Research On Surface Scratch Detection Of Automobile Body Parts Based On Machine Vision

Panel is one of the most important parts for cars, which directly represents the overall quality and the aesthetics of car. Scratch is a common defect on panel and an important factor which will influence the quality of panel. Nowadays, manual work is still the most common method to detect scratches on panel.Regarding the disadvantages of manual work such as high missing rate and low efficiency, an automatic scratch detection method based on machine vision has been proposed in this dissertation after the comparison among some common scratch detection methods. To avoid the strong specular reflection from panel, this article has designed a general illumination system based on the indirect even diffusion light. Meanwhile, a special curved design has been employed according to the structural characteristic of panel. Besides, an industrial camera with high resolution ratio has been employed for the image acquisition of panel. By utilizing the significant and morphologic characteristic of scratch on the image, an image processing method based on smooth band-pass filter has been proposed to implement the standard detection of scratches with different characteristics.Finally, the machine vision system designed for scratch detection has been applied to the actual measurement. The experiment results of different samples indicate that the system has high accuracy and good compatibility, which makes automatic detection a feasible technology.