An Automatic Detection Method for Coal–Rock Interface Based on YOLOv8 and Cubic Spline Interpolation

It is crucial to accurately extract the continuous coal–rock interface in order to achieve automatic height control of the coal mining machine. This paper proposes an automatic extraction method that combines YOLOv8 detection and cubic spline interpolation to address the problem of discrete output and difficulty in directly generating coal–rock interfaces in existing deep learning detection methods. First, YOLOv8 is used to efficiently perceive coal–rock images and output a discrete sequence of bounding box centre points. Then, through cubic spline interpolation, it is fitted into a smooth, continuous coal–rock interface that conforms to the shape of the rock layer. Experiments have shown that this method can achieve end-to-end processing and performs well in typical complex underground scenes. The cubic spline interpolation strategy improves the vertical positioning accuracy of the boundary line, with a 14.9% increase in vertical positioning accuracy compared to linear interpolation. It significantly improves the engineering usability and control accuracy of the output curve, providing a reliable technical solution for intelligent mining in coal mines.