LK光流法和三帧差分法的运动目标检测算法

Moving target detection algorithm based on LK optical flow and three-frame difference method

三帧差分法是目前较为常见的运动目标检测算法之一。它的执行速度较快,但是它会存在各种干扰以及易受到环境噪声的影响,而且容易在检测到的运动目标内部产生较大的空洞,以致影响到最后的检测效果。针对这些问题,将Lucas-Kanade光流法与三帧差分法进行结合。利用Lucas-Kanade光流法计算得到运动目标的大致矩形区域。在确定的区域内外通过选取不同的阈值利用三帧差分算法提取运动目标,构成一种分级阈值的三帧差分法。并且利用前面光流法计算得到的角点来完善目标轮廓。这样将传统三帧差分算法的阈值分割转换成阈值分割与区域分割相结合的模式。试验结果表明,该改进算法具有良好的抗噪性,能够得到比原算法更好的检测效果。

The three-frame difference method is one of the most common moving target detection algorithms at pres-ent. Its execution is quite fast, and there inevitable exists various disturbances, and it is susceptible to the environ-mental noise. This method is also likly to form large cavities inside the detected moving targets, which affects the fi-nal result of the detection. To solve these problems, this article combines the three-frame difference method with theLucas-Kanade optical flow method. The Lucas-Kanade optical flow method is used to calculate and get the generalrectangular areas containing the moving targets. Different thresholds are selected inside and outside the determinedregions to extract the moving targets by the three-frame difference method, and then to constitute a kind of three-frame difference method that has rated thresholds. The corners calculated by the optical flow method are used to im-prove the contours of the targets. In this way, the threshold segmentation of the traditional three-frame differencemethod is converted into another mode, which combines the threshold segmentation with the region segmentation.The experimental results show that, the improved algorithm has good noise immunity and can get better detection re-sults than the three-frame difference method.