Lithium (Li) is an essential resource for energy storage; however, traditional flotation processes for spodumene are inefficient and environmentally expensive. This study aims to develop an intelligent, low-carbon pre-selection technology based on the photoluminescence properties of spodumene under 365 nm ultraviolet (UV) light. An improved MT-YOLOv11 deep-learning algorithm that integrates wavelet transform convolution and a dynamic detection head is proposed to accurately distinguish spodumene from gangue minerals in UV fluorescence images. Ablation and comparative experiments demonstrated that MT-YOLOv11 achieved superior detection performance, with a precision of 93.5%, recall of 79.2%, and mean Average Precision (mAP) of 88.5%, outperforming other classical detection algorithms. The model was deployed in a semi-industrial UV fluorescence sorting system at the Dahongliutan Mine in Xinjiang. It showed stable operation, increased concentrate grade, and enhanced tailings rejection compared with conventional X-ray sorting. The proposed method is practical and efficient for intelligent, green, and low-carbon Li mining.
Yu et al. (Fri,) studied this question.