基于正则化卷积神经网络的目标跟踪算法
|
摘要
为了提高目标物体的跟踪鲁棒性和稳定性,文中将L2正则化最小二乘法和卷积神经网络(CNN)相互结合,提出了一种基于正则化卷积神经网络的目标跟踪算法。通过L2跟踪器来评估目标无题被遮挡的程度,利用两层CNN对目标进行目标表示,去除了大部分无关样本,降低了算法的复杂度。实验结果表明,当目标物体发生姿态变化或旋转等剧烈的外观变化时,所提算法具有较强的鲁棒性和稳定性,并且比其他经典的跟踪算法具有更高的精度。
In order to improve the tracking robustness and stability of the target object,a target tracking algorithm based on regularized convolution neural network( CNN) was proposed in this paper,which combines L2 regularized least squares method with convolution neural network( CNN). L2 tracker was used to evaluate the degree of occlusion of the target. Two-layer CNN was used to represent the target robustly. Most of the irrelevant samples were removed and the complexity of the algorithm was reduced.The experimental results show that the proposed algorithm has strong robustness and stability,and has higher accuracy than other classical tracking algorithms when the object changes its formation dramatically.
In order to improve the tracking robustness and stability of the target object,a target tracking algorithm based on regularized convolution neural network( CNN) was proposed in this paper,which combines L2 regularized least squares method with convolution neural network( CNN). L2 tracker was used to evaluate the degree of occlusion of the target. Two-layer CNN was used to represent the target robustly. Most of the irrelevant samples were removed and the complexity of the algorithm was reduced.The experimental results show that the proposed algorithm has strong robustness and stability,and has higher accuracy than other classical tracking algorithms when the object changes its formation dramatically.
引文
[1]黄健,郭志波,林科军.一种基于核相关滤波的视觉跟踪算法[J].计算机科学,2018,45(S2):230-233.
[2]管皓,薛向阳,安志勇.在线单目标视频跟踪算法综述[J].小型微型计算机系统,2017,38(1):147-153.
[3]李慧霞,李临生,闫庆森,等.基于Mean Shift算法的目标跟踪综述[J].计算机与现代化,2017(1):65-70.
[4]王洪,牛晓灵.基于l2正则化回声状态网络的模拟电路故障诊断[J].电子器件,2017,40(5):1283-1286.
[5]郭彦麟,张冬. lp范数正则化视觉跟踪[J].小型微型计算机系统,2017,38(7):1648-1652.
[6]Kwak N. Principal component analysis by L_{p}-norm maximization[J]. IEEE Transactions on Cybernetics,2014,44(5):594-609.
[7] Tsagkarakis N,Markopoulos P P,Sklivanitis G,et al. L1-norm principal-component analysis of complex data[J]. IEEE Transactions on Signal Processing,2018,66(12):3256-3267.
[8]Li Y,Su H,Qi C R,et al. Joint embeddings of shapes and images via cnn image purification[J]. ACM Transactions on Graphics(TOG),2015,34(6):234-239.
[9]冯珂垚,饶鹏,陆福星,等.基于神经网络的高分辨率快速目标检测方法[J].电子设计工程,2018(22):169-173.
[10]晁安娜,刘坤.基于卷积神经网络的遥感图像飞机目标识别[J].微型机与应用,2017,36(22):66-69,73.
[11]阮威.基于动态粒子群优化与K均值聚类的图像分割算法[J].信息技术,2018,42(10):126-130.
[12]Zhao Q,Zhang B,Lyu S,et al. A CNN-SIFT hybrid pedestrian navigation method based on First-Person vision[J]. Remote Sensing,2018,10(8):1229.
[13]Luo M,Nie F,Chang X,et al. Avoiding optimal mean2,1-Norm Maximization-Based robust PCA for reconstruction[J]. Neural computation,2017,29(4):1124-1150.
[14] He Z,Yi S,Cheng Y M,et al. Robust object tracking via key patch sparse representation[J]. IEEE transactions on cybernetics,2017,47(2):354-364.
[15]Milan A,Schindler K,Roth S. Multi-target tracking by discretecontinuous energy minimization[J]. IEEE transactions on pattern analysis and machine intelligence,2016,38(10):2054-2068.
[16]Wu Y,Lim J,Yang M H. Object tracking benchmark[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2015,37(9):1834-1848.
[17]茅正冲,黄舒伟.基于结构化判别稀疏表示的目标跟踪[J].南京理工大学学报,2018,42(3):271-277.
[18]郭克友,暴启超.改进的核相关滤波器的目标跟踪算法[J].计算机工程与设计,2018,39(3):769-773,797.
[19] Xiao Z,Lu H,Wang D. L2-RLS-Based Object Tracking[J].IEEE Trans. Circuits Syst. Video Techn,2014,24(8):1301-1309.
[2]管皓,薛向阳,安志勇.在线单目标视频跟踪算法综述[J].小型微型计算机系统,2017,38(1):147-153.
[3]李慧霞,李临生,闫庆森,等.基于Mean Shift算法的目标跟踪综述[J].计算机与现代化,2017(1):65-70.
[4]王洪,牛晓灵.基于l2正则化回声状态网络的模拟电路故障诊断[J].电子器件,2017,40(5):1283-1286.
[5]郭彦麟,张冬. lp范数正则化视觉跟踪[J].小型微型计算机系统,2017,38(7):1648-1652.
[6]Kwak N. Principal component analysis by L_{p}-norm maximization[J]. IEEE Transactions on Cybernetics,2014,44(5):594-609.
[7] Tsagkarakis N,Markopoulos P P,Sklivanitis G,et al. L1-norm principal-component analysis of complex data[J]. IEEE Transactions on Signal Processing,2018,66(12):3256-3267.
[8]Li Y,Su H,Qi C R,et al. Joint embeddings of shapes and images via cnn image purification[J]. ACM Transactions on Graphics(TOG),2015,34(6):234-239.
[9]冯珂垚,饶鹏,陆福星,等.基于神经网络的高分辨率快速目标检测方法[J].电子设计工程,2018(22):169-173.
[10]晁安娜,刘坤.基于卷积神经网络的遥感图像飞机目标识别[J].微型机与应用,2017,36(22):66-69,73.
[11]阮威.基于动态粒子群优化与K均值聚类的图像分割算法[J].信息技术,2018,42(10):126-130.
[12]Zhao Q,Zhang B,Lyu S,et al. A CNN-SIFT hybrid pedestrian navigation method based on First-Person vision[J]. Remote Sensing,2018,10(8):1229.
[13]Luo M,Nie F,Chang X,et al. Avoiding optimal mean2,1-Norm Maximization-Based robust PCA for reconstruction[J]. Neural computation,2017,29(4):1124-1150.
[14] He Z,Yi S,Cheng Y M,et al. Robust object tracking via key patch sparse representation[J]. IEEE transactions on cybernetics,2017,47(2):354-364.
[15]Milan A,Schindler K,Roth S. Multi-target tracking by discretecontinuous energy minimization[J]. IEEE transactions on pattern analysis and machine intelligence,2016,38(10):2054-2068.
[16]Wu Y,Lim J,Yang M H. Object tracking benchmark[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2015,37(9):1834-1848.
[17]茅正冲,黄舒伟.基于结构化判别稀疏表示的目标跟踪[J].南京理工大学学报,2018,42(3):271-277.
[18]郭克友,暴启超.改进的核相关滤波器的目标跟踪算法[J].计算机工程与设计,2018,39(3):769-773,797.
[19] Xiao Z,Lu H,Wang D. L2-RLS-Based Object Tracking[J].IEEE Trans. Circuits Syst. Video Techn,2014,24(8):1301-1309.