Abstract: The structure of the high and steep slope dangerous rock mass is complex, and the critical sliding surface is in a relatively static state of ultimate equilibrium. There are problems such as unclear boundaries and ranges of the dangerous rock mass, and low accuracy in risk identification. A high and steep slope dangerous rock mass recognition algorithm was proposed, based on unmanned aerial vehicle red green blue (RGB) image segmentation. The optimal segmentation scale was identified by calculating the heterogeneity coefficient of objects in the RGB images, and the characteristics weight was evaluated by Relief algorithm based on key features of inclination angle of the rear wall and maximum height difference. Then the dangerous rock mass was precisely recognized with support vectors. The experimental results show that the optimal segmentation scale for the experimental slope is 320, and the dangerous rock mass can be effectively identified by this method, with accurate delineation of the boundary and scope of the dangerous rock mass. The research method could improve the accuracy of identifying dangerous rock mass on high and steep slopes to 100% under different terrain conditions, providing references for the prevention and control of instability of high and steep slopes.