The mechanical arm needs to carry out feature recognition for machine tool parts before automatic and accurate grinding of machine tool parts.The traditional laser lattice scanning has high accuracy,but the equipment cost is high and it is difficult to scale application.Fixed projection was introduced to generate scanning line map,checkerboard map and random gray scale map as measurement aids,and binocular stereo vision measurement was combined to detect part defects,by which the stereo matching of traditional binocular vision for the features that were difficult to recognize for smooth plane parts was improved.The accuracy of the analysis results was verified with the laser lattice scanning results.At the same time,the accuracy of five classical local stereo matching algorithms was compared with the proposed method,and the accuracy correlation analysis of the key parameters of the auxiliary image was carried out.Finally,an adaptive algorithm was introduced to optimize the proposed method to improve the recognition accuracy.The results show that the random gray scale image can significantly improve the matching accuracy,the scanning line optimization and sub-pixel fitting can further improve the matching accuracy,and the accuracy of the scanning image and checkerboard method after optimization is greatly improved.The new method can effectively identify the defect features of plane parts,and the detection accuracy reaches the standard area accounting for more than 98%.