A Novel 3D Defect Detection System for Complex-Shaped Components

Accurate defect detection in high-end equipment components is essential for ensuring reliability and performance. This paper addresses the challenges of inspecting complex-shaped components, such as aviation core components, by developing an intelligent defect detection system based on human-like imaging principles. The system integrates a high-precision five-axis motion mechanism, a microscopic imaging system and a multi-spectral illumination system, enabling comprehensive and precise data acquisition. Additionally, we have constructed a small-scale 3D point cloud dataset specifically tailored for complex-shaped components. Furthermore, we propose a novel 3D anomaly detection method that combines coarse and fine registration strategies for point cloud alignment, utilizes the PointMamba for feature extraction, and employs a dual memory bank approach for defect identification. Experimental results demonstrate significant improvements in detection efficiency and accuracy, providing robust technical support for the high-end equipment manufacturing industry.