利用角点历史信息的异常行为识别算法

针对视频监控场景中的异常行为事件，如突然的奔跑、人群的异常聚集等现象，提出一种利用角点运动历史图策略的行为识别算法，即首先通过角点提取算法进行场景角点提取；然后通过时间累积获取角点的历史图，通过角点历史图将场景中的角点划分为静态角点和动态角点；最后通过动态角点分析完成监控场景异常行为分析识别。新算法充分利用了图像的时空信息，并且克服了场景光照影响，增强了异常行为检测与识别的准确性。通过真实场景实验可以看出，新算法能够对不同监控场景的异常行为进行准确检测，并且其检测速度快，满足实际应用需求。

Abnormal behavior recognition based on corner motion history

In video surveillance scenes, the abnormal events, such as a sudden run, the crowd abnormal aggregation phenomenon, are studied. A corner motion history image strategy is used for behavior recognition. Firstly, the corner extraction algorithm is used as scene corner extraction. Secondly, through time accumulated, the corner historical image is constructed. And all corners are divided into static and dynamic corner point. Finally, through the dynamic corner analysis, the abnormal behavior is under analysis and recognition. The new algorithm makes full use of image information, overcomes the illumination effect, and enhances the abnormal behavior detection and recognition accuracy. Through the real scene experiments show that the new algorithm can be used for different scenes. It can get accurate detection for abnormal behavior, and the detection speed meets the needs of practical application.