Leaf area measurement of green wall plants using deep learning with terrestrial laser

• Published in: Journal of the Japanese Society of Revegetation Technology Vol. 48; no. 1; pp. 9 - 14
• Main Authors: ASANO, Ryota, HIKOSAKA, Shoko, UEYANAGI, Ryohei, KURIKI, Shigeru, YAMAGUCHI, Jun, OSHIMA, Kaori, KATO, Akira
• Format: Journal Article
• Language: Japanese
• Published: Tokyo JAPANESE SOCIETY OF REVEGETATION TECHNOLOGY 31.08.2022; Japan Science and Technology Agency
• Subjects: carbon neutral; Data acquisition; Deep learning; energy efficient building; Green infrastructure; green wall; Heat islands; Image acquisition; Leaf area; Leaves; Plant structures; Plants; terrestrial laser; Urban areas; Urban heat islands
• ISSN: 0916-7439; 0916-7439
• DOI: 10.7211/jjsrt.48.9

Green wall plants have various functionality such as mitigating heat island, creating green space, and absorbing carbon in an urban area. A method to evaluate the various functionality is needed for construction field. This study focuses on the plant structure and deep learning was introduced to estimate the leaf area and counting leaves automatically. To achieve this, the panorama distant image was created from 3D data acquired by terrestrial laser to prepare for the deep learning model. The model for the presence or absence of leaves had 90% accuracy and the counting leaves had 72% accuracy and leaf area estimated through deep learning had 27% error compared to the destructively sampled data.