Detection of moving vehicles with WorldView-2 satellite data

Traffic monitoring in urban areas is a complex issue and recent Remote Sensing technologies can play an important role in planning and monitoring the urban environment. In this study a semi-automatic object-oriented workflow was designed to detect moving vehicles and their speed from single pass WorldView-2 multispectral data. The time lag in data recording between each spectral band causes a small image displacement of moving objects and this discrepancy is used to detect moving vehicles, their speed and direction of travel. The method proposed was applied to a very complex study area in the historical core of city of Multan, in the Pakistani southern province of Punjab, where very small and extremely dense built-up old style houses are mixed together with narrow roads and bazaar streets. First results show interesting applications of this new technology, with achieved accuracies of about 67% evaluated comparing automatic detection vs. manual interpretation.