3D location of gangue by point cloud segmentation with RG-TCF.

The existing method of gangue location, primarily relying on 2D coordinates and simplified 3D coordinates, often results in distorted position information, leading to failures in gangue sorting. In this paper, we propose region growing with two-component feature (RG-TCF) algorithm to segment the complete and uncut point cloud of coal and gangue for accurate 3D gangue location. Firstly, the workflow of RG-TCF was developed by the advantage of fast point feature histograms (FPFH) over the angle between two normal vectors used in RG (region growing). Secondly, the extraction, validation and test sets were built based on the production and annotation of point cloud. Thirdly, after eliminating noise and redundant points with the proposed down-sampling based on key point (DS-KP), segmentation thresholds of two-component feature were also worked out by histogram analysis. Finally, the performance of RG-TCF was validated and tested by segmentation and location experiments. It could be concluded that RG-TCF improved the under-segmentation effectively; it increased <italic>Dice</italic> coefficient and location precision by 10.8% and 9.2% compared with those of the popular segmentation algorithms, respectively. [ABSTRACT FROM AUTHOR]