A LiDAR-based approach for real-time bulk coal extraction in cargo holds

Accurate bulk coal extraction in cargo holds is crucial for ensuring the operational safety and efficiency of automated cleaning systems. This paper presents a LiDAR-based approach tailored for real-time coal feature identification in cargo holds. First, a self-adaptive mechanism is proposed to filter redundant point cloud data, dynamically adjusting to the dimensions of the cargo hold. Next, an angle prediction method is developed based on the structural characteristics of the cargo hold, ensuring precise heading estimation. Subsequently, point cloud registration is performed to determine the robot’s pose and generate comprehensive global point cloud data. Finally, the registered data are projected into spherical coordinates, enabling the identification of coal point clouds based on their angle of repose. The proposed method achieves a data processing rate of 10 Hz. Experimental validation carried out in both simulated and real-world environments, including Dushan Port, demonstrates the capability of the method for accurate and efficient coal extraction, effectively addressing the challenges posed by uneven coal surfaces.