Research and Development of a Machine Vision-Based System for Surface Defect Detection on Titanium Alloy Bars

In order to solve the problems of low efficiency, time-consuming and labor-intensive problems in the current methods of manual visual inspection and ultrasonic testing of surface defects of titanium alloy bars, a set of surface defect detection system for titanium alloy bars based on machine vision was proposed and designed. Firstly, in order to collect a complete image of the surface of the titanium rod, the line scan camera was used to collect the image with the rotation of the titanium rod, and the encoder was used to trigger the method to judge whether the titanium rod rotated for one time. Then, in order to construct a high-quality dataset, the collected images are denoised and enhanced, and the dataset is enriched by data augmentation to prevent over-fitting problems. Then, the improved YOLOv8s algorithm was used to detect and locate defects in the acquired titanium rod images. Finally, the QT program development framework is used to design a simple system UI interface to realize the operation of the titanium rod defect detection system and the display of defect information. Experimental results show that the system can effectively detect and classify the surface defects of titanium rods, and improve the accuracy and detection speed of surface defects of titanium rods.