Estimating Tea Plantation Area Based On Multi-source Satellite Data

China is the origin of tea and the first country in the world to discover and plant tea trees.It is also the world's largest producer,exporter and consumer of tea.Tea is cultivated in about 60 countries around the world,and China's tea plantation accounts for more than 60% of the global tea planting area.Tea plantation area in China are widely distributed,mainly concentrated in the southern regions,such as Zhejiang,Fujian,Yunnan,Anhui and Hunan province.The tea industry has an important impact on China's economy and rural development.In major tea-growing provinces such as Zhejiang,tea planting has become a pillar industry for the revitalization of some towns and villages,farmers out of poverty and the government's precision poverty alleviation.Therefore,it is also of great significance for the classification of tea plantation area by remote sensing.The traditional method of obtaining tea plantation area is mainly from field survey data of rural technicians.This method requires a huge amount of manpower,material and financial resources to complete a large-scale census of tea plantation area,and because of the subjectivity in the manual survey,it is impossible to obtain continuous data on the planting area and spatial distribution of tea plantation.At the same time,due to the development of remote sensing,more and more scholars use this technology to remotely monitor crop planting information.Due to the unique planting method of tea trees,tea plantation area present unique texture features on remote sensing images.Therefore,this paper chooses Gabor texture features as classification features.Because tea gardens have unique phenological characteristics,this paper selects NDVI as the vegetation index feature and combines the spectral features to construct a classification feature dataset for tea plantation area remote sensing classification.In this paper,four different remote sensing data are used to extract the planted area of the tea garden.The specific research contents are:(1)Explore the influence of topographic correction and different classifiers on the extraction of tea plantation area.In this paper,the SCS + C method is used to correct the terrain of the remote sensing image,and the support vector machine and random forest classifier are selected.The results show that adding terrain correction to the remote sensing image preprocessing step and choosing a random forest classifier can improve the accuracy of tea plantation classification.(2)Explore the influence of different data and red edge band on the accuracy of extracting tea plantation area.Three types remote sensing data with different resolutions(Landsat8 OLI,GF-1 WFV and Sentinel-2 MSI)were selected to extract tea plantation area.Using vegetation index with red edge band instead of NDVI in feature data set,the results shows that Sentinel-2 MSI data and red edge band can effectively improve the accuracy of tea garden remote sensing classification.(3)The research based on the fusion of multi-source remote sensing data is to use the ordinary least squares method to relatively correct the classification features of HJ-1 CCD data,GF-1 WFV data and Sentinel-2 MSI data to the Landsat8 OLI data set,and construct multi-temporal dataset.The classification feature dataset of Landsat8 OLI data is classified using a random forest classifier.The result shows that the overall classification accuracy of tea plantation area is 93.14% and the Kappa coefficient is 0.91.