Topological mapping and navigation based on visual sensor network

In this paper, a novel topological mapping approach applied to the navigation mission is proposed based on visual sensor network. Firstly, the natural objects in the scene monitored by visual sensor network are recognized through the inference model based on structural features. Then, topological nodes are designed according to the recognized objects and their spatial structure information, and topological mapping will be fulfilled. On this basis, global localization for mobile robot in camera view can be implemented with HOG feature based detection and tracking. Finally, the robot can make motion decisions for navigation. Our approach does not rely on artificial landmarks and is tested in large-scale office scene. The effectiveness of the proposed approach is verified.