A novel and efficient auto-navigation system based on machine vision for an unknown environment is developed in this study. A 3D model using only the measurements of a single image is implemented for 3D object measuring and map-building. For path-planning, the A~* algorithm is combined with the Floyd's shortest path to determine the optimal sub-goals in the image sensing range. The performance of the present method is compared to those of the Bug algorithm, virtual force field (VFF) path planning, genetic algorithm (GA) path planning and fuzzy logic control (FLC) path planning and shows overall better results in a cluttered environment. Further, the 3D map-building of the environment map is successfully accomplished.
展开▼
