Abstract: To address the issues of low efficiency and poor quality consistency in the manual fabrication of power cable terminals in urban distribution networks, as well as the problem where automatic cutting devices struggle to accurately obtain the actual thickness of each covering layer—often resulting in conductor damage or residual semiconductive layers—this paper proposes a nondestructive cable covering thickness detection technology that integrates physical-geometric modeling with an intelligent compensation algorithm. This technology enables the intelligent perception of covering thickness and the precise acquisition of the optimal cutting depth for cables awaiting processing. The approach constructs a heterogeneous sensor rotary scanning array that fuses eddy current and laser information, establishing a dynamic differential thickness solution model based on a polar coordinate system.To tackle the nonlinear temperature drift of sensors in field environments, a temperature compensation model based on an Improved Sparrow Search Algorithm (ISSA) optimized Kernel Extreme Learning Machine (KELM) is proposed. The improved SSA algorithm, enhanced by Cauchy mutation and Logistic mapping, efficiently obtains the optimal regularization and kernel parameters for the KELM. This establishes a nonlinear mapping relationship between the sensor output voltage, ambient temperature, and true displacement. Experimental results show that this technology effectively eliminates interference from cable geometric deformation and temperature fields, significantly improves the thermal stability of displacement perception, and provides reliable depth feedback for the automated precision cutting of cable terminals.