基于视频的生猪行为跟踪和识别对于实现精细化养殖具有重要价值。为了应对群养生猪多目标跟踪任务中由猪只外观相似、遮挡交互等因素带来的挑战,研究提出了基于PigsTrack跟踪器的群养生猪多目标跟踪方法。PigsTrack跟踪器利用高性能YOLO...基于视频的生猪行为跟踪和识别对于实现精细化养殖具有重要价值。为了应对群养生猪多目标跟踪任务中由猪只外观相似、遮挡交互等因素带来的挑战,研究提出了基于PigsTrack跟踪器的群养生猪多目标跟踪方法。PigsTrack跟踪器利用高性能YOLOX网络降低目标误检与漏检率,采用Transformer模型获取具有良好区分特性的目标外观特征;基于OC-SORT(observation-centric sort)的思想,通过集成特征匹配、IoU匹配和遮挡恢复匹配策略实现群养生猪的准确跟踪。基于PBVD(pigs behaviours video dataset)数据集的试验结果表明,PigsTrack跟踪器的HOTA(higher order tracking accuracy),MOTA(multiple object tracking accuracy)和IDF1得分(identification F1 score)分别为85.66%、98.59%和99.57%,相较于现有算法的最高精度,分别提高了3.71、0.03和2.05个百分点,证明了PigsTrack跟踪器在解决外观相似和遮挡交互引起的跟踪过程中身份跳变问题方面的有效性。随后,利用Slowfast网络对PigsTrack跟踪器的跟踪结果进行了典型行为统计,结果显示PigsTrack在群养生猪个体行为统计方面更准确。此外,通过在ABVD(aggressive-behavior video)数据集上的试验,PigsTrack跟踪器的HOTA、MOTA和IDF1得分分别为69.14%、94.82%和90.11%,相对于现有算法的最高精度,提高了5.33、0.57和8.60个百分点,验证了PigsTrack跟踪器在群养生猪跟踪任务中的有效性。总而言之,PigsTrack跟踪器能够有效应对外观相似和遮挡交互等挑战,实现了准确的生猪多目标跟踪,并在行为统计方面展现出更高的准确性,为生猪养殖领域的研究和实际应用提供了有价值的指导。展开更多
Abstract:Sparse coding(SC)based visual tracking(l1-tracker)is gaining increasing attention,and many related algorithms are developed.In these algorithms,each candidate region is sparsely represented as a set of target...Abstract:Sparse coding(SC)based visual tracking(l1-tracker)is gaining increasing attention,and many related algorithms are developed.In these algorithms,each candidate region is sparsely represented as a set of target templates.However,the structure connecting these candidate regions is usually ignored.Lu proposed an NLSSC-tracker with non-local self-similarity sparse coding to address this issue,which has a high computational cost.In this study,we propose an Euclidean local-structure constraint based sparse coding tracker with a smoothed Euclidean local structure.With this tracker,the optimization procedure is transformed to a small-scale l1-optimization problem,significantly reducing the computational cost.Extensive experimental results on visual tracking demonstrate the eectiveness and efficiency of the proposed algorithm.展开更多
文摘基于视频的生猪行为跟踪和识别对于实现精细化养殖具有重要价值。为了应对群养生猪多目标跟踪任务中由猪只外观相似、遮挡交互等因素带来的挑战,研究提出了基于PigsTrack跟踪器的群养生猪多目标跟踪方法。PigsTrack跟踪器利用高性能YOLOX网络降低目标误检与漏检率,采用Transformer模型获取具有良好区分特性的目标外观特征;基于OC-SORT(observation-centric sort)的思想,通过集成特征匹配、IoU匹配和遮挡恢复匹配策略实现群养生猪的准确跟踪。基于PBVD(pigs behaviours video dataset)数据集的试验结果表明,PigsTrack跟踪器的HOTA(higher order tracking accuracy),MOTA(multiple object tracking accuracy)和IDF1得分(identification F1 score)分别为85.66%、98.59%和99.57%,相较于现有算法的最高精度,分别提高了3.71、0.03和2.05个百分点,证明了PigsTrack跟踪器在解决外观相似和遮挡交互引起的跟踪过程中身份跳变问题方面的有效性。随后,利用Slowfast网络对PigsTrack跟踪器的跟踪结果进行了典型行为统计,结果显示PigsTrack在群养生猪个体行为统计方面更准确。此外,通过在ABVD(aggressive-behavior video)数据集上的试验,PigsTrack跟踪器的HOTA、MOTA和IDF1得分分别为69.14%、94.82%和90.11%,相对于现有算法的最高精度,提高了5.33、0.57和8.60个百分点,验证了PigsTrack跟踪器在群养生猪跟踪任务中的有效性。总而言之,PigsTrack跟踪器能够有效应对外观相似和遮挡交互等挑战,实现了准确的生猪多目标跟踪,并在行为统计方面展现出更高的准确性,为生猪养殖领域的研究和实际应用提供了有价值的指导。
基金National Natural Foundation of China under Grant(61572085,61502058)
文摘Abstract:Sparse coding(SC)based visual tracking(l1-tracker)is gaining increasing attention,and many related algorithms are developed.In these algorithms,each candidate region is sparsely represented as a set of target templates.However,the structure connecting these candidate regions is usually ignored.Lu proposed an NLSSC-tracker with non-local self-similarity sparse coding to address this issue,which has a high computational cost.In this study,we propose an Euclidean local-structure constraint based sparse coding tracker with a smoothed Euclidean local structure.With this tracker,the optimization procedure is transformed to a small-scale l1-optimization problem,significantly reducing the computational cost.Extensive experimental results on visual tracking demonstrate the eectiveness and efficiency of the proposed algorithm.
