基于颜色与深度信息特征融合的一种多目标跟踪新算法  

A novel multi-object tracking algorithm based on RJMCMC with RGB and depth information fusion

作　　者：姜明新[1,2] 邱天爽[1] 迟宏波[3] 

机构地区：[1]大连理工大学电子信息与电气工程学部,辽宁大连116024 [2]大连民族学院信息与通信工程学院,辽宁大连116600 [3]首都航天机械公司,北京100076

出　　处：《光电子．激光》2015年第7期1342-1348,共7页Journal of Optoelectronics·Laser

基　　金：国家自然科学基金(6143060);中国博士后科学基金(2014M551081;2015T80249)资助项目

摘　　要：提出了一种基于颜色与深度信息特征融合的可逆跳的马尔科夫链蒙特长洛(RJMCMC)多目标跟踪算法。首先,融合彩色信息和深度信息对运动目标进行检测;然后,根据多目标检测的结果建立观测似然模型,并构建合理的状态转移模型;最后,通过RJMCMC粒子滤波算法实现多目标跟踪。实验结果表明,本文提出的多目标跟踪算法具有较强的鲁棒性,能够稳定的跟踪多目标,具有较高的准确率。Multi-object tracking technique is one of the challenging issues in computer vision.In this paper,we propose a multi-object tracking algorithm based on RJMCMC with RGB and depth information fusion.Firstly,the moving objects are detected by fusing color information and depth information.Then,the observation likelihood model is established according to the result of object detection,and a reasonable state transition model is also established.Finally,the multi-object tracking result is obtained using RJMCMC particle filtering algorithm.Experimental results show that the proposed method has strong robustness.The algorithm can steadily track multiple targets and has higher accuracy.

关 键 词：RGB-D数据 多目标检测 多目标跟踪 可逆跳的马尔科夫链蒙特长洛(RJMCMC)粒子滤波 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]