Construction method of 3D mine mining model based on multi-source point cloud data

This study proposes a deep convolutional neural network (CNN) classification model that combines multi-source point cloud data (MSPCD) fusion to address the limitations of traditional model construction methods in terms of accuracy, efficiency, and large-scale data processing. For the 3D mining model, a richer point cloud dataset is created through data fusion, and the CNN is utilised for classification and feature extraction. Experimental results show that this model outperforms existing methods in terms of processing time (12.678 seconds), memory usage (102.345 MB), and accuracy (97.35%). Additionally, it exhibits lower model distortion at various scales, particularly at 1,000 m and 5,000 m. This research offers a novel approach for constructing high-precision mining models, especially in complex terrain and large-scale data scenarios, demonstrating significant practical value.