关键词: 混合高斯模型, 边缘检测, 目标检测, 三帧差分, 光流法

Abstract: Gaussian mixture model has been widely used for background modeling. However, the detection result is easily affected by noise and illumination mutation. In order to solve this problem, this paper proposes to combine the improved Gaussian mixture model with edge information. Once the method of three frame difference detects changes of the environment, the learning rate will be adjusted adaptively. The improved Gaussian mixture model is applied to extracting edge images and foreground images of moving objects. After dilating edge images, the result is obtained by computing the intersection of edge images and foreground images, and filling the hollow part based on the information of optical flow. Experimental results indicate that the proposed method has great capacity in restraining noise and dealing with illumination mutation, and improves the performance of object detection. It is more efficient than the traditional method.

Key words: Gaussian mixture model, edge detection, object detection, three frame difference, optical flow