| ชื่อบทความ | Utilizing Neural Network Models for Detecting Anti-Social Activities in Surveillance Monitoring within the Internet of Things (IoT) |
|---|---|
| ประเภทการตีพิมพ์ | วารสารวิชาการระดับนานาชาติ |
| ชื่องานประชุมวิชาการ/วารสาร | Tuijin Jishu/Journal of Propulsion Technology |
| ผู้แต่ง |
สุรเชษฐ์ จันทร์งาม เกียรติศักดิ์ จันทร์แก้ว ต่วนนูรีซันน์ สุกิจจานันท์ เสกสรร ชะนะ ธภัทร ชัยชูโชค |
| วันที่ตีพิมพ์/นำเสนอ | 15 ก.ย. 2566 |
| ปีที่ | No.3 (2023 |
| ฉบับที่ | Vol.44 |
| หมายเลขหน้า | 423-432 |
| ลักษณะบทความ | |
| Abstract | The use of sensors and electronic devices for remote monitoring has become increasingly prevalent across various sectors, including traffic management, forestry, military operations, commerce, and medical applications. These monitoring systems are designed to detect and respond to abnormalities in real-time, providing valuable insights and enhancing safety and security. However, in the realm of surveillance videos, the computational complexity of computer vision-based video processing systems has posed a significant challenge. To address this, recent research has developed a novel approach known as the Slow-Fast Convolutional Neural Network (SF-CNN). This innovative CNN architecture is designed to automatically learn from video frames and effectively classify anomalous behavior in surveillance videos. What sets SF-CNN apart is its adaptability, as it can learn at varying speeds depending on the frame rate, enabling it to capture both spatial and temporal information. By identifying humans, vehicles, and animals and handling normal and aberrant activities in different scenarios, SF-CNN offers a robust solution to address the limitations associated with detecting anomalous motions end-to-end. In testing on benchmark datasets, this proposed method achieved a remarkable accuracy rate of 99.6%, surpassing the performance of previous approaches. |