Underground infrastructure constitutes critical urban structures However, they may experience structural deformation or damage during construction due to the combined effects of earth pressure, water pressure, and external loads, potentially compromising operational safety and leading to severe accidents. Consequently, deformation monitoring of underground infrastructure serves as a vital safeguard for structural integrity. This paper proposes an intelligent deformation monitoring method for underground infrastructure based on inertial-visual measurement, addressing the limitations of conventional visual measurement techniques affected by environmental vibrations. The methodology involves collecting tunnel target images using an inertial-visual measurement system, extracting target pixel coordinates through image processing algorithms, and establishing a deformation calculation model incorporating reference point calibration. Engineering validations demonstrate that this approach achieves high-frequency, high-precision real-time monitoring in long-span tunnel settlement observations, with measurement accuracy exceeding 1mm and showing a mutual discrepancy of 0.27 mm compared to total station measurements. Experimental results indicate that the inertial-visual measurement system exhibits superior precision and stability, demonstrating strong consistency with total station monitoring data and promising application prospects.