Research on Global Vision-Based Navigation for Unmanned Vehicles in Complex Environments

In response to the issues of delayed dynamic obstacle perception and low timeliness in local path replanning under traditional navigation methods in complex construction site scenarios, this paper proposes an unmanned vehicle navigation method based on a global vision camera. First, the environment's depth information is captured by a global depth camera, generating a 3D point cloud which is then converted into a dynamic 2D grid map, updating the obstacle distribution to achieve environmental topology reconstruction; when the depth camera detects long-distance obstacles causing blockage of the original path, this paper introduces a path optimization mechanism that triggers local path adjustment or global path replanning based on the updated real-time grid map. Experimental results show that this method can effectively predict obstacle interference in advance and promptly adjust the path. Compared with traditional methods, the global vision method reduces the path length by 32.3 % and shortens the path planning time to 43.5 % of that of traditional methods in sudden obstacle tests.