講演詳細

経営部門[Forest Management]

日付	2024年3月10日
開始時刻	ポスター発表
会場名	531
講演番号	PD-65
発表題目	Forest/non-forest mapping with StriX X-band SAR images based on semantic segmentation Forest/non-forest mapping with StriX X-band SAR images based on semantic segmentation
所属	株式会社Synspective
要旨本文	Satellite remote sensing promises to revolutionize forest monitoring by reducing the required manpower. Synthetic Aperture Radar (SAR) technology is increasingly favored over optical sensors due to its ability to provide observations during nighttime or cloudy conditions. X-band satellites like StriX, advanced through miniaturization and constellation development, are particularly promising for high-frequency observations. However, their shorter wavelength compared to L or C bands hinders forest penetration, making forest/non-forest distinction challenging. Here, we addressed this issue by using a deep learning-based segmentation model for SAR images, which takes backscatter intensities and incidence angles as inputs. With extensive training on StriX images in multiple locations, our model achieves an average IoU score exceeding 90% across various sites. Our approach represents an advancement towards frequent forest monitoring, contributing to global forest conservation efforts.
著者氏名	○Takushi Uda ・ Clement Barras ・ Asuka Wachi
著者所属	Synspective Inc.
キーワード	SAR, Deep learning, Segmentation, Forestry, Geophysical image processing
Key word	SAR, Deep learning, Segmentation, Forestry, Geophysical image processing