摘要
针对目前在大范围内难以快速地实现人体目标检测的问题,提出了一种基于全方位传感器(ODVS)的快速检测人体的理论与方法.在人体建模上,首先将人体简化为矩形,针对人体在不同区域半径,通过ODVS标定来建立不同形状大小的人体矩形模型;最后,对前景运动目标对象进行面积属性与形状属性的判断,对于在区域半径中不符合标定的矩形模型的前景运动目标区域,作垂直方向的投影以及人脸肤色的二次综合判断,从而实现快速的人体检测.提出的人体检测的方案可方便的实现在大监控领域中检测是否有人、有多少人以及在什么方位,为全方位的捕捉人体信息、提高人体检测精度提供了一种新手段.
Against the defect of existing human detection, which can not detect rapidly in the large field at present, this paper presents a rapid human detection theory and method based on omni- directional vision sensor （ODVS）. From the view of the human model, firstly it simplifies human model as rectangle. In the different region radius where the human exists, different shape and size of the rectangular models are established in accord with the calibration of ODVS. At last, the foreground moving target is judged by area and shape attributes. If the moving targets do not conform to the Calibration rectangular model of region, it uses the methods of the vertical direction projection and color of the face detection as the secondary judgment, thus achieving rapid detection of the human body. The human detection method that the paper presents, can facilitate to realize surveillance control in the large field, which also can get information , such as whether or not the human exists, the number of people, and the position of human. So it provides a new method which can capture the human information from all-direction and improve the human detection accuracy.
出处
《浙江工业大学学报》
CAS
2008年第4期359-364,共6页
Journal of Zhejiang University of Technology
基金
浙江省科技厅重大科技资助项目(2006C11202)
关键词
全方位视觉
人体矩形模型标定
面积属性与形状属性
人体投影
人脸肤色
omni-directional vision sensor （ODVS） and shape attribute
human projection
human rectangular model calibration
area human face color
