In recent years, Transformer-based object trackers have demonstrated exceptional performance in object tracking. However, traditional methods often employ single-scale pixel-level attention mechanisms to compute the correlation between templates and search regions, disrupting object's integrity and positional information. To address these issues, we introduce a cyclic-shift mechanism to expand the diversity of sample positions and replace the traditional single-scale pixel-level attention mechanism with a multi-scale window-level attention mechanism. This approach not only preserves the object's integrity but also enriches the diversity of samples. Nevertheless, the introduced cyclic-shift operation heavily burdens storage and computation. To this end, we treat the attention computation of shifted and static windows in the spatial domain as convolution. By leveraging the convolution theorem, we transform the attention computation of cyclic shift samples from the spatial domain to element-wise multiplication in the frequency domain. This approach enhances computational efficiency and reduces data storage requirements. We conducted extensive experiments on the proposed module. The results demonstrate that the proposed module outperforms multiple existing tracking algorithms regarding performance. Moreover, ablation studies show that the method effectively reduces the storage and computational burden without compromising performance.
Citation: Huanyu Wu, Yingpin Chen, Changhui Wu, Ronghuan Zhang, Kaiwei Chen. A multi-scale cyclic-shift window Transformer object tracker based on fast Fourier transform[J]. Electronic Research Archive, 2025, 33(6): 3638-3672. doi: 10.3934/era.2025162
[1] | Jiani Jin, Haokun Qi, Bing Liu . Hopf bifurcation induced by fear: A Leslie-Gower reaction-diffusion predator-prey model. Electronic Research Archive, 2024, 32(12): 6503-6534. doi: 10.3934/era.2024304 |
[2] | Fengrong Zhang, Ruining Chen . Spatiotemporal patterns of a delayed diffusive prey-predator model with prey-taxis. Electronic Research Archive, 2024, 32(7): 4723-4740. doi: 10.3934/era.2024215 |
[3] | Ruizhi Yang, Dan Jin . Dynamics in a predator-prey model with memory effect in predator and fear effect in prey. Electronic Research Archive, 2022, 30(4): 1322-1339. doi: 10.3934/era.2022069 |
[4] | Yujia Xiang, Yuqi Jiao, Xin Wang, Ruizhi Yang . Dynamics of a delayed diffusive predator-prey model with Allee effect and nonlocal competition in prey and hunting cooperation in predator. Electronic Research Archive, 2023, 31(4): 2120-2138. doi: 10.3934/era.2023109 |
[5] | Xiaowen Zhang, Wufei Huang, Jiaxin Ma, Ruizhi Yang . Hopf bifurcation analysis in a delayed diffusive predator-prey system with nonlocal competition and schooling behavior. Electronic Research Archive, 2022, 30(7): 2510-2523. doi: 10.3934/era.2022128 |
[6] | Kimun Ryu, Wonlyul Ko . Global existence of classical solutions and steady-state bifurcation in a prey-taxis predator-prey system with hunting cooperation and a logistic source for predators. Electronic Research Archive, 2025, 33(6): 3811-3833. doi: 10.3934/era.2025169 |
[7] | Yichao Shao, Hengguo Yu, Chenglei Jin, Jingzhe Fang, Min Zhao . Dynamics analysis of a predator-prey model with Allee effect and harvesting effort. Electronic Research Archive, 2024, 32(10): 5682-5716. doi: 10.3934/era.2024263 |
[8] | Mengting Sui, Yanfei Du . Bifurcations, stability switches and chaos in a diffusive predator-prey model with fear response delay. Electronic Research Archive, 2023, 31(9): 5124-5150. doi: 10.3934/era.2023262 |
[9] | Yuan Tian, Yang Liu, Kaibiao Sun . Complex dynamics of a predator-prey fishery model: The impact of the Allee effect and bilateral intervention. Electronic Research Archive, 2024, 32(11): 6379-6404. doi: 10.3934/era.2024297 |
[10] | Wenbin Zhong, Yuting Ding . Spatiotemporal dynamics of a predator-prey model with a gestation delay and nonlocal competition. Electronic Research Archive, 2025, 33(4): 2601-2617. doi: 10.3934/era.2025116 |
In recent years, Transformer-based object trackers have demonstrated exceptional performance in object tracking. However, traditional methods often employ single-scale pixel-level attention mechanisms to compute the correlation between templates and search regions, disrupting object's integrity and positional information. To address these issues, we introduce a cyclic-shift mechanism to expand the diversity of sample positions and replace the traditional single-scale pixel-level attention mechanism with a multi-scale window-level attention mechanism. This approach not only preserves the object's integrity but also enriches the diversity of samples. Nevertheless, the introduced cyclic-shift operation heavily burdens storage and computation. To this end, we treat the attention computation of shifted and static windows in the spatial domain as convolution. By leveraging the convolution theorem, we transform the attention computation of cyclic shift samples from the spatial domain to element-wise multiplication in the frequency domain. This approach enhances computational efficiency and reduces data storage requirements. We conducted extensive experiments on the proposed module. The results demonstrate that the proposed module outperforms multiple existing tracking algorithms regarding performance. Moreover, ablation studies show that the method effectively reduces the storage and computational burden without compromising performance.
In this paper, we consider the Schrödinger operators
L=−△+V(x),x∈Rn,n≥3, |
where Δ=∑ni=1∂2∂2xi and V(x) is a nonnegative potential belonging to the reverse Hölder class RHq for some q≥n2. Assume that f is a nonnegative locally Lq(Rn) integrable function on Rn, then we say that f belongs to RHq (1<q≤∞) if there exists a positive constant C such that the reverse Hölder's inequality
(1|B(x,r)|∫B(x,r)|f(y)|qdy)1q≤C|B(x,r)|∫B(x,r)|f(y)|dy |
holds for x in Rn, where B(x,r) denotes the ball centered at x with radius r<∞ [1]. For example, the nonnegative polynomial V∈RH∞, in particular, |x|2∈RH∞.
Let the potential V∈RHq with q≥n2, and the critical radius function ρ(x) is defined as
ρ(x)=supr>0{r:1rn−2∫B(x,r)V(y)dy≤1},x∈Rn. | (1.1) |
We also write ρ(x)=1mV(x),x∈Rn. Clearly, 0<mV(x)<∞ when V≠0, and mV(x)=1 when V=1. For the harmonic oscillator operator (Hermite operator) H=−Δ+|x|2, we have mV(x)∼(1+|x|).
Thanks to the heat diffusion semigroup e−tL for enough good function f, the negative powers L−α2(α>0) related to the Schrödinger operators L can be written as
Iαf(x)=L−α2f(x)=∫∞0e−tLf(x)tα2−1dt,0<α<n. | (1.2) |
Applying Lemma 3.3 in [2] for enough good function f holds that
Iαf(x)=∫RnKα(x,y)f(y)dy,0<α<n, |
and the kernel Kα(x,y) satisfies the following inequality
Kα(x,y)≤Ck(1+|x−y|(mV(x)+mV(y)))k1|x−y|n−α. | (1.3) |
Moreover, we have Kα(x,y)≤C|x−y|n−α,0<α<n.
Shen [1] obtained Lp estimates of the Schrödinger type operators when the potential V∈RHq with q≥n2. For Schrödinger operators L=−Δ+V with V∈RHq for some q≥n2, Harboure et al. [3] established the necessary and sufficient conditions to ensure that the operators L−α2(α>0) are bounded from weighted strong and weak Lp spaces into suitable weighted BMOL(w) space and Lipschitz spaces when p≥nα. Bongioanni Harboure and Salinas proved that the fractional integral operator L−α/2 is bounded form Lp,∞(w) into BMOβL(w) under suitable conditions for weighted w [4]. For more backgrounds and recent progress, we refer to [5,6,7] and references therein.
Ramseyer, Salinas and Viviani in [8] studied the fractional integral operator and obtained the boundedness from strong and weak Lp(⋅) spaces into the suitable Lipschitz spaces under some conditions on p(⋅). In this article, our main interest lies in considering the properties of fractional integrals operator L−α2(α>0), related to L=−Δ+V with V∈RHq for some q≥n2 in variable exponential spaces.
We now introduce some basic properties of variable exponent Lebsegue spaces, which are used frequently later on.
Let p(⋅):Ω→[1,∞) be a measurable function. For a measurable function f on Rn, the variable exponent Lebesgue space Lp(⋅)(Ω) is defined by
Lp(⋅)(Ω)={f:∫Ω|f(x)s|p(x)dx<∞}, |
where s is a positive constant. Then Lp(⋅)(Ω) is a Banach space equipped with the follow norm
‖f‖Lp(⋅)(Ω):=inf{s>0:∫Ω|f(x)s|p(x)dx≤1}. |
We denote
p−:=essinfx∈Ωp(x) and p+:=esssupx∈Ωp(x). |
Let P(Rn) denote the set of all measurable functions p on Rn that take value in [1,∞), such that 1<p−(Rn)≤p(⋅)≤p+(Rn)<∞.
Assume that p is a real value measurable function p on Rn. We say that p is locally log-Hölder continuous if there exists a constant C such that
|p(x)−p(y)|≤Clog(e+1/|x−y|),x,y∈Rn, |
and we say p is log-Hölder continuous at infinity if there exists a positive constant C such that
|p(x)−p(∞)|≤Clog(e+|x|),x∈Rn, |
where p(∞):=lim|x|→∞p(x)∈R.
The notation Plog(Rn) denotes all measurable functions p in P(Rn), which states p is locally log-Hölder continuous and log-Hölder continuous at infinity. Moreover, we have that p(⋅)∈Plog(Rn), which implies that p′(⋅)∈Plog(Rn).
Definition 1.1. [8] Assume that p(⋅) is an exponent function on Rn. We say that a measurable function f belongs to Lp(⋅),∞(Rn), if there exists a constant C such that for t>0,
∫Rntp(x)χ{|f|>t}(x)dx≤C. |
It is easy to check that Lp(⋅),∞(Rn) is a quasi-norm space equipped with the following quasi-norm
‖f‖p(⋅),∞=inf{s>0:supt>0∫Rn(ts)p(x)χ{|f|>t}(x)dx≤1}. |
Next, we define LipLα,p(⋅) spaces related to the nonnegative potential V.
Definition 1.2. Let p(⋅) be an exponent function with 1<p−≤p+<∞ and 0<α<n. We say that a locally integrable function f∈LipLα,p(⋅)(Rn) if there exist constants C1,C2 such that for every ball B⊂Rn,
1|B|αn‖χB‖p′(⋅)∫B|f(x)−mBf|dx≤C1, | (1.4) |
and for R≥ρ(x),
1|B|αn‖χB‖p′(⋅)∫B|f(x)|dx≤C2, | (1.5) |
where mBf=1|B|∫Bf. The norm of space LipLα,p(⋅)(Rn) is defined as the maximum value of two infimum of constants C1 and C2 in (1.4) and (1.5).
Remark 1.1. LipLα,p(⋅)(Rn)⊂Lα,p(⋅)(Rn) is introduced in [8]. In particular, when p(⋅)=C for some constant, then LipLα,p(⋅)(Rn) is the usual weighted BMO space BMOβL(w), with w=1 and β=α−np [4].
Remark 1.2. It is easy to see that for some ball B, the inequality (1.5) leads to inequality (1.4) holding, and the average mBf in (1.4) can be replaced by a constant c in following sense
12‖f‖LipLα,p(⋅)≤supB∈Rninfc∈R1|B|αn‖χB‖p′(⋅)∫B|f(x)−c|dx≤‖f‖LipLα,p(⋅). |
In 2013, Ramseyer et al. in [8] studied the Lipschitz-type smoothness of fractional integral operators Iα on variable exponent spaces when p+>αn. Hence, when p+>αn, it will be an interesting problem to see whether or not we can establish the boundedness of fractional integral operators L−α2(α>0) related to Schrödinger operators from Lebesgue spaces Lp(⋅) into Lipschitz-type spaces with variable exponents. The main aim of this article is to answer the problem above.
We now state our results as the following two theorems.
Theorem 1.3. Let potential V∈RHq for some q≥n/2 and p(⋅)∈Plog(Rn). Assume that 1<p−≤p+<n(α−δ0)+ where δ0=min{1,2−n/q}, then the fractional integral operator Iα defined in (1.2) is bounded from Lp(⋅)(Rn) into LipLα,p(⋅)(Rn).
Theorem 1.4. Let the potential V∈RHq with q≥n/2 and p(⋅)∈Plog(Rn). Assume that 1<p−≤p+<n(α−δ0)+ where δ0=min{1,2−n/q}. If there exists a positive number r0 such that p(x)≤p∞ when |x|>r0, then the fractional integral operator Iα defined in (1.2) is bounded from Lp(⋅),∞(Rn) into LipLα,p(⋅)(Rn).
To prove Theorem 1.3, we first need to decompose Rn into the union of some disjoint ball B(xk,ρ(xk))(k≥1) according to the critical radius function ρ(x) defined in (1.1). According to Lemma 2.6, we establish the necessary and sufficient conditions to ensure f∈LipLα,p(⋅)(Rn). In order to prove Theorem 1.3, by applying Corollary 1 and Remark 1.2, we only need to prove that the following two conditions hold:
(ⅰ) For every ball B=B(x0,r) with r<ρ(x0), then
∫B|Iαf(x)−c|dx≤C|B|αn‖χB‖p′(⋅)‖f‖p(⋅); |
(ⅱ) For any x0∈Rn, then
∫B(x0,ρ(x0))Iα(|f|)(x)dx≤C|B(x0,ρ(x0))|αn‖χB(x0,ρ(x0))‖p′(⋅)‖f‖p(⋅). |
In order to check the conditions (ⅰ) and (ⅱ) above, we need to find the accurate estimate of kernel Kα(x,y) of fractional integral operator Iα (see Lemmas 2.8 and 2.9, then use them to obtain the proof of this theorem; the proof of the Theorem 1.4 proceeds identically).
The paper is organized as follows. In Section 2, we give some important lemmas. In Section 3, we are devoted to proving Theorems 1.3 and 1.4.
Throughout this article, C always means a positive constant independent of the main parameters, which may not be the same in each occurrence. B(x,r)={y∈Rn:|x−y|<r}, Bk=B(x0,2kR) and χBk are the characteristic functions of the set Bk for k∈Z. |S| denotes the Lebesgue measure of S. f∼g means C−1g≤f≤Cg.
In this section, we give several useful lemmas that are used frequently later on.
Lemma 2.1. [9] Assume that the exponent function p(⋅)∈P(Rn). If f∈Lp(⋅)(Rn) and g∈Lp′(⋅)(Rn), then
∫Rn|f(x)g(x)|dx≤rp‖f‖Lp(⋅)(Rn)‖g‖Lp′(⋅)(Rn), |
where rp=1+1/p−−1/p+.
Lemma 2.2. [8] Assume that p(⋅)∈Plog(Rn) and 1<p−≤p+<∞, and p(x)≤p(∞) when |x|>r0>1. For every ball B and f∈Lp(⋅),∞ we have
∫B|f(x)|dx≤C‖f‖Lp(⋅),∞‖χB‖Lp′(⋅), |
where the constant C only depends on r0.
Fo the following lemma see Corollary 4.5.9 in [10].
Lemma 2.3. Let p(⋅)∈Plog(Rn), then for every ball B⊂Rn we have
‖χB‖p(⋅)∼|B|1p(x),if|B|≤2n,x∈B, |
and
‖χB‖p(⋅)∼|B|1p(∞),if|B|≥1. |
Lemma 2.4. Assume that p(⋅)∈Plog(Rn), then for all balls B and all measurable subsets S:=B(x0,r0)⊂B:=B(x1,r1) we have
‖χS‖p′(⋅)‖χB‖p′(⋅)≤C(|S||B|)1−1p−, ‖χB‖p′(⋅)‖χS‖p′(⋅)≤C(|B||S|)1−1p+. | (2.1) |
Proof. We only prove the first inequality in (2.1), and the second inequality in (2.1) proceeds identically. We consider three cases below by applying Lemma 2.3, and it holds that
1) if |S|<1<|B|, then ‖χS‖p′(⋅)‖χB‖p′(⋅)∼|S|1p′(xS)|B|1p′(∞)≤(|S||B|)1(p′)+=(|S||B|)1−1p−;
2) if 1≤|S|<|B|, then ‖χS‖p′(⋅)‖χB‖p′(⋅)∼|S|1p′(∞)|B|1p′(∞)≤(|S||B|)1(p′)+=(|S||B|)1−1p−;
3) if |S|<|B|<1, then ‖χS‖p′(⋅)‖χB‖p′(⋅)∼|S|1p′(xS)|B|1p′(xS)|B|1p′(xS)−1p′(xB)≤C(|S||B|)1(p′)+=C(|S||B|)1−1p−, where xS∈S and xB∈B.
Indeed, since |xB−xS|≤2r1, by using the local-Hölder continuity of p′(x) we have
|1p′(xS)−1p′(xB)|log1r1≤log1r1log(e+1|xS−xB|)≤log1r1log(e+12r1)≤C. |
We end the proof of this lemma.
Remark 2.1. Thanks to the second inequality in (2.1), it is easy to prove that
‖χ2B‖p′(⋅)≤C‖χB‖p′(⋅). |
Lemma 2.5. [1] Suppose that the potential V∈Bq with q≥n/2, then there exists positive constants C and k0 such that
1) ρ(x)∼ρ(y) when |x−y|≤Cρ(x);
2) C−1ρ(x)(1+|x−y|ρ(x))−k0≤ρ(y)≤Cρ(x)(1+|x−y|ρ(x))k0/(k0+1).
Lemma 2.6. [11] There exists a sequence of points {xk}∞k=1 in Rn such that Bk:=B(xk,ρ(xk)) satisfies
1) Rn=⋃kBk,
2) For every k≥1, then there exists N≥1 such that card {j:4Bj∩4Bk≠∅}≤N.
Lemma 2.7. Assume that p(⋅)∈P(Rn) and 0<α<n. Let sequence {xk}∞k=1 satisfy the propositions of Lemma 2.6. Then a function f∈LipLα,p(⋅)(Rn) if and only if f satisfies (1.4) for every ball, and
1|B(xk,ρ(xk))|αn‖χB(xk,ρ(xk))‖p′(⋅)∫B(xk,ρ(xk))|f(x)|dx≤C,forallk≥1. | (2.2) |
Proof. Let B:=B(x,R) denote a ball with center x and radius R>ρ(x). Noting that f satisfies (1.4), and thanks to Lemma 2.6 we obtain that the set G={k:B∩Bk≠∅} is finite.
Applying Lemma 2.5, if z∈Bk∩B, we get
ρ(xk)≤Cρ(z)(1+|xk−z|ρ(xk))k0≤C2k0ρ(z)≤C2k0ρ(x)(1+|x−z|ρ(x))k0k0+1≤C2k0ρ(x)(1+Rρ(x))≤C2k0R. |
Thus, for every k∈G, we have Bk⊂CB.
Thanks to Lemmas 2.4 and 2.6, it holds that
∫B|f(x)|dx=∫B⋂⋃kBk|f(x)|dx=∫⋃k∈G(B⋂Bk)|f(x)|dx≤∑k∈G∫B∩Bk|f(x)|dx≤∑k∈G∫Bk|f(x)|dx≤C∑k∈G|Bk|αn‖χBk‖p′(⋅)≤C|B|αn‖χB‖p′(⋅). |
The proof of this lemma is completed.
Corollary 1. Assume that p(⋅)∈P(Rn) and 0<α<n, then a measurable function f∈LipLα,p(⋅) if and only if f satisfies (1.4) for every ball B(x,R) with radius R<ρ(x) and
1|B(x,ρ(x))|αn‖χB(x,ρ(x))‖p′(⋅)∫B(x,ρ(x))|f(x)|dx≤C. | (2.3) |
Let kt(x,y) denote the kernel of heat semigroup e−tL associated to L, and Kα(x,y) be the kernel of fractional integral operator Iα, then it holds that
Kα(x,y)=∫∞0kt(x,y)tα2dt. | (2.4) |
Some estimates of kt are presented below.
Lemma 2.8. [12] There exists a constant C such that for N>0,
kt(x,y)≤Ct−n/2e−|x−y|2Ct(1+√tρ(x)+√tρ(y))−N,x,y∈Rn. |
Lemma 2.9. [13] Let 0<δ<min(1,2−nq). If |x−x0|<√t, then for N>0 the kernel kt(x,y) defined in (2.4) satisfies
|kt(x,y)−kt(x0,y)|≤C(|x−x0|√t)δt−n/2e−|x−y|2Ct(1+√tρ(x)+√tρ(y))−N, |
for all x,y and x0 in Rn.
In this section, we are devoted to the proof of Theorems 1.3 and 1.4. To prove Theorem 1.3, thanks to Corollary 1 and Remark 1.2, we only need to prove that the following two conditions hold:
(ⅰ) For every ball B=B(x0,r) with r<ρ(x0), then
∫B|Iαf(x)−c|dx≤C|B|αn‖χB‖p′(⋅)‖f‖p(⋅); |
(ⅱ) For any x0∈Rn, then
∫B(x0,ρ(x0))Iα(|f|)(x)dx≤C|B(x0,ρ(x0))|αn‖χB(x0,ρ(x0))‖p′(⋅)‖f‖p(⋅). |
We now begin to check that these conditions hold. First, we prove (ⅱ).
Assume that B=B(x0,R) and R=ρ(x0). We write f=f1+f2, where f1=fχ2B and f2=fχRn∖2B. Hence, by the inequality (1.3), we have
∫BIα(|f1|)(x)dx=∫BIα(|fχ2B|)(x)dx≤C∫B∫2B|f(y)||x−y|n−αdydx. |
Applying Tonelli theorem, Lemma 2.1 and Remark 1.2, we get the following estimate
∫BIα(|f1|)(x)dx≤C∫2B|f(y)|∫Bdx|x−y|n−αdy≤CRα∫2B|f(y)|dy≤C|B|αn‖χB‖p′(⋅)‖f‖p(⋅). | (3.1) |
To deal with f2, let x∈B and we split Iαf2 as follows:
Iαf2(x)=∫R20e−tLf2(x)tα2−1dt+∫∞R2e−tLf2(x)tα2−1dt:=I1+I2. |
For I1, if x∈B and y∈Rn∖2B, we note that |x0−y|<|x0−x|+|x−y|<C|x−y|. By Lemma 2.8, it holds that
I1=|∫R20∫Rn∖2Bkt(x,y)f(y)dytα2−1dt|≤C∫R20∫Rn∖2Bt−n2e−|x−y|2t|f(y)|dytα2−1dt≤C∫R20t−n+α2−1∫Rn∖2B(t|x−y|2)M/2|f(y)|dydt≤C∫R20tM−n+α2−1dt∫Rn∖2B|f(y)||x0−y|Mdy, |
where the constant C only depends the constant M.
Applying Lemma 2.1 to the last integral, we get
∫Rn∖2B|f(y)||x0−y|Mdy=∞∑i=1∫2i+1B∖2iB|f(y)||x0−y|Mdy≤∞∑i=1(2iR)−M∫2i+1B|f(y)|dy≤C∞∑i=1(2iR)−M‖χ2i+1B‖p′(⋅)‖f‖p(⋅). |
By using Lemma 2.4, we arrive at the inequality
∫Rn∖2B|f(y)||x0−y|Mdy≤C∞∑i=1(R)−M(2i)n−np+−M‖χB‖p′(⋅)‖f‖p(⋅)≤CR−M‖f‖p(⋅)‖χB‖p′(⋅). | (3.2) |
Here, the series above converges when M>n−np+. Hence, for such M,
|∫R20e−tLf2(x)tα2−1dt|≤CR−M‖f‖p(⋅)‖χB‖p′(⋅)∫R20tM−n+α2−1dt≤C|B|αn−1‖f‖p(⋅)‖χB‖p′(⋅). |
For I2, thanks to Lemma 2.8, we may choose M as above and N≥M, then it holds that
|∫∞R2e−tLf2(x)tα−22dt|=|∫∞R2∫Rn∖2Bkt(x,y)f(y)dytα−22dt|≤C∫∞R2∫Rn∖2Btα−n−N−22ρ(x)Ne−|x−y|2t|f(y)|dydt≤Cρ(x)N∫∞R2tα−n−N−22∫Rn∖2B(t|x−y|2)M/2|f(y)|dydt. |
As x∈B, thanks to Lemma 2.5, ρ(x)∼ρ(x0)=R. Hence we have
|∫∞R2e−tLf2(x)tα2−1dt|≤CRN∫∞R2tM+α−n−N2−1dt∫Rn∖2B|f(y)||x0−y|Mdy. |
Since M+α−n−N<0, the integral above for variable t converges, and by applying inequality (3.2) we have
|∫∞R2e−tLf2(x)tα2−1dt|≤C|B|αn−1‖f‖p(⋅)‖χB‖p′(⋅), |
thus we have proved (ⅱ).
We now begin to prove that the condition (ⅰ) holds. Let B=B(x0,r) and r<ρ(x0). We set f=f1+f2 with f1=fχ2B and f2=fχRn∖2B. We write
cr=∫∞r2e−tLf2(x0)tα2−1dt. | (3.3) |
Thanks to (3.1), it holds that
∫B|Iα(f(x))−cr|≤∫BIα(|f1|)(x)dx+∫B|Iα(f2)(x)−cr|dx≤C|B|αn−1‖χB‖p′(⋅)‖f‖p(⋅)+∫B|Iα(f2)(x)−cr|dx. |
Let x∈B and we split Iαf2(x) as follows:
Iαf2(x)=∫r20e−tLf2(x)tα2−1dt+∫∞r2e−tLf2(x)tα2−1dt:=I3+I4. |
For I3, by the same argument it holds that
I3=|∫r20e−tLf2(x)tα2−1dt|≤C|B|αn−1‖f‖p(⋅)‖χB‖p′(⋅). |
For I4, by Lemma 2.9 and (3.3), it follows that
|∫∞r2e−tLf2(x)tα2−1dt−cr|≤∫∞r2∫Rn∖2B|kt(x,y)−kt(x0,y)||f(y)|dytα2−1dt≤Cδ∫∞r2∫Rn∖2B(|x−x0|√t)δt−n/2e−|x−y|2Ct|f(y)|dytα2−1dt≤Cδrδ∫Rn∖2B|f(y)|∫∞r2t−(n−α+δ)/2e−|x−y|2Ctdttdy. |
Let s=|x−y|2t, then we obtain the following estimate
|∫∞r2e−tLf2(x)tα2−1dt−cr|≤Cδrδ∫Rn∖2B|f(y)||x−y|n−α+δdy∫∞0sn−α+δ2e−sCdss. |
Notice that the integral above for variable s is finite, thus we only need to compute the integral above for variable y. Thanks to inequality (3.2), it follows that
|∫∞r2e−tLf2(x)tα2−1dt−cr|≤Cδrδ∫Rn∖2B|f(y)||x−y|n−α+δdy≤C∞∑i=1Rα−n(2i)α−np+−δ‖χB‖p′(⋅)‖f‖p(⋅)≤C|B|α−nn‖f‖p(⋅)‖χB‖p′(⋅), |
so (ⅰ) is proved.
Remark 3.1. By the same argument as the proof of Theorem 1.3, thanks to Lemma 2.2 we immediately obtained that the conclusions of Theorem 1.4 hold.
The authors declare they have not used Artificial Intelligence (AI) tools in the creation of this article.
Ping Li is partially supported by NSFC (No. 12371136). The authors would like to thank the anonymous referees for carefully reading the manuscript and providing valuable suggestions, which substantially helped in improving the quality of this paper. We also thank Professor Meng Qu for his useful discussions.
The authors declare there are no conflicts of interest.
[1] |
Y. Li, X. Yuan, H. Wu, L. Zhang, R. Wang, J. Chen, CVT-track: Concentrating on valid tokens for one-stream tracking, IEEE Trans. Circuits Syst. Video Technol., 34 (2024), 321–334. https://doi.org/10.1109/TCSVT.2024.3452231 doi: 10.1109/TCSVT.2024.3452231
![]() |
[2] |
S. Zhang, Y. Chen, ATM-DEN: Image inpainting via attention transfer module and decoder-encoder network, SPIC, 133 (2025), 117268. https://doi.org/10.1016/j.image.2025.117268 doi: 10.1016/j.image.2025.117268
![]() |
[3] |
F. Chen, X. Wang, Y. Zhao, S. Lv, X. Niu, Visual object tracking: A survey, Comput. Vision Image Underst., 222 (2022), 103508. https://doi.org/10.1016/j.cviu.2022.103508 doi: 10.1016/j.cviu.2022.103508
![]() |
[4] |
F. Zhang, S. Ma, Z. Qiu, T. Qi, Learning target-aware background-suppressed correlation filters with dual regression for real-time UAV tracking, Signal Process., 191 (2022), 108352. https://doi.org/10.1016/j.sigpro.2021.108352 doi: 10.1016/j.sigpro.2021.108352
![]() |
[5] |
S. Ma, B. Zhao, Z. Hou, W. Yu, L. Pu, X. Yang, SOCF: A correlation filter for real-time UAV tracking based on spatial disturbance suppression and object saliency-aware, Expert Syst. Appl., 238 (2024), 122131. https://doi.org/10.1016/j.eswa.2023.122131 doi: 10.1016/j.eswa.2023.122131
![]() |
[6] |
J. Lin, J. Peng, J. Chai, Real-time UAV correlation filter based on response-weighted background residual and spatio-temporal regularization, IEEE Geosci. Remote Sens. Lett., 20 (2023), 1–5. https://doi.org/10.1109/LGRS.2023.3272522 doi: 10.1109/LGRS.2023.3272522
![]() |
[7] |
J. Cao, H. Zhang, L. Jin, J. Lv, G. Hou, C. Zhang, A review of object tracking methods: From general field to autonomous vehicles, Neurocomputing, 585 (2024), 127635. https://doi.org/10.1016/j.neucom.2024.127635 doi: 10.1016/j.neucom.2024.127635
![]() |
[8] |
X. Hao, Y. Xia, H. Yang, Z. Zuo, Asynchronous information fusion in intelligent driving systems for target tracking using cameras and radars, IEEE Trans. Ind. Electron., 70 (2023), 2708–2717. https://doi.org/10.1109/TIE.2022.3169717 doi: 10.1109/TIE.2022.3169717
![]() |
[9] |
L. Liang, Z. Chen, L. Dai, S. Wang, Target signature network for small object tracking, Eng. Appl. Artif. Intell., 138 (2024), 109445. https://doi.org/10.1016/j.engappai.2024.109445 doi: 10.1016/j.engappai.2024.109445
![]() |
[10] |
R. Yao, L. Zhang, Y. Zhou, H. Zhu, J. Zhao, Z. Shao, Hyperspectral object tracking with dual-stream prompt, IEEE Trans. Geosci. Remote Sens., 63 (2025), 1–12. https://doi.org/10.1109/TGRS.2024.3516833 doi: 10.1109/TGRS.2024.3516833
![]() |
[11] |
N. K. Rathore, S. Pande, A. Purohit, An efficient visual tracking system based on extreme learning machine in the defence and military sector, Def. Sci. J., 74 (2024), 643–650. https://doi.org/10.14429/dsj.74.19576 doi: 10.14429/dsj.74.19576
![]() |
[12] |
Y. Chen, Y. Tang, Y. Xiao, Q. Yuan, Y. Zhang, F. Liu, et al., Satellite video single object tracking: A systematic review and an oriented object tracking benchmark, ISPRS J. Photogramm. Remote Sens., 210 (2024), 212–240. https://doi.org/10.1016/j.isprsjprs.2024.03.013 doi: 10.1016/j.isprsjprs.2024.03.013
![]() |
[13] | W. Cai, Q. Liu, Y. Wang, HIPTrack: Visual tracking with historical prompts, in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), (2024), 19258–19267. https://doi.org/10.1109/CVPR52733.2024.01822 |
[14] |
L. Sun, J. Zhang, D. Gao, B. Fan, Z. Fu, Occlusion-aware visual object tracking based on multi-template updating Siamese network, Digit. Signal Process., 148 (2024), 104440. https://doi.org/10.1016/j.dsp.2024.104440 doi: 10.1016/j.dsp.2024.104440
![]() |
[15] |
Y. Chen, L. Wang, eMoE-Tracker: Environmental MoE-based transformer for robust event-guided object tracking, IEEE Robot. Autom. Lett., 10 (2025), 1393–1400. https://doi.org/10.1109/LRA.2024.3518305 doi: 10.1109/LRA.2024.3518305
![]() |
[16] |
Y. Sun, T. Wu, X. Peng, M. Li, D. Liu, Y. Liu, et al., Adaptive representation-aligned modeling for visual tracking, Knowl. Based Syst., 309 (2025), 112847. https://doi.org/10.1016/j.knosys.2024.112847 doi: 10.1016/j.knosys.2024.112847
![]() |
[17] |
J. Wang, S. Yang, Y. Wang, G. Yang, PPTtrack: Pyramid pooling based Transformer backbone for visual tracking, Expert Syst. Appl., 249 (2024), 123716. https://doi.org/10.1016/j.eswa.2024.123716 doi: 10.1016/j.eswa.2024.123716
![]() |
[18] |
C. Wu, J. Shen, K. Chen, Y. Chen, Y. Liao, UAV object tracking algorithm based on spatial saliency-aware correlation filter, Electron. Res. Arch., 33 (2025), 1446–1475. https://doi.org/10.3934/era.2025068 doi: 10.3934/era.2025068
![]() |
[19] |
A. Lukežič, T. Vojíř, L. Čehovin, J. Matas, M. Kristan, Discriminative correlation filter with channel and spatial reliability, Int. J. Comput. Vision, 126 (2018), 671–688. https://doi.org/10.1007/s11263-017-1061-3 doi: 10.1007/s11263-017-1061-3
![]() |
[20] |
T. Xu, Z. Feng, X. Wu, J. Kittler, Learning adaptive discriminative correlation filters via temporal consistency preserving spatial feature selection for robust visual object tracking, IEEE Trans. Image Process., 28 (2019), 5596–5609. https://doi.org/10.1109/TIP.2019.2919201 doi: 10.1109/TIP.2019.2919201
![]() |
[21] |
J. F. Henriques, R. Caseiro, P. Martins, J. Batista, High-speed tracking with kernelized correlation filters, IEEE Trans. Pattern Anal. Mach. Intell., 37 (2015), 583–596. https://doi.org/10.1109/TPAMI.2014.2345390 doi: 10.1109/TPAMI.2014.2345390
![]() |
[22] |
E. O. Brigham, R. E. Morrow, The fast Fourier transform, IEEE Spectrum, 4 (1967), 63–70. https://doi.org/10.1109/MSPEC.1967.5217220 doi: 10.1109/MSPEC.1967.5217220
![]() |
[23] | H. K. Galoogahi, A. Fagg, S. Lucey, Learning background-aware correlation filters for visual tracking, in IEEE International Conference on Computer Vision (ICCV), (2017), 1144–1152. https://doi.org/10.1109/ICCV.2017.129 |
[24] | Z. Zhang, H. Peng, J. Fu, B. Li, W. Hu, Ocean: Object-aware anchor-free tracking, in European Conference on Computer Vision (ECCV), (2020), 771–787. https://doi.org/10.1007/978-3-030-58589-1_46 |
[25] |
Y. Zhang, H. Pan, J. Wang, Enabling deformation slack in tracking with temporally even correlation filters, Neural Networks, 181 (2025), 106839. https://doi.org/10.1016/j.neunet.2024.106839 doi: 10.1016/j.neunet.2024.106839
![]() |
[26] |
Y. Chen, H. Wu, Z. Deng, J. Zhang, H. Wang, L. Wang, et al., Deep-feature-based asymmetrical background-aware correlation filter for object tracking, Digit. Signal Process., 148 (2024), 104446. https://doi.org/10.1016/j.dsp.2024.104446 doi: 10.1016/j.dsp.2024.104446
![]() |
[27] |
K. Chen, L. Wang, H. Wu, C. Wu, Y. Liao, Y. Chen, et al., Background-aware correlation filter for object tracking with deep CNN features, Eng. Lett., 32 (2024), 1353–1363. https://doi.org/10.1016/j.dsp.2024.104446 doi: 10.1016/j.dsp.2024.104446
![]() |
[28] |
J. Zhang, Y. He, W. Chen, L. D. Kuang, B. Zheng, CorrFormer: Context-aware tracking with cross-correlation and transformer, Comput. Electr. Eng., 114 (2024), 109075. https://doi.org/10.1016/j.compeleceng.2024.109075 doi: 10.1016/j.compeleceng.2024.109075
![]() |
[29] | L. Bertinetto, J. Valmadre, J. F. Henriques, A. Vedaldi, P. H. Torr, Fully-convolutional siamese networks for object tracking, in European Conference on Computer Vision (ECCV), (2016), 850–865. https://doi.org/10.1007/978-3-319-48881-3_56 |
[30] | Q. Guo, W. Feng, C. Zhou, R. Huang, L. Wan, S. Wang, Learning dynamic siamese network for visual object tracking, in IEEE International Conference on Computer Vision (ICCV), (2017), 1781–1789. https://doi.org/10.1109/ICCV.2017.196 |
[31] | B. Li, J. Yan, W. Wu, Z. Zhu, X. Hu, High performance visual tracking with siamese region proposal network, in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), (2018), 8971–8980. |
[32] | B. Li, W. Wu, Q. Wang, F. Zhang, J. Xing, J. Yan, SiamRPN++: Evolution of siamese visual tracking with very deep networks, in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), (2019), 4277–4286. |
[33] |
L. Zhao, C. Fan, M. Li, Z. Zheng, X. Zhang, Global-local feature-mixed network with template update for visual tracking, Pattern Recognit. Lett., 188 (2025), 111–116. https://doi.org/10.1016/j.patrec.2024.11.034 doi: 10.1016/j.patrec.2024.11.034
![]() |
[34] |
F. Gu, J. Lu, C. Cai, Q. Zhu, Z. Ju, RTSformer: A robust toroidal transformer with spatiotemporal features for visual tracking, IEEE Trans. Human Mach. Syst., 54 (2024), 214–225. https://doi.org/10.1109/THMS.2024.3370582 doi: 10.1109/THMS.2024.3370582
![]() |
[35] |
O. Abdelaziz, M. Shehata, DMTrack: Learning deformable masked visual representations for single object tracking, SIViP, 19 (2025), 61. https://doi.org/10.1007/s11760-024-03713-0 doi: 10.1007/s11760-024-03713-0
![]() |
[36] | A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, et al., Attention is all you need, in the 31st International Conference on Neural Information Processing Systems (NIPS), (2017), 6000–6010. |
[37] |
O. C. Koyun, R. K. Keser, S. O. Şahin, D. Bulut, M. Yorulmaz, V. Yücesoy, et al., RamanFormer: A Transformer-based quantification approach for raman mixture components, ACS Omega, 9 (2024), 23241–23251. https://doi.org/10.1021/acsomega.3c09247 doi: 10.1021/acsomega.3c09247
![]() |
[38] | H. Fan, X. Wang, S. Li, H. Ling, Joint feature learning and relation modeling for tracking: A one-stream framework, in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), (2022), 341–357. https://doi.org/10.1007/978-3-031-20047-2_20 |
[39] |
H. Zhang, J. Song, H. Liu, Y. Han, Y. Yang, H. Ma, AwareTrack: Object awareness for visual tracking via templates interaction, Image Vision Comput., 154 (2025), 105363. https://doi.org/10.1016/j.imavis.2024.105363 doi: 10.1016/j.imavis.2024.105363
![]() |
[40] |
Z. Wang, L. Yuan, Y. Ren, S. Zhang, H. Tian, ADSTrack: Adaptive dynamic sampling for visual tracking, Complex Intell. Syst., 11 (2025), 79. https://doi.org/10.1007/s40747-024-01672-0 doi: 10.1007/s40747-024-01672-0
![]() |
[41] | X. Chen, B. Yan, J. Zhu, D. Wang, X. Yang, H. Lu, Transformer tracking, in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), (2021), 8122–8131. https://doi.org/10.1109/CVPR46437.2021.00803 |
[42] | A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, et al., An image is worth 16x16 words: Transformers for image recognition at scale, preprint, arXiv: 2010.11929. https://doi.org/10.48550/arXiv.2010.11929 |
[43] | B. Yan, H. Peng, J. Fu, D. Wang, H. Lu, Learning spatio-temporal transformer for visual tracking, in IEEE International Conference on Computer Vision (ICCV), (2021), 10428–10437. https://doi.org/10.1109/ICCV48922.2021.01028 |
[44] | Z. Liu, Y. Lin, Y. Cao, H. Hu, Y. Wei, Z. Zhang, et al., Swin Transformer: Hierarchical vision transformer using shifted windows, in IEEE/CVF International Conference on Computer Vision (ICCV), (2021), 10012–10022. https://doi.org/10.1109/ICCV48922.2021.00986 |
[45] | L. Lin, H. Fan, Z. Zhang, Y. Xu, H. Ling, SwinTrack: A simple and strong baseline for transformer tracking, in Advances in Neural Information Processing Systems (NIPS), 35 (2022), 16743–16754. |
[46] | Z. Song, J. Yu, Y. P. P. Chen, W. Yang, Transformer tracking with cyclic shifting window attention, in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), (2022), 8781–8790. https://doi.org/10.1109/CVPR52688.2022.00859 |
[47] |
Y. Chen, K. Chen, Four mathematical modeling forms for correlation filter object tracking algorithms and the fast calculation for the filter, Electron. Res. Arch., 32 (2024), 4684–4714. https://doi.org/10.3934/era.2024213 doi: 10.3934/era.2024213
![]() |
[48] | H. Fan, L. Lin, F. Yang, P. Chu, G. Deng, S. Yu, et al., LaSOT: A high-quality benchmark for large-scale single object tracking, in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), (2019), 5369–5378. https://doi.org/10.1109/CVPR.2019.00552 |
[49] | Y. Wu, J. Lim, M. -H. Yang, Online object tracking: A benchmark, in IEEE Conference on Computer Vision and Pattern Recognition (CVPR), (2013), 2411–2418. https://doi.org/10.1109/CVPR.2013.312 |
[50] | M. Mueller, N. Smith, B. Ghanem, A benchmark and simulator for UAV tracking, in Computer Vision–ECCV 2016, (2016), 445–461. https://doi.org/10.1007/978-3-319-46448-0_27 |
[51] |
Y. Huang, Y. Chen, C. Lin, Q. Hu, J. Song, Visual attention learning and antiocclusion-based correlation filter for visual object tracking, J. Electron. Imaging, 32 (2023), 23. https://doi.org/10.1117/1.JEI.32.1.013023 doi: 10.1117/1.JEI.32.1.013023
![]() |
[52] | K. He, X. Zhang, S. Ren, J. Sun, Deep residual learning for image recognition, in IEEE Conference on Computer Vision and Pattern Recognition (CVPR), (2016), 770–778. |
[53] | N. Carion, F. Massa, G. Synnaeve, N. Usunier, A. Kirillov, S. Zagoruyko, End-to-end object detection with transformers, in European Conference on Computer Vision (ECCV), (2020), 213–229. https://doi.org/10.1007/978-3-030-58452-8_13 |
[54] | D. P. Kingma, J. Ba, Adam: A method for stochastic optimization, preprint, arXiv: 1412.6980. https://doi.org/10.48550/arXiv.1412.6980 |
[55] |
Y. Cui, C. Jiang, G. Wu, L. Wang, MixFormer: End-to-end tracking with iterative mixed attention, IEEE Trans. Pattern Anal. Mach. Intell., 46 (2024), 4129–4146. https://doi.org/10.1109/TPAMI.2024.3349519 doi: 10.1109/TPAMI.2024.3349519
![]() |
[56] |
J. Shen, Y. Liu, X. Dong, X. Lu, F. S. Khan, S. Hoi, Distilled siamese networks for visual tracking, IEEE Trans. Pattern Anal. Mach. Intell., 44 (2022), 8896–8909. https://doi.org/10.1109/TPAMI.2021.3127492 doi: 10.1109/TPAMI.2021.3127492
![]() |
[57] |
X. Dong, J. Shen, F. Porikli, J. Luo, L. Shao, Adaptive siamese tracking with a compact latent network, IEEE Trans. Pattern Anal. Mach. Intell., 45 (2023), 8049–8062. https://doi.org/10.1109/TPAMI.2022.3230064 doi: 10.1109/TPAMI.2022.3230064
![]() |
[58] |
Z. Cao, Z. Huang, L. Pan, S. Zhang, Z. Liu, C. Fu, Towards real-world visual tracking with temporal contexts, IEEE Trans. Pattern Anal. Mach. Intell., 45 (2023), 15834–15849. https://doi.org/10.1109/TPAMI.2023.3307174 doi: 10.1109/TPAMI.2023.3307174
![]() |
[59] |
Y. Yang, X. Gu, Attention-based gating network for robust segmentation tracking, IEEE Trans. Circuits Syst. Video Technol., 35 (2025), 245–258. https://doi.org/10.1109/TCSVT.2024.3460400 doi: 10.1109/TCSVT.2024.3460400
![]() |
[60] |
W. Han, X. Dong, Y. Zhang, D. Crandall, C. Z. Xu, J. Shen, Asymmetric Convolution: An efficient and generalized method to fuse feature maps in multiple vision tasks, IEEE Trans. Pattern Anal. Mach. Intell., 46 (2024), 7363–7376. https://doi.org/10.1109/TPAMI.2024.3400873 doi: 10.1109/TPAMI.2024.3400873
![]() |
[61] |
X. Zhu, Y. Wu, D. Xu, Z. Feng, J. Kittler, Robust visual object tracking via adaptive attribute-aware discriminative correlation filters, IEEE Trans. Multimedia, 23 (2021), 2625–2638. https://doi.org/10.1109/TMM.2021.3050073 doi: 10.1109/TMM.2021.3050073
![]() |
[62] | M. Danelljan, H. Gustav, F. Shahbaz Khan, M. Felsberg, Learning spatially regularized correlation filters for visual tracking, in IEEE International Conference on Computer Vision (ICCV), (2015), 4310–4318. https://doi.org/10.1109/ICCV.2015.490 |
[63] | J. Valmadre, L. Bertinetto, J. Henriques, A. Vedaldi, P. H. S. Torr, End-to-end representation learning for correlation filter based tracking, in IEEE Conference on Computer Vision and Pattern Recognition (CVPR), (2017), 5000–5008. https://doi.org/10.1109/CVPR.2017.531 |
[64] | G. Bhat, M. Danelljan, L. V. Gool, R. Timofte, Learning discriminative model prediction for tracking, in IEEE/CVF International Conference on Computer Vision (ICCV), (2019), 6182–6191. https://doi.org/10.1109/ICCV.2019.00628 |
[65] | N. Wang, W. Zhou, J. Wang, H. Li, Transformer meets tracker: exploiting temporal context for robust visual tracking, in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), (2021), 1571–1580. https://doi.org/10.1109/CVPR46437.2021.00162 |
[66] | Z. Chen, B. Zhong, G. Li, S. Zhang, R. Ji, Siamese box adaptive network for visual tracking, in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), (2020), 6668–6677. https://doi.org/10.1109/CVPR42600.2020.00670 |
[67] | Y. Guo, H. Li, L. Zhang, L. Zhang, K. Deng, F. Porikli, SiamCAR: Siamese fully convolutional classification and regression for visual tracking, in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), (2020), 1176–1185. https://doi.org/10.1109/CVPR42600.2020.00630 |
[68] | D. Xing, N. Evangeliou, A. Tsoukalas, A. Tzes, Siamese transformer pyramid networks for real-time UAV tracking, in IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), (2022), 1898–1907. https://doi.org/10.1109/WACV51458.2022.00196 |
1. | Ying Yu, Yahui Chen, You Zhou, Cross-Diffusion-Induced Turing Instability in a Two-Prey One-Predator System, 2023, 11, 2227-7390, 2411, 10.3390/math11112411 | |
2. | Debjit Pal, Dipak Kesh, Debasis Mukherjee, Pattern dynamics in a predator–prey model with Smith growth function and prey refuge in predator poisoned environment, 2024, 92, 05779073, 366, 10.1016/j.cjph.2024.09.015 | |
3. | Xiaoyan Zhao, Liangru Yu, Xue-Zhi Li, Dynamics analysis of a predator-prey model incorporating fear effect in prey species, 2025, 10, 2473-6988, 12464, 10.3934/math.2025563 |