The detection of steel surface anomalies has become an industrial challenge due to variations in production equipment, processes, and steel characteristics. To alleviate the problem, this paper proposes a detection and localization method combining 3D depth and 2D RGB features. The framework comprises three stages: defect classification, defect location, and warpage judgment. The first stage uses a data-efficient image Transformer model, the second stage utilizes reverse knowledge distillation, and the third stage performs feature fusion using 3D depth and 2D RGB features. Experimental results show that the proposed algorithm achieves relatively high accuracy and feasibility, and can be effectively used in industrial scenarios.

Vision-based measurement technology benefits high-quality manufacturers through improved dimensional precision, enhanced geometric tolerance, and increased product yield. The monocular 3D structured light visual sensing method is popular for detecting online parts since it can reach micron-meter depth accuracy. However, the line-of-sight requirement of a single viewpoint vision system often fails when hiding occurs due to the object’s surface structure, such as edges, slopes, and holes. To address this issue, a multi-view 3D structured light vision system is proposed in this paper to achieve high accuracy, i.e., Z-direction repeatability, and reduce hiding probability during mechanical dimension measurement. The main contribution of this paper includes the use of industrial cameras with high resolution and high frame rates to achieve high-precision 3D reconstruction. Moreover, a multi-wavelength (heterodyne) phase expansion method is employed for high-precision phase calculation. By leveraging multiple industrial cameras, the system overcomes field of view occlusions, thereby broadening the 3D reconstruction field of view. Finally, the system achieves a Z-axis repetition accuracy of 0.48 μm.