基于改进Bytetrack的群体机器人跟踪算法  

Tracking algorithm for swarm robots based on improved Bytetrack

作　　者：李正龙 雷斌[1,2] 蒋林 唐雄[1,2] LI Zhenglong;LEI Bin;JIANG Lin;TANG Xiong(College of Machinery and Automation,Wuhan University of Science and Technology,Wuhan 430081,Hubei,China;Institute of Robotics and Intelligent Systems,Wuhan University of Science and Technology,Wuhan 430081,Hubei,China)

机构地区：[1]武汉科技大学机械自动化学院,湖北武汉430081 [2]武汉科技大学机器人与智能系统研究院,湖北武汉430081

出　　处：《农业装备与车辆工程》2024年第2期107-111,共5页Agricultural Equipment & Vehicle Engineering

基　　金：国家重点研发计划(2019YFB1310000);湖北省自然科学基金(2018CFB626)。

摘　　要：针对群体机器人在实际跟踪场景中常常受到遮挡、目标密集、尺度变换等因素的影响导致漏检、轨迹中断和ID频繁跳变等问题,基于Bytetrack跟踪算法,改进了卡尔曼滤波的状态变量,提出了噪声尺度自适应卡尔曼算法(NASA-Kalman),并在卡尔曼滤波中引入加速度参数(AMA)提高跟踪的准确性。实验表明,在MOTA、MOTP方面相较于原算法均有所提高。为了进一步验证跟踪算法的有效性,在MOT20数据集上对算法进行了评估,在MOTA、MOTP方面分别提高了0.65%和1.26%。In practical tracking scenarios,swarm robots are often affected by factors such as occlusion,target density,and scale transformation,resulting in missed detection,trajectory interruption,and frequent ID jumps.Based on the Bytetrack tracking algorithm,the state variables of the Kalman filter was improved,a noise scale adaptive Kalman algorithm(NASA-Kalman)was proposed,and acceleration parameters(AMA)were introduced into Kalman filter to improve the tracking accuracy.Compared to the original algorithm,MOTA and MOTP were improved.In order to further verify the effectiveness of the tracking algorithm,the algorithm was evaluated on the MOT20 dataset,and improved by 0.65%and 1.26%in MOTA and MOTP respectively.

关 键 词：群体机器人 目标跟踪 卡尔曼滤波 加速度参数 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]