摘要
针对智能视频监控系统中基于不变矩的人形识别进行研究。建立一个含有201个人形样本、67个动物样本的图库,采用最小距离分类器对目标Hu矩和Zernike矩分类识别。实验验证表明该方法是有效且可靠的,一般场景下人形识别率可以达到89%以上。
In this article we study the moment invariants-based human shape recognition in intelligent video monitoring system, and establish an image library which contains 201 human shape samples and 67 animal samples. Minimum distance classifier is used to classify and recognise the Hu moment invariants and Zernike moment invariants. Experimental verification demonstrates that method is effective and reliable, in general scenarios the human shape recognition rate can achieve 89% and higher.
出处
《计算机应用与软件》
CSCD
北大核心
2013年第8期162-165,共4页
Computer Applications and Software
基金
陕西省工业攻关计划项目(2011K09-36)
陕西省教育厅科研计划项目(12JK0528)
陕西省科技攻关计划项目(2012K06-16)
关键词
HU矩
ZERNIKE矩
人形识别
最小距离分类器
Hu moment invariants Zernike moment invariants Human shape recognition Minimum distance classifier
