DEVELOPMENT OF TRESPASSING DETECTION SYSTEM ON SMART VIDEO SURVEILLANCE SYSTEM BASED ON EDGE COMPUTING FOR LEVEL CROSSING

• Main Author: Putra Pratama, Rian
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/70731
• Institution: Institut Teknologi Bandung

A city needs innovative, efficient and integrated solutions. Safe and secure aspects are part of the development of a smart city. One of the problems that has a high potential for violations and accidents in a city is level crossings. The surveillance system applied at level crossings is currently still conventional. The smart video surveillance system is part of the smart city development as a solution. The edge computing approach is required to be implemented on systems that require real- time processing characteristics. Aspects of computing accuracy and performance are a challenge in implementing smart video surveillance systems on devices with limited computing to achieve ideal conditions, so optimization is required to develop effective and efficient solutions. This research will develop a prototype of a smart video surveillance system by implementing offloading computing on limited computing resource devices to improve system computing performance in identifying and collecting data on violations at level crossings as a form of implementing surveillance of potential accident locations at level crossings. The results of tests carried out on system development resulted in an average increase in the speed of the computing process of about 1.5 times faster with an accuracy value of 89.4%. Meanwhile, the GPU temperature computational performance results decreased by around 5.50 °C, decreased by 44.05% in the GPU utilization process and decreased by around 301 Mb in the memory utilization process and 2.28 watts in power consumption.