Aiming at the problem that it was difficult to achieve large-scale and high-precision 3D reconstruction of land form of geomorphology by visual method in deep-sea environment, a design scheme of line structured light 3D reconstruction vision sensor for deep-sea topography features based on Jiaolong manned submersible was presented. According to the principle of 3D reconstruction of linear structured light, firstly, the Steger algorithm was improved to extract the center line of laser fringes quickly and accurately, then direct calibration method was used to solve the 2D spatial coordinates of topographic features, and the influence of image distortion on reconstruction was overcame. Then, the obtained 2D point cloud was fused with multi-sensor data, and the offset was calculated in real time according to the coordinates of the submersible, and the image was corrected according to the change of sensor position and attitude. Finally, the 3D point cloud data of landform was obtained. According to the reconstruction principle, the hardware and software system of the vision sensor was designed, and an experimental simulation platform was built to verify the feasibility of the scheme. Through underwater experiments, the complete reconstruction of simulated geomorphology can be achieved, with an accuracy of 96.9%, which meets the reconstruction requirements.