Acceso usuarios
Shanglin Wang, Celso. Co
During the stamping process, various defects may occur due to various factors such as materials, molds, and processes, such as cracks, dents, wrinkles, etc. Therefore, it is of great practical significance to quickly and accurately detect and identify stamping defects in automotive panels. This article aims to study a fine-grained recognition method for stamping defects in automotive panels based on deep learning. By constructing a deep learning model, automatic detection and classification of surface defects in stamped parts can be achieved. This study designed a multi-stage defect recognition model. The model first extracts deep features of the image through convolutional neural networks, and then uses a classifier to classify the features, achieving coarse-grained recognition of defects. Subsequently, for each type of defect, this study further designed a fine-grained recognition module, which improved the model’s ability to capture defect details by introducing attention mechanisms and feature fusion techniques. Through experimental verification, the proposed deep learning based fine-grained recognition method for automotive panel stamping defects has achieved significant results. Compared with traditional methods, this method not only improves detection efficiency and accuracy, but also reduces the degree of manual intervention and enhances the objectivity of detection. In addition, this study also conducted an in-depth analysis of the performance of the model, exploring the impact of different network structures, parameter settings, and other factors on model performance, providing a basis for subsequent model optimization and improvement.
