摘要
[目的]使用窄波段植被指数、纹理信息等特征对影像进行分类,探究植被指数和纹理信息对于云阴影下树种分类的潜力。[方法]使用经过大气校正后的高光谱影像进行窄波段植被指数的计算、纹理分析以及主成分分析,并对计算的结果进行波段组合。用于计算纹理信息的波段通过最佳指示因子进行选择,选取的波段数为31(0.67 nm),51(0.86 nm),55(0.89nm) 3个波段。结合高分辨率的航空相片进行训练样本的选择,采用Support Vector Machine(SVM)方法对经过大气校正后的反射率影像和重组后的特征影像分别进行分类,使用样地实测的树种信息对分类结果进行验证,使用总体精度和Kappa系数作为分类精度的评价指标。[结果]相对于直接使用反射率影像进行分类,使用窄波段植被指数以及纹理信息可以显著地提高云阴影下地物的分类精度,其分类精度和Kappa系数分别为90.4%和0.88,比直接使用反射率影像的分类精度和Kappa系数分别提高了18%和0.2。[结论]使用重新组合后的影像进行树种分类比直接使用反射率影像进行分类,其分类精度更高,说明窄波段植被指数与纹理特征可以提高云阴影区树种分类的精度。使用波段重组后的影像对云阴影下地物分类,其对于单个地物的分类精度也有明显的提高。
[Objective] Using narrow-band vegetation index and texture information to classify images, and to explore the potential of vegetation index and texture information for tree species classification under cloud shadows.[Method] Vegetation indices and texture features were used to recombine a new image. Reflectance images and recombination images were classified by support vector machine classifier. By comparing the classification results, the potential of narrow band vegetation indices, texture information and other characteristics in the classification of forest in cloud shadow was explored. The band used to calculate the texture information was selected by the optimum index factor(OIF), and the number of bands selected were band 31(0.67 μm), 51(0.86 μm) and 55(0.89 μm). Tree species training samples were selected based on high resolution aerial photographs. The Support Vector Machine(SVM) method was adopted to classify the reflectivity images and the feature images after recombination. The classification results were verified by filed data, the overall accuracy and Kappa coefficient were used as the evaluation indices for classification accuracy.[Result] Compared with the classification result of reflectance image, the combination of vegetation index and texture information significantly improved the classification accuracy. The overall accuracy and Kappa coefficient were 90.4% and 0.88, which increased by 18% and 0.2 respectively. The classification accuracy of individual tree species was also significantly improved. It can be seen from the confusion matrix that when using the reflectance image for classification, the Pinus koraiensis is misclassified as P. sylvestris. However, using vegetation index, the error was significantly reduced.[Conclusion] It is concluded that the forest in cloud shadow area can be classified based on the narrow band vegetation index(NDVI 705, mSR 705, mNDVI 705, VOG1, VOG2, REP) and texture information and the classification result is better than the reflectance image does.
作者
李军玲
庞勇
李增元
荚文
LI Jun-ling;PANG Yong;LI Zeng-yuan;JIA Wen(Research Institute of Forest Resources Information Techniques, Chinese Academy of Forestry, Beijing 1000917 China)
出处
《林业科学研究》
CSCD
北大核心
2019年第5期136-141,共6页
Forest Research
基金
中央级公益性科研院所基本科研业务费专项“黑龙江省多尺度森林生物量和碳储量估计(CAFYBB2016ZD004)”
国家重点研发计划“多尺度落叶松人工林生长预测(2017YFD0600404)”