An automatic segmentation method for coal gangue based on improved region growing algorithm

Accurate segmentation of coal gangue contours during intelligent coal gangue sorting substantially reduces the occurrence of coal gangue dropping and misgrasping by robotic manipulators, thus enhancing sorting efficiency. Precise segmentation using computer algorithms effectively extracts coal gangue contours, thereby enhancing intelligent sorting efficiency. The key to achieving accurate segmentation lies in enhancing algorithmic performance. Therefore, we propose an improved region growing algorithm for automatic coal gangue segmentation. This algorithm introduces improvements in coarse contour acquisition, seed point expansion mechanisms, and automatic threshold updating. The final segmentation result is achieved through the combination of multiple segmentation outcomes. We conducted 20 experiments to evaluate the performance of the improved region growing algorithm, comparing it with four used segmentation algorithms. Experimental results show that the proposed algorithm achieved an average Dice coefficient of 0.988 and an average Jaccard distance of 0.023. These findings demonstrate that the proposed algorithm can automatically segment coal gangue contours with high accuracy and robustness.