To determine out-of-plane displacement, it is challenging to simultaneously implement high resolution, wide range, and real-time measurement. This study proposes a method for displacement compensation based on the Michelson single-point displacement measurement system. The direction of fringe movement and amplitude of the object’s displacement was calculated and converted into a feedback-tracking signal. The compensator was driven by the feedback signal to change the reference optical path to keep the fringes stable, and a method to detect fringe movement was developed. A convolutional neural network model was set to distinguish the direction of fringe movement, and a backpropagation neural network was used to calculate the amplitude of the movement using simple image processing at a high speed. The system’s resolution was 10 nm in the range 210 µm, and the tracking time step was smaller than 200 ms. This provides an effective solution for high precision, real-time, and wide range measurement.