基于哨兵2号时序影像的冬小麦空间分布研究

Study on Spatial Distribution of Winter Wheat on Sentinel-2 Time Series Image

为获取当地精准、详实的冬小麦空间信息,为冬小麦种植区域规划提供技术支撑,本文基于多时间序列的归一化植被指数构建模型,估算了2018-2021年归一化植被指数(Normalize Difference Vegetation Index,NDVI)模型提取的冬小麦面积,获取其时空分布情况,并参考农业数据进行精度评定,研究了多时相指数模型与黄土高原冬小麦的相关性,探讨了多时相归一化植被指数模型对冬小麦提取的可行性。结果表明:①基于多时相NDVI模型,利用随机森林算法提取冬小麦具有较高精度。每年2、4、6月(越冬期、拔节期、乳熟期)为冬小麦识别的关键期,其构建的指数模型提取效果最佳。②单期遥感影像容易受到同期植被的干扰,多时相指数模型可以有效提高冬小麦提取的精度。分别对2018-2021年冬小麦提取,对提取结果进行检验,总体精度分别为91.16%、90.35%、94.26%。③基于Sentinel-2影像数据的甘谷县提取结果,近3年冬小麦种植面积整体呈现平稳态势。种植的时空分布上看,冬小麦主要集中于甘谷县中部断陷河谷地区,南北山区向中部种植区域逐渐增加,且分布在海拔在2036 m以下山地的特点。因此,基于合成指数模型对冬小麦耕种范围提取方法具有可行性,可以有效提取冬小麦的空间信息,证明了利用该方法可有效获取黄土高原冬小麦空间分布情况,并为当地的冬小麦遥感种植时空变化监测研究提供借鉴和参考。

To obtain accurate and detailed spatial information on winter wheat and also provides technical support for planting regional planning of winter wheat.Based on the multi-time series normalized vegetation index model,the winter wheat area extracted by the Normalize Difference Vegetation Index(NDVI)model during 2018-2021 was estimated,the spatial-temporal distribution was obtained,and the accuracy was evaluated by reference to the agricultural data.The correlation between the multi-temporal index model and winter wheat in Loess Plateau was studied and the feasibility of a multi-temporal normalized vegetation index model for winter wheat extraction was discussed.The result shows that:(1)Based on the multi-temporal NDVI model,the random forest algorithm was used to extract winter wheat with high accuracy.February,April,and June of each year(overwintering stage,jointing stage,and milk ripening stage)are the key periods for winter wheat identification,and the index model constructed by them has the best extraction effect.(2)The single-phase remote sensing image is susceptible to the interference of vegetation at the same time.The multi-temporal index model can improve the accuracy of winter wheat extraction.The results of winter wheat extraction from 2018 to 2021 were tested,and the overall accuracy was 91.16%,90.35%,and 94.26%,respectively.(3)The extraction results of Sentinel-2 image data in Gangu County showed that the planting area of winter wheat was stable in recent three years.In terms of the temporal and spatial distribution of planting,winter wheat was mainly concentrated in the fault valley area of central Gangu County,and the planting area gradually increased from the north and south mountain areas to the central area,and the wheat was distributed in the mountains below 2036 meters above sea level.Therefore,the study showed that the method based on the synthetic index model was feasible to extract the spatial information of winter wheat,which proved that this method could effectively obtain the spatial distribution of winter wheat on the Loess Plateau,and a provide reference for monitoring the spatiotemporal changes of local winter wheat planting by remote sensing.