Real-time automated surveillance for counteracting automobile larceny (RASCAL) is envisioned to be an automated monitoring and vehicle identification system that utilizes neural network technology to provide a non-intrusive detection system for identifying stolen vehicles on the nation's highways. Specific facets of the system that are presented include automatic scene analysis, vehicle classification, and a potential commercial off the shelf license plate reader for plate identification. The general approach is envisioned to include the generation of a target vehicle profile that can be downloaded to each of the remote video surveillance sites on the highway. The profile will be a predefined characterization containing the make, model, year, color and license plate of the target vehicle. The intelligent video processor, located at each remote camera site, will utilize the stolen vehicle profile and advanced neural network classification techniques to search the visual scene for a potential candidate of the stolen vehicle. An automated license plate reading system is used to confirm the identification of the vehicle. Once a potential stolen vehicle has been identified a snapshot of the vehicle, along with the vehicle profile, is transmitted back to the central control facility where law enforcement officials can take appropriate action. For this effort, a three sensor distributed configuration is envisioned.