講演詳細

経営部門[Forest Management]

日付	2024年3月8日
開始時刻	ポスター発表
会場名	531
講演番号	PD-20（学生ポスター賞審査対象）
発表題目	Detecting high-value hardwood trees using deep learning algorithm with unmanned aerial vehicle (UAV) imagery Detecting high-value hardwood trees using deep learning algorithm with unmanned aerial vehicle (UAV) imagery
所属	The University of Tokyo
要旨本文	This study explored the applicability of UAV imagery and a deep learning algorithm to detect high-value deciduous hardwood tree crowns of Japanese oak (Quercus crispula) in an uneven-aged mixed forest in northern Japan. UAV images were collected in September and October 2022 before and after changing leaf color of Japanese oak, to determine the appropriate timing of UAV image collection. We analyzed RGB information of UAV images through a ResU-Net (U-Net with ResNet101 backbone) deep learning algorithm. Our results showed both datasets provided reliable F1 scores above 0.80. While September dataset was better at describing Japanese oak tree crowns, October dataset was better at distinguishing between Japanese oak and other hardwood species. Moreover, our case study highlights a potential methodology to offer transferable solutions for the resource assessment of high-value timber species in other regions.
著者氏名	○Nyo Me Htun1 ・ Toshiaki Owari2 ・ Satoshi Tsuyuki1 ・ Takuya Hiroshima1
著者所属	1The University of Tokyo ・ 2The University of Tokyo
キーワード	high-value timber species, uneven-aged mixed forest, ResU-Net algorithm, UAV imagery
Key word	high-value timber species, uneven-aged mixed forest, ResU-Net algorithm, UAV imagery