基于帧间差分背景模型的运动物体检测与跟踪

针对背景差分算法中在复杂背景下参考帧的提取问题,提出了一种新的背景提取方法;该算法用帧间差分法将帧中的背景象素点检测出来,再确立出背景帧;由于排除了帧中运动物体的影响,因而提取出的背景干净,效果很好,然后运用背景差分检测出场景中的物体,最后采用一种新的运动物体跟踪算法,实现了运动物体和静止物体的识别,克服了以往检测算法中的误检和空洞问题,实验结果表明,该方法快速有效,能够满足实时性的要求.

Moving Objects Detection and Tracking Based on Two Consecutive Frames Subtraction Background Model

Aimed at the problem that how to attain background frame in complex condition, a new method to acquire background is presented. Two consecutive frames is subtracted to pick up background pixel in this method, so, background frame is built. The cleanly background was attained quickly, because the moving object in frame did not affect the process of attaining. Then background subtraction was used to detect objects in frame. At last, moving objects and still objects were distinguished by using a new tracking algorithm. The algorithm can overcome the mistake and imperfection which appeared in old moving object detection algorithm. The results showed that this algorithm combines the advantages of veracity and runtime, and fit for real time detection.