基于Barnard特征和LoG算子的月球车图像特征匹配  

An Image Match Algorithm of Lunar Rover Based on Barnard and LoG

作　　者：宋毅[1] 崔平远[1] 陈阳舟[1] 居鹤华[1] 

机构地区：[1]北京工业大学电控学院,北京100022

出　　处：《计算机仿真》2006年第7期56-59,共4页Computer Simulation

基　　金：国家自然科学基金资助项目(60374067)

摘　　要：月球车视觉系统对匹配速度和鲁棒性要求较高,提出一种基于图像特征的特征点匹配算法。特征点的提取采用Bar-nard算法和LoG算子共同作用得到。原图像同时采用Barnard算法进行特征提取和LoG算子进行噪声滤波和边缘提取,然后将二者的计算结果进行"与"操作得到最终的特征图像。然后通过阈值进一步处理噪声和选择待匹配点。最后用计算图像序列中像素差的平方和的方法来计算每对待匹配点的匹配程度,得到匹配点对。仿真结果表明,单点匹配和多点匹配并将无匹配点处理后得到的结果都令人满意。A point matching algorithm based on the image features is proposed for the stereo vision system of a lunar rover, which needs high robustness and matching speed. The image features are computed by Barnard algorithm and the operator LoG. Fir.st the images are computed by both the Barnard algorithm and LoG at the same time. Then an operation ＂and＂ is applied to the results of two steps to get the final feature image. Then a threshold is applied to reduce noise and choose the location of pixels. At last the matches are found using the sum of squared differences. The simulation results are satisfactory.

关 键 词：月球车 匹配 特征提取 像素差的平方和 

分 类 号：TP242.6[自动化与计算机技术—检测技术与自动化装置]