基于五帧帧差和混合高斯模型的运动目标检测

Moving Object Detection Based on Five-frame Difference and Gaussian Mixture Model

鉴于传统的帧差法检测准确率不高,而且在光照变化、噪声干扰时鲁棒性不高,容易造成检测错误等问题,提出了一种改进的视频序列运动目标检测算法.该算法是将混合高斯模型与改进的五帧差分算法相结合:首先改进五帧差分是将当前帧与前2帧、后2帧进行差分二值运算,然后将4个差分的结果轮廓填充,最后进行先"与"再"或"运算;通过将混合高斯建模后得到的运动目标与改进的五帧差分算法得到的运动目标,进行逻辑"与"操作,最后再通过形态学处理检测出运动目标.从实验结果证明,改进的算法既能适应光照的变化,又能有效克服空洞的现象,与同类的算法相比具有更高的鲁棒性和准确率.

In view of the problem of low detection accurancy to traditional frame difference method and low robustness to noise interference and light changes, this paper proposed an improved video sequence motion target detection algorithm. The algorithm combines the Gaussian mixture model with the improved five-fra-me difference algorithm. Firstly, the improved five-frame difference is the difference between the current frame and the first two frames and the second frame. Secondly, we should fill out the four differential result contours. Finally we take the union after the intersection; and the intersection between the moving target obtained by mixing the Gaussian model and the moving ob-ject obtained by the improved five-frame difference algorithm, and then the motion target is detected by morphological proc-ssing. Based on the experimental results, it is proved that the improved algorithm can adapt to the change of illumination and overcome the phenomenon of void effectively, and it has higher robustness and accuracy than the similar algorithm.