Development of a visual inspection system and the corresponding algorithm for the detection and subsequent classification of paint defects on car bodies in the automotive industry

Increasing expectations of customers on the appearance of cars force the automotive paint shops to inspect all car bodies with great care. Currently, skilled workers visually inspect each car body to detect and repair occurring paint defects. However, humans can neither detect nor judge the blemishes objectively and reliably over a longer period. Hence, this paper focuses on the development and validation of an algorithm for a surface inspection system, which improves the accuracy of detecting paint defects through an image processing system. Once they are detected, a reliable classifier is necessary to gain further information about the paint defects. The generated data can be utilized for improving the paint application process and for identifying the perpetrator of the paint defects once the defect classification is operational. As a consequence, the quality control circuit can be shortened, the surface quality of different paint formulae can be evaluated objectively, and actions can be taken in order to reduce the occurrences of certain defect types. This paper presents the physical setup of the visual inspection system and a detailed description of the algorithm for detecting the defects in the acquired images. Further research is essential to generate a classifier to differentiate the types of paint defects.