Fusion of visual and compass sensors for location recognition

This paper presents a fusing methodology through an illustration of how visual and compass sensors for location and route recognition can be combined. The data acquisition platform consists of two cameras, a compass sensor and several peripheral inputs such as a GPS and an accelerometer. With just the visual features and compass directions, a Fast Learning Artificial Neural Network (FLANN) is used to learn and recognize locations. Through video sequence training and corresponding compass information, the system was able to automatically learn key visual and compass information from key locations and routes. This combination of visual and compass information provides the possibility for visual navigation. While there are only two sensor-type dimensions, this serves as a blue print for extensions onto other sensor fusing capabilities.