Automatic recognition of erosion area on the slope of tailings dam using region growing segmentation algorithm

The precise recognition of the deformation region is of importance for the early warning of tailings dam erosion failure. A region-growing segmentation algorithm based on the combination of multiple sub-pixels and point cloud coordinates is proposed in this paper for the above purpose. This method is implemented using Python and OpenCV open-source libraries. Through image cropping, threshold segmentation, point cloud coordinate extraction, and region growing segmentation algorithm, the erosion and failure area of the tailings dam was recognized. A physical modeling experiment was used to simulate the deformation process of the tailings dam failure for validation of our method. Compared to the current mainstream superpixel SEEDS segmentation method, our method can effectively improve the image recognition accuracy in complex scenes. The results show that the average identification error in X - and Y -directions by using the new method reduced significantly (3.744 and 4.910% in the current method; 8.302 and 9.976% in the superpixel method). In addition, the form and mechanism of erosion failure are analyzed from the two factors of rainfall infiltration and surface water infiltration. The erosion failure of the dam body is mainly caused by the material sliding downward and sinking inward, while the different forces will produce different failure effects. This study probably applies to identify the landslides deformation like the tailings dam failure in complex scenes with high precision.