Surface tiny defects detection based on deep learning:A review
Surface tiny defects are characterized by small scale,low contrast,and insufficient sample size,which lead to low detection accuracy and high missed detection rate based on deep learning.Therefore,visual detection of surface tiny de-fects remains a challenging task.Research indicates that enhancing the feature extraction capabilities of network models,re-ducing feature loss or gradient disappearance,and employing attention mechanisms to focus on important regions in images can significantly improve detection accuracy for surface tiny defects.This paper systematically analyzes network structures such as ResNet,DenseNet,and FPN that can effectively improve the detection accuracy of surface tiny defects,summari-zes the application of attention mechanisms in the detection of surface tiny defects,analyzes the solutions and specific appli-cations of generative adversarial networks(GAN)for the problem of insufficient samples of surface tiny defects,and com-prehensively summarizes effective network structures and solving mechanisms in the detection of surface tiny defects.
万方数据
维普
