NJII Defense: Enhancing Police BodyCam Technology with AI

NJII Defense and Homeland Security has teamed with the Newark Police Department to conduct research on how to advance policing technologies, specifically in the form of body cams. Thanks to Newark PD’s partnership with IBM, NJII’s Public Safety branch was given access to Watson, a versatile and long-standing artificial intelligence platform that uses natural language processing to solve problems, answer questions, and conduct experiments relating to machine learning and computer vision. 

This project, known as the Newark project, has been in the works since 2019, and was initiated on a vision of leveraging existing state of the art technology in AI and machine learning to enhance the monitoring role and provide real-time feedback on officer behaviors. In November of 2020, Governor Phil Murphy passed a law requiring law enforcement officers to wear body cams while on duty; the bill was enacted in June of 2021, and the state appropriated $58 million to help local PDs fund the equipment, resulting in 487 departments receiving grants. However, body cams alone are not enough to increase police accountability, as stated in a 2017 Washington D.C. study that examined over 2,000 police officers; a 2018 study of 504 officers from Milwaukee yielded similar results. Artificial intelligence has the potential to increase the impact of body cams, and if implemented properly, could lead to a safer and more advanced police force. 

AI has been utilized in policing before, one example being Truleo, a Chicago-based company that uses language processing to transcribe audio from body cams. The system was adopted by the Seattle police department as well as 12 California PDs, including Alameda, CA, where use of force dropped by 36% after using the program. It also resulted in a 30% drop in unprofessional language, and a 12% reduction in civilian non-compliance. However, what’s being tested in Newark relies on computer vision, another branch of AI based on analyzing video in real-time through object and location detection, face recognition, and other techniques. In a 2021 paper from Visual Computing for Industry, Biomedicine, and Art, the authors claim that computer vision allows for more accurate crime forecasting, with prediction rates up to 87% depending on the model used. The paper goes into immense detail about the various models used, as well as the crime data sets necessary to train a predictive model; the authors conclude that computer vision has the potential to improve policing, but results would depend mostly on implementation. 

Between IBM, NJII, NJIT, and all the other resources that Newark PD have at their disposal, our community will have special opportunities to experiment and improve in this field that others will not. NJII Defense and Public Safety are committed to the implementation of new technology into local and federal law enforcement, as well as the improvement of safety procedures that will benefit officers and citizens alike; machine learning and computer vision are simply the next step in our goal.