快速混合高斯模型的运动目标检测

Fast Gaussian Mixture Model of Moving Target Detection

针对经典混合高斯模型算法在实际应用中计算量大实时性差,且对光线变化和运动物体速度敏感的缺点,提出一种改进的快速检测算法.通过选取合适的间距,先用帧间差分法提取出完整的运动区域和背景区域,只对前者进行混合高斯模型匹配,来降低计算量.对背景图像不同区域采用不同背景更新率,及时响应背景变化.最后引入一个光线突变参数,来预防光线突变给检测带来的干扰.通过实验,证明本算法在实时性,鲁棒性,稳定性等上有了很大的改善,能够很好的检测出运动目标.

In this paper, we propose an improved fast detecting algorithm to solve the disadvantages which are the high computation, sensitive to light changes and the speed of moving object of the classical Gaussian mixture model. We extract the complete regions of the motion and the background by using frame different method of choosing the appropriate space. The algorithm reduces the amount of calculation because it just needs to calculate part of pixels. We use different background update rate in the different area of the background to respond the changes of the background timely. Finally, we introduce an environment mutation parameter to detect mutations. Through experiments, the algorithm has made a lot of improvement in the aspects of the iustantaneity, the robustness and the stability. The moving target can be detected very well.