A New Method for Forest Canopy Hemispherical Photography Segmentation based on Deep Learning
<i>Research Highlights:</i> This paper proposes a new method for hemispherical forest canopy image segmentation. The method is based on a deep learning methodology and provides a robust and fully automatic technique for the segmentation of forest canopy hemispherical photography (CHP) an...
Main Authors: | , , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2020-12-01
|
Series: | Forests |
Subjects: | |
Online Access: | https://www.mdpi.com/1999-4907/11/12/1366 |
id |
doaj-fc44760628d14d7bb9289caf4a92016e |
---|---|
record_format |
Article |
spelling |
doaj-fc44760628d14d7bb9289caf4a92016e2020-12-20T00:02:34ZengMDPI AGForests1999-49072020-12-01111366136610.3390/f11121366A New Method for Forest Canopy Hemispherical Photography Segmentation based on Deep LearningKexin Li0Xinwang Huang1Jingzhe Zhang2Zhihu Sun3Jianping Huang4Chunxue Sun5Qiancheng Xie6Wenlong Song7School of Mechanical and Electrical Engineering, Northeast Forestry University, Harbin 150040, ChinaSchool of Mechanical and Electrical Engineering, Northeast Forestry University, Harbin 150040, ChinaDepartment of Nursing, Heilongjiang Vocational College of Winter Sports, Harbin 150040, ChinaSchool of Forestry, Northeast Forestry University, Harbin 150040, ChinaSchool of Mechanical and Electrical Engineering, Northeast Forestry University, Harbin 150040, ChinaSchool of Mechanical and Electrical Engineering, Northeast Forestry University, Harbin 150040, ChinaSchool of Mechanical and Electrical Engineering, Northeast Forestry University, Harbin 150040, ChinaSchool of Mechanical and Electrical Engineering, Northeast Forestry University, Harbin 150040, China<i>Research Highlights:</i> This paper proposes a new method for hemispherical forest canopy image segmentation. The method is based on a deep learning methodology and provides a robust and fully automatic technique for the segmentation of forest canopy hemispherical photography (CHP) and gap fraction (GF) calculation. <i>Background and Objectives</i>: CHP is widely used to estimate structural forest variables. The GF is the most important parameter for calculating the leaf area index (LAI), and its calculation requires the binary segmentation result of the CHP. <i>Materials and Methods:</i> Our method consists of three modules, namely, northing correction, valid region extraction, and hemispherical image segmentation. In these steps, a core procedure is hemispherical canopy image segmentation based on the U-Net convolutional neural network. Our method is compared with traditional threshold methods (e.g., the Otsu and Ridler methods), a fuzzy clustering method (FCM), commercial professional software (WinSCANOPY), and the Habitat-Net network method. <i>Results:</i> The experimental results show that the method presented here achieves a Dice similarity coefficient (DSC) of 89.20% and an accuracy of 98.73%. <i>Conclusions:</i> The method presented here outperforms the Habitat-Net and WinSCANOPY methods, along with the FCM, and it is significantly better than the Otsu and Ridler threshold methods. The method takes the original canopy hemisphere image first and then automatically executes the three modules in sequence, and finally outputs the binary segmentation map. The method presented here is a pipelined, end-to-end method.https://www.mdpi.com/1999-4907/11/12/1366canopy hemisphere photographyimage segmentationgap fractiondeep learningU-Net |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Kexin Li Xinwang Huang Jingzhe Zhang Zhihu Sun Jianping Huang Chunxue Sun Qiancheng Xie Wenlong Song |
spellingShingle |
Kexin Li Xinwang Huang Jingzhe Zhang Zhihu Sun Jianping Huang Chunxue Sun Qiancheng Xie Wenlong Song A New Method for Forest Canopy Hemispherical Photography Segmentation based on Deep Learning Forests canopy hemisphere photography image segmentation gap fraction deep learning U-Net |
author_facet |
Kexin Li Xinwang Huang Jingzhe Zhang Zhihu Sun Jianping Huang Chunxue Sun Qiancheng Xie Wenlong Song |
author_sort |
Kexin Li |
title |
A New Method for Forest Canopy Hemispherical Photography Segmentation based on Deep Learning |
title_short |
A New Method for Forest Canopy Hemispherical Photography Segmentation based on Deep Learning |
title_full |
A New Method for Forest Canopy Hemispherical Photography Segmentation based on Deep Learning |
title_fullStr |
A New Method for Forest Canopy Hemispherical Photography Segmentation based on Deep Learning |
title_full_unstemmed |
A New Method for Forest Canopy Hemispherical Photography Segmentation based on Deep Learning |
title_sort |
new method for forest canopy hemispherical photography segmentation based on deep learning |
publisher |
MDPI AG |
series |
Forests |
issn |
1999-4907 |
publishDate |
2020-12-01 |
description |
<i>Research Highlights:</i> This paper proposes a new method for hemispherical forest canopy image segmentation. The method is based on a deep learning methodology and provides a robust and fully automatic technique for the segmentation of forest canopy hemispherical photography (CHP) and gap fraction (GF) calculation. <i>Background and Objectives</i>: CHP is widely used to estimate structural forest variables. The GF is the most important parameter for calculating the leaf area index (LAI), and its calculation requires the binary segmentation result of the CHP. <i>Materials and Methods:</i> Our method consists of three modules, namely, northing correction, valid region extraction, and hemispherical image segmentation. In these steps, a core procedure is hemispherical canopy image segmentation based on the U-Net convolutional neural network. Our method is compared with traditional threshold methods (e.g., the Otsu and Ridler methods), a fuzzy clustering method (FCM), commercial professional software (WinSCANOPY), and the Habitat-Net network method. <i>Results:</i> The experimental results show that the method presented here achieves a Dice similarity coefficient (DSC) of 89.20% and an accuracy of 98.73%. <i>Conclusions:</i> The method presented here outperforms the Habitat-Net and WinSCANOPY methods, along with the FCM, and it is significantly better than the Otsu and Ridler threshold methods. The method takes the original canopy hemisphere image first and then automatically executes the three modules in sequence, and finally outputs the binary segmentation map. The method presented here is a pipelined, end-to-end method. |
topic |
canopy hemisphere photography image segmentation gap fraction deep learning U-Net |
url |
https://www.mdpi.com/1999-4907/11/12/1366 |
work_keys_str_mv |
AT kexinli anewmethodforforestcanopyhemisphericalphotographysegmentationbasedondeeplearning AT xinwanghuang anewmethodforforestcanopyhemisphericalphotographysegmentationbasedondeeplearning AT jingzhezhang anewmethodforforestcanopyhemisphericalphotographysegmentationbasedondeeplearning AT zhihusun anewmethodforforestcanopyhemisphericalphotographysegmentationbasedondeeplearning AT jianpinghuang anewmethodforforestcanopyhemisphericalphotographysegmentationbasedondeeplearning AT chunxuesun anewmethodforforestcanopyhemisphericalphotographysegmentationbasedondeeplearning AT qianchengxie anewmethodforforestcanopyhemisphericalphotographysegmentationbasedondeeplearning AT wenlongsong anewmethodforforestcanopyhemisphericalphotographysegmentationbasedondeeplearning AT kexinli newmethodforforestcanopyhemisphericalphotographysegmentationbasedondeeplearning AT xinwanghuang newmethodforforestcanopyhemisphericalphotographysegmentationbasedondeeplearning AT jingzhezhang newmethodforforestcanopyhemisphericalphotographysegmentationbasedondeeplearning AT zhihusun newmethodforforestcanopyhemisphericalphotographysegmentationbasedondeeplearning AT jianpinghuang newmethodforforestcanopyhemisphericalphotographysegmentationbasedondeeplearning AT chunxuesun newmethodforforestcanopyhemisphericalphotographysegmentationbasedondeeplearning AT qianchengxie newmethodforforestcanopyhemisphericalphotographysegmentationbasedondeeplearning AT wenlongsong newmethodforforestcanopyhemisphericalphotographysegmentationbasedondeeplearning |
_version_ |
1724377351032143872 |