Tree Species Identification Based on FCN Using the Visible Images Obtained from an Unmanned Aerial Vehicle

Pengqin Dai; Lixia Ding; Lijuan Liu; Luofan Dong; Yiting Huang

doi:10.3788/LOP57.101001

Laser & Optoelectronics Progress, Volume. 57, Issue 10, 101001(2020)

Tree Species Identification Based on FCN Using the Visible Images Obtained from an Unmanned Aerial Vehicle

Pengqin Dai^1,2,3、*, Lixia Ding^1,2,3、**, Lijuan Liu^1,2,3, Luofan Dong^1,2,3, and Yiting Huang³

Author Affiliations

¹State Key Laboratory of Subtropical Silviculture, Hangzhou, Zhejiang 311300, China

²Key Laboratory of Carbon Cycling in Forest Ecosystems and Carbon Sequestration of Zhejiang Province, Hangzhou, Zhejiang 311300, China

³School of Environmental and Resources Science, Zhejiang A & F University, Hangzhou, Zhejiang 311300, China;

show less

Abstract Get PDF(in Chinese)

Figures & Tables(12)

Fig. 1. Res-U-Net model

Download full size

Fig. 2. Flow chart of proposed method

Download full size

Fig. 3. Global average score for each segmentation scale

Download full size

Fig. 4. Vegetation image and object-oriented segmentation results in the study area

Download full size

Fig. 5. Label making. (a) Label map; (b) rotation of the sample and corresponding label

Download full size

Fig. 6. Relationship between n_tree and OOB error

Download full size

Fig. 7. Average drop accuracy rate of each characteristic variable

Download full size

Fig. 8. Loss and accuracy change during training. (a) Loss; (b) accuracy

Download full size

Fig. 9. Classification results by different methods. (a) Groundtruth map; (b) FCN method with fused data including original RGB data, VDVI data and ExG-ExR data; (c) FCN method with original RGB data; (d) RF method with 33 feature variables

Download full size

Fig. 10. Correction of FCN classification results by object-oriented segmentation. (a) Before object-oriented amendments; (b) after object-oriented amendments

Download full size

Table 1. Accuracy evaluation of classification of UAV images

View table

Table 1. Accuracy evaluation of classification of UAV images

Vegetation	FCN method withfeature variables		FCN method withoutfeature variable		RF method with33 feature variables
Vegetation	Producer accuracy	User accuracy	Producer accuracy	User accuracy	Producer accuracy	User accuracy
Bamboo	0.996	0.994	0.986	0.987	0.707	0.992
Camphor tree	0.993	0.997	0.991	0.980	0.990	0.999
Fern	0.961	0.943	0.919	0.839	0.911	0.916
Fir	0.966	0.950	0.896	0.945	0.983	0.705
Masson pine	0.920	0.944	0.792	0.791	0.725	0.885
Overall accuracy	0.978		0.954		0.890
Kappa	0.970		0.937		0.850

Table 2. Correction of FCN classification results by object-oriented segmentation
View table
Table 2. Correction of FCN classification results by object-oriented segmentation
Vegetation Bamboo Camphortree Fern Fir Massonpine
Produceraccuracy 1.000 0.997 0.989 0.977 0864
Useraccuracy 0.996 0.999 0.965 0.972 0.964
Overallaccuracy 0.987
Kappa 0.983

Tools

Get Citation

Copy Citation Text

Pengqin Dai, Lixia Ding, Lijuan Liu, Luofan Dong, Yiting Huang. Tree Species Identification Based on FCN Using the Visible Images Obtained from an Unmanned Aerial Vehicle[J]. Laser & Optoelectronics Progress, 2020, 57(10): 101001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites