| Abstrakt: |
In the face of burgeoning population growth, ensuring security during public events, familial gatherings, and in high-traffic areas has become increasingly challenging. The manual monitoring of these areas, though facilitated by closed-circuit television (CCTV) cameras, often proves laborious and error-prone, leading to potential oversight of suspicious activities within crowds. To ameliorate this issue, an intelligent system for crowd monitoring and suspicious activity detection has been developed, utilizing deep learning algorithms. Specifically, the combined use of Fully Convolutional Networks (FCN) and Long Short-Term Memory (LSTM) was employed in the analysis of crowd behavior. Although previous attempts have been made to address this issue, the accuracy of such systems has remained a concern, often marred by false alarms and overlooked incidents. However, the present system exhibits a marked reduction in both false positives and negatives, boasting an accuracy of 97.84%, a significant improvement over existing model. This research proposes an effective solution to the problem of manual crowd monitoring, offering enhanced security outcomes through intelligent, automated surveillance. The high accuracy achieved underlines the potential of deep learning techniques in revolutionizing the field of surveillance, with further implications for crowd management and public safety. [ABSTRACT FROM AUTHOR] |
