Research on online identification technology for coal gangue based on multimodal data fusion

Abstract To improve coal sorting efficiency, reduce gangue transportation costs, and minimize environmental pollution, this paper proposes an online gangue identification method based on multimodal data fusion. The method integrates three types of characteristic data, which are visual images, near-infrared spectra, and vibration spectra. It employs a multimodal feature extraction and fusion strategy to construct an efficient gangue identification model. The study first collected multimodal data from coal and gangue, then preprocessed the data to eliminate noise and interference. Subsequently, a feature-level fusion method was employed to extract and fuse key features from different modalities. Finally, machine learning algorithms were used to train the identification model and optimize its performance. Experimental results demonstrated that the method achieved a high recognition rate (>95%) in gangue identification tasks, outperforming single-modality identification methods. Multimodal data fusion effectively addresses the limitations of single data sources, enhancing the robustness and adaptability of the recognition system. This study provides a feasible technical solution for online coal gangue sorting, offering practical application value for the clean and efficient utilization of coal.