摘要
利用植被物候特征进行植被信息提取是植被遥感分类的重要研究内容。连云港云台山林场的樱桃花期与其他植被类型的花期存在明显差异,该差异使得樱桃花在盛开和凋落过程中,其影像光谱特征显著且独特。尤其是樱桃花盛开时,其光谱特征的变化为利用遥感技术进行植被面积提取提供了较好的技术手段。选择2020年3月樱桃花期无人机RGB高分辨率影像为数据源,利用最邻近法(nearest neighbour method,NNM)、随机森林法(random forest,RF)、支持向量机法(support vector machine,SVM)对云台山林场樱桃种植面积进行分类提取。针对樱桃树特定植被类型,探索了其在开花时节的遥感影像提取方法。结果表明:结合樱桃树遥感影像的特征,采用不同分类方法显著提升了樱桃树的分类精度,最高提取精度达到98%。对分类提取后的影像进行精度评价分析:随机森林法在添加了植被指数、光谱特征和纹理特征后樱桃花提取质量最好,其次是最邻近法添加植被指数、光谱特征和纹理特征。该研究不仅填补了樱桃树物候特征提取的研究空白,也为其他具有独特物候特征的植被类型提供了可借鉴的分类方法。
Utilizing vegetation phenological features for vegetation information extraction is an important research content in vegetation remote sensing classification.The cherry blossom season in Yuntai Mountain Forest Farm exhibits significant differences from other vegetation types,enabling the spectral features of cherry blossoms to be distinct and unique during their blooming and fading stages.Especially during the peak of cherry blossoms,the changes in their spectral characteristics provide a favorable technical means for extracting vegetation areas using remote sensing techniques.This study selects high-resolution RGB drone images of the cherry blossom season in March 2020 as the data source and utilizes the nearest neighbour method(NNM),random forest(RF),and support vector machine(SVM)to classify and extract the cherry planting areas in Yuntai Mountain Forest Farm.Focusing on the specific vegetation type of cherry trees,this study explores remote sensing image extraction methods during the blooming season.This research not only fills the gap in the study of cherry tree phenological feature extraction but also provides a referential classification method for other vegetation types with unique phenological features.The research results indicate that combining the characteristics of cherry tree remote sensing images,the use of different classification methods significantly improves the classification accuracy of cherry trees,with the highest extraction accuracy reaching 98%.Accuracy evaluation analysis of the classified and extracted images reveal that the random forest method achieves the best cherry tree flower extraction quality when vegetation indices,spectral features,and texture features are added,followed by the nearest neighbour method with the addition of these features.
作者
常瑶
费鲜芸
王圳
高亚军
杨民书
文陈昊
CHANG Yao;FEI Xianyun;WANG Zhen;GAO Yajun;YANG Minshu;WEN Chenhao(School of Marine Technology and Geomatics,Jiangsu Ocean University,Lianyungang 222005,China;Lianyungang Forestry Technical Guidance Station,Lianyungang 222002,China)
出处
《江苏海洋大学学报（自然科学版）》
CAS
2024年第2期89-96,共8页
Journal of Jiangsu Ocean University:Natural Science Edition
关键词
樱桃树提取
物候特征
分类方法对比
添加多特征
cherry tree extraction
phenological characteristics
comparative analysis of classification methods
addition of multiple features
