Learning traffic camera locations using vehicle re-identification

In its effort to provide necessary intelligence and analysis, the National Geospatial-Intelligence Agency (NGA) utilizes extensive traffic camera systems. However, the large amount of data overwhelms both analysts and existing processing methods. In order to provide a better understanding and reduce the search space for common problems such as target tracking, it is necessary to extract the camera network layout. Areté proposes the CAMEra LOcalization using Traffic (CAMELOT) solution to accomplish this by using a vehicle re-identification algorithm along with robust statistical network analysis to extract relative camera locations from traffic cameras. To do so, we will leverage Areté’s previously developed deep convolutional neural networks to create a robust solution for vehicle re-identification that considers both novel vehicle types and camera viewing geometries. The resulting vehicle re-identifications between cameras will be used to build a camera correlation matrix that defines the relative location of each camera.