TY - GEN
T1 - Power Line Detection Using Unmanned Aerial Vehicle with Spherical Shell
AU - Sumagayan, Moheddin
AU - Mangorsi, Rohanni
AU - Aleluya, Earl Ryan
AU - Salaan, Carl John
AU - Premachandra, Chinthaka
N1 - Funding Information:
ACKNOWLEDGMENT This work thanks the Department of Science and Technology - Engineering Research and Development for Technology.
Publisher Copyright:
© 2022 IEEE.
PY - 2022
Y1 - 2022
N2 - Power lines are one of the components in the electrical grid that needs persistent monitoring. Detecting the power lines can help UAV-based computer vision systems identify faults within the line, such as the damaged strand in the cable. Aside from fault detection, detecting the power lines also aids the UAV in its autonomous navigation and planning. However, with the protection of the shell to UAV, its presence in the images has become a challenge. In this paper, the authors investigate the feasibility of a deep neural network on an image dataset with occlusion. The paper enhances the Point Instance network to address the challenges of occlusion. The network is trained using synthetic images and tested on the image dataset obtained from the UAV prototype. The experimental results showed no significant difference in the network performance, even with or without the occlusion. The future work of this study involves further data acquisition of images using the prototype, especially at varying inspection sites.
AB - Power lines are one of the components in the electrical grid that needs persistent monitoring. Detecting the power lines can help UAV-based computer vision systems identify faults within the line, such as the damaged strand in the cable. Aside from fault detection, detecting the power lines also aids the UAV in its autonomous navigation and planning. However, with the protection of the shell to UAV, its presence in the images has become a challenge. In this paper, the authors investigate the feasibility of a deep neural network on an image dataset with occlusion. The paper enhances the Point Instance network to address the challenges of occlusion. The network is trained using synthetic images and tested on the image dataset obtained from the UAV prototype. The experimental results showed no significant difference in the network performance, even with or without the occlusion. The future work of this study involves further data acquisition of images using the prototype, especially at varying inspection sites.
KW - deep learning
KW - machine vision
KW - power distribution lines
KW - unmanned aerial vehicles
UR - http://www.scopus.com/inward/record.url?scp=85128918958&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85128918958&partnerID=8YFLogxK
U2 - 10.1109/ICARC54489.2022.9753854
DO - 10.1109/ICARC54489.2022.9753854
M3 - Conference contribution
AN - SCOPUS:85128918958
T3 - ICARC 2022 - 2nd International Conference on Advanced Research in Computing: Towards a Digitally Empowered Society
SP - 160
EP - 164
BT - ICARC 2022 - 2nd International Conference on Advanced Research in Computing
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2nd International Conference on Advanced Research in Computing, ICARC 2022
Y2 - 23 February 2022 through 24 February 2022
ER -