基于帧间差分法的动体特征速度聚类分析

针对智能视频监控中快速、准确的检测和识别运动物体的问题，提出了一种依据运动物体特征速度来检测识别动体以及解读其语义含义的算法。该方法以相对帧间差分法为基础，通过对预处理后的二值斑块图像的标记，计算斑块的像素长度作为其特征速度，并依据斑块特征速度的众数进行聚类分析，从斑块特征速度得到运动物体的特征速度语义解读和运动物体的检测识别。实验结果表明，斑块的特征速度不仅可以实现对运动物体的检测，而且通过聚类分析可以准确的得出动体特征的语义解读。用特征速度和众数聚类分析方法实现对运动物体的检测识别和语义解读相对于其他统计算法简单有效，便于智能摄像机的嵌入式开发。

Clustering Analysis of Moving Objects Characteristic Speed Based on Inter-Frame Difference Method

For detection moving objects in intelligent video surveillance fast and accurately, we proposed an algorithm based on characteristic velocity to detect and identify moving objects, then interpret the semantic meaning of mobile objects. Based on relative inter-frame difference method, this method computed the pixel length of plaques as its characteristic velocity, clustered analyze according to the mode of the characteristic velocity, got semantic interpretation of the characteristic velocity and identified the mobile by labeling the pre-processed binary plaque images. Experimental results show that the plaque which has a characteristic velocity can not only detect moving objects, but also can accurately obtain the semantic interpretation of the mobile. The way that is using characteristic velocity and mode clustering analysis method to realize the moving object detection and semantic interpretation is simple and effective. It has the advantage of facilitating intelligent camera embedded development.