作　　者： ;

机构地区： 广东工业大学华立学院

出　　处： 《计算机技术与发展》 2019年第11期123-127,共5页

摘　　要： 通过人体行走步态特征智能检测识别应用在安防监控系统,提出一种基于卷积神经网络(CNN)的人体行走步态特征识别方法。采用视频和红外成像技术进行步态图像采集,采用三维区域轮廓扫描方法进行安防监控人体步态图像的三维轮廓检测,采用相关滤波跟踪算法进行步态红外图像的形状特征提取和信息增强处理,突出安防监控人体步态的类别属性特征点,在步态监控区域内对人体行走轨迹进行直方图均衡化处理,实现安防监控人体步态的个性化特征点提取。对提取的个性化特征点采用卷积神经网络进行分类训练,实现人体行走步态特征智能识别。选取大量安防监控图像进行实验,仿真结果表明,采用该方法进行人体行走步态特征识别的成功率较高,输出人体行走步态个性化匹配特征点总数较多,步态的跟踪识别的偏移像素较小,能有效实现个体化的步态识别,实现安防监控。 Based on the application of intelligent detection and recognition of human walking gait features in security monitoring system,a method of human walking gait feature recognition based on convolutional neural network(CNN)is proposed.The gait image is collected by video and infrared imaging technology,and the 3D contour detection of human gait image is carried out by using 3D area contour scanning method.The correlation filter tracking algorithm is used to deal with the shape features and information enhancement of the infrared gait image,which highlights the category attribute feature points of the human gait monitored by security and equalizes the human walking track in the gait monitoring area.The personalized feature point extraction of human gait is realized.The extracted personalized feature points are classified and trained by convolution neural network to realize intelligent recognition of human walking gait features.A large number of infrared security surveillance images are selected for experiments.The simulation shows that the method has a high success rate in recognition of human walking gait features,and the total number of personalized matching feature points for human walking gait is more.It can effectively realize individual gait recognition and security monitoring.

关 键 词： 步态 特征提取 安防监控 图像 识别

领　　域： [自动化与计算机技术—计算机应用技术] [自动化与计算机技术—计算机科学与技术] [电子电信—通信与信息系统] [电子电信—信息与通信工程]