Machine Part Defect Detection based on Deep Learning:Key Issues Research

Due to the lack of noise removal of parts in conventional detection methods, the accuracy and real-time performance of detection are affected. To this end, this project proposes to study machine part defect detection methods based on deep learning. First, the impact noise detection and filtering in mechanical part images based on median filtering will be studied; secondly, using deep learning techniques, a multi-level classifier will be established to implement feature extraction of the entire connectivity hierarchy on this basis; finally, the posterior probability of the target image will be obtained according to the output results of the classifier. On this basis, a fault diagnosis method based on decision criteria is proposed, and the results of fault classification are used as the basis for fault diagnosis. Experiments prove that the algorithm proposed in this paper has better detection accuracy and detection time than conventional detection method