The future of this project is very bright as there is a complete version currently implemented and going through testing.   

 

In addition, testing has not been completed as high-fidelity traffic camera data is extremely difficult to find for free. To attempt to get actual high resolution, high frame rate traffic camera footage, requests have been made to the Dayton Police Department. The University of Dayton Research Institute also has access to traffic camera footage in Dayton, Ohio, and efforts are being made to test with using this footage to test the system. In the future, collaboration will also occur with MADD to make it easier for research institutions to get traffic camera footage.  

 

The system is also being tested by manually creating anomalous car footage and running it through the system to ensure that anomalies are detected, while also reducing false-positives.   

 

The system is multi-threaded, and designed to work with a single CPU. However, in the future, GPU based processing is being investigated. In addition, Spark is being investigated to use clusters to run the system. Spark allows code to be written, with minimal change, to run on multiple computers and run directly on memory, utilizing multiple processors, to radically reduce compute time. The current run time is around 20 seconds per frame, and work is being done to bring it to 15 Frames per Second, a speed at which the system can be deployed for real use.

 

In the future, work will also be conducted to use the results of the Computer Vision system through the Haar classifier, which will allow the system to learn and get more accurate over time. Car databases for Haar are currently too small, so an autonomous training system will radically improve detection rates.

 

Further systems to create more complex models of drivers, to have more accurate anomaly detection are also being investigated.  

 

The primary focus of future development of TCD3 is to upgrade the code to production level and smooth out bugs, and begin running trials with real traffic camera data. TCD3 will eventually be merged with the Footprint product to assist Police Departments.