在表面缺陷检测中，针对光照不均或有纹理的产品上微小缺陷难于识别的问题，提出了一种新的视觉识别方法。该方法首先计算产品表面图像中每行和每列的灰度标准差，然后根据标准差的相对变化量判别缺陷，并确定缺陷的坐标。实验结果表明，该方法能准确识别和定位产品表面的微小缺陷。

In surface defect inspection, a novel approach was introduced for identifying small defects on the object with non-uniform illumination or homogeneous texture. Firstly, it was to calculate the grey-level standard deviations of each row and each column in the surface image. Secondly, based on the relative variance of standard deviation, it could be confirmed whether defects were occurring or not, and the coordinate of the defect might be fixed on too. Experimental results indicated that the small defect in surface image of product could be identified accurately by this method.