Intelligent 3D Visual defect Detection for Industrial Applications

Abstract

Currently, new energy industries such as electric vehicles and energy storage batteries are experiencing rapid growth throughout the world. As a recognized ideal energy storage element, lithium batteries have also attracted considerable interest. To improve image contrast, the industry employs image preprocessing algorithms. Due to the high requirement of image consistency in the template matching method, which is difficult to meet in practice, and the large apparent difference between defects, it is challenging to design features. Therefore, the framework for defect detection is based on a method of deep learning with strong feature expression capability. In light of the severe imbalance in the number of samples in defect detection, data augmentation and generation methods are used to simulate real samples in order to improve the training effect of deep neural networks and alleviate the burden of data annotation to some extent.