Intruder Detection by Using Faster R-CNN in Power Substation

Srijakkot, Krit; Kanjanasurat, Isoon; Wiriyakrieng, Nuttakan; Lartwatechakul, Mayulee; Benjangkaprasert, Chawalit

doi:10.1007/978-3-030-44044-2_16

Intruder Detection by Using Faster R-CNN in Power Substation

Author(s)

Srijakkot, Krit

Kanjanasurat, Isoon

Wiriyakrieng, Nuttakan

Lartwatechakul, Mayulee

Benjangkaprasert, Chawalit

Date Issued

January 1, 2020

Type

Conference Paper

DOI

10.1007/978-3-030-44044-2_16

Abstract

This paper presents the intruder detection by using the Faster R-CNN model and administrator system for the power substation in Khon Kaen substation 4 of the Electricity Generating Authority of Thailand (EGAT). There are two processes of intruder detection-detecting the intruder and sending a notification to the system administrator of EGAT through Line application. The Faster R-CNN model of intruder detection was trained and tested by using the Open Image Dataset and our dataset. We collected our dataset of 1,500 images from a different condition from the real environment. There are two conditions, including distance and light intensity. Our system used a high-performance computer by using GPU: Nvidia Titan RTX 24 GB to support the object detection system from using five cameras at the same time. The performance of intruder detection achieved by greater than 95%.

Citation

Advances in Intelligent Systems and Computing, 1149 AISC, 159-167, 2020

Subjects

Faster R-CNN

Intruder detection

Object detection