An OGM-Based Visual Navigation System with Reduced Homographies and Accurate Computations for Mobile Robots

摘要

This article presents a novel approach to calculate and accelerate the Occupancy Grid Mapping (OGM) for real-time mobile robot navigation. Here, we have improved the main process involved in the construction of OGMs in two ways. On one hand, we propose a new method to perform the camera calibration (CC) process, in which we reduce the number of intermediate steps (homographies) when the reference system is transformed from the real world into a navigation map, reducing consequently the computational costs. On the other hand, the OGM is constructed by an online probabilistic method, which simultaneously performs both the camera calibration and maps construction, thus reducing cumulative errors when these two processes are separately treated. Furthermore, the proposed system can be easily parallelized and mapped onto a digital system (e.g., Field Programmable Gate Arrays, FPGAs) for real-time robot navigation.