Research on a Region-Growing-Based Compensation Algorithm for Shape and Density Values in Deeply Overlapping Ore X-ray Transmission Images

Abstract: [Background]: X-ray transmission (XRT) sorting technology can improve the concentration of target minerals, thereby enhancing resource utilization and production efficiency. However, in practice, ore images often exhibit adhesion and deep overlapping, leading to sorting errors and resource waste. [Purpose]:To address this, a shape and density compensation algorithm is proposed for deeply overlapping ore images with pronounced shadow regions. [Methods]:The method employs a region-growing algorithm to extract binary images of overlapping areas, which are then combined with individual ore binary images via a logical OR operation to restore ore shapes. Density compensation is performed using contour lines based on varying circumscribed circle radii. [Results]:Experimental results show that the processed ore images better approximate actual shapes and density values. Compared with uncorrected images, the centroid accuracy improves by 73.58%, average density compensation in overlapping areas reaches 91.35%, and similarity to real ore structures increases by 82.71%. [Conclusions]:The proposed method effectively enhances shape restoration and density compensation, significantly improving mineral sorting efficiency.