Abstract: The improved tangent angle early warning model is an effective landslide early warning method, but the mechanism of the improved tangent angle model is based on the accurate identification of the uniform deformation stage and its rate. The uniform deformation rate of the landslide fluctuates with the change of the disaster prone environment and the landslide evolution stage. Therefore, the tangent angles obtained by the same model under different uniform deformation rates are quite different, which is unfavorable for landslide early warning. Based on this, the paper proposes to use curve convex concave characteristics and filtering technology to adaptively obtain the landslide uniform deformation stage and average rate, and establish rate reliability evaluation rules, so as to achieve real–time correction of the improved tangent angle early warning model and ensure the accuracy of landslide early warning. The research of this paper provides technical support for intelligent and automatic real–time processing of monitoring and early warning data of landslide monitoring and early warning.