【摘要】 车灯灯罩表面缺陷部分进行检测分割采用机器视觉的方法，首先，对常见光源以及打光方式进行调研分析，选取LED光源折射光面阵式条形打光的方式，能够解决灯罩表面的透光特性和高反光性特性，使得灯罩表面的缺陷部分能够在数字图像上显影，之后，采用Mean Shift算法和模糊C均值算法相融合的算法框架，该算法框架通过LUV颜色空间的特征均值偏移将条纹区域与非条纹区域分离，模糊C均值算法通过计算每一个候选像素的隶属度值将属于缺陷部分的像素分离，完成缺陷区域的定位与分割。实验部分，该算法与分水岭算法和最大类间方差算法在分割效果上做了对比，并采用精确率与召回率定量的评价三种算法的分割性能，该算法在单幅图像上精确率和召回率分别达到了95.25%和91.13%，在多幅图像上精确率和召回率达到94.96%和93.03%，均超过另外两种算法，且满足于工业质检要求。更多还原

【Abstract】 Lamp shade the part surface defect detection segmentation using machine vision method, first of all, the research mode of common light source and lighting analysis, select the LED light refraction smooth formation strip lighting, can solve the pervious to light on the surface of the lamp shade features with the features of high reflective make lampshade defects on the surface of the part in the digital image enhancement,after that, Using the Mean Shift algorithm and the fuzzy c-means algorithm fusion algorithm framework, and the algorithm framework by LUV color space characteristic of the Mean Shift will stripe area separated from the fringe area, the fuzzy c-means algorithm through the calculation of membership degree of each candidate pixels value will belong to defect parts of the pixels separation, complete the defect area of location and segmentation. In the experimental part, the segmentation effect of the algorithm is compared with watershed algorithm and maximum between class variance algorithm, and the segmentation performance of the three algorithms is quantitatively evaluated by precision rate and recall rate.The accuracy rate and recall rate of the algorithm on a single image are 95.25% and 91.13%, respectively. The accuracy and recall of multiple images are 94.96% and 93.03%, which are higher than the other two algorithms and meet the requirements of industrial quality inspection.更多还原