Defect Recognition Method based on Fusion Learning of Multi-layer Image Features

Aiming at the problem of scarce defect data set in industrial product detection, a defect recognition method based on fusion learning of multi-layer image feature is proposed in this paper. The method introduces three Gaussian pyramids with the same size and different resolution to obtain more defect information, and establishes a Defect Recognition Convolutional Network to extract the information. The experimental results show that our model is more capable of extracting features, and can obtain superior accuracy, recall, and precision, as well as deal with industrial fuzzy sampling.