High-precision visual inspection defect detection system on CFRP using small aperture optical device and computer vision

Defects such as delamination, burrs, and splits in small-aperture drilled holes can compromise quality and accuracy, affecting assembly performance and jeopardizing the safety of structural components in the precision processing of carbon fiber reinforced plastic (CFRP) . However, detecting these defects presents significant challenges, highlighting the need for effective detection methods. In our work, we tackle these challenges by introducing an automated defect detection system designed for the inner walls of small-diameter holes, integrating a specialized detection device and a computer vision-based algorithm. The device features a self-rotating optical inspection mechanism, equipped with an isosceles triangular prism for image refraction and a closed-loop electronic control platform for precise 360° automatic imaging. The algorithm applies image pre-processing techniques, including rotation, cropping, and enhancement to improve data quality. An enhanced YOLOv5s-CCE model was also developed to boost defect detection speed and accuracy, enabling high-precision identification of inner-wall defects. The system demonstrated exceptional robust performance across diverse hole diameters and depths, achieving mAP50 (mean Average Precision at 50% IOU) of 94.7%. By enabling high-precision micrometer-level visual detection, the system not only advances precision visual detection technologies but also delivers reliable quality assurance solutions for advanced manufacturing applications.