结合多尺度改进颜色特征应对遮挡的跟踪算法
|
摘要
针对目标跟踪所面临难题中的遮挡及尺度变化等问题,在相关滤波算法CN的基础上提出了一种在多尺度下改进颜色特征的目标跟踪算法。首先,在检测阶段提取搜索区域CN的同时提取梯度直方图特征,生成多通道特征,自适应的融合两种特征信息,并对图像特征信息采用降维压缩技术,来提升跟踪速度。其次,采用尺度金字塔原理对目标遮挡及尺度变化的目标进行搜索和自适应跟踪。实验结果表明:改进后的算法在非语义遮挡、语义遮挡及尺度变化等情况下均有较高的鲁棒性。因此,融合多种特征可以更好地描述目标图像信息,提高跟踪器的精度和鲁棒性,而多尺度比较可以实现在发生目标遮挡后在一定范围内搜索目标,保证跟踪效果。
In order to solve the problems of occlusion and scale change in target tracking,a target tracking algorithm based on correlation filtering algorithm CN was proposed to improve color features at multiple scales. The CN of the search area was extracted while the gradient histogram feature was extracted in the detection stage The multi-channel feature was generated,the two feature information were adaptively fused,and the image feature information was reduced in dimension to improve the tracking speed. The principle of scale pyramid was used to search and adaptively track targets with occlusion and scale change.Experimental results show that the improved algorithm has high robustness in non-semantic occlusion,semantic occlusion and scale change. Therefore,the fusion of multiple features can better describe the target image information and improve the accuracy and robustness of the tracker,while the multi-scale comparison can realize searching the target within a certain range after the target occlusion occurs and ensure the tracking effect.
In order to solve the problems of occlusion and scale change in target tracking,a target tracking algorithm based on correlation filtering algorithm CN was proposed to improve color features at multiple scales. The CN of the search area was extracted while the gradient histogram feature was extracted in the detection stage The multi-channel feature was generated,the two feature information were adaptively fused,and the image feature information was reduced in dimension to improve the tracking speed. The principle of scale pyramid was used to search and adaptively track targets with occlusion and scale change.Experimental results show that the improved algorithm has high robustness in non-semantic occlusion,semantic occlusion and scale change. Therefore,the fusion of multiple features can better describe the target image information and improve the accuracy and robustness of the tracker,while the multi-scale comparison can realize searching the target within a certain range after the target occlusion occurs and ensure the tracking effect.
引文
[1] FELSBERG M,KRISTAN M,MATAS J,et al. The Thermal Infrared Visual Object Tracking VOT-TIR2016 Challenge Results[M]. Computer Vision-ECCV 2016 Workshops.Springer International Publishing,2016:824-849.
[2] BOLME D S,BEVERIDGE J R,DRAPER B A,et al. Visual object tracking using adaptive correlation filters[C]//Computer Vision and Pattern Recognition. IEEE,2010:2544-2550.
[3] HENRIQUES J F,RUI C,MARTINS P,et al. Exploiting the Circulant Structure of Tracking-by-Detection with Kernels[J]. Lecture Notes in Computer Science,2012,7575(1):702-715.
[4] HENRIQUES J F,RUI C,MARTINS P,et al. High-Speed Tracking with Kernelized Correlation Filters[J]. IEEE Transactions on Pattern Analysis&Machine Intelligence,2014,37(3):583.
[5] DANELLJAN M,KHAN F S,FELSBERG M,et al. Adaptive Color Attributes for Real-Time Visual Tracking[C]//Computer Vision and Pattern Recognition. IEEE,2014:1090-1097.
[6] DANELLJAN M,HAGER G,KHAN F S. Accurate scale estimation for robust visual tracking[C]//British Machine Vision Conference. 2014.
[7]张昊.基于多尺度金字塔特征块提取HOG特征的新型人脸识别算法[D].长春:吉林大学,2017.
[8] WU Y,LIM J,YANG M H. Object tracking benchmark[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2015,37(9):1834-1848.
[9]王暐,王春平,李军,等.特征融合和模型自适应更新相结合的相关滤波目标跟踪[J].光学精密工程,2016,24(8):2059-2066.
[10]沈秋,严小乐,刘霖枫,等.基于自适应特征选择的多尺度相关滤波跟踪[J].光学学报,2017(5):174-183.
[2] BOLME D S,BEVERIDGE J R,DRAPER B A,et al. Visual object tracking using adaptive correlation filters[C]//Computer Vision and Pattern Recognition. IEEE,2010:2544-2550.
[3] HENRIQUES J F,RUI C,MARTINS P,et al. Exploiting the Circulant Structure of Tracking-by-Detection with Kernels[J]. Lecture Notes in Computer Science,2012,7575(1):702-715.
[4] HENRIQUES J F,RUI C,MARTINS P,et al. High-Speed Tracking with Kernelized Correlation Filters[J]. IEEE Transactions on Pattern Analysis&Machine Intelligence,2014,37(3):583.
[5] DANELLJAN M,KHAN F S,FELSBERG M,et al. Adaptive Color Attributes for Real-Time Visual Tracking[C]//Computer Vision and Pattern Recognition. IEEE,2014:1090-1097.
[6] DANELLJAN M,HAGER G,KHAN F S. Accurate scale estimation for robust visual tracking[C]//British Machine Vision Conference. 2014.
[7]张昊.基于多尺度金字塔特征块提取HOG特征的新型人脸识别算法[D].长春:吉林大学,2017.
[8] WU Y,LIM J,YANG M H. Object tracking benchmark[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2015,37(9):1834-1848.
[9]王暐,王春平,李军,等.特征融合和模型自适应更新相结合的相关滤波目标跟踪[J].光学精密工程,2016,24(8):2059-2066.
[10]沈秋,严小乐,刘霖枫,等.基于自适应特征选择的多尺度相关滤波跟踪[J].光学学报,2017(5):174-183.