监控图像中的人体识别

Human body recognition method in monitoring image

在变电站的安防范围中有不少无人危险区域,采取视频监控的方式发现闯入危险区域的人并做出预警与警告是主要的监控安防手段.监控图像对人体的识别是一种动态模式识别,采用模糊聚类的方法将视频图像中的静态图像元素过滤,针对动态的像素区域采用对图像识别鲁棒性好的卷积神经网络来识别人体形态,能够快速地对形态多变的人体进行有效识别.

In the security range of the substation,there are many uninhabited areas.In those area the video monitoring method is used to find the person who entered the danger zone and make warning.Monitoring image recognition of the human body is a kind of dynamic pattern recognition.This article adopts the method of fuzzy clustering to filter the static image elements in the video image.In view of the dynamic pixel area of image recognition,the better robustness convolutional neural network is used to identify the human form,which can quickly form variables to effectively recognize the human body.Keywords:dynamic body recognition;CNN;FCM