基于预测和加权匹配的目标快速定位方法  

Fast locating algorithm based on predicting object and matching weights

作　　者：董宇青[1] 段发阶[1] 段晓杰[1] 韩芳芳[1] 

机构地区：[1]天津大学精密测试技术与仪器国家重点实验室,天津300072

出　　处：《激光与红外》2012年第6期718-722,共5页Laser & Infrared

基　　金：国家自然科学基金项目(No.50375110)资助

摘　　要：基于模板匹配算法,提出了一种适用于高速流水线检测的目标定位改进算法。该算法利用已知目标位置预测未知被测物体,并用加权模板匹配实现精确定位。首先对运动的物方空间进行标定使得连续采集的图片有连续统一的坐标系,从而实现图像间物体位置预测。同时,提出一种对模板不同特征区域加权的方法,在不影响算法复杂度的前提下,正确选择权值,能使目标物体区域内的匹配相似度有更好的显著度和单调性。实验结果表明,结合预测位置和加权求相似度的模板匹配具有更高的快速性和鲁棒性。Based on the template matching, this paper proposes an improved object location algorithm which can be ap- plied to high-speed pipeline. This algorithm predicts the object＇s approximate position by known information, in addi- tion,weighted template matching achieves precise positioning. On one hand,the camera calibrates athletic real space, so continuous images collected are in the same coordinate. On the other hand, a matching method with different weights in different characteristic region is presented. Appropriate weight values make matching similarity in target ar- ea more monotonously and significantly. The experimental results show that the template matching which combines pre- dicting position and weighting for similarity can is more rapid and robust.

关 键 词：模板匹配 预测目标模型 加权匹配 鲁棒寻优 

分 类 号：TP931.4[自动化与计算机技术]