Research article Special Issues

Optimized YOLOv7-tiny model for smoke detection in power transmission lines


  • Received: 08 August 2023 Revised: 09 September 2023 Accepted: 17 September 2023 Published: 17 October 2023
  • Fire incidents near power transmission lines pose significant safety hazards to the regular operation of the power system. Therefore, achieving fast and accurate smoke detection around power transmission lines is crucial. Due to the complexity and variability of smoke scenarios, existing smoke detection models suffer from low detection accuracy and slow detection speed. This paper proposes an improved model for smoke detection in high-voltage power transmission lines based on the improved YOLOv7-tiny. First, we construct a dataset for smoke detection in high-voltage power transmission lines. Due to the limited number of real samples, we employ a particle system to randomly generate smoke and composite it into randomly selected real scenes, effectively expanding the dataset with high quality. Next, we introduce multiple parameter-free attention modules into the YOLOv7-tiny model and replace regular convolutions in the Neck of the model with Spd-Conv (Space-to-depth Conv) to improve detection accuracy and speed. Finally, we utilize the synthesized smoke dataset as the source domain for model transfer learning. We pre-train the improved model and fine-tune it on a dataset consisting of real scenarios. Experimental results demonstrate that the proposed improved YOLOv7-tiny model achieves a 2.61% increase in mean Average Precision (mAP) for smoke detection on power transmission lines compared to the original model. The precision is improved by 2.26%, and the recall is improved by 7.25%. Compared to other object detection models, the smoke detection proposed in this paper achieves high detection accuracy and speed. Our model also improved detection accuracy on the already publicly available wildfire smoke dataset Figlib (Fire Ignition Library).

    Citation: Chen Chen, Guowu Yuan, Hao Zhou, Yutang Ma, Yi Ma. Optimized YOLOv7-tiny model for smoke detection in power transmission lines[J]. Mathematical Biosciences and Engineering, 2023, 20(11): 19300-19319. doi: 10.3934/mbe.2023853

    Related Papers:

    [1] Yi Tian . Approximate solution of initial boundary value problems for ordinary differential equations with fractal derivative. Mathematical Modelling and Control, 2022, 2(2): 75-80. doi: 10.3934/mmc.2022009
    [2] Abduljawad Anwar, Shayma Adil Murad . On the Ulam stability and existence of Lp-solutions for fractional differential and integro-differential equations with Caputo-Hadamard derivative. Mathematical Modelling and Control, 2024, 4(4): 439-458. doi: 10.3934/mmc.2024035
    [3] Ravindra Rao, Jagan Mohan Jonnalagadda . Existence of a unique solution to a fourth-order boundary value problem and elastic beam analysis. Mathematical Modelling and Control, 2024, 4(3): 297-306. doi: 10.3934/mmc.2024024
    [4] Zhen Xin, Yuhe Yang, Qiaoxia Li . Controllability of nonlinear ordinary differential equations with non-instantaneous impulses. Mathematical Modelling and Control, 2022, 2(1): 1-6. doi: 10.3934/mmc.2022001
    [5] K. Venkatachalam, M. Sathish Kumar, P. Jayakumar . Results on non local impulsive implicit Caputo-Hadamard fractional differential equations. Mathematical Modelling and Control, 2024, 4(3): 286-296. doi: 10.3934/mmc.2024023
    [6] Iman Malmir . Novel closed-loop controllers for fractional nonlinear quadratic systems. Mathematical Modelling and Control, 2023, 3(4): 345-354. doi: 10.3934/mmc.2023028
    [7] Hui Li, Nana Jin, Yu Zhang . Existence of nonoscillatory solutions for higher order nonlinear mixed neutral differential equations. Mathematical Modelling and Control, 2024, 4(4): 417-423. doi: 10.3934/mmc.2024033
    [8] Mrutyunjaya Sahoo, Dhabaleswar Mohapatra, S. Chakraverty . Wave solution for time fractional geophysical KdV equation in uncertain environment. Mathematical Modelling and Control, 2025, 5(1): 61-72. doi: 10.3934/mmc.2025005
    [9] Kexin Ouyang, Xinmin Qu, Huiqin Lu . Sign-changing and signed solutions for fractional Laplacian equations with critical or supercritical nonlinearity. Mathematical Modelling and Control, 2025, 5(1): 1-14. doi: 10.3934/mmc.2025001
    [10] Zhenguo Luo, Liping Luo . New criteria for oscillation of damped fractional partial differential equations. Mathematical Modelling and Control, 2022, 2(4): 219-227. doi: 10.3934/mmc.2022021
  • Fire incidents near power transmission lines pose significant safety hazards to the regular operation of the power system. Therefore, achieving fast and accurate smoke detection around power transmission lines is crucial. Due to the complexity and variability of smoke scenarios, existing smoke detection models suffer from low detection accuracy and slow detection speed. This paper proposes an improved model for smoke detection in high-voltage power transmission lines based on the improved YOLOv7-tiny. First, we construct a dataset for smoke detection in high-voltage power transmission lines. Due to the limited number of real samples, we employ a particle system to randomly generate smoke and composite it into randomly selected real scenes, effectively expanding the dataset with high quality. Next, we introduce multiple parameter-free attention modules into the YOLOv7-tiny model and replace regular convolutions in the Neck of the model with Spd-Conv (Space-to-depth Conv) to improve detection accuracy and speed. Finally, we utilize the synthesized smoke dataset as the source domain for model transfer learning. We pre-train the improved model and fine-tune it on a dataset consisting of real scenarios. Experimental results demonstrate that the proposed improved YOLOv7-tiny model achieves a 2.61% increase in mean Average Precision (mAP) for smoke detection on power transmission lines compared to the original model. The precision is improved by 2.26%, and the recall is improved by 7.25%. Compared to other object detection models, the smoke detection proposed in this paper achieves high detection accuracy and speed. Our model also improved detection accuracy on the already publicly available wildfire smoke dataset Figlib (Fire Ignition Library).



    Nonlinear equations arise always in electroanalytical chemistry with complex and esoteric nonlinear terms[1,2], though there are some advanced analytical methods to deal with nonlinear problems, for examples, the Gamma function method[3], Fourier spectral method[4], the reproducing kernel method[5], the perturbation method[6], the homotopy perturbation method[7,8], He's frequency formulation[9,10,11] and the dimensional method[12], chemists are always eager to have a simple one step method for nonlinear equations. This paper introduces an ancient Chinese algorithm called as the Ying Buzu algorithm[13] to solve nonlinear differential equations.

    We first introduce the Taylor series method[14]. Considering the nonlinear differential equation:

    d2udx2+F(u)=0. (0.1)

    The boundary conditions are

    dudx(a)=α, (0.2)
    u(b)=β. (0.3)

    If u(a) is known, we can use an infinite Taylor series to express the exact solution[14]. We assume that

    u(a)=c. (0.4)

    From (0.1), we have

    u(a)=F(u(a))=F(c),
    u(a)=F(c)uu(a)=αF(c)u.

    Other higher order derivatives can be obtained with ease, and its Taylor series solution is

    u(x)=u(a)+(xa)u(a)+12!(xa)2u(a)+13!(xa)3u(a)+...+1N!(xa)Nu(N)(a),

    the constant c can be determined by the boundary condition of (0.3).

    The Ying Buzu algorithm[15,16] was used to solve differential equations in 2006[13], it was further developed to He's frequency formulation for nonlinear oscillators[13,17,18,19,20,21,22,23] and Chun-Hui He's algorithm for numerical simulation[24].

    As c in (0.4) is unknown, according to the Ying Buzu algorithm[13,15,16], we can assume two initial guesses:

    u1(a)=c1,u2(a)=c2. (0.5)

    where c1 and c2 are given approximate values.

    Using the initial conditions given in (0.2) and (0.5), we can obtain the terminal values:

    u(b,c1)=β1,u(b,c2)=β2.

    According to the Ying Buzu algorithm[6,7,8,9,10,11,12], the initial guess can be updated as

    u(a)est=c3=c1(ββ2)c2(ββ1)(ββ2)(ββ1),

    and its terminal value can be calculated as

    u(b,c3)=β3.

    For a given small threshold, ε, |ββ3|ε, we obtain u(a)=c3 as an approximate solution.

    Here, we take Michaelis Menten dynamics as an example to solve the equation. Michaelis Menten reaction diffusion equation is considered as follows[25,26]:

    d2udx2u1+u=0. (0.6)

    The boundary conditions of it are as follows:

    dudx(0)=0,u(1)=1. (0.7)

    We assume

    u(0)=c.

    From (0.6), we have

    u(0)=c1+c,
    u(0)=0, (0.8)
    u(4)=c(1+c)3.

    The 2nd order Taylor series solution is

    u(x)=u(0)+u(0)1!x+u(0)2!x2=c+c2(1+c)x2.

    In view of the boundary condition of (0.7), we have

    u(1)=c+c2(1+c)=1, (0.9)

    solving c from (0.9) results in

    c=0.7808.

    So we obtain the following approximate solution

    u(x)=0.7808+0.2192x2.

    Similarly the fourth order Taylor series solution is

    u(x)=c+c2!(1+c)x2+c4!(1+c)3x4.

    Incorporating the boundary condition, u(1)=1, we have

    c+c2!(1+c)+c4!(1+c)3=1. (0.10)

    We use the Ying Buzu algorithm to solve c, and write (0.10) in the form

    R(c)=c+c2(1+c)+c24(1+c)31.

    Assume the two initial solutions are

    c1=0.8,c2=0.5.

    We obtain the following residuals

    R1(0.8)=0.0279,R2(0.5)=0.3271.

    By the Ying Buzu algorithm, c can be calculated as

    c=R2c1R1c2R2R1=0.0279×0.5+0.3271×0.80.0279+0.3271=0.7764.

    The exact solution of (0.10) is

    c=0.7758.

    The 4th order Taylor series solution is

    u(x)=0.7758+0.2192x2+0.0057x4.

    Figure 1 shows the Taylor series solutions, which approximately meet the requirement of the boundary condition at x=1.

    Figure 1.  Taylor series solution.

    Now we use the Ying Buzu algorithm by choosing two initial guesses

    u1(0)=0.5,u2(0)=1,

    which lead to u1=0.6726 and u2=1.2550, respectively, see Figure 2 (a) and (b).

    Figure 2.  The shooting processes with different initial guesses.

    It is obvious that the terminal value at x=1 deviates from u(1)=1 for each guess, according to the Ying Buzu algorithm, the initial guess can be updated as

    u3(0)=0.5×(11.2550)1×(10.6726)(11.2550)(10.6726)=0.7810. (0.11)

    The shooting process using (0.11) results in

    u3(1)=1.0058,

    which deviates the exact value of u(1)=1 with a relative error of 0.5%, see Figure 3.

    Figure 3.  The shooting processes with an updated initial guess of u(0)=0.7810..

    We can continue the iteration process to obtain a higher accuracy by using two following two guesses u1(0)=0.5, u3(0)=0.7810:

    u4(0)=0.5×(11.0058)0.7810×(10.6726)(11.0058)(10.6726)=0.7761.

    Using this updated initial value, the shooting process leads to the result

    u(1)=1.0001,

    so the approximate u(0)=0.7761 has only a relative error of 0.01%.

    The above solution process couples the numerical method, and the ancient method can also be solved independently.

    We assume that solution is

    u(x)=c+(1c)x2. (0.12)

    Equation (0.12) meets all boundary conditions.

    The residual equation is

    R(x)=d2udx2u1+u.

    It is easy to find that

    R(0)=2(1c)c1+c.

    We choose two guesses:

    c1=0.5,c2=1.

    We obtain the following residuals

    R1(0)=2(10.5)0.51+0.5=23,
    R2(0)=2(11)11+1=12.

    The Ying Buzu algorithm leads to the updated result:

    c=c2R1(0)c1R2(0)R1(0)R2(0)=23×1+12×0.523+12=0.7857.

    The relative error is 1.2%, and the process can continue if a higher accuracy is still needed.

    The ancient Chinese algorithm provides a simple and straightforward tool to two-point boundary value problems arising in chemistry, and it can be used for fast insight into the solution property of a complex problem.

    The authors declare that they have no conflicts of interest to this work.



    [1] Z. B. Zhao, Z. G. Jiang, Y. X. Li, Y. C. Qi, Y. J. Zhai, W. Q. Zhao, et al., Overview of visual defect detection of transmission line components, J. Image Graphics, 26 (2021), 2545–2560. https://doi.org/10.11834/jig.200689 doi: 10.11834/jig.200689
    [2] Y. Sui, P. F. Ning, P. J. Niu, C. Y. Wang, D. Zhao, W. L. Zhang, et al, Review on mounted UAV for transmission line inspection, Power Syst. Technol., 45 (2021), 3636–3648. http://doi.org/10.13335/j.1000-3673.pst.2020.1178 doi: 10.13335/j.1000-3673.pst.2020.1178
    [3] Z. Y. Liu, X. R. Miu, J. Chen, H. Jiang, Review of visible image intelligent processing for transmission line inspection, Power Syst. Technol., 44 (2020), 1057–1069. http://doi.org/10.13335/j.1000-3673.pst.2019.0349 doi: 10.13335/j.1000-3673.pst.2019.0349
    [4] S. Khan, K. Muhammad, S. Mumtaz, S. W. Baik, V. H. C. Albuquerque, Energy-efficient deep CNN for smoke detection in foggy IoT environment, IEEE Internet Things J., 6 (2019), 9237–9245. http://doi.org/10.1109/JIOT.2019.2896120 doi: 10.1109/JIOT.2019.2896120
    [5] H. Yin, Y. R. Wei, An improved algorithm based on convolutional neural network for smoke detection, in 2019 IEEE International Conferences on Ubiquitous Computing & Communications (IUCC) and Data Science and Computational Intelligence (DSCI) and Smart Computing, Networking and Services (SmartCNS), IEEE, (2019), 207–211. http://doi.org/10.1109/IUCC/DSCI/SmartCNS.2019.00063
    [6] C. H. Li, B. Yang, H. Ding, H. L. Shi, X. P. Jiang, J. Sun, Real-time video-based smoke detection with high accuracy and efficiency, Fire Saf. J., 117 (2020), 103184. http://doi.org/10.1016/j.firesaf.2020.103184 doi: 10.1016/j.firesaf.2020.103184
    [7] M. H. Jiang, Y. X. Zhao, F. Yu, C. L. Zhou, T. Peng, A self-attention network for smoke detection, Fire Saf. J., 129 (2022), 103547. http://doi.org/10.1016/j.firesaf.2022.103547 doi: 10.1016/j.firesaf.2022.103547
    [8] Z. Q. Li, A. Khananian, R. H. Fraser, J. Cihlar, Automatic detection of fire smoke using artificial neural networks and threshold approaches applied to AVHRR imagery, IEEE Trans. Geosci. Remote Sens., 39 (2001), 1859–1870. http://doi.org/10.1109/36.951076 doi: 10.1109/36.951076
    [9] K. Muhammad, J. Ahmad, I. Mehmood, S. Rho, S. W. Baik, Convolutional neural networks based fire detection in surveillance videos, IEEE Access, 6 (2018), 18174–18183. http://10.1109/ACCESS.2018.2812835 doi: 10.1109/ACCESS.2018.2812835
    [10] W. B. Cai, C. Y. Wang, H. Huang, T. Z. Wang, A real-time smoke detection model based on YOLO-smoke algorithm, in 2020 Cross Strait Radio Science & Wireless Technology Conference (CSRSWTC), IEEE, (2020), 1–3. http://10.1109/CSRSWTC50769.2020.9372453
    [11] F. R. Zhou, G. Wen, Y. Ma, Y. F. Wang, Y. T. Ma, G. F. Wang, et al., Multilevel feature cooperative alignment and fusion for unsupervised domain adaptation smoke detection, Front. Phys., 11 (2023), 81. https://doi.org/10.3389/fphy.2023.1136021 doi: 10.3389/fphy.2023.1136021
    [12] S. G. Zhang, F. Zhang, Y. Ding, Y. Li, Swin-YOLOv5: Research and application of fire and smoke detection algorithm based on YOLOv5, Comput. Intell. Neurosci., 2022 (2022). https://doi.org/10.1155/2022/6081680 doi: 10.1155/2022/6081680
    [13] C. Y. Wang, A. Bochkovskiy, H. Y. M. Liao, YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors, in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, (2023), 7464–7475. https://doi.org/10.48550/arXiv.2207.02696
    [14] Y. C. Zhou, L. H. Fang, X. Y. Zheng, X. L. Chen, Virtual battlefield smoke effect simulation based on particle system, Comput. Simul., 32 (2015), 417–420. https://doi.org/10.3969/j.issn.1006-9348.2015.07.093 doi: 10.3969/j.issn.1006-9348.2015.07.093
    [15] A. Bochkovskiy, C. Y. Wang, H. Y. M. Liao, YOLOv4: Optimal speed and accuracy of object detection, preprint, arXiv: 2004.10934. https://doi.org/10.48550/arXiv.2004.10934
    [16] G. Jocher, A. Stoken, J. Borovec, L. Changyu, A. Hogan, L. Diaconu, et al., ultralytics/yolov5: v3. 0, Zenodo, 2020. Available from: https://ui.adsabs.harvard.edu/abs/2020zndo...3983579J/abstract.
    [17] Y. Liu, X. Wang, Sar ship detection based on improved YOLOv7-tiny, in 2022 IEEE 8th International Conference on Computer and Communications (ICCC), IEEE, (2022), 2166–2170. https://doi.org/10.1109/ICCC56324.2022.10065775
    [18] T. Y. Lin, P. Dollár, R. Girshick, K. He, B. Hariharan, S. Belongie, Feature pyramid networks for object detection, in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, (2017), 2117–2125. https://doi.org/10.1109/CVPR.2017.106
    [19] L. Yang, R. Y. Zhang, L. Li, X. Xie, Simam: A simple, parameter-free attention module for convolutional neural networks, in International Conference on Machine Learning, (2021), 11863–11874.
    [20] R. Sunkara, T. Luo, No more strided convolutions or pooling: A new CNN building block for low-resolution images and small objects, in Machine Learning and Knowledge Discovery in Databases, Springer Nature, Cham, Switzerland, (2023), 443–459. https://doi.org/10.1007/978-3-031-26409-2_27
    [21] Q. Tian, R. Hu, Z. Li, Y. Cai, Z. Yu, Insulator detection based on se-YOLOv5s, Chin. J. Intell. Sci. Technol., 3 (2021), 312–321. https://doi.org/10.11959/j.issn.2096-6652.202132 doi: 10.11959/j.issn.2096-6652.202132
    [22] B. S. Webb, N. T. Dhruv, S. G. Solomon, C. Tailby, P. Lennie, Early and late mechanisms of surround suppression in striate cortex of macaque, J. Neurosci., 25 (2005), 11666–11675. https://doi.org/10.1523/JNEUROSCI.3414-05.2005 doi: 10.1523/JNEUROSCI.3414-05.2005
    [23] J. Hu, L. Shen, G. Sun, Squeeze-and-excitation networks, in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, (2018), 7132–7141. https://doi.org/10.1109/TPAMI.2019.2913372
    [24] Q. L. Wang, B. G. Wu, P. F. Zhu, P. H. Li, W. M. Zuo, Q. H. Hu, ECA-Net: Efficient channel attention for deep convolutional neural networks, in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, (2020), 11531–11539. https://doi.org/10.1109/CVPR42600.2020.01155
    [25] S. Woo, J. Park, J. Y. Lee, I. S. Kweon, CBAM: convolutional block attention module, in Proceedings of the European Conference on Computer Vision (ECCV), (2018), 3–19. https://doi.org/10.48550/arXiv.1807.06521
    [26] A. Dewangan, Y. Pande, H. W. Braun, F. Vernon, I. Perez, I. Altintas, et al., Figlib & smokeynet: Dataset and deep learning model for real-time wildland fire smoke detection, Remote Sens., 14 (2022), 1007. https://doi.org/10.3390/rs14041007 doi: 10.3390/rs14041007
    [27] K. Govil, M. L. Welch, J. T. Ball, C. R. Pennypacker, Preliminary results from a wildfire detection system using deep learning on remote camera images, Remote Sens., 12 (2020), 166. https://doi.org/10.3390/rs12010166 doi: 10.3390/rs12010166
  • This article has been cited by:

    1. Yi Tian, Approximate solution of initial boundary value problems for ordinary differential equations with fractal derivative, 2022, 2, 2767-8946, 75, 10.3934/mmc.2022009
    2. Guang-Qing Feng, Jing-Yan Niu, The analysis for the dynamic pull-in of a micro-electromechanical system, 2022, 1461-3484, 146134842211455, 10.1177/14613484221145588
    3. YONG-GANG KANG, SHUAI-JIA KOU, SI-REN SONG, YU-ZHEN CHANG, AN-YANG WANG, YONG-GANG CHEN, VIBRATION ANALYSIS OF HEAVY WEAPONS IN TRANSIT BY AIRCRAFT IN FRACTAL SPACE CONSIDERING LOCATION DEVIATION, 2024, 32, 0218-348X, 10.1142/S0218348X2450018X
    4. Jing-Yan Niu, Guang-Qing Feng, A mini-review on ancient mathematics’ modern applications with an emphasis on the old Babylonian mathematics for MEMS systems, 2024, 12, 2296-424X, 10.3389/fphy.2024.1532630
  • 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(2673) PDF downloads(98) Cited by(4)

Figures and Tables

Figures(12)  /  Tables(6)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog