Findings and methodology are reported for a project that developed and examined the effect of computer analytics for public-space surveillance camera systems.
Videos from public-space surveillance cameras have the potential to aid in post-event investigations and facilitate intervention in live criminal incidents by flagging them as they occur; however, when they are left unmonitored or are poorly integrated into police departments, these surveillance cameras are of little benefit to law enforcement. In an effort to address this untapped potential, the current project developed computer vision analytics for large surveillance camera networks and installed them in a Public Safety Visual Analytics Workstation (PSVAW) at the Orlando Police Department (OPD) in Florida. This project report focuses on the tasks performed in four areas during the project period (January 2016 - June 2018). The project first developed and tested algorithms related to action and event detection in videos, video anomaly detection, video summarization, object and attribute retrieval, and hashing for efficient information retrieval. Second, the project developed a computer vision system GUI that incorporates a set of computer vision modules for the PSVAW. These computer vision modules included anomaly detection, face attribute prediction, body attribute predictions, and action detection. Third, the project installed in the OPD the equipment needed to operate the PSVAW. Fourth, the project developed this report on the current status of this placement in collaboration with the OPD. Currently, OPD is considering means to continue to staff the workstation. 37 figures, 18 tables, 13 references, and a listing of project publications
Report (Grant Sponsored)
Date Published: October 1, 2018