融合物候及纹理特征的茉莉花种植区提取研究

为充分挖掘图像时序物候信息、几何空间、纹理特征在农作物监测提取研究上的优势,本研究通过构建物候参数与纹理特征的方法对广西横州市茉莉花（Jasminum sambac）种植区进行提取,该方法通过增强型时空自适应反射率融合模型（ESTARFM）融合年度内“高空间-低时间”“高时间-低空间”的遥感图像,重建生成年度中低尺度的卫星影像,获取研究区内不同植被的物候信息参数,并结合地物光谱信息提取地物纹理特征。选取物候参数、纹理特征作为茉莉花种植区分类提取信息,对不同的参数特征进行组合,开展分类对比和综合精度量化评价。结果表明,重建时序归一化植被指数（NDVI）数据结合物候参数及纹理特征的茉莉花种植区提取方法,总体分类精度达到88.36%,Kappa系数为0.863 3,在中低尺度多光谱影像遥感数据“时空信息”的缺失下对茉莉花种植区提取具有良好的效果。

Study on Extraction of Jasminum sambac Planting Areas by Combing Phenological and Texture Features

In order to fully exploit the advantages of image timing phenological information, geometric space and texture features in crop monitoring and extraction research, this study extracts Jasminum sambac planting area in Hengzhou City, Guangxi by constructing phenological parameters and texture features. In this study, the Enhanced Spatio-Temporal Adaptive Reflectance Fusion Model (ESTARFM) was used to fuse the annual "high space-low time" and "high time-low space" remote sensing images, and reconstructed to generate the annual medium and low-scale satellite imagery to obtain the parameters of phenological information of different vegetation in the study area. The phenological information parameters of different vegetation in the study area are obtained, and the texture features of the ground objects are extracted combined with the spectral information of the ground objects. The phenological parameters and texture features were selected as the classification and extraction information of J.sambac planting areas, and different parameter features were combined to carry out classification comparison and quantitative evaluation of the comprehensive accuracy. The results show that the overall classification accuracy of J.sambac planting areas extraction method based on the reconstruction of time-series Normalized Difference Vegetation Index (NDVI) data combined with phenological parameters and texture features reaches 88.36%, and the Kappa coefficient was 0.863 3. The extraction of J.sambac planting area has a good effect under the lack of "spatio-temporal information" of medium and low-scale multi-spectral image remote sensing data.