Document of bibliographic reference 344537

BibliographicReference record

Type

Bibliographic resource

Type of document

Book chapters

BibLvlCode

AMS

Title

Patch-wise semantic segmentation of sedimentation from high-resolution satellite images using deep learning

Abstract

n recent times, satellite data availability has increased significantly, helping researchers worldwide to explore, analyze and approach different problems using the most recent techniques. The segmentation of sediment load in coastal areas using satellite imagery can be considered as a cost-efficient process as sediment load analysis can be costly and time-consuming if done hands on. In this work, we created dataset of Bangladesh marine area for segmenting sediment load and showed the applicability of deep learning technique to segment sedimentation into 5 different classes (Land, Hight Sediment, Moderate Sediment, Low Sediment and No Sediment) using deep neural network called U-Net. As our collected satellite image is enormous, we showed how patch-wise learning technique can be an effective solution in the context of batch-wise training. Highest dice coefficient of 86% and validation dice coefficient of 87% has been acquired for Dec-2019 time frame data. The highest 77% of pixel accuracy and 78% of validation pixel accuracy was achieved on the same time frame data.

WebOfScience code

https://www.webofscience.com/wos/woscc/full-record/WOS:000696173400041

Bibliographic citation

Pranto, T.H.; Noman, A.A.; Noor, A.; Deepty, U.H.; Rahman, R.M. (2021). Patch-wise semantic segmentation of sedimentation from high-resolution satellite images using deep learning, in: Rojas, I. et al. Advances in computational intelligence: 16th International Work-Conference on Artificial Neural Networks, IWANN 2021, Virtual Event, June 16–18, 2021, Proceedings, Part I. Lecture Notes in Computer Science, 12861: pp. 498-509. https://dx.doi.org/10.1007/978-3-030-85030-2_41

Is peer reviewed

true