基于Hu矩和纹理特征结合的人体异常行为识别

为了提高人体异常行为识别的准确率,采用了一种将多特征结合的异常行为识别算法,主要包括对步行、快跑、慢跑、拳击、双手挥舞、鼓掌六种异常行为进行识别。首先从视频流中提取出人体轮廓,然后从所得的轮廓中提取Hu矩特征与纹理特征。最后通过模板匹配的方法,采用马氏距离度量所需识别的当前行为特征向量与标准模板行为的特征向量之间的相似性,并通过设置相应的阈值判定该行为所属类别。实验证明,该方法比提取单一特征的方法识别率高,且具有一定的实用价值。

Recognition of Human Abnormal Action Based on Hu-Moment and Texture Feature

In order to improve the accuracy of human abnormal behavior recognition, an action recognition algorithm using multiple features is employed in this paper, actions mainly including walking, jogging, running, boxing, hand waving, hand clapping. Firstly, human silhouette is extracted from video flowing. Then, Hu-moment features and texture features are extracted from this silhouette. Finally, the similarity between current behavior feature vectors and feature vectors of standard template is calculated using Mahalanobis distance. Experiment results show that this method has a higher recognition rate than that which extracts unique feature and it is of a great practical value.