本文提出一种基于目标检测的多维假设多目标检测和跟踪方法,此算法对于序列图像的照明变化和遮蔽现象具有很高的鲁棒性。首先,对序列图像进行背景抑制、时域滤波和杂波剔除的预处理,得到单帧的初始目标检测结果。对于初始检测中存在的漏检和误检现象,采用基于假设理论的跟踪概率模型优化初始检测轨迹;将目标的跟踪信息反馈于目标检测模块,形成一闭环自适应跟踪系统,达到多目标的最优检测和跟踪。实验结果表明了所提出的方法在多目标检测和跟踪中的可行性和有效性。

A multiple hypothesis method for tracking multiple motion objects is proposed based on the object detection,which is robust to changes illumination and shadow conditions in image sequence. After video preprocessing including background suppression,temporal filter and clutter removal,the initial results of object detection is obtained in each frame image. Based on the detection results which may have missing and/or false detection,the multiple object tracking algorithm accepts the probabilities of initial object detection and keeps multiple hypotheses of object in a graph structure. Through extending and pruning the graph based on the previous image information,the multiple object tracking method gives feedbacks which are predictions of object locations to the object detection module. Therefore,the method integrates object detection and tracking tightly. The most possible hypothesis provides the multiple motion object tracking results. The experiment and evaluation results illustrate the validity and advantage of the proposed algorithm.