Abstract: In order to meet the requirements of intelligent construction for borehole robots in the future, the coal and rock identification technology is studied by using the image characteristics of the returned slag, which is characterized by only cuttings and cinders in the returned slag. The results show that the coal and rock interface recognition methods mainly include image preprocessing, threshold segmentation and image recognition. In the image preprocessing, HSV color space is first used for conversion to extract the brightness component, then Gaussian filter is used for image denoising, and finally Laplace method is used to realize image enhancement. The preprocessed image was segmented with a fixed single threshold. Taking the experimental data as an example, the fixed threshold value of the sample image was determined to be 115 by using the maximum variance method. Finally, based on the binary image of image segmentation, coal and rock were identified by calculating the proportion of coal cinder and cuttings in the total number of image pixels. Then, by analyzing the large number of samples collected in the scene, threshold values of coal and rock boundary were set under different coal seam conditions, so as to realize image recognition.