Remote Sensing Technology and Application

Research on UAV Hyperspectral of Tree Species Classification based on Machine Learning Algorithms and Spatial Resolution Adjustment

Xiangshan ZHOU, Wunian YANG, Ke LUO, Hongyi PIAO, Tao ZHOU, Jie ZHOU, Xiaolu TANG

Remote Sensing Technology and Application, 2024, 39(4): 880-896. DOI: 10.11873/j.issn.1004-0323.2024.4.0880

分类方法	空间分辨率	平均整体精度	平均Kappa
支持向量机	0.12 m	0.55	0.50
	0.5 m	0.66	0.62
	1 m	0.71	0.68
	1.5 m	0.73	0.70
	2 m	0.73	0.70
	2.5 m	0.74	0.71
	3 m	0.75	0.72
	3.5 m	0.73	0.71
	4 m	0.71	0.68
随机森林	0.12 m	0.54	0.49
	0.5 m	0.63	0.58
	1 m	0.67	0.63
	1.5 m	0.68	0.65
	2 m	0.70	0.66
	2.5 m	0.72	0.69
	3 m	0.73	0.70
	3.5 m	0.72	0.68
	4 m	0.70	0.66

Table 8 Average classification results of 10， 15， and 20 tree species at different scales

Other figure/table from this article

Fig. 1 Overview of Chengdu botanical garden and distribution of sample points
Table 1 List of 20 tree types sampl in the study area
Table 2 Original band information of hyperspectral image
Fig. 2 Technical route of UAV hyperspectral tree species classification method based on machine learning algorithm
Table 3 Calculation of hyperspectral image vegetation index
Fig.3 Removal of spectral envelope lines for 20 tree species
Fig.4 First derivative of spectra for 20 tree species
Table 4 Classification results of 20 tree species under original spatial resolution
Table 5 Impact of different spatial resolutions on the classification results of 10 tree species
Table 6 Impact of different spatial resolutions on the classification results of 15 tree species
Table 7 Impact of different spatial resolutions on the classification results of 20 tree species
Fig.5 Classification accuracy trends of random forests and support vector machine algorithms based on different spatial resolutions of images for different tree species in the research area