To address the problems of labor-intensive operation and poor real-time performance in traditional excavation monitoring, a high-precision displacement detection method integrating deep learning and stereo vision is proposed. Firstly, a calibrated stereo camera is used to acquire displacement images of the target, and the captured stereo images are rectified using calibration parameters. Secondly, a trained deep learning model is applied to rapidly locate the target region. Then, the localization result is input into a stereo vision processing module, where target feature points are extracted through image processing and stereo matching, and the three-dimensional spatial coordinates of the target are calculated based on the stereo geometric model to achieve high-precision displacement measurement. Experimental results show that when the distance between the stereo camera and the target is 3 m, the mean error is less than 0.002 mm and the mean absolute error is less than 0.01 mm. When the distance increases to 8 m, the accuracy loss is compensated by enlarging the stereo baseline, and the mean error remains below 0.002 mm with a mean absolute error less than 0.02 mm. The results indicate that the proposed method can achieve high-precision excavation displacement monitoring and meets engineering application requirements.