基于GF-1PMS影像的柠檬种植面积估算

发布时间：2018-06-01 07:31

本文选题：GF- + PMS影像　；参考：《中国农业资源与区划》2016年11期

【摘要】：为应用国产高空间分辨率影像快速、准确估算丘陵区柠檬种植面积,文章基于GF-1 PMS影像使用不同数据预处理及分类法估算柠檬种植面积。通过对影像进行大气校正、数据融合、滤波等处理,分别得到光谱反射率数据、融合影像和纹理特征数据(分辨率分别为8m、2m、2m)。通过可分离性分析,发现荒草地和未成林柠檬的可分性最差,二者在光谱反射率数据、融合影像和纹理图像中的可分性数值均小于1.8,是影响柠檬分类精度的主要因素。基于最大似然法的分类精度评价结果显示,纹理图像数据估算精度好于8m分辨率的多光谱反射率数据和2m分辨率的融合图像,适当的影像预处理有助于提高监督分类精度。对比基于最大似然法的精度,基于面向对象法的柠檬面积估算精度达95.09%,高于监督分类法的,使用GF-1 PMS影像估算柠檬面积最优方法为面向对象法。该研究为应用国产高分辨率遥感数据快速、准确估算丘陵地区果树种植面积提供了相关参考。
[Abstract]:In order to estimate the planting area of lemon in hilly area accurately by using high spatial resolution images made in China, different data preprocessing and classification methods were used to estimate the planting area of lemon based on GF-1 PMS image. The spectral reflectivity data are obtained by atmospheric correction, data fusion, filtering and so on. The fused image and texture feature data (the resolution is 8mt / 2mt / m respectively) are obtained. Through the analysis of separability, it was found that the fractionability of barren grassland and non-forest lemons was the worst. The main factors affecting the classification accuracy of lemons were spectral reflectivity data and the values of separability in fused image and texture image were all less than 1.8. The results of classification accuracy evaluation based on maximum likelihood method show that the estimation accuracy of texture image data is better than that of multi-spectral reflectance data with 8m resolution and fusion image with 2m resolution. Proper image preprocessing is helpful to improve the supervised classification accuracy. Compared with the precision of maximum likelihood method, the accuracy of lemon area estimation based on object oriented method is 95.09, which is higher than that of supervised classification method. The best method of using GF-1 PMS image to estimate lemon area is object oriented method. This study provides a reference for estimating the planting area of fruit trees in hilly area by using high resolution remote sensing data.
【作者单位】：四川省农业科学院遥感应用研究所;
【基金】：四川省农业科学院青年基金项目“基于数据同化的成都平原水稻估产方法研究”(2015QNJJ-022) 四川省财政创新能力提升工程专项资金项目“基于遥感技术的四川盆地主要农作物估产系统研究”(2016GXTZ-012)
【分类号】：S666.5;S127

【参考文献】