CCTV as an automated sensor for firearms detection: human-derived performance as a precursor to automatic recognition

CCTV operators are able to detect firearms, via CCTV, but their capacity for surveillance is limited. Thus, it is desirable to automate the monitoring of CCTV cameras for firearms using machine vision techniques. The abilities of CCTV operators to detect concealed and unconcealed firearms in CCTV footage were quantified within a signal detection framework. Additionally, the visual search strategies adopted by the CCTV operators were elicited and their efficacies indexed with respect to signal detection performance, separately for concealed and unconcealed firearms. Future work will automate effective, human visual search strategies using image processing algorithms.