Moving Target Detection And Tracking Algorithm Based On The Sequence Of Images

Detection and tracking of moving object in image sequence is one important area in the domain of digital image processing and computer vision. Its applications include fields like robot navigation, vision-based supervision, security surveillance, medical image analysis, industrial automation, video processing, etc..The study object of this paper is image sequence. Each image is composed of moving objects and background. Normally there are three cases: the background is static and simple, the background is complex with small change, and the background is complex with vast change. Most published researches focus on detection and tracking of moving object in the first case. So far, there is limited research done on the third case. This paper has tried to analyze both the first and the second cases. In the research, a digital image capturing system is constructed. The hardware and software part of the system are chosen according to practical needs. The presently common methods on detection and tracking of moving objects are introduced first. For moving object detection, three methods, including 'differential between adjacent frames', 'background subtractction' and 'optical flow', are discussed. Through comparison and analysis, the merits, weaknesses and application field of each method are concluded. For moving object tracking, there are three feature matching algorithms: 'absolute-balance searching', 'normalized cross-correlation matching', and 'histogram matching'. Two fast matching methods, 'sequential similarity detection algorithm' and 'pyramid hierarchical searching algorithm' are also introduced. Based on the studies of these algorithms, this paper designs and implements a practical system for sequence image capturing and moving object detection and tracking, using Visual C++ and OpenCV. This system can detect and track moving object more accurately and less time-consuming. In order to improve the accuracy and enhance the real-time ability of tracking, a round-off method is used to remove the potential erroneous points, and a fast matching program based on pyramid hierarchical searching algorithm is included. The algorithm and methods applied in this research can be applied in fields such as vision-based safety-monitoring, air surveillance, and industrial detection.