摘要
动态视频对象的识别与检测是计算机图像处理领域中的一个前沿课题。该文提出了一种利用实时视频图像识别技术 ,对无规则行进队伍的人员进行检测与统计的系统模型和算法设计。基于颜色模型转换、减影、膨胀、腐蚀、聚类、匹配、跟踪等技术 ,实现了对特定检测区域内运动对象的识别、检测和计数。着重讨论了系统的设计原理和实现方法 ,以及对象运动的物理模型和多目标识别跟踪算法。该系统于1998年开始成功地应用于毛主席纪念堂 ,对每日的瞻仰人数进行实时统计 ,计数误差小于 3% 。
An image processing system was developed to recognize and trace video objects to automatically count people. This paper describes the system design principles including the color model transparency, the physical models of the object motion, multi object tracing, counting algorithms, and reasonable simplifications. The system detects and counts moving objects in a specific area. The counting system error is less than 3%. This system has been authenticated, and is now installed in the Memorial Hall of Chairman Mao.
出处
《清华大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2001年第7期61-64,共4页
Journal of Tsinghua University(Science and Technology)
