摘要
为了获取准确的茶园空间分布信息,利用珠海一号和Sentinel-2号影像数据以最小距离法、支持向量机(SVM)和随机森林(RF)分类方法,结合光谱和纹理特征,对普洱市思茅街道的茶园分布进行识别。结果显示,仅以光谱特征识别,珠海一号和Sentinel-2的最佳分类效果均为随机森林构建的分类模型:OA分别为82.98%和91.29%,Kappa系数分别为0.58和0.79;将其与纹理特征相结合,珠海一号和Sentinel-2的最佳分类效果均为随机森林构建的分类模型:OA分别为95.07%和95.48%,Kappa系数分别为0.84和0.88。表明在不同的影像数据源中,相比仅以光谱特征识别茶园,将光谱特征和纹理特征二者相结合,可以极大提高茶园的识别精度。
In order to obtain accurate information on the spatial distribution of tea plantations,the image data of Zhuhai-1 and Sentinel-2 were utilized to identify the distribution of tea plantations in Simao Street,Pu’er City,using the minimum distance method,Support Vector Machine(SVM)and Random Forest(RF)classification methods,combined with spectral and texture features.The results showed that,with only spectral features,the best classification results of both Zhuhai-1 and Sentinel-2 were the classification models constructed by Random Forest:the OA was 82.98%and 91.29%,and the Kappa coefficients were 0.58 and 0.79,respectively;combining them with texture features,the best classification results of both Zhuhai-1 and Sentinel-2 were the classification models constructed by Random Forest:the OA was 82.98%and 91.29%,and the Kappa coefficients were 0.58 and 0.79,respectively.classification model:the OA is 95.07%,95.48%,and the Kappa coefficient is 0.84 and 0.88,respectively,indicating that the combination of both spectral and texture features can greatly improve the recognition accuracy of the tea plantation compared to recognizing the tea plantation by spectral features only in different image data sources.
作者
韩颖
王泽华
吕杰
HAN Ying;WANG Zehua;LYU Jie(Faculty of Land Resources Engineering,Kunming University of Science and Technology,Kunming 650000,China;City College,Kunming University of Science and Technology,Kunming 650051,China)
出处
《城市勘测》
2024年第2期70-75,共6页
Urban Geotechnical Investigation & Surveying
基金
国家自然科学基金项目(62266026)
教育部产学合作协同育人项目(202101096033)
昆明理工大学课程思政内涵式建设项目(109620220216)
2022年度昆明理工大学分析测试基金(2022T20140090,2022M20212201154)。
