Surface Defect Detection of Corrugated Sheets Based on Photometric Stereo Vision

Plate heat exchangers are extensively employed in industrial processes for thermal energy transfer and waste heat recovery. The core component, the cor-rugated sheets, possesses a complex metallic surface geometry that is critical to the ex-changer's performance. Corrugated sheets are formed through die stamp-ing, dur-ing which manufacturing defects may occur due to process limitations. Detecting defects prior to use is crucial for ensuring the safety of industrial op-erations. Due to the complex curved surface geometry of the corrugated sheets, images captured by a single-viewpoint imaging system often suffer from shading caused by the plate’s undulating structure. As a result, defects tend to appear within shadowed regions of the surface, making it difficult to enhance their con-trast through illu-mination alone, thereby posing significant challenges for defect detection. The paper, based on photometric stereo vision, develops a photomet-ric stereo imaging system tailored for corrugated sheets and proposes a feature map extraction method based on Gauss spherical distance mean convolution. Based on the Gauss spherical distance feature map, defects are enhanced, and defect detection is per-formed using a method that fills the gray-scale concave segments of the feature map and marks the extrema points. Experimental results show that the defect de-tection accuracy in both the flow guide and heat ex-change regions of the corru-gated sheets exceeds 95%.