多特征联合的薄壁零件表面缺陷检测方法

Surface Defect Detection Method for Thin-wall Parts Based on Multi-feature Joint

采用基于机器视觉的无接触检测方式对飞机制造中蒙皮、机翼缘条以及角片等薄壁零件表面缺陷进行自动检测,使用VMS-4030G影像仪采集零件表面信息,提出多特征联合检测方法检测缺陷。该方法主要包括图像Tamura纹理特征提取、图像局部二值模式(LBP)直方图和LBP下的灰度梯度共生矩阵特征(GGCM)提取。根据缺陷特性选择提取特征,对得到的特征应用主成分分析法(PCA)进行降维以及支持向量机(SVM)分类,最终得到检测结果。为了验证所提方法可行性,以带铆接孔的6061铝合金板代替飞机薄壁零件进行数据采集和检测。试验结果表明,该检测方法对毛刺、裂纹、凹陷及划痕的检测率均大于92%,明显优于单一特征提取的检测方法。

A contactless detection method based on machine vision is used to detect the surface defects of thin-walled parts,such as skin,wing flange and laminate in aircraft manufacturing.Multi-features joint detection method,including image Tamura texture feature,image local binary pattern histogram and gray gradient co-occurrence matrix feature in LBP,is proposed for defect detection based on the surface information collected by VMS-4030G vision measuring machine.The optimal parameters are selected based on the characteristics which is extracted from the defeats,and applied for principal component analysis dimensionality reduction,which is classified by support vector machine.In order to verify the performance of the proposed method,6061 aluminum alloy plate with riveting holes is used to replace aircraft′s thin-walled parts for data acquisition and detection.The results show that the detection rate of burr,crack,pit and scratch is higher than 92%,which is obviously better than the detection method based on the single feature.