基于自适应窗口的多信息立体匹配算法

Research of Multi-information Stereo Matching Algorithms Based on Adaptive Window

针对焊接图像焊缝的立体定位问题,提出一种基于自适应窗口的多信息立体匹配算法。首先,改进了基于canny算子的边缘特征点提取算法。其次,利用初始焊接点给出匹配点的搜索范围,根据外极线约束确定候选匹配点集合。然后根据边缘差分强度和梯度方向相似性约束准则缩小候选匹配点集合。最后利用候选匹配点邻域信息构建自适应支持窗口并进行动态增长,根据窗口代价函数确定唯一正确的匹配点。仿真实验表明:采用笔者提出的算法,立体匹配的准确度较高、时效性较好。

Considering weld's stereo-location in welding image,an adaptive window-based multi-information stereo matching algorithm was proposed. In which,having the canny operator-based edge feature point extraction algorithm improved and the initial weld point used to determine the search range of matching points as well as the epipolar constraint based to fix the set of candidate matching points,including having the set of candidate matching points reduced according to constraints and similarity criteria of the edge difference strength and gradient direction and the neighborhood information of matching points used to build adaptive support window and to implement dynamic growth so as to determine the only correct matching point according to the cost function of windows. The simulation result shows that this algorithm proposed has higher accuracy and timeliness.