Estimating urban activity on high-resolution thermal image sequences aided by large scale vector maps

In this contribution we describe a method to assess the activity of vehicles based on airborne image sequences from an infrared camera. At the resolution of approximately one meter vehicles appear as elongated spots. In urban areas many additional other objects have the same property. To discriminate vehicles from these objects we fuse information from IR-images and vector-maps. Also, evidence from overlapping areas of frames or an IR-sequence is accumulated in the scene. We also perform grouping of vehicles into rows along the margins of the roads. These tasks can be automatically performed by a production net. The generic productions for grouping become feasible through the use of map knowledge as context.