Abstract:
This paper proposes a pedestrian video detection algorithm based on convex hull clipping in order to solve the problem of pedestrian crowd video detection. The algorithm uses part convex hull technology to search pedestrian outlines, clips out contour curves of pedestrians by using concave points mining technology. It gets rid of non-head curves in clipped lines depending on rules, fast ellipse detection based on least square. Experimental results indicate that this algorithm can exactly detect overlap or connected heads and exclude non-head objects, and it has fast processing speed, high practical application value.
Key words:
convex hull,
concave point mining,
pedestrian detection
摘要: 为了解决行人群体视频检测的难题,提出一种基于凸包裁剪的行人视频检测算法。该算法采用局部凸包技术搜索行人外部轮廓,利用凹点挖掘技术裁剪轮廓曲线,建立相应规则排除非头部线段,通过最小二乘拟合法对头部曲线进行快速椭圆检测。实验结果表明,该算法能准确地检测出重叠或连通的头部,排除非头部物体,且处理速度快,实际应用价值高。
关键词:
凸包,
凹点挖掘,
行人检测
CLC Number:
LI Jiang; SUN Li-jun. Pedestrian Video Detection Algorithm Based on Convex Hull Clipping[J]. Computer Engineering, 2010, 36(2): 173-175.
李 江;孙立军. 基于凸包裁剪的行人视频检测算法[J]. 计算机工程, 2010, 36(2): 173-175.