基于实测冠层光谱数据的三明市13种树种识别研究
Tree Species Discrimination in Subtropical Area:Based on Situ Hyperspectral Data at Canopy Level
摘要
针对亚热带地区树种丰富,树种间相似度大对树种识别带来的问题,本研究以福建省三明市莘口镇格氏栲自然保护区附近13种常见树种的冠层实测高光谱数据为例,采用一种分层分析方法,探讨不同光谱类别的树种识别精度以及树种识别的最佳波段。首先,对原始光谱进行变换处理,包括一阶微分、二阶微分、对数一阶微分、包络线去除和植被指数;其次,通过分析选择出13种树种各光谱类别的差异显著波段;最后,利用逐步判别法对选择的差异显著波段进一步降维,判断不同光谱类别的树种识别精度并找出识别13种树种的最佳波段。结果表明:光谱变换能有效地提高树种的识别精度,尤其是对数一阶微分光谱,总识别精度高达98.7%;对于原始光谱,近红外波段(760~1300 nm)的树种识别能力更强,对于变换光谱,可见光波段(350~760 nm)的树种识别效果更佳;不同光谱类别之间具体的显著性差异波段存在很大差别,原始光谱与变换光谱之间仅在绿光波段(500~600 nm)有少量相同的显著性差异波段,此研究成果可为亚热带地区树种识别提供参考。
Taking situ hyperspectral data of 13 tree species near C. kawakamii Nature Reserve in Sanyuan District, Sanming City, Fujian Province as study area, a hierarchical method was used to ex-plore species recognition accuracies of different spectral classes and the optimal bands for tree species discrimination. First, five types of transformation were applied to original spectral data, including first derivative, second derivative, logarithmic first derivative, continuum removal and vegetable indices. Then, the curves of original spectrum and transform spectrum of the species were analyzed and com-pared to select out the bands with bigger differences. Finally, stepwise discriminant analysis method was used to identify the species recognition accuracy and the most sensitive bands for 13 tree species discrimination. Results showed that spectrum transform could improve the recognition accuracies effec-tively, especially the overall recognition accuracy of logarithmic first derivative reflectance, up to 98. 7%. In terms of tree species recognition capability, the near-infrared (760 ~1 300 nm) bands are more powerful for original spectrum, while bands in visible region (350~760 nm) are more powerful for transform spectrum. The significantly different bands of different spectral classes vary greatly. Only in the green (500~600 nm) bands did original spectral data and transform spectral data have a little in common in the significantly different bands.
Author:
JIANG Chun-mei
CHEN Wen-hui
HUANG Chuan-yin
作者单位:
福建师范大学 地理科学学院,福州,350007
福建师范大学 地理科学学院,福州350007
黎明职业大学 信息与电子工程学院,福建 泉州362000
年,卷(期):
2016, 11(2)
分类号:
S771.8
机标分类号:
S66
S127
TP79
在线出版日期:
2016年11月3日
基金项目:
2012CB722203:国家重点基础研究发展计划(973计划),