LAUSR

Abstract Vision-based intelligent detection is widely used in the quality detection for steel plate owing to the advantages of online contactless detection and visual images. However, there are still many difficult challenges for complex surface regions of interest (ROI) detection of steel plate. To solve these challenges, a new detection method fusing the gray image and 3D depth information (GIDI) is proposed. In the proposed method, a graph is constructed to get the potential relationship of the ROI objects. To obtain a coarse ROI objects image, the ROI coarse compactness metric fusing the depth information is proposed. Then, a refined ROI saliency image is obtained using the ROI seeds extraction and local contrast operation. Meanwhile, a ROI segmentation method is proposed to extract the ROI objects. Experimental results demonstrate that the proposed method presents the good performance for detecting surface depressed objects.