There may be scratches, dislocations, color, and texture changes in mineral materials such as wood, steel pipes, and fabrics in the mine production process, which influences the production efficiency of mine to some extent. To solve this problem, better detect anomalies and improve production efficiency, this paper proposes a detection algorithm for mineral material anomalies based on visual saliency. In our algorithm, first, we perform morphological processing and median filtering on the input image of mineral materials with anomalies, highlighting the area that has anomalies. Secondly, we divide the processed image into multiple overlapping image blocks to obtain the feature template of the image, use the template to obtain the visual saliency of each part of the image blocks, stitch the images together, form a visual saliency map, and initially locate the anomaly area. Finally, we combine the visual saliency map with superpixel segmentation to pinpoint the anomaly area further. The experiments prove that our algorithm has a high recognition accuracy and good recognition effect. The algorithm can be directly applied to the real-time monitoring of mines.

mineral materials, anomaly detection, visual saliency, superpixel segmentation