摘要
总结了变轮廓运动目标的特点,并将其应用到方向盘上操作手数的检测中;提出了先进行方向盘自动定位,再快速检测其上操作手数目的技术路线;预定位中运用Haar特征的AdaBoost分类器进行初检,得到包含目标轮廓的图像;利用HOG特征的Real-AdaBoost分类器进行精确检测,并确定操作手位置点集;对取得的操作手质心点坐标集进行奇异值分解并拟合椭圆,获取图像中方向盘位置,最终实现操作手的快速准确检测;算法在保证了原算法的实时性和准确性外,提高了检测系统应用的灵活性。
It summarized the characteristics of variable contour moving targets,and applied it to detect operating hands＇ position on steering wheel.It raised a means by automaticly detecting the position of steering wheel firstly,and after that detecting the operating hands.For gathering the operating hands＇ positions,this paper used AdaBoost on Haar to detect preliminarily and Real-AdaBoost on HOG to detect precisely.It fitted ellipse using SVD based on mass center coordinates of operating hands to simulate steering wheel and detected the operating hands quickly and accurately.Algorithm not only ensures the real-time and accuracy of the original one,but improves flexibility.
出处
《计算机测量与控制》
2016年第3期263-266,共4页
Computer Measurement &Control
基金
交通运输部信息化科技项目(2012-364-835-110)
北京工商大学科研能力计划项目
关键词
智能交通
操作手
特征提取
奇异值分解
intelligent transportation
operating hands
feature extraction
SVD
