Recognition of Coal Gangue via Stereo Vision and Density: A Feasibility Study

Coal gangue recognition is essential for improving the efficiency of coal utilization and reducing environmental pollution in mining operations. This paper presents a recognition method that integrates stereo vision and density estimation to classify coal and gangue. The proposed system captures left–right image pairs of the target object using a binocular stereo camera, followed by image preprocessing and feature point extraction using the Speeded-Up Robust Features (SURF) algorithm. 3D reconstruction is then performed based on stereo matching to estimate the object volume. Combined with mass information obtained from an external weighing device, the object’s density is calculated and used as a basis for classification. Experimental results on two representative samples demonstrate the feasibility of the proposed method. The approach shows potential to distinguish coal from gangue even when surface dust is present, highlighting its robustness compared to traditional texture- and grayscale-based methods. However, as a proof-of-concept study, further large-scale validation is required before industrial deployment.