Automated tree-crown and height detection in a young forest plantation using mask region-based convolutional neural network (Mask R-CNN)

被引:133
作者
Hao, Zhenbang [1 ,2 ]
Lin, Lili [2 ,3 ,4 ]
Post, Christopher J. [3 ]
Mikhailova, Elena A. [3 ]
Li, Minghui [1 ,2 ]
Chen, Yan [1 ,2 ]
Yu, Kunyong [1 ,2 ]
Liu, Jian [1 ,2 ,4 ]
机构
[1] Fujian Agr & Forestry Univ, Coll Forestry, Fuzhou 350002, Fujian, Peoples R China
[2] Univ Key Lab Geomat Technol & Optimized Resourc, 15 Shangxiadian Rd, Fuzhou 350002, Fujian, Peoples R China
[3] Clemson Univ, Dept Forestry & Environm Conservat, Clemson, SC 29634 USA
[4] Fujian Agr & Forestry Univ, Coll Landscape Architecture, Fuzhou 350002, Fujian, Peoples R China
基金
中国国家自然科学基金;
关键词
Deep learning; Instance segmentation; Tree-crown delineation; Tree height; UAV imagery; Plantation forest; AERIAL VEHICLE UAV; SPECIES CLASSIFICATION; IMAGERY; DELINEATION; LIDAR;
D O I
10.1016/j.isprsjprs.2021.06.003
中图分类号
P9 [自然地理学];
学科分类号
0705 ; 070501 ;
摘要
Tree-crown and height are primary tree measurements in forest inventory. Convolutional neural networks (CNNs) are a class of neural networks, which can be used in forest inventory; however, no prior studies have developed a CNN model to detect tree crown and height simultaneously. This study is the first-of-its-kind that explored training a mask region-based convolutional neural network (Mask R-CNN) for automatically and concurrently detecting discontinuous tree crown and height of Chinese fir (Cunninghamia lanceolata (Lamb) Hook) in a plantation. A DJI Phantom4-Multispectral Unmanned Aerial Vehicle (UAV) was used to obtain high-resolution images of the study site, Shunchang County, China. Tree crown and height of Chinese fir was manually delineated and derived from this UAV imagery. A portion of the ground-truthed tree height values were used as a test set, and the remaining measurements were used as the model training data. Six different band combinations and derivations of the UAV imagery were used to detect tree crown and height, respectively (Multi band-DSM, RGB-DSM, NDVI-DSM, Multi band-CHM, RGB-CHM, and NDVI-CHM combination). The Mask R-CNN model with the NDVI-CHM combination achieved superior performance. The accuracy of Chinese fir's individual tree-crown detection was considerable (F1 score = 84.68%), the Intersection over Union (IoU) of tree crown delineation was 91.27%, and tree height estimates were highly correlated with the height from UAV imagery (R-2 = 0.97, RMSE = 0.11 m, rRMSE = 4.35%) and field measurement (R-2 = 0.87, RMSE = 0.24 m, rRMSE = 9.67%). Results demonstrate that the input image with an CHM achieves higher accuracy of tree crown delineation and tree height assessment compared to an image with a DSM. The accuracy and efficiency of Mask R-CNN has a great potential to assist the application of remote sensing in forests.
引用
收藏
页码:112 / 123
页数:12
相关论文
共 70 条
[1]   Mapping tall shrub biomass in Alaska at landscape scale using structure-from-motion photogrammetry and lidar [J].
Alonzo, Michael ;
Dial, Roman J. ;
Schulz, Bethany K. ;
Andersen, Hans-Erik ;
Lewis-Clark, Eric ;
Cook, Bruce D. ;
Morton, Douglas C. .
REMOTE SENSING OF ENVIRONMENT, 2020, 245
[2]   Tree Crown Delineation Algorithm Based on a Convolutional Neural Network [J].
Braga, Jose R. G. ;
Peripato, Vinicius ;
Dalagnol, Ricardo ;
Ferreira, Matheus P. ;
Tarabalka, Yuliya ;
Aragao, Luiz E. O. C. ;
de Campos Velho, Haroldo E. ;
Shiguemori, Elcio H. ;
Wagner, Fabien H. .
REMOTE SENSING, 2020, 12 (08)
[3]   An unexpectedly large count of trees in the West African Sahara and Sahel [J].
Brandt, Martin ;
Tucker, Compton J. ;
Kariryaa, Ankit ;
Rasmussen, Kjeld ;
Abel, Christin ;
Small, Jennifer ;
Chave, Jerome ;
Rasmussen, Laura Vang ;
Hiernaux, Pierre ;
Diouf, Abdoul Aziz ;
Kergoat, Laurent ;
Mertz, Ole ;
Igel, Christian ;
Gieseke, Fabian ;
Schoning, Johannes ;
Li, Sizhuo ;
Melocik, Katherine ;
Meyer, Jesse ;
Sinno, Scott ;
Romero, Eric ;
Glennie, Erin ;
Montagu, Amandine ;
Dendoncker, Morgane ;
Fensholt, Rasmus .
NATURE, 2020, 587 (7832) :78-+
[4]   Estimating Individual Conifer Seedling Height Using Drone-Based Image Point Clouds [J].
Castilla, Guillermo ;
Filiatrault, Michelle ;
McDermid, Gregory J. ;
Gartrell, Michael .
FORESTS, 2020, 11 (09)
[5]   Automatic Delineation and Height Measurement of Regenerating Conifer Crowns under Leaf-Off Conditions Using UAV Imagery [J].
Chadwick, Andrew J. ;
Goodbody, Tristan R. H. ;
Coops, Nicholas C. ;
Hervieux, Anne ;
Bater, Christopher W. ;
Martens, Lee A. ;
White, Barry ;
Roeser, Dominik .
REMOTE SENSING, 2020, 12 (24) :1-26
[6]   Remote Sensing Image Scene Classification: Benchmark and State of the Art [J].
Cheng, Gong ;
Han, Junwei ;
Lu, Xiaoqiang .
PROCEEDINGS OF THE IEEE, 2017, 105 (10) :1865-1883
[7]   Tree crown delineation and tree species classification in boreal forests using hyperspectral and ALS data [J].
Dalponte, Michele ;
Orka, Hans Ole ;
Ene, Liviu Theodor ;
Gobakken, Terje ;
Naesset, Erik .
REMOTE SENSING OF ENVIRONMENT, 2014, 140 :306-317
[8]   Early Detection of Invasive Exotic Trees Using UAV and Manned Aircraft Multispectral and LiDAR Data [J].
Dash, Jonathan P. ;
Watt, Michael S. ;
Paul, Thomas S. H. ;
Morgenroth, Justin ;
Pearse, Grant D. .
REMOTE SENSING, 2019, 11 (15)
[9]   Multi-scale object detection in remote sensing imagery with convolutional neural networks [J].
Deng, Zhipeng ;
Sun, Hao ;
Zhou, Shilin ;
Zhao, Juanping ;
Lei, Lin ;
Zou, Huanxin .
ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING, 2018, 145 :3-22
[10]   A light and faster regional convolutional neural network for object detection in optical remote sensing images [J].
Ding, Peng ;
Zhang, Ye ;
Deng, Wei-Jian ;
Jia, Ping ;
Kuijper, Arjan .
ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING, 2018, 141 :208-218