论文部分内容阅读
针对智能视频监控提出了一种融合两种运动信息的分级运动检测算法。算法分别在像素级、区域级和帧级进行处理。像素级处理中,运用了改进的基于混合高斯模型的背景相减法,并提出了一个能够及时适应背景快速变化的累积时间差分法,两种方法检测得到两种不同的运动信息,区域级处理融合了这两种运动信息,使算法既能适应背景的缓变,又能适应背景的快变,解决了背景物体运动的问题。帧级处理主要解决全局光照突然变化的问题。实验结果表明,这个算法能够实现稳健可靠的运动检测。
Aiming at intelligent video surveillance, a hierarchical motion detection algorithm based on two kinds of motion information is proposed. Algorithm at the pixel level, regional level and frame level for processing. In pixel-level processing, an improved background subtraction method based on Gaussian mixture model is proposed, and a cumulant time difference method which can adapt to the rapid change of background is proposed. Two methods can detect two kinds of motion information, and the fusion of area-level processing The two kinds of motion information make the algorithm not only adapt to the gradual change of the background but also adapt to the rapid change of the background and solve the problem of the movement of the background object. Frame-level processing to solve the sudden change in global illumination issues. Experimental results show that this algorithm can achieve robust and reliable motion detection.