基于直线间结构信息的立体视觉图像动态匹配方法  

Image dynamic matching for stereo vision based on structural information between lines

作　　者：崔巍[1] 强文义[1] 陈兴林[1] 

机构地区：[1]哈尔滨工业大学控制科学与工程系,黑龙江哈尔滨150001

出　　处：《控制与决策》2003年第5期633-636,共4页Control and Decision

摘　　要：针对立体视觉匹配问题,介绍一种改进的动态规划图像匹配方法,它将边缘直线相似测度分为局部相似测度和全局相似测度,在后者中加入图像边缘直线之间的结构关系信息,并在动态搜索最优匹配路径的过程中利用结构关系约束删除不合理的匹配路径。仿真实验结果证明,采用该方法解决立体视觉中边缘线段的匹配问题,不仅提高了匹配的准确率,而且大大减少了匹配时间。The stereo correspondence problem is the problem of matching two images of the same scene from different viewing positions. A dynamic programming approach is used to solve the correspondence problem in stereo vision. The similarity measure between the straight lines in left and right image is divided into local similarity measure (LSM) and global similarity measure (GSM). The structural information is added to GSM. Some rules based on the structural network of each image are used to eliminate the incorrect matching between edge lines of left image and right image. The simulation result shows that the image matching achieves a high success rate and needs less time when improved dynamic programming is used for stereo matching.

关 键 词：立体视觉匹配 全局相似测度 动态规划匹配 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]