摘要
提出距离相关的变N值区域采样目标检测方法和距离无关的自相关目标识别方法用于检测识别特定公路段行驶的车辆.特点是不用对摄像机进行标定,使用方便,算法简单,运算速度快,识别距离远.用不同的N值对不同距离处公路宽度进行水平等分并且在车辆目标位置形成区域采样网格.由于模型图像采样网格密度也由N值决定,利用网格上的点集自相关方法将实际序列图像中不同大小目标区域与同一模型进行匹配.
. An approach to vehicle detection and recognition is introduced on vision. A method of N-divided sub-sampling detection on distance-variant and a method of distance-invariant self-relative recognition are summrized. The features of the approach are calibration-free camera,convenient to use,quick computing and far distance recognition. The road width is divided by different N value according to the road distortion in images at different distance and the sub-sampling grid formed. As the modal′s sub-sampling grid is also formed by N, the per sub-sampling on object in actual serial images matches modal images with the self-relative method.
出处
《大连铁道学院学报》
2004年第4期35-38,共4页
Journal of Dalian Railway Institute
