power supply, availability of raw water and resources/labor.

    ● The "Cost" includes medium, high and very high.

    ● The parameter "Geographical Conditions" affects the site selection, and it includes hydro geology, soil type and rock exposure.

    The final judgments of experts about the alternatives, as in terms of the favorable parameters, are presented in Table 2 in the form of 3F decision matrix.

    Table 2.  3F decision matrix.
    E1 E2 E3 E4 E5
    S1 (0.3, 0.5, 0.8) (0.6, 0.9, 0.5) (0.8, 0.5, 0.4) (0.7, 0.4, 0.2) (0.5, 0.7, 0.3)
    S2 (0.5, 0.7, 0.8) (0.4, 0.8, 0.7) (0.6, 0.4, 0.2) (0.7, 0.6, 0.9) (0.6, 0.4, 0.7)
    S3 (0.8, 0.4, 0.9) (0.4, 0.8, 0.4) (0.7, 0.8, 0.6) (0.3, 0.5, 0.7) (0.8, 0.6, 0.5)
    S4 (0.7, 0.4, 0.5) (0.9, 0.7, 0.6) (0.8, 0.6, 0.5) (0.7, 0.4, 0.6) (0.6, 0.8, 0.5)
    S5 (0.8, 0.5, 0.4) (0.7, 0.2, 0.5) (0.9, 0.4, 0.8) (0.7, 0.9, 0.7) (0.8, 0.3, 0.6)
    S6 (0.5, 0.7, 0.4) (0.6, 0.8, 0.7) (0.8, 0.4, 0.6) (0.1, 0.8, 0.9) (0.6, 0.7, 0.3)
    S7 (0.8, 0.3, 0.7) (0.8, 0.7, 0.3) (0.5, 0.9, 0.8) (0.7, 0.6, 0.5) (0.4, 0.5, 0.8)
    S8 (0.7, 0.6, 0.2) (0.5, 0.9, 0.1) (0.8, 0.2, 0.5) (0.6, 0.7, 0.9) (0.9, 0.7, 0.5)

     | Show Table
    DownLoad: CSV

    In view of the government officials, the team of experts assign a weight-vector to the set of parameters as follows:

    Υ1=0.24,Υ2=0.35,Υ3=0.10,Υ4=0.25andΥ5=0.06.

    Since 5t=1Υt=1. We now compute the most suitable ranking between the available sites for an oil refinery with the help of developed AOs, i.e., the: mFYWA and mFYWG operators:

    Step Ⅰ: When σ=4, by implementing the mFYWA operator, we compute the values ~di of the alternatives Si,i=1,2,,8 in terms of the ranking of sites for an oil refinery.

    ~d1=(0.6630,0.7925,0.6722),~d2=(0.6063,0.7256,0.8022),~d3=(0.6980,0.7265,0.7671),~d4=(0.8161,0.6510,0.5731),~d5=(0.7734,0.7577,0.6546),~d6=(0.6293,0.7656,0.7746),~d7=(0.7621,0.7145,0.6722),~d8=(0.7064,0.8028,0.7574).

    Step Ⅱ: Find the score values \mathfrak{S}(\tilde{\mathfrak{d}}_i) of 3 F numbers \tilde{\mathfrak{d}}_i, \; (i = 1, 2, \ldots, 5) of the alternatives \mathcal{S}_i :

    \begin{align*} \mathfrak{S}(\tilde{\mathfrak{d}}_1)& = 0.7092,\; \; \; \; \; \; \mathfrak{S}(\tilde{\mathfrak{d}}_2) = 0.7114,\; \; \; \; \; \; \mathfrak{S}(\tilde{\mathfrak{d}}_3) = 0.7305,\\ \mathfrak{S}(\tilde{\mathfrak{d}}_4)& = 0.6800,\; \; \; \; \; \; \mathfrak{S}(\tilde{\mathfrak{d}}_5) = 0.7286,\; \; \; \; \; \; \mathfrak{S}(\tilde{\mathfrak{d}}_6) = 0.7232,\\ \mathfrak{S}(\tilde{\mathfrak{d}}_7)& = 0.7162,\; \; \; \; \; \; \mathfrak{S}(\tilde{\mathfrak{d}}_8) = 0.7555. \end{align*}

    Step Ⅲ: Compute the ranking of alternatives using the scores obtained in the previous step: \mathcal{S}_8 > \mathcal{S}_3 > \mathcal{S}_5 > \mathcal{S}_6 > \mathcal{S}_7 > \mathcal{S}_2 > \mathcal{S}_1 > \mathcal{S}_4 . Step Ⅳ: The alternative \mathcal{S}_8 has the highest score; thus, it is the most suitable suitable site for the construction of an oil refinery in Pakistan.

    We now apply the m FYWG operator to compute a suitable option.

    Step Ⅰ: For \sigma = 4 , by using the m FYWG operator, we find the values \tilde{\mathfrak{d}_i} of the alternatives \mathcal{S}_i, \; i = 1, 2, \ldots, 8 in terms of the ranking of sites for an oil refinery.

    \begin{align*} \widehat{\widetilde{{{\mathfrak{d}}_{1}}}} = (0.5342,0.5501,0.4426),\; \; \; \; \; \; \; \tilde{\mathfrak{d}_2} = (0.5135,0.6199,0.6443),\\ \tilde{\mathfrak{d}_3} = (0.4762,0.5640,0.5564),\; \; \; \; \; \; \; \tilde{\mathfrak{d}_4} = (0.7338,0.5187,0.5564),\\ \tilde{\mathfrak{d}_5} = (0.7331,0.4177,0.5354)\; \; \; \; \; \; \; \; \tilde{\mathfrak{d}_6} = (0.4599,0.6840,0.5745),\\ \tilde{\mathfrak{d}_7} = (0.6744,0.5416,0.4873)\; \; \; \; \; \; \; \; \tilde{\mathfrak{d}_8} = (0.5977,0.6174,0.3510). \end{align*}

    Step Ⅱ: Find the score values \mathfrak{S}(\tilde{\mathfrak{d}}_i) of 3 F numbers \tilde{\mathfrak{d}}_i, \; (i = 1, 2, \ldots, 8) of the alternatives \mathcal{S}_i :

    \begin{align*} \mathfrak{S}(\tilde{\mathfrak{d}}_1)& = 0.5090,\; \; \mathfrak{S}(\tilde{\mathfrak{d}}_2) = 0.5926,\; \; \mathfrak{S}(\tilde{\mathfrak{d}}_3) = 0.5322,\\ \mathfrak{S}(\tilde{\mathfrak{d}}_4)& = 0.6030,\; \; \mathfrak{S}(\tilde{\mathfrak{d}}_5) = 0.5621,\; \; \mathfrak{S}(\tilde{\mathfrak{d}}_6) = 0.5728,\\ \mathfrak{S}(\tilde{\mathfrak{d}}_7)& = 0.5678,\; \; \mathfrak{S}(\tilde{\mathfrak{d}}_8) = 0.5220. \end{align*}

    Step Ⅲ: Compute the ranking of alternatives using the scores \mathfrak{S}(\tilde{\mathfrak{d}}_i), \; (i = 1, 2, \ldots, 8) determined in previous step: \mathcal{S}_4 > \mathcal{S}_2 > \mathcal{S}_6 > \mathcal{S}_7 > \mathcal{S}_5 > \mathcal{S}_3 > \mathcal{S}_8 > \mathcal{S}_1 .

    Step Ⅳ: The alternative \mathcal{S}_4 has the highest score; thus, it is the most suitable option for the construction of an oil refinery in Pakistan.

    The method used to solve the above MCDM application is displayed in Figure 1.

    Figure 1.  Flowchart diagram.

    In this section, we give both qualitative and quantitative comparative analyses of the initiated m F Yager AOs with m F Dombi AOs [26], m F Hamacher AOs [24] and some Yager's operation-based AOs to prove their cogency and efficiency. Further, we discuss the validity of the proposed AOs through the use of three effectiveness tests which have been introduced by Wang and Triantaphyllou [45].

    To effectively deal with m F information, the existing Yager's operation-based AOs, including Fermatean fuzzy Yager AOs [28], complex Pythagorean fuzzy Yager AOs [30] and q -rung picture fuzzy Yager AOs [29] are not useful; therefore, m F Yager AOs have been proposed in this study. In the literature, m F information is aggregated via Dombi and Hamacher TNs and TCoNs with very hard calculations. Yager's operations are simpler than Dombi and Hamacher TNs and TCoNs. This is another reason that has motivated us to select Yager's TN and TCoN in the current work.

    We now discuss the comparison between the results of initiated m F Yager AOs and existing m F Dombi AOs [26] and m F Hamacher AOs [24]. For this, we applied these AOs to a daily-life scenario, and the computed results are provided in Tables 3 and 4 (for more detail see Figure 2). From Tables 3 and 4, we can easily see that the optimal object (i.e., \mathcal{S}_2 ) is the same by applying m FHWA and m FHWG [24] operators but it is not similar to the optimal object " \mathcal{S}_4 " which is obtained by applying the proposed m FYWG operator. Besides, the optimal object (i.e., \mathcal{S}_8 ) is the same by applying existing m FDWA and m FDWG [26] operators and the proposed m FYWA operator. Thus, to deal with m F MCDM situations effectively, our proposed m F Yager AOs are much more versatile and generalized than certain existing MCDM tools, including m FDWA and m FDWG [26] operators, and m FHWA and m FHWG [24] AOs.

    Table 3.  Comparison of m F Yager AOs with m F Dombi AOs [26] and m F Hamacher AOs [24].
    Operators \mathfrak{S}(\tilde{\mathfrak{d}}_1) \mathfrak{S}(\tilde{\mathfrak{d}}_2) \mathfrak{S}(\tilde{\mathfrak{d}}_3) \mathfrak{S}(\tilde{\mathfrak{d}}_4) \mathfrak{S}(\tilde{\mathfrak{d}}_5) \mathfrak{S}(\tilde{\mathfrak{d}}_6) \mathfrak{S}(\tilde{\mathfrak{d}}_7) \mathfrak{S}(\tilde{\mathfrak{d}}_8)
    m FHWA [24] 0.5892 0.6589 0.6214 0.6402 0.6231 0.6385 0.6328 0.6123
    m FHWG [24] 0.5508 0.6361 0.5887 0.6300 0.5895 0.5957 0.6066 0.5479
    m FDWA [26] 0.7731 0.7517 0.7930 0.7149 0.8035 0.7808 0.7803 0.8506
    m FDWG [26] 0.6306 0.5761 0.5661 0.4361 0.5205 0.6307 0.5700 0.6704
    Proposed m FYWA 0.7092 0.7114 0.7305 0.6800 0.7286 0.7232 0.7162 0.7555
    Proposed m FYWG 0.5090 0.5926 0.5322 0.6030 0.5621 0.5728 0.5678 0.5220

     | Show Table
    DownLoad: CSV
    Table 4.  Comparison between the ranking results of m F Yager AOs and m F Dombi AOs [26] and m F Hamacher AOs [24].
    Operators Ranking Order Best Option
    m FHWA [24] \mathcal{S}_{2} > \mathcal{S}_{4} > \mathcal{S}_{6} > \mathcal{S}_7 > \mathcal{S}_{5} > \mathcal{S}_{3} > \mathcal{S}_{8} > \mathcal{S}_{1} \mathcal{S}_{2}
    m FHWG [24] \mathcal{S}_{2} > \mathcal{S}_{4} > \mathcal{S}_{7} > \mathcal{S}_6 > \mathcal{S}_{5} > \mathcal{S}_{3} > \mathcal{S}_{1} > \mathcal{S}_{1} \mathcal{S}_{2}
    m FDWA [26] \mathcal{S}_{8} > \mathcal{S}_{5} > \mathcal{S}_{3} > \mathcal{S}_6 > \mathcal{S}_{7} > \mathcal{S}_{1} > \mathcal{S}_{2} > \mathcal{S}_{4} \mathcal{S}_{8}
    m FDWG [26] \mathcal{S}_{8} > \mathcal{S}_{6} > \mathcal{S}_{1} > \mathcal{S}_2 > \mathcal{S}_{7} > \mathcal{S}_{3} > \mathcal{S}_{5} > \mathcal{S}_{4} \mathcal{S}_{8}
    Proposed m FYWA \mathcal{S}_{8} > \mathcal{S}_{3} > \mathcal{S}_{5} > \mathcal{S}_6 > \mathcal{S}_{7} > \mathcal{S}_{2} > \mathcal{S}_{1} > \mathcal{S}_{4} \mathcal{S}_{8}
    Proposed m FYWG \mathcal{S}_{4} > \mathcal{S}_{2} > \mathcal{S}_{6} > \mathcal{S}_7 > \mathcal{S}_{5} > \mathcal{S}_{3} > \mathcal{S}_{8} > \mathcal{S}_{1} \mathcal{S}_{4}

     | Show Table
    DownLoad: CSV
    Figure 2.  Comparison of m F Yager AOs with existing m F Dombi and Hamacher AOs on the application in Section 3.2.

    In the following, the feasibility and productiveness of the proposed algorithm based on m FYWA and m FYWG operators is justified via three tests criteria, which have been introduced by Wang and Triantaphyllou [45] to check the validity of MCDM methods.

    Test-Ⅰ: When the belongingness degrees of a sub-optimal alternative are replaced with worse belongingness degrees without changing the criteria, then the decision object should be invariant.

    Test-Ⅱ: The MCDM method should verify the transitive law.

    Test-Ⅲ: If a given problem is resolved into different small portions by removing alternatives and the same MCDM approach is applied, then the ranking of alternatives should be the same as the original.

    Now, we discuss the effectiveness of the proposed MCDM method under the conditions of m F Yager AOs by means of the above validity tests.

    1) Validity checking by Test Ⅰ: The proposed MCDM approach with Yager AOs verifies this test because when we replace the belongingness degrees of the object \mathcal{S}_1 with \mathcal{S}_1^{'} and the object \mathcal{S}_4 with \mathcal{S}_4^{'} in Table 2 (i.e., 3F decision matrix), then by applying the developed m FYWA operator to the new decision matrix, which is provided by Table 5, the scores of the alternatives \mathcal{S}_1 and \mathcal{S}_6 are \mathfrak{S}(\mathfrak{d}_1) = 0.4383 and \mathfrak{S}(\mathfrak{d}_6) = 0.4858 , respectively. Clearly, \mathcal{S}_8 is again the best alternative, which is the same as the original decision object. In a similar manner, if we apply the m FYWG operator, the scores of the objects \mathcal{S}_1 and \mathcal{S}_6 are \mathfrak{S}(\mathfrak{d}_1) = 0.3158 and \mathfrak{S}(\mathfrak{d}_6) = 0.4051 , respectively. Clearly, \mathcal{S}_4 is the best alternative which is the same as the original. Thus, the optimal alternatives were the same as that of the original ranking when we changed the sub-optimal alternatives belongingness values. Thus, the developed algorithm is reliable under the validity test criterion Ⅰ.

    Table 5.  3 F decision matrix.
    \mathcal{E}_1 \mathcal{E}_2 \mathcal{E}_3 \mathcal{E}_4 \mathcal{E}_5
    \mathcal{S}_1^{'} (0.1, 0.3, 0.5) (0.4, 0.6, 0.2) (0.3, 0.4, 0.1) (0.1, 0.3, 0.1) (0.4, 0.6, 0.2)
    \mathcal{S}_2 (0.5, 0.7, 0.8) (0.4, 0.8, 0.7) (0.6, 0.4, 0.2) (0.7, 0.6, 0.9) (0.6, 0.4, 0.7)
    \mathcal{S}_3 (0.8, 0.4, 0.9) (0.4, 0.8, 0.4) (0.7, 0.8, 0.6) (0.3, 0.5, 0.7) (0.8, 0.6, 0.5)
    \mathcal{S}_4 (0.6, 0.3, 0.2) (0.7, 0.5, 0.3) (0.4, 0.2, 0.4) (0.5, 0.2, 0.4) (0.2, 0.6, 0.3)
    \mathcal{S}_5 (0.8, 0.5, 0.4) (0.7, 0.2, 0.5) (0.9, 0.4, 0.8) (0.7, 0.9, 0.7) (0.8, 0.3, 0.6)
    \mathcal{S}_6^{'} (0.6, 0.3, 0.2) (0.7, 0.5, 0.3) (0.4, 0.2, 0.4) (0.5, 0.2, 0.4) (0.2, 0.6, 0.3)
    \mathcal{S}_7 (0.8, 0.3, 0.7) (0.8, 0.7, 0.3) (0.5, 0.9, 0.8) (0.7, 0.6, 0.5) (0.4, 0.5, 0.8)
    \mathcal{S}_8 (0.7, 0.6, 0.2) (0.5, 0.9, 0.1) (0.8, 0.2, 0.5) (0.6, 0.7, 0.9) (0.9, 0.7, 0.5)

     | Show Table
    DownLoad: CSV

    2) Validity checking by Tests Ⅱ and Ⅲ: These test criteria also hold for our proposed MCDM approach with m F Yager AOs because when we remove some objects in the developed application (Section 3) and apply the developed m FYWA and m FYWG operators, we obtain similar ranking orders between the alternatives and the original. This is why, the overall ranking order of the alternatives will not be changed and the transitive property holds. Thus, the proposed algorithm is reliable under the validity checking tests Ⅱ and Ⅲ.

    Nowadays, due to the existence of multipolar data and multiple attributes in several real-world problems, the fuzzification of multipolar information with AOs is emerging as a very popular mathematical topic for the unification of various inputs into a single useful output because traditional MCDM approaches fail to deal with complex decision-making problems. With the motivation to remove these issues of existing MCDM methods, and we have integrated m F numbers with Yager's TN and TCoN operations, and have presented some new Yager AOs in an m F environment, namely, m FYWA, m FYOWA, m FYHA, m FYWG, m FYOWG and m FYHG operators, which are respectively explained with illustrative numerical examples. Further, we have applied different results of the proposed AOs. To prove the feasibility and reliability of the developed m F AOs, we have implemented them to a daily-life problem, that is, the selection of an appropriate site for the construction of an oil refinery. Subsequently, we performed a comparative analysis of the initiated m F Yager AOs with existing m F Dombi [26] and m F Hamacher AOs [24]. From the comparative analysis (Tables 3 and 4), we have clearly observed that the optimal object (that is, \mathcal{S}_8 ) is the same by applying m FDWA and m FDWG [26] operators and the proposed m FYWA operator. On the other hand, the optimal object (that is, \mathcal{S}_2 ) is the same by applying m FHWA and m FHWG [24] operators, but it is not similar to the optimal object " \mathcal{S}_4 " which is obtained by applying the proposed m FYWG operator. In the end, we have verified the effectiveness of the developed MCDM method by applying validity tests that were presented by Wang and Triantaphyllou [45].

    The literature analysis revealed that the existing AOs have both pros and cons. Because of this, we noticed that our initiated AOs also have some limitations. The developed m F Yager AOs are not useful in the case of multipolar information from opposite sources because they only deal with multi-valued membership-based information. It may not be easy to compute final ranking results in the case of a big number of attributes without using mathematical software, including MAPAL, MATLAB, Mathematica, etc.

    In the future, our presented work can be extended to the following

    \bullet m F Yager prioritized AOs,

    \bullet m F soft Yager AOs,

    \bullet Hesitant m F Yager AOs,

    \bullet m F Yager Bonferroni mean operators,

    \bullet Possibility m F Yager AOs,

    \bullet Rough m F Yager AOs.

    The third author extends his appreciation to the Deanship of Scientific Research at King Khalid University for funding this study through the General Research Project under grant number R.G.P.2/48/43.

    The authors declare no conflict of interest.

    Proof. By utilizing the mathematical induction method we can easily prove it.

    1). By putting n = 1 in Eq (2.4), we get

    \begin{align*} mF&DWA_\Upsilon(\tilde {\eta_1},,\tilde {\eta_2},\ldots,\tilde {\eta_n}) = \Upsilon_1{\tilde \eta_1} = {\tilde \eta_1}, \; ({\rm since} \; \Upsilon_1 = 1)\\ & = \Bigg(\sqrt{\min\bigg(1,\big((\mathfrak{p}_1\circ \eta_1)^{2\sigma}\big)^{\frac{1}{\sigma}}\bigg)},\ldots,\sqrt{\min\bigg(1,\big((\mathfrak{p}_m\circ \eta_1)^{2\sigma}\big)^{\frac{1}{\sigma}}\bigg)}\; \Bigg). \end{align*}

    Thus, Eq (2.4) holds for n = 1 .

    2). Now let us suppose that Eq (2.4) holds when n = r , where r\in \mathbb{N} (set of natural numbers); then,

    \begin{align} mFY&WA_\Upsilon(\tilde {\eta_1},\tilde {\eta_2},\ldots,\tilde {\eta_r}) = \bigoplus_{t = 1}^r (\Upsilon_t{\tilde \eta_t}),\\ & = \Bigg(\sqrt{\min\bigg(1,\big(\sum\limits_{t = 1}^{r}\Upsilon_t(\mathfrak{p}_1\circ \eta_t)^{2\sigma}\big)^{\frac{1}{\sigma}}\bigg)},\ldots,\sqrt{\min\bigg(1,\big(\sum\limits_{t = 1}^{r}\Upsilon_t(\mathfrak{p}_m\circ \eta_t)^{2\sigma}\big)^{\frac{1}{\sigma}}\bigg)}\; \Bigg). \end{align} (A.1)

    For n = r+1 ,

    \begin{array}{l} mFY\;WA_\Upsilon(\tilde {\eta_1},\tilde {\eta_2},\ldots,\tilde {\eta_r},{\tilde \eta}_{r+1}) = \bigoplus_{t = 1}^r (\Upsilon_t{\tilde \eta_t})\oplus (\Upsilon_{r+1}{\tilde \eta}_{r+1}),\\ = \Bigg(\sqrt{\min\bigg(1,\big(\sum\limits_{t = 1}^{r}\Upsilon_t(\mathfrak{p}_1\circ \eta_t)^{2\sigma}\big)^{\frac{1}{\sigma}}\bigg)},\ldots,\sqrt{\min\bigg(1,\big(\sum\limits_{t = 1}^{r}\Upsilon_t(\mathfrak{p}_m\circ \eta_t)^{2\sigma}\big)^{\frac{1}{\sigma}}\bigg)}\; \Bigg)\oplus\\ \Bigg(\sqrt{\min\bigg(1,\big(\Upsilon_{r+1}(\mathfrak{p}_1\circ \eta_{r+1})^{2\sigma}\big)^{\frac{1}{\sigma}}\bigg)},\ldots,\sqrt{\min\bigg(1,\big(\Upsilon_{r+1}(\mathfrak{p}_m\circ \eta_{r+1})^{2\sigma}\big)^{\frac{1}{\sigma}}\bigg)}\; \Bigg)\\ = \Bigg(\sqrt{\min\bigg(1,\big(\sum\limits_{t = 1}^{r+1}\Upsilon_t(\mathfrak{p}_1\circ \eta_t)^{2\sigma}\big)^{\frac{1}{\sigma}}\bigg)},\ldots,\sqrt{\min\bigg(1,\big(\sum\limits_{t = 1}^{r+1}\Upsilon_t(\mathfrak{p}_m\circ \eta_t)^{2\sigma}\big)^{\frac{1}{\sigma}}\bigg)}\; \Bigg). \end{array}

    Thus, Eq (2.4) holds for n = r+1 . Consequently, Eq (2.4) verifies for all n\in \mathbb{N} .

    Proof. Since \tilde \eta_t = (\mathfrak{p}_1\circ \eta_t, \ldots, \mathfrak{p}_m\circ \eta_t) = \tilde \eta, where t = 1, \ldots, n . Then, by Eq (2.4),

    \begin{align*} mFY&WA_\Upsilon(\tilde {\eta_1},\tilde {\eta_2},\ldots,\tilde {\eta_n}) = \bigoplus_{t = 1}^n (\Upsilon_t{\tilde \eta_t}),\\ & = \Bigg(\sqrt{\min\bigg(1,\big(\sum\limits_{t = 1}^{n}\Upsilon_t(\mathfrak{p}_1\circ \eta_t)^{2\sigma}\big)^{\frac{1}{\sigma}}\bigg)},\ldots,\sqrt{\min\bigg(1,\big(\sum\limits_{t = 1}^{n}\Upsilon_t(\mathfrak{p}_m\circ \eta_t)^{2\sigma}\big)^{\frac{1}{\sigma}}\bigg)}\; \Bigg),\\ & = \Bigg(\sqrt{\min\bigg(1,\big((\mathfrak{p}_1\circ \eta)^{2\sigma}\big)^{\frac{1}{\sigma}}\bigg)},\ldots,\sqrt{\min\bigg(1,\big((\mathfrak{p}_m\circ \eta)^{2\sigma}\big)^{\frac{1}{\sigma}}\bigg)}\; \Bigg),\\ & = (\mathfrak{p}_1\circ \eta,\ldots, \mathfrak{p}_m\circ \eta),\; {\rm for }\; \sigma = 1\\ & = \tilde \eta. \end{align*}

    Hence, mFYWA_\Upsilon(\tilde {\eta_1}, \tilde {\eta_2}, \ldots, \tilde {\eta_n}) = \tilde \eta holds if \tilde \eta_t = \tilde \eta, {when ` t ' varies from 1 to n }.



    [1] X. Shao, H. Fan, G. Lu, J. Xu, An improved infrared dim and small target detection algorithm based on the contrast mechanism of human visual system, Infrared Phys. Technol., 55 (2012), 403–408. https://doi.org/10.1016/j.infrared.2012.06.001 doi: 10.1016/j.infrared.2012.06.001
    [2] M. Fan, S. Tian, K. Liu, J. Zhao, Y. Li, Infrared small target detection based on region proposal and CNN classifier, Signal, Image Video Process., 15 (2021), 1927–1936. https://doi.org/10.1007/s11760-021-01936-z doi: 10.1007/s11760-021-01936-z
    [3] Y. Zhang, L. Zheng, Y. Zhang, Small infrared target detection via a Mexican-Hat distribution, Appl. Sci., 9 (2019). https://doi.org/10.3390/app9245570 doi: 10.3390/app9245570
    [4] X. Cao, C. Rong, X. Bai, Infrared small target detection based on derivative dissimilarity measure, IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens., 12 (2019), 3101–3116. https://doi.org/10.1109/JSTARS.2019.2920327 doi: 10.1109/JSTARS.2019.2920327
    [5] C. Chen, H. Li, Y. Wei, T. Xia, Y. Tang, A local contrast method for small infrared target detection, IEEE Trans. Geosci. Remote Sens., 52 (2014), 574–581. https://doi.org/10.1109/TGRS.2013.2242477 doi: 10.1109/TGRS.2013.2242477
    [6] J. Han, Y. Ma, B. Zhou, F. Fan, K. Liang, Y. Fang, A robust infrared small target detection algorithm based on human visual system, IEEE Geosci. Remote Sens. Lett., 11 (2014), 2168–2172. https://doi.org/10.1109/LGRS.2014.2323236 doi: 10.1109/LGRS.2014.2323236
    [7] Y. Qin, B. Li, Effective infrared small target detection utilizing a novel local contrast method, IEEE Geosci. Remote Sens. Lett., 13 (2016), 1890–1894. https://doi.org/10.1109/LGRS.2016.2616416 doi: 10.1109/LGRS.2016.2616416
    [8] L. Wu, Y. Ma, F. Fan, M. Wu, J. Huang, A double-neighborhood gradient method for infrared small target detection, IEEE Geosci. Remote Sens. Lett., 18 (2021), 1476–1480. https://doi.org/10.1109/LGRS.2020.3003267 doi: 10.1109/LGRS.2020.3003267
    [9] Y. Lu, L. Dong, T. Zhang, W. Xu, A robust detection algorithm for infrared maritime small and dim targets, Sensors, 20 (2020). https://doi.org/10.3390/s20041237 doi: 10.3390/s20041237
    [10] X. Wang, Z. Peng, D. Kong, Y. He, Infrared dim and small target detection based on stable multisubspace learning in heterogeneous scene, IEEE Trans. Geosci. Remote Sens., 55 (2017), 5481–5493. https://doi.org/10.1109/TGRS.2017.2709250 doi: 10.1109/TGRS.2017.2709250
    [11] L. Dong, B. Wang, M. Zhao, W. Xu, Robust infrared maritime target detection based on visual attention and spatiotemporal filtering, IEEE Trans. Geosci. Remote Sens., 55 (2017), 3037–3050. https://doi.org/10.1109/TGRS.2017.2660879 doi: 10.1109/TGRS.2017.2660879
    [12] L. Ding, X. Xu, Y. Cao, G. Zhai, F. Yang, L. Qian, Detection and tracking of infrared small target by jointly using SSD and pipeline filter, Digital Signal Process., 110 (2021). https://doi.org/10.1016/j.dsp.2020.102949 doi: 10.1016/j.dsp.2020.102949
    [13] Q. Guo, Z. Li, W. Song, W. Fu, Parallel computing based dynamic programming algorithm of track-before-detect, Symmetry, 11 (2019). https://doi.org/10.3390/sym11010029 doi: 10.3390/sym11010029
    [14] Q. Hu, J. Yang, P. Qin, S. Fong, Towards a context-free machine universal grammar (CF-MUG) in natural language processing, IEEE Access, 8 (2020). https://doi.org/10.1109/ACCESS.2020.3022674 doi: 10.1109/ACCESS.2020.3022674
    [15] A. Ilioudi, A. Dabiri, B. J. Wolf, B. De Schutter, Deep learning for object detection and segmentation in videos: toward an integration with domain knowledge, IEEE Access, 10 (2022), 34562–34576. https://doi.org/10.1109/ACCESS.2022.3162827 doi: 10.1109/ACCESS.2022.3162827
    [16] R. Girshick, J. Donahue, T. Darrell, J. Malik, Rich feature hierarchies for accurate object detection and semantic segmentation, in 2014 IEEE Conference on Computer Vision and Pattern Recognition, (2014), 580–587. https://doi.org/10.1109/CVPR.2014.81
    [17] R. Girshick, Fast R-CNN, in 2015 IEEE International Conference on Computer Vision (ICCV), (2015), 7–13. https://doi.org/10.1109/ICCV.2015.169
    [18] S. Ren, K. He, R. Girshick, J. Sun, Faster R-CNN: Towards real-time object detection with region proposal networks, IEEE Trans. Pattern Anal. Mach. Intell., 39 (2017), 1137–1149. https://doi.org/10.1109/TPAMI.2016.2577031 doi: 10.1109/TPAMI.2016.2577031
    [19] W. Liu, A. Dragomir, E. Dumitru, S. Christian, R. Scott, C. Fu, et al., SSD: Single shot multibox detector, in Lecture Notes in Computer Science, (2016), 21–37. https://doi.org/10.1007/978-3-319-46448-0_2
    [20] J. Redmon, A. Farhadi, YOLOV3: An incremental improvement, arXiv preprints, (2018), arXiv: 1804.02767. https://doi.org/10.48550/arXiv.1804.02767
    [21] T. Y. Lin, P. Goyal, R. Girshick, K. He, P. Dollár, Focal loss for dense object detection, in 2017 IEEE International Conference on Computer Vision (ICCV), (2017), 2999–3007. https://doi.org/10.1109/ICCV.2017.324
    [22] M. Ju, J. Luo, P. Zhang, M. He, H. Luo, A simple and efficient network for small target detection, IEEE Access, 7 (2019), 85771–85781. https://doi.org/10.1109/ACCESS.2019.2924960 doi: 10.1109/ACCESS.2019.2924960
    [23] H. Liang, J. Yang, M. Shao, FE-RetinaNet: small target detection with parallel multi-scale feature enhancement, Symmetry, 13 (2021). https://doi.org/10.3390/sym13060950 doi: 10.3390/sym13060950
    [24] Q. Hou, Z. Wang, F. Tan, Y. Zhao, H. Zheng, W. Zhang, RISTDnet: Robust infrared small target detection network, IEEE Geosci. Remote Sens. Lett., 19 (2021), 1–5. https://doi.org/10.1109/LGRS.2021.3050828 doi: 10.1109/LGRS.2021.3050828
    [25] Y. Li, S. Li, H. Du, L. Chen, D. Zhang, Y. Li. YOLO-ACN: Focusing on small target and occluded object detection, IEEE Access, 8 (2020), 227288–227303. https://doi.org/10.1109/ACCESS.2020.3046515 doi: 10.1109/ACCESS.2020.3046515
    [26] Z. Song, Y. Zhang, Y. Liu, K. Yang, M. Sun, MSFYOLO: Feature fusion-based detection for small objects, IEEE Lat. Am. Trans., 20 (2022), 823–830. https://doi.org/10.1109/TLA.2022.9693567 doi: 10.1109/TLA.2022.9693567
    [27] K. Wang, S. Li, S. Niu, K. Zhang, Detection of infrared small targets using feature fusion convolutional network, IEEE Access, 7 (2019), 146081–146092. https://doi.org/10.1109/ACCESS.2019.2944661 doi: 10.1109/ACCESS.2019.2944661
    [28] M. Ju, J. Luo, G. Liu, H. Luo, ISTDet: An efficient end-to-end neural network for infrared small target detection, Infrared Phys. Technol., 114 (2021). https://doi.org/10.1016/j.infrared.2021.103659 doi: 10.1016/j.infrared.2021.103659
    [29] S. Yao, Q. Zhu, T. Zhang, W. Cui, P. Yan, Infrared image small-target detection based on improved FCOS and spatio-temporal features, Electronics, 11 (2022). https://doi.org/10.3390/electronics11060933 doi: 10.3390/electronics11060933
    [30] X. Tong, B. Sun, J. Wei, Z. Zuo, S. Su, EAAU-Net: Enhanced asymmetric attention U-Net for infrared small target detection, Remote Sens., 13 (2021). https://doi.org/10.3390/rs13163200 doi: 10.3390/rs13163200
    [31] Q. Xu, H. Huang, C. Zhou, X. Zhang, Research on real-time infrared image fault detection of substation high-voltage lead connectors based on improved YOLOv3 network, Electronics, 10 (2021). https://doi.org/10.3390/electronics10050544 doi: 10.3390/electronics10050544
    [32] C. Deng, M. Wang, L. Liu, Y. Liu, Y. Jiang, Extended feature pyramid network for small object detection, IEEE Trans. Multimedia, 24 (2021), 1968–1979. https://doi.org/10.1109/TMM.2021.3074273 doi: 10.1109/TMM.2021.3074273
    [33] G. Huang, Z. Liu, L. Van Der Maaten, K. Q. Weinberger, Densely connected convolutional networks, in 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), (2017), 2261–2269. https://doi.org/10.1109/CVPR.2017.243
    [34] Y. Dai, Y. Wu, F. Zhou, K. Barnard, Asymmetric contextual modulation for infrared small target detection, in 2021 IEEE Winter Conference on Applications of Computer Vision (WACV), (2021), 949–958. https://doi.org/10.1109/WACV48630.2021.00099
    [35] Q. Wang, B. Wu, P. Zhu, P. Li, W. Zuo, Q. Hu, ECA-Net: Efficient channel attention for deep convolutional neural networks, in 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), (2020), 11531–11539. https://doi.org/10.1109/CVPR42600.2020.01155
    [36] M. Pastorino, G. Moser, S. B. Serpico, J. Zerubia, Fully convolutional and feedforward networks for the semantic segmentation of remotely sensed images, in 2022 IEEE International Conference on Image Processing (ICIP), (2022), 1876–1880. https://doi.org/10.1109/ICIP46576.2022.9897336
    [37] M. Lin, Q. Chen, S. Yan, Network in network, arXiv preprints, (2013), arXiv: 1312.4400. https://doi.org/10.48550/arXiv.1312.4400
    [38] B. Hui, Z. Song, H. Fan, P. Zhong, W. Hu, X. Zhang, et al., A dataset for infrared image dim-small aircraft target detection and tracking under ground/air background, Science Data Bank, (2019). https://doi.org/10.11922/sciencedb
    [39] X. Zhou, D. Wang, P. Krähenbähl, Objects as points, arXiv preprint, (2019), arXiv: 1904.07850. http://arXiv.org/abs/1904.07850
  • This article has been cited by:

    1. Tahir Mahmood, Zeeshan Ali, Multi-attribute decision-making methods based on Aczel–Alsina power aggregation operators for managing complex intuitionistic fuzzy sets, 2023, 42, 2238-3603, 10.1007/s40314-023-02204-1
    2. Mohammed M. Ali Al-Shamiri, Adeel Farooq, Muhammad Nabeel, Ghous Ali, Dragan Pamučar, Integrating TOPSIS and ELECTRE-Ⅰ methods with cubic m -polar fuzzy sets and its application to the diagnosis of psychiatric disorders, 2023, 8, 2473-6988, 11875, 10.3934/math.2023601
    3. Lin Sha, Yabin Shao, Fermatean Hesitant Fuzzy Choquet Integral Aggregation Operators, 2023, 11, 2169-3536, 38548, 10.1109/ACCESS.2023.3267512
    4. Zia Ur Rahman, Ghous Ali, Muhammad Asif, Yufeng Chen, Muhammad Zain Ul Abidin, Identification of desalination and wind power plants sites using m-polar fuzzy Aczel–Alsina aggregation information, 2024, 14, 2045-2322, 10.1038/s41598-023-50397-6
    5. Ghous Ali, Kholood Alsager, Asad Ali, Novel linguistic q -rung orthopair fuzzy Aczel-Alsina aggregation operators for group decision-making with applications, 2024, 9, 2473-6988, 32328, 10.3934/math.20241551
    6. Muhammad Saeed Raja, Khizar Hayat, Adeeba Munshi, Tahir Mahmood, Rawish Sheraz, Iqra Matloob, Aggregation operators on group-based generalized q-rung orthopair fuzzy N-soft sets and applications in solar panel evaluation, 2024, 10, 24058440, e27323, 10.1016/j.heliyon.2024.e27323
    7. Inayat Ullah, Muhammad Akram, Tofigh Allahviranloo, Multi-criteria decision making with Hamacher aggregation operators based on multi-polar fuzzy Z-numbers, 2025, 694, 00200255, 121707, 10.1016/j.ins.2024.121707
    8. Ghous Ali, Kholood Alsager, Novel Heronian mean based m -polar fuzzy power geometric aggregation operators and their application to urban transportation management, 2024, 9, 2473-6988, 34109, 10.3934/math.20241626
    9. Ghous Ali, Muhammad Anwar, Bander Almutairi, Muhammad Faheem, Sabeeha Kanwal, Novel Wind Power Station Site Selection Framework Based on Multipolar Fuzzy Schweizer-Sklar Aggregation Operators, 2024, 12, 2169-3536, 194030, 10.1109/ACCESS.2024.3516825
    10. Inayat Ullah, Muhammad Akram, Tofigh Allahviranloo, An outranking method with Dombi aggregation operators based on multi-polar fuzzy Z-numbers for selection of best rehabilitation center, 2025, 2452414X, 100781, 10.1016/j.jii.2025.100781
  • Reader Comments
  • © 2023 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(2595) PDF downloads(127) Cited by(2)

Figures and Tables

Figures(14)  /  Tables(4)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog