Solder Joint Inspection Using Digital Model Priors for Industrial Applications

Solder joint inspection is a critical step in manufacturing electronic and vehicle assemblies. Defective joints can cause operational failures, making automated defect detection necessary. This project develops an inspection system based on deep learning models PSFNET and CDANET, which use digital model priors to improve recognition accuracy. The system is trained on the vehicle-solder-joint-partly-dataset containing annotated images of acceptable and defective solder joints. A web application is built using HTML, CSS, and JavaScript for the front-end, with Flask as the back-end framework and SQLite as the database. The application includes modules for home page, user registration, login, dashboard, image detection, inspection history, and logout. Users upload solder joint images through the interface; the system processes each image using both PSFNET and CDANET, then returns a combined defect classification result. All detection records are stored per user for later review. The integrated approach demonstrates that digital model priors enhance defect detection performance. The system provides a structured environment for research into automated optical inspection methods.

Keywords: Solder joint inspection, digital model priors, PSFNET, CDANET, defect detection, deep learning, Flask web application, SQLite database, image classification, vehicle manufacturing.