智能交通监控系统中多目标跟踪算法

在视频监控系统中,以自适应背景差分法提取运动目标,并以单连接区域增长法标记各运动目标区域.针对在城市交通监控系统的特点,建立视频流中帧间图像的约束条件,利用相邻两帧目标之间的距离、平均灰度以及面积差等特征建立匹配距离测度函数,通过求全局最优解使得距离测度达最小,进而确定最佳匹配目标,并分析了目标运动过程中的遮挡现象及提出了解决方法.实验结果表明,该方法跟踪运动目标准确率达98%.

Multi-Target Tracking Algorithm in Intelligent Traffic Surveillance System

In video surveillance system, moving targets are extracted by adaptive background subtraction algorithm, then every region of moving target can be labeled by single linking region growing algorithm. Aiming at some features of urban traffic surveillance system, some restricted conditions are formed, some eigenvalues, such as distance, average values of gray and difference of areas, are used to establish measure function for matching in moving targets between two adjacent images, the minimum of measurement distance is obtained by finding global optimization, then the best matching objects are determined, the appearance of occlusion is analyzed in targets moving, and the solving method is proposed. Experimental results show the correct ratio of tracking moving targets is up to 98%.