基于物候特征和混合光谱信息的春玉米种植面积提取

发布时间：2018-10-15 17:21

【摘要】：快速、准确地获取玉米种植面积信息对国家粮食安全和现代信息农业发展有重要的现实意义。遥感技术在作物种植面积提取中具有一定优势,尤其是作物物候信息和光谱数据相结合的方法更是目前发展的趋势之一。选取辽宁省3县市为研究区,利用Savitzky-Golay滤波重构后的2014-2015年MODIS-NDVI时间序列数据,提取作物物候特征和其他主要地物的NDVI曲线变化规律,结合水稻移栽期的NDVI、LSWI数据与大豆鼓粒期的近红外波段反射率数据,训练地物分类阈值、构建决策树提取2015年春玉米种植面积。考虑到耕地的破碎化和土地覆盖类型的多样性,利用MOD09A1反射率影像提取春玉米及混合地物的端元波谱,基于线性光谱混合模型进行混合像元分解获取春玉米丰度,根据决策树分类结果与春玉米丰度信息精确提取春玉米种植面积。利用Landsat8 OLI监督分类影像对种植面积解译结果进行精度评价,结果显示研究区春玉米种植面积提取精度在81%以上,利用统计数据验证得到仅利用物候信息经决策树分类方法提取的春玉米种植面积精度为88.416%,结合混合光谱信息后的提取精度提高到92.382%。春玉米种植面积提取结果较好地反应了其地理分布,表明重构后的生长季NDVI曲线可以准确地反应作物生长变化规律,运用中分辨率MODIS-NDVI时间序列数据获取物候信息,快速、准确提取春玉米种植面积具有可行性。物候特征耦合混合光谱信息的方法突破了传统研究中的像元限制,将作物种植面积识别水平提高到亚像元尺度,效果好于传统的像元尺度MODIS时间序列信息分类方法,能够有效提高作物种植面积估算精度,对于加快数字农业进程、提升农业信息化水平具有重要作用。
[Abstract]:It is of great practical significance to obtain corn planting area information quickly and accurately for national food security and the development of modern information agriculture. Remote sensing technology has some advantages in crop acreage extraction, especially the method of combining crop phenological information with spectral data is one of the developing trends. Three counties and cities in Liaoning Province were selected as study areas. The MODIS-NDVI time series data of 2014-2015 were reconstructed by using Savitzky-Golay filter to extract the phenological characteristics of crops and the NDVI curve changes of other main features. Combined with the NDVI,LSWI data of rice transplanting period and the near infrared band reflectance data of soybean seed filling stage, the classification threshold of ground objects was trained, and the decision tree was constructed to extract the planting area of spring corn in 2015. Considering the fragmentation of cultivated land and the diversity of land cover types, MOD09A1 reflectance image was used to extract the end-component spectra of spring maize and its mixed features, and the abundance of spring maize was obtained by mixed pixel decomposition based on the linear spectral mixing model. The planting area of spring maize was extracted accurately according to the classification result of decision tree and the abundance information of spring corn. Landsat8 OLI supervised classification image was used to evaluate the precision of the interpretation of planting area. The results showed that the precision of spring maize area extraction was more than 81%. The precision of planting area of spring corn extracted by decision tree classification method was 88.416, and the precision of mixed spectral information was improved to 92.382. The result of spring maize planting area extraction reflected its geographical distribution well, which indicated that the reconstructed NDVI curve of growing season could accurately reflect the changing rule of crop growth, and the phenological information could be obtained by using medium resolution MODIS-NDVI time series data. It is feasible to extract the planting area of spring corn accurately. The method of phenological characteristics coupling mixed spectral information breaks through the pixel limitation in traditional research and improves the recognition level of crop planting area to sub-pixel scale, which is better than the traditional pixel scale MODIS time series information classification method. It can effectively improve the precision of crop planting area estimation and play an important role in speeding up the process of digital agriculture and improving the level of agricultural informatization.
【作者单位】：北京林业大学水土保持学院;
【基金】：国家科技支撑计划项目(2015BAD07B03)
【分类号】：S127;S513

【相似文献】