Light weight method for solar cell defect detection based on data enhancement

Abstract

The invention discloses a lightweight method for solar cell defect detection based on data enhancement, and the method comprises the following steps: S1, carrying out the screening of collected fault pictures of a cell, and obtaining a true fault sample; the types of the pictures are manually marked; s2, building an improved ACGAN network, and generating a false fault sample and a corresponding category by adopting a generator according to the randomly generated noise data and type information; the false fault sample and the real picture are input into a discrimination network together, the discrimination network determines the authenticity, the following actions are circulated, and the generation network and the discrimination network are trained in sequence; according to the method, a data set is expanded based on a data enhancement algorithm of the ACGAN network, the problem of overfitting caused by insufficient training data is relieved, the structure of the convolutional neural network is optimized, lightweight processing is carried out, network training parameters are reduced, the model is compressed, the time for model training and defect detection is shortened, and the defect detection efficiency is improved.

Description

technical field [0001] The invention relates to a lightweight method, in particular to a lightweight method for defect detection of solar cells based on data enhancement, and belongs to the technical field of defect detection of solar cells. Background technique [0002] With the decline of traditional energy sources, research on various new energy sources has begun to emerge, among which solar power generation is the best among them. However, with the development of solar power generation technology, the requirements for the detection of solar cells and components are getting higher and higher, because even if there are small defects in the solar cells, it will have an adverse effect on them, and even affect their lifespan. Therefore, it is particularly important to detect it accurately, sensitively, quickly and efficiently. [0003] At present, the mainstream method of using images to achieve fault detection is deep learning, but the application effect on solar cells is n...

AI Technical Summary

Problems solved by technology

Due to the difficulty of collecting fault pictures in the actual operation of the equipment, the obtained fault picture data is less. If the conventional method is used to solve the problem, the pictures are flipped, cut, amplified, translated, etc., or the corresponding noise is added. , the image processing can only obtain less information, which cannot solve the problem well, and the generative confrontation network can only generate as real images as possible, often ignoring the real key local information

Images