A visual defect detection method based on a multi-spectral depth convolution neural network

Abstract

The invention relates to a visual defect detection method based on a multi-spectral depth convolution neural network. The method is used for the appearance defect detection of a photovoltaic cell sheet. By designing the multi-spectral neural network model, the effectiveness and accuracy of the extraction of a plurality of spectral features by the model are enhanced, and the decoupling of the features and the background is realized. By analyzing the characteristics of defects in multi-spectra and using the method of image multi-spectral information feature separation and extraction, the abilityof the model to extract multi-spectral image information features is enhanced. Compared with the LBP + HOG-SVM and Gabor-SVM surface defect detection methods, the designed multi-spectral neural network model has increase about 10% at the three indicators (precision, recall, F-measurement), but also can effectively solve the surface of the complex background texture, defect characteristics, shaperandom problems, and has the defect recognition accuracy of 94.30%.

Description

technical field [0001] The invention relates to the technical field of photovoltaic cell defect detection, and mainly relates to a method for visual defect detection based on a multispectral deep convolutional neural network. Background technique [0002] At present, image-based intelligent visual inspection methods have become an important technical component of solar cell surface quality control. Doing a good job in solar surface quality inspection can not only improve the service life of battery components, but also improve the power generation efficiency of solar cells. [0003] Solar cells are divided into monocrystalline silicon and polycrystalline silicon in terms of production materials. Monocrystalline silicon has a simple background texture and high power generation efficiency, but the production cost is relatively high; polycrystalline silicon cells contain a large number of lattice particles of random shapes and sizes on the surface. Randomly distributed in diffe...

AI Technical Summary

Problems solved by technology

[0008] In order to solve the problem of multi-category defect detection under the background of non-uniform and complex texture on the surface of solar cells, the present invention provides a visual defect detection method based on multi-spectral deep convolutional neural network, which is used for defect detection of the appearance of photovoltaic cells. It can detect defects with random shapes and complex backgrounds on the surface of solar cells, and has higher accuracy and adaptability. The accuracy of defect recognition reaches 94.30%.

Images