零件形状误差的MTS测量识别方法

Research for MTS classification methods of parts shape error

针对工业生产中的对机器人视觉及产品实时质量监控,提出了机械零件的MTS(Mahalnobis-TaguchiSystem)判别分类方法。并以Taguchi方法对判别特征进行优化筛选,减少了无用特征和“噪音”的干扰,以马氏距离(MahalanobisDistance)作为判别分类的判据。实验表明该方法能有效提高判别分类的速度和精确度。

In modern industry fields,such as the quality inspection,the classification of parts and the machine vision of robot,we often need to recognize the mechanical parts in order to perform the auto classify,the auto assemble and the online quality inspection of mechanical parts.In this paper,a method based on the MTS(Mahalanobis-Taguchi System)is developed for images classification.The color images are converted to gray images and a feature extraction is applied before classification.The amount of pixel of each gray scale is computed as the feature for classification.A feature reduction process using taguchi method is applied to the feature before classification.Then the Mahalanobis Distance between the source image and the target image is used as the symbol for classification. The MTS method performs better than using only the mahalanobis method.