摘要
多个运动目标跟踪是计算机视觉研究中的一个难点,具有较大的挑战性。目前研究较多的是单个运动目标跟踪,相比之下,多个运动目标跟踪的难度更高。多目标跟踪会因各运动目标之间相互遮挡,而造成跟踪漂移的问题,最终无法完成目标跟踪。针对此问题,本文将深度学习框架应用于多运动目标跟踪,提出一种基于深度学习抗遮挡的多目标跟踪算法,用于智能交通视频多目标跟踪场景中。实验结果表明,基于深度学习抗遮挡的多目标跟踪算法,能够较好地解决跟踪漂移问题,提高了多目标跟踪的准确性。
Multiple moving object tracking is a difficult point in computer vision research,which is more challenging.At present single moving object tracking is more studied.In contrast,multiple moving object tracking is more difficult.Multi-target tracking will cause tracking drift due to mutual occlusion between moving objects,and eventually cannot complete target tracking.Because deep learning Network model has made great achievements in tracking single moving object,this paper applies deep learning framework to multi-target tracking and proposes a multi-target tracking algorithm based on deep learning and anti occlusion,which is applied to multi-target tracking scene of intelligent transportation video.The experimental results showthat the multi-target tracking algorithm based on deep learning anti occlusion can achieve better performance.The problem of tracking drift is solved and the accuracy of multi-target tracking is improved.
作者
左国才
苏秀芝
陈明丽
匡林爱
吴小平
ZUO Guocai;SU Xiuzhi;CHEN Mingli;KUANG Lin'ai;WU Xiaoping(School of Software and Information Engineering,Hunan Software Vocational Institute,Xiangtan Hunan 411100,China;Xinhua Chuyi Industrial School,Xinhua Hunan 417600,China)
出处
《智能计算机与应用》
2020年第7期239-242,共4页
Intelligent Computer and Applications
基金
湖南省自然科学基金(2020JJ7007)
关键词
深度学习
多目标跟踪
抗遮挡
Deep Learning
multi-target tracking
anti occlusion
