Application of Deep Learning for Defect Detection of Paint Film

Paint film can protect and decorate the metallic workpieces. Automatic defect detection of paint film is important and challenging for intelligent manufacturing. In this paper, the surface quality of paint film is detected using defection system combining with deep learning. The image set of paint film is first collected for training and testing, which includes 2000 images. We use both SSD and Faster R-CNN methods to detect the defects of paint film. Finally, the framework of defect detection program is implemented through virtual instrument and python. Experiments indicate that the SSD and Faster R-CNN methods can get good experimental data in the paint film detection. As opposed to previous visual detection methods, the defect detection method using deep learning does not have to manually design defect characteristics previously, which is more robust. Moreover, the proposed framework can be easily applied to other industrial application fields, such as detection of wood defects, surface defect detection of parts and weld defect detection.