Abstract: Aiming at the problem of lacking accurate and efficient difference analysis methods for different versions of instrumentation and control (I&C) drawings in the iterative design process of nuclear power I&C systems, this paper proposes an intelligent difference analysis method for nuclear power I&C drawings integrating image semantics. This method employs the Hash algorithm to rapidly analyze identical images, uses image segmentation algorithms to focus on effective content, and combines pixel comparison with semantic understanding to accurately identify real differences, achieving fast and highly accurate intelligent analysis of drawing differences. The results of case studies show that this method achieves a precision of 89.7% and a recall of 98.6% on the validation dataset, with an analysis speed of 16.89 FPS (Frames Per Second), striking a good balance between recall and precision indicators. This method provides a new idea for difference analysis of I&C drawings. Carrying out difference analysis of I&C drawings in a human-computer collaboration manner can meet engineering application requirements, and its analysis efficiency is significantly better than manual analysis, demonstrating remarkable engineering application value.