Improving Geolocation by Combining GPS with Image Analysis

The Global Positioning System (GPS) provides geolocation to a considerable number of applications in domains such as agriculture, commerce, transportation and tourism. Operational factors such as signal noise or the lack of direct vision from the receiver to the satellites, reduce the GPS geolocation accuracy. Urban canyons are a good example of an environment where continuous GPS signal reception may fail. For some applications, the lack of geolocation accuracy, even if happening for a short period of time, may lead to undesired results. For instance, consider the damages caused by the failure of the geolocation system in a city tour-bus transportation that shows location-sensitive data (historical/cultural data, publicity) in its screens as it passes by a location. This work presents an innovative approach for keeping geolocation accurate in mobile systems that rely mostly on GPS, by using computer vision to help providing geolocation data when the GPS signal becomes temporarily low or even unavailable. Captured frames of the landscape surrounding the mobile system are analysed in real-time by a computer vision algorithm, trying to match it with a set of geo-referenced images in a preconfigured database. When a match is found, it is assumed that the mobile system current location is close to the GPS location of the corresponding matched point. We tested this approach several times, in a real world scenario, and the results achieved evidence that geolocation can effectively be improved for scenarios where GPS signal stops being available.