Lithium battery surface defect detection method based on depth field adaptation

Abstract

The invention provides a lithium battery surface defect detection method based on depth domain adaptation, and the method comprises the steps: designing an adaptation layer based on statistics such asmaximum mean value difference MMD and KL divergence in a classifier, and designing a domain discriminator for adversarial discrimination of which domain the extracted features come from after featureextraction. On the one hand, a complementary mechanism of the two modes can enable the extracted public features of the two domains to be more sufficient; on the other hand, the adaptive layer designbased on the statistics can enable the target domain data to participate in the training of the classifier, thereby enabling the model to have better generalization capability on the target domain. According to the model, a simple and effective multi-scale feature fusion strategy is designed in a feature extraction network, and a good recognition effect can be achieved on small defects. Accordingto the method, an efficient detection effect is achieved, the dependence of deep learning on label data is relieved, and the trained model has better generalization ability for target domain data.

Description

technical field [0001] The invention relates to the technical field of machine vision, and specifically provides a detection method for lithium battery surface defects based on depth field adaptation. Background technique [0002] At present, lithium-ion batteries are more and more widely used, such as mobile phones, notebooks, electric vehicles, etc., forming a huge industrial group. However, some defects produced in the production process seriously affect the life and safety factor of lithium batteries. Such as edge folds, electrode sheet scratches, exposed foil, particles, perforations, dark spots, foreign objects, and surface dents, stains, bulges, coding deformation, etc. The traditional method for battery defect detection is manual measurement and judgment. However, the results of battery testing are affected by human factors such as the subjective wishes, emotions, and visual fatigue of the testing personnel, resulting in missed and false detections. The detection ...

AI Technical Summary

Problems solved by technology

[0004] However, the existing deep field adaptation methods cannot be directly applied to the surface defect detection of lithium batteries, and the recognition rate is not ideal

Images