Journal of Frontiers of Computer Science and Technology ›› 2020, Vol. 14 ›› Issue (10): 1762-1775.DOI: 10.3778/j.issn.1673-9418.1908068

Previous Articles     Next Articles

Detection Optimized Multi-Bernoulli Algorithm for Visual Multi-target Tracking

YANG Jinlong, CHENG Xiaoxue, MIAO Jiani, ZHANG Guangnan   

  1. 1. School of Internet of Things Engineering, Jiangnan University, Wuxi, Jiangsu 214122, China
    2. School of Information Engineering, Chang'an University, Xi??an 710064, China
  • Online:2020-10-01 Published:2020-10-12

检测优化的多伯努利视频多目标跟踪算法

杨金龙程小雪缪佳妮张光南   

  1. 1. 江南大学 物联网工程学院,江苏 无锡 214122
    2. 长安大学 信息工程学院,西安 710064

Abstract:

For the problems of the Multi-Bernoulli filter (MBF) in the multi-target tracking filed, such as the newborn targets cannot be detected, the tracking accuracy decreases or the targets are underestimated when the targets are occluded, and the underestimated targets cannot be associated with their previous tracks when they are redetected. The YOLOv3 method is introduced to predetect the targets under the framework of MBF, and then the predetected targets are presented by the convolution features and their similarity matrix are calculated between the adjacent frames. The detection strategy is proposed to realize the discrimination of newborn target. The reiden-tifiable strategy is proposed to achieve the continuous estimation of underestimated targets. In addition, when the template is updated, the high-confidence detection is integrated, and the occluded target processing mechanism is proposed to effectively improve the accuracy of target tracking. Finally, the proposed algorithm is tested on some challenging video sequences from the public standard dataset. The results show that the proposed algorithm has a good performance on detecting the newborn targets and the underestimated targets, realizing continuous tracking of visual multi-target trajectory.

Key words: multi-Bernoulli filter, tracking-by-detection, newborn target, similarity matrix

摘要:

针对多伯努利滤波方法在多目标跟踪时,难以检测新生目标,且当目标出现互相遮挡等干扰时,跟踪精度下降,甚至出现目标漏跟,以及当漏跟目标被重新跟踪后,与之前运动轨迹难以关联等问题,在多伯努利滤波框架下,引入YOLOv3检测算法,并采用卷积特征对目标进行描述,计算相邻帧目标的相似度矩阵,设计新生目标识别和漏跟目标的重识别策略,以实现对目标新生判别和漏跟目标的连续估计;此外,在模板更新时,融合高置信度检测框,提出遮挡目标处理机制,有效提高目标跟踪精度。最后,采用标准数据集中具有挑战性的视频序列进行算法测试,结果表明,提出算法能有效识别新生目标和漏跟目标,实现对视频多目标轨迹的连续跟踪。

关键词: 多伯努利滤波, 检测跟踪, 目标新生, 相似度矩阵