Abstract: In large coal mine stockyards, the presence of numerous coal piles with complex surface structures and irregular shapes makes traditional manual methods for measuring coal pile volume difficult to implement and often inaccurate.To address this issue, an incremental learning-optimized coal pile volume measurement algorithm is proposed.The method uses LiDAR to scan the irregular surfaces of coal piles, acquiring a series of three-dimensional point cloud data covering the surfaces at specific intervals.The Delaunay triangulation algorithm is then adopted to process the point cloud data, while an incremental learning algorithm automatically adjusts the density and distribution of the triangulation to adapt to the complexity of the coal pile surfaces.This enables the reconstruction of three-dimensional models of the coal piles and accurate volume measurement, ensuring the completeness of the 3D models and the precision of the volume calculations.To validate the effectiveness of this method, simulated coal pile point cloud data were used to test the accuracy of the volume measurements.The test results demonstrate that this approach enhances the quality of Delaunay triangulation through incremental learning, significantly improving the accuracy and reliability of coal pile volume measurements.