基于纹理特征的机械零件图像快速匹配方法

The Fast Matching Method for Work Piece Images Based on Texture Features

为了实现机械零件图像的匹配,同时保证具有较高的计算速度,提出了一种针对条形纹理表面工件的匹配方法。首先,在工件图像中取若干圆形区域(称为纹理圆),通过纹理圆的统计特征来计算工件表面纹理方向,确定2幅匹配图像间的大致角度差。然后根据角度差旋转其中1幅图像,保证了2幅图像角度差在相关系数法作用范围内。最后用相关系数法对2幅图像进行灰度特征匹配。实验结果表明:基于纹理特征的机械零件图像匹配,解决了条形纹理表面工件图像匹配问题,同时具有较高的计算速度。

In order to realize the work piece images matching, it proposes a new method for the work piece of strip texture surface. First, the rude angle is gotten by extracting some circle areas called texture circle from the image of work piece, and getting the strip texture direction of work piece based on the texture circle. Then based on the rude angle rotating one of the two matching images makes sure the matching method of correlation coefficient do not lose its role. Finally gray character matching is presented based on the method of correlation coefficient. Experimental results indicate that the method solves the matching problem of work piece with strip texture surface and makes sure fast computing speed.