基于网格的运动目标检测与跟踪处理

为提高运动目标的检测与跟踪处理速度,设计了一个基于网格计算的解决方案,应用网格计算以分布并行方式来处理图像序列. 网格计算节点上执行的一个任务对应处理图像序列中的一个帧图像单元,一个帧图像单元包含了每次处理过程中所涉及的一帧或多帧图像,因而网格计算中的任务数即为图像序列中的帧图像单元数. 利用Condor系统搭建了一个网格计算试验台,开发了一个用户交互界面和若干中间件服务模块. 以基于相邻帧差法和模板匹配法的运动背景下的目标检测和跟踪算法为例进行了试验. 试验结果表明,该方案具有可行性,并能大幅度缩减计算时间,提高处理效率. 

Grid-Based Moving Object Detection and Tracking

To reduce the computing time of detection and tracking of moving object in video sequence, a grid-based processing scheme was designed. The grid computing technology was applied to deal with the frames of a video sequence in distributed-parallel manner. One or more adjacent frames of the video sequence treated at a time were regarded as a frame-unit, and the whole treatment of a frame-unit was considered as a job executed on the grid node machine. The number of jobs which were submitted to grid nodes, therefore, was equal to that of the frame-unit of the video sequence. A Condor-based grid test-bed was constructed to verify the feasibility of the proposed scheme, and a graphical user interface (GUI) program and several service modules were developed for it. The combination of the frame difference detection method and the template matching tracking method was implemented on the test-bed for a moving object detection and tracking in a moving background. Experiment results show that the scheme is feasible and able to improve the execution efficiency greatly.