Improved SSD-based transmission tower detection in SAR images

• Main Authors: Yuan Gao, Wei Yang, Chunsheng Li, Fei Zou
• Format: Article
• Language: English
• Published: Wiley 2019-09-01
• Series: The Journal of Engineering
• Subjects: radar imaging; learning (artificial intelligence); synthetic aperture radar; object detection; poles and towers; radar computing; convolutional neural nets; maintenance engineering; convolutional neural network; shallow feature; transmission system maintenance; object detection method; transmission system monitoring; sar transmission tower detection; sar images; deep learning-based technique; improved ssd-based transmission tower detection method
• Online Access: https://digital-library.theiet.org/content/journals/10.1049/joe.2019.0568

SAR transmission tower detection is essential to transmission system monitoring and maintenance. Various methods have been proposed to detect transmission tower by using shallow feature. With the development of the deep learning based on convolutional neural network, this study proposed an improved SSD-based transmission tower detection method by introducing deep learning-based technique. First, SSD, a state-of-the-art object detection method, is analysed in detail. Then, the algorithm is modified to fit authors’ specific domain. According to the experiment results in real SAR images, it is showed that the modified method gains better detection performance. Finally, ablation experiments are also performed to confirm each modification step produces better result.