Development of Deep Learning Based Technique for Iceberg Detection with 6SD of Polarametric SAR Data

• Main Authors: Singh, Vatsala, Singh, Gulab, Maurya, Ajay
• Format: Conference Proceeding
• Language: English
• Subjects: Artificial Neural Network; Convolutional neural networks; Deep learning; iceberg; Model-based Decomposition; Object recognition; PolSAR; Radar polarimetry; Scattering; Shape; Temperature distribution
• ISSN: 2153-7003
• DOI: 10.1109/IGARSS46834.2022.9884585

Icebergs have been a major concern to the environmentalists, researchers and maritime workers since decades. Especially with the temperatures rising globally the rate of calving of icebergs has increased and thus increasing their probability of them drifting into the major ship lanes posing various threats to people all across the world. Being an open hazard to the ocean, monitoring the iceberg behaviour is critical to ensure the safety of maritime activities. Synthetic Aperture Radar (SAR) images prove to be of major help in studying these icebergs since they strongly influence the SAR backscattering. However due to similarities in scattering behaviour of icebergs and background clutter because of their irregular shapes and sizes, it becomes challenging to accurately classify/identify them. Although the current state of the art techniques like decompositions, model-based scattering power decomposition and eigenvalue/eigenvector decomposition are quite helpful but they come with their own set of limitations. Therefore, the objective of this paper is to explore the application of Deep learning on PolSAR data with Six-component scattering matric power decomposition for efficient identification and classification of the icebergs.