Abstract: Remote real-time monitoring of surface subsidence is a critical requirement for effective risk management in underground engineering, particularly concerning surface subsidence and geological landslide disasters. This study presents an innovative method that combines digital image correlation (DIC) with deep learning to address the challenges of environmental complexity, accuracy reduction, and rapid monitoring in engineering sites. The proposed approach enables swift computer vision tracking and precise settlement calculations for multiple target measurement points. Utilizing YOLO V8 along with custom-developed targets facilitates multi-target localization and matching. DIC provides accurate tracking of measurement points, while sub-pixel search algorithms with quadratic surface fitting enhance monitoring accuracy. Additionally, a homography transformation method is employed to derive true deformation values. Based on this algorithm, we have developed a comprehensive computer vision remote real-time monitoring system for surface subsidence and built a complete monitoring system that integrates remote image acquisition, real-time image transmission, real-time settlement calculation, data management based on WebGIS, and risk warning. This system is both practical and scalable. The functionality of the system was validated through both indoor simulations and field experiments. Results from the simulations demonstrated that the sub-pixel search algorithm significantly improves monitoring accuracy, while the homography transformation effectively mitigates errors arising from excessive angles between the camera axis and the measurement target surface. The integrated DIC tracking algorithm, augmented by deep learning, successfully minimizes occlusion and accurately monitors large deformations. Field experiments further confirmed that our multi-target measurement approach outperforms traditional DIC algorithms by reducing the effects of strong light and fog, thus exhibiting greater adaptability to natural environmental conditions. These findings contribute positively to the advancement of new technologies for surface subsidence monitoring in engineering disaster management.