A Novel Transfer Learning-Based Classification Method for Painting Defect Detection of Balance Beam of Port Machinery

A novel transfer learning-based classification method for steel structure painting defects recognition is designed, which is used to solve the problem of identifying painting defects in balance beams of port machinery. This paper proposes an AlexNet model based on transfer learning. Firstly, the weight of the original pre-training model is fine-tuned, and then the last full connection layer (FCL) in the AlexNet model is replaced. Finally, the target image data set is taken as the new input, and the weight of the model is fine-tuned through the back propagation mode, so as to realize the model migration. The experimental results show that the success rate of classification with and without defects can reach more than 93%; the success rate of point, line and surface defect classification reaches more than 88%;the accuracy rate for orange peel test reaches more than 91%. It can meet the requirements of defect detection and classification.