AIDNet: Detecting Insulators and Defects from Satellite Remote Sensing Images: An Exploration
- Resource Type
- Conference
- Authors
- Zhou, Fangrong; Ma, Yi; Wen, Gang; Ma, Yutang; Lan, Zhicai; Zhang, Zhengde
- Source
- 2022 8th Annual International Conference on Network and Information Systems for Computers (ICNISC) ICNISC Network and Information Systems for Computers (ICNISC), 2022 8th Annual International Conference on. :347-353 Sep, 2022
- Subject
- Computing and Processing
Computers
Satellites
Adaptive systems
Filtering
Object detection
Insulators
Task analysis
deep learning
neural networks
insulators
defects
satellite remote sensing images
detection
mixed pixels
- Language
The insulator has a diameter of about 200 mm and occupies only 0.4 pixels in the best satellite remote sensing image (SRSI). Detecting insulators and self-explosion defects from SRSI is a challenging task due to the extremely low spatial resolution and mixed pixels issues. Algorithm design that improves the required spatial resolution (rSR) is one way to advance this task. This paper proposes a two-stage detection method combining an adaptive network and a detection network, using multiple filtering and adaptive enhancement methods to achieve end-to-end detection, which can effectively improve the rSR. The idea proposed in this paper can promote the realization of goals that detect insulators and defects from SRSI. The proposed algorithm can realize the detection at relatively poor spatial resolution, and can be applied to small targets detection in general object detection tasks.