摘要
提出了智能视频监控中基于机器学习的自动人数统计系统。该系统通过机器学习的方法对视频序列中人的头肩部位进行准确地检测,克服了传统检测方法如连通域分析和简单模板匹配的不足,对光线变化和人群拥挤等问题具有较好的稳健性,在对具体场景的初步测试中取得了较满意的效果。
An automatic people counting system based on machine learning in intelligent video surveillance is proposed in this paper. The system detects head-shoulder component based on machine learning, which overcomes the disadvantages of traditional detection methods such as connected component analysis or simple template matching. The system demonstrates its robustness to illumination change and crowd in practical applications.
出处
《电视技术》
北大核心
2009年第4期78-81,共4页
Video Engineering
基金
国家自然科学基金(60872084)
教育部高等学校博士学科点专项科研基金(20060003102)
关键词
智能视频监控
人数统计
目标检测与跟踪
机器学习
intelligent video surveillance
people counting
object detection and tracking
machine learning