摘要
圈养东北虎日常生活中遮挡、非线性运动以及模糊态等复杂情况频发,造成多目标跟踪效果较差,易出现ID切换的问题。在Bytetrack跟踪算法的基础上,采用扩展卡尔曼滤波(EKF)能更好地应对东北虎运动过程中的非线性动态特性。采用插值轨迹的方法易应对东北虎在目标检测过程中容易出现漏检的问题。为了处理东北虎活动中出现的遮挡与重叠现象,提出重叠轨迹概念。结果显示,改进后Bytetrack算法的IDS为68,减少了97,MOTA为82.5%,提高了5.2%,IDF1为91.6%,提高了4.2%,这表明在遇到遮挡等复杂场景时,改进后Bytetrack算法的跟踪精度和鲁棒性得到了提升。
In the daily life of captive Northeast tigers,complex situations such as occlusion,nonlinear motion,and blurry states occur frequently,resulting in poor multi-target tracking performance and easy ID switching problems.Based on the Bytetrack tracking algorithm,the Extended Kalman Filter(EKF)is adopted to better cope with the nonlinear dynamic characteristics of the Northeast tiger during its movement.The interpolation trajectory method is adopted to easily address the problem of missed detections in the target detection process of Northeast tigers.In order to address the occlusion and overlap phenomena that occur during the activities of the Northeast tiger,the concept of overlapping trajectories is proposed.The results showed that the improved Bytetrack algorithm had an IDS of 68,a reduction of 97,a MOTA of 82.5%,an increase of 5.2%,and an IDF1 of 91.6%,an increase of 4.2%.This indicates that the improved Bytetrack algorithm has improved tracking accuracy and robustness in complex scenes such as occlusion.
作者
姜正海
马光凯
JIANG Zhenghai;MA Guangkai(School of Computer and Control Engineering,Northeast Forestry University,Harbin,Heilongjiang 150040,China)
出处
《自动化应用》
2025年第22期94-98,共5页
Automation Application
