摘要
介绍了一种实时人形识别方法,使用这个方法,可以自动监测进入摄像机镜头拍摄范围内的物体,并识别其是否是人类。其主要步骤包括,首先预处理其中包含了除背景之外的其它物体的图像,然后提取其个不变矩,利用贝叶斯分类器作主要手段对不变矩进行分类,7最后采用辅助手段修正分类结果,识别该物体的形状是否为人形,从而判断出是否有人类经过。
This paper presents a real-time human's shape recognition methodwhich can monitor the object moving in the shooting ,scope of the camera lensand recognizes the object being human or not. This paper firstly pretreats the image, which includes other ,object besides backgroundthat is ,getting rid of background image and obtaining the shape image of the object. Second, it calculates the seven moment invariants of that image. Finally it uses Beyas assorting approach as main method to assort the moment invariants and uses other assistant methods to correct the result in order to recognize the human shape and to judge whether human has passed or not.
出处
《计算机工程》
CAS
CSCD
北大核心
2002年第9期80-83,共4页
Computer Engineering
基金
国家自然科学基金资助项目 (60072034)
