针对固定场景中运动目标检测遇到的运动目标状态突变,非运动目标干扰以及阴影等问题,提出了一种背景差分和帧间差分相结合的运动目标检测算法;该算法首先通过平均法背景模型确立背景,使用背景差分得到一幅二值化前景图像,然后将通过连续的多帧图像进行相邻帧差分得到的多幅二值化前景图像进行逻辑或运算,最后将运算结果同背景差分所得到的二值化前景图像进行逻辑与运算,得到最终运动目标区域;实验表明,该算法运算速度快,准确率高,并能满足实时检测的需要;不足之处是在摄像机与运动物体夹角很小的情况下,检测效果较差,但可以通过多台摄像机协同操作来达到理想的效果.%According to the problems of moving objects detection in the settled scene, such as state mutations from moving objects, interfences from non—moving objects and shadow, etc. , an algorithm based on frames subtraction and background subtraction is presented in the paper. At first, the background model for the background subtraction is estabished with the average method, then, one binary foreground is obtained with the background. The other one is gotten by continual frames subtraction. The moving objects are obtained with the logical operations between them. The experiment shows that the algorithm is a real—time one with the high speed and accuracy, the bad effect caused by the small angle between the camera and moving objects can be resolved by the cooperation with several cameras.
展开▼
