Research on intelligent identification method of coal gangue based on ECT-YOLOv8n

Identification of coal gangue is a crucial step in the deep processing and utilization of coal, as the presence of gangue significantly limits the efficiency of clean coal utilization. Accurate discrimination between coal and gangue remains a key technical challenge, directly impacting both the advancement of clean coal technologies and the effective reuse of gangue in construction materials. To address this challenge, this study presents a differential electrode electrical capacitance tomography (ECT) sensor and develops an ECT system specifically tailored for coal-gangue imaging. Through integrated data acquisition, signal processing, image reconstruction, and systematic analysis, the system enables synchronous and independent visualization of multiple gangue particles located at different positions within the sensing field. A dataset containing 600 fully labeled ECT images of various coal-gangue distributions is established, based on which nine YOLO-series models (from YOLOv8n to YOLOv12n) are trained, achieving recognition accuracies ranging from 93.2% to 96.7%. The proposed coal-gangue identification method using the ECT-YOLOv8n model offers a theoretical foundation for intelligent coal-gangue separation, thereby promoting higher levels of clean coal utilization and enhancing suitability for downstream deep processing applications.