Recognition on ore zone separation points target detection and identification in mineral processing shaking table based on improved YOLOv5 algorithm

The operation of shaking tables in mineral processing is influenced by multiple parameters, including feed rate, feed concentration, feed grade, and feed particle size, which cause variations in the position, color, and width of the ore bands on the table surface. To ensure the quality of the concentrate, it's necessary for workers to timely adjust the position where the concentrate is collected, maintaining the stability of the concentrate grade. Varying experience and skills of operators often lead to the fluctuations in production indicators. To reduce the labor intensity of operators and enhance the level of automation in mineral sorting with shaking tables, this paper introduces an improved YOLOv5 target detection algorithm, which successfully extracts the boundary points (ore band separation points) and marker information of both the concentrate band and the middling band on the shaking table. Compared with other algorithms such as YOLOv5, SSD, and Faster-RCNN, the improved YOLOv5 algorithm demonstrates the best detection performance with the highest precision, achieving an average precision of 98.3%.