Development of an AI Model for Detecting Risk Situations in Surveillance Camera Images

Abstract

Security cameras have become increasingly common in everyday life worldwide. Along with the growth in the number of devices, there has been significant technological advancement, resulting in improved image quality and the incorporation of features such as night vision, audio capture, cloud storage, and remote communication, enabling real-time remote monitoring. Major cities have integrated video surveillance systems into their public safety strategies. Despite these advances, traditional monitoring still presents notable limitations, particularly the reliance on a large number of human operators. In many cases, cameras are used solely for image recording, with subsequent analysis conducted for identification purposes or as evidentiary support. In this context, this study employs the Edge Impulse platform to evaluate the implementation of artificial intelligence aimed at the automatic recognition of hazardous situations. The proposed system seeks to alert operators and security authorities, while enabling the expansion of monitoring programs without the need to increase control room capacity or personnel.