一种利用点基元灰度的快速精确图像匹配算法

A Fast and Accurate Image Matching Algorithm Based on Point Grayscale

为了能够让基于双目视觉的机械手获取轮胎测温孔精确的三维坐标,并且满足机械手测温实时性要求,目前常用的算法是采用基于外极线约束的区域匹配来确定两个对应点间的匹配。但该算法要求将目标点沿极线上每一点都进行区域灰度相关计算,计算量大且匹配准确率不高。针对此问题,基于两对应点之间灰度值相近的原则,提出一种将目标点与外极线上各像素点先进行点之间灰度比较,然后只对与目标点灰度值相近的像素点进行区域灰度相关计算的改进算法。该算法将外极线上与目标点灰度相差较大的点先排除,以减少区域相关匹配计算点的数量和降低误匹配率。

In order to improve the coordinate accuracy of thermometer hole of tire for the manipulator based on stereovision and meet the requirement that manipulator gets the temperature of tire in real-time,an available algorithm is using region matching based on epipolar line restriction to determine two matching points,but the algorithm requiring to calculate region grayscale for all points on the epipolar line will lead to huge calculation and low matching accuracy. Based on the principle of the corresponding points possessing similar grayscale value,an improved algorithm comparing the target pixel point with each pixel point on the epipolar line in the aspect of grayscale firstly and then calculating region grayscale for the points possessing similar grayscale value to target point was proposed.The algorithm excludes some non-matching points on the epipolar line,which can reduce the amount of computation and drop the false match rate.