North Mississippi Regional Law Enforcement Technology Project

Recently there has been a surge in use of surveillance technologies, such as Closed-Circuit Television (CCTV) cameras, body worn cameras, and in-car video cameras used by law enforcement in attempt to prevent, reduce, and resolve crime cases (Piza, 2021). While most of the literature has focused on the effectiveness of these surveillance technologies regarding crime prevention, law enforcement primarily uses surveillance technologies to resolve crime cases (Dowling et al., 2019). The study of surveillance technologies for improving clearance rates is an emerging literature identifying some areas of improvement such as increasing video quality, improving video coverage, and reducing the number of hours spent surveilling cameras by enhancing video analytics (La Vigne et al., 2011; Shukla, Lawrence, & Peterson, 2020). With rural policing departments often being underfunded and understaffed (La Vigne et al., 2011), we propose to improve video analytics through artificial intelligence. Specifically, we aim to 1) Create an Automated Video Annotation and Search Tool (AVAST) to assist crime analysts with improving detection of anomalies in alleged criminal activity; 2) Deliver and evaluate effectiveness of training to law enforcement partners to assist them in using AVAST in investigations; 3) Distribute AVAST to academic and law enforcement practitioner communities. To build the AVAST, we will develop a system to convert images from video at regular intervals and create automated annotations of these images using computer vision and artificial intelligence methods. Further annotations will then be added by subject matter experts, and the system will be retrained based on added information. The annotations and their associations to videos will be stored in a database that can be retrieved via a search engine that will be designed for use by law enforcement personnel. To evaluate the effectiveness of the AVAST, we will create pre- and post- test surveys, conduct semi-structured interviews, and conduct usability testing using think-aloud techniques. We will conduct quantitative and qualitative data analyses developing extensive codebooks for variable creation. Data will be collected across police agency roles, including Chief of Police, camera operators, investigators, law enforcement officers, and crime analysts. Results will be used in an iterative design process and eventually employed in other nearby regions. We will disseminate our results widely through presenting at top conferences and submitting manuscripts to journals across Computer Science, Sociology, Criminology, and Cognitive Science. Additionally, we will seek to reach a wider audience by writing blogs and creating infographics to share across online platforms.