IMPROVING LOCAL GOVERNMENT RESPONSE THROUGH AI-DRIVEN SPATIO-TEMPORAL CROWD DETECTION AND TRACKING

Abstract

               Unusual crowd behavior in public places can quickly turn into dangerous situations if it goes unnoticed. For local governments, being able to detect and respond to such events in time is vital for protecting citizens and maintaining public safety. Although many video-based systems have been developed over the years to monitor crowds, most are still limited in their ability to recognize abnormal activities in real-world, real-time conditions. To address this gap, our study presents a Spatio-Temporal Crowd Detection and Tracking (STCDT) approach that analyzes movement patterns to identify unusual behaviors. The method applies clear rules to detect events such as crowd merging, sudden running, or splitting, which often signal risk. By providing timely insights, this system can support local authorities in making faster and more effective decisions during emergencies. We test the approach on well-known datasets including UMN, Avenue, and UCSD, and the results show its strong potential to improve local government response in managing public spaces safely.