基于时间序列的冬小麦信息提取及灌溉信息识别方法研究

Research on Method of Winter Wheat Information Extraction and Irrigation Information Recognition Based on Time Series

河北是我国主要农业大省,一直是农业干旱多发区,其原因一方面在于区域水资源短缺严重,另一方面是仍以粗放型灌溉为主,造成水资源浪费。解决上述问题的关键在于提高我国农田灌溉用水效率,要做好该项工作,充分掌握农作物实际种植结构和灌溉信息至关重要。选取邯郸市邱县为研究区域,基于Landsat 8、高分1号影像数据以及无人机数据,结合冬小麦物候特征,构建冬小麦生长季时间序列数据集,采用支持向量机分类法进行了冬小麦信息提取;在此基础上,根据灌溉前后遥感指数变化规律,反演了地表温度(LST)、植被供水指数(VSWI)、温度植被干旱指数(TVDI)等因子,进行区域农作物灌溉信息提取模型实验。研究结果表明:①遥感影像时间序列数据SVM分类方法信息提取精度较高,Kappa系数为0.92;②Landsat 8影像LST、VSWI、TVDI灌溉面积提取结果对比发现,三类指数提取结果占冬小麦面积的比例均在60%以上,具有较好的一致性,调查证实基于VSWI反演灌溉面积效果最好;③三类指数提取结果叠加分析得到面积为128.357 km^(2),其中VSWI指数与此面积重叠率达88.48%。上述研究方法较准确的识别出冬小麦种植面积及其灌溉信息,可以作为区域水资源调度、种植结构调整、干旱防治的理论支撑和决策参考。

Hebei is a major agricultural province in China,and has always been an agricultural drought prone area,which is caused by the serious shortage of regional water resources on the one hand,and the waste of water resources caused by extensive irrigation on the other hand.The key to solve the above problems is to improve the efficiency of irrigation water.To do this work well,it is very important to fully grasp the actual planting structure and irrigation information of crops.In this paper,Qiu County of Handan City was selected as the study ar-ea.Based on Landsat 8,GF-1 image data and UAV data,combined with the phenological characteristics of winter wheat,the winter wheat growing season time series data set was constructed,and winter wheat information was extracted by Support Vector Machine(SVM)classifi-cation method.On this basis,the land surface temperature(LST),vegetation water supply index(VSWI)and temperature vegetation drought index(TVDI)were retrieved according to the changes of remote sensing index before and after irrigation,and the regional crop irri-gation information extraction model experiment was carried out.The results show that:①The SVM classification method of remote sensing image time series data has high information extraction accuracy,and the Kappa coefficient is 0.92;②The comparison of the results of LST,VSWI and TVDI irrigated area extraction in Landsat 8 images shows that the three indexes accounted for more than 60%of the winter wheat area,which has a good consistency.The investigation confirms that the inversion of irrigated area based on VSWI is the best.③The superpo-sition analysis of the extraction results of the three indexes obtains an area of 128.357 km^(2),in which the overlap rate between VSWI index and this area reaches 88.48%.The above research methods can accurately identify the planting area and irrigation information of winter wheat,which can be used as theoretical support and decision-making reference for regional water resource scheduling,planting structure ad-justment,and drought prevention and control.