MODIS与TM冬小麦种植面积遥感测量一致性研究——小区域实验研究

针对大范围冬小麦种植面积遥感测量业务化运行中所存在的不同空间分辨率遥感影像的获取能力、空间分辨率与测量精度之间相互制约的现实问题,选择典型实验区,采取全覆盖的多时相低分辨率MODIS数据与中高分辨率TM样区数据相结合的方法,以支持向量机（SVM）为主要技术手段,通过选取不同比例的TM样本量,对MODIS进行混合像元分解,并对MODIS冬小麦测量结果与TM测量结果进行一致性分析,进而提出了一套可用于大范围冬小麦种植面积遥感测量业务化推广的识别和精度检验方法。研究结果表明：当TM样本量大于5%时,相对于TM识别结果而言,MODIS像元精度可以稳定在82%以上;当TM样本量大于40%时,区域精度可以稳定在97%。因此,从理论上讲,在实际业务化运行中,只要能够获得监测区40%的中高分辨率的影像,采用本研究提出的多尺度冬小麦种植面积测量方法,基本上可以满足业务化运行的精度要求。

Consistency Study between MODIS and TM on Winter Wheat Plant Area Monitoring——A Case in Small Area

Accurate and timely updated-information for plant area of wheat is essential to yield estimation,agricultural management,optimizing the spatial pattern of winter wheat and food security.On account of the mutual restricting problems among the acquisition ability of RS images,spatial resolution and accuracy in the measurement of wheat plant area in large scale,this paper introduced the technology of support vector machine(SVM),and developed the model of consistency check used in operational measurement and accuracy verification by integrating the multi-temporal full-covered data(MODIS) and medium resolution sample data(TM).The paper chose the middle part of Henan province in China as a typical experimental region,where winter wheat is the main crop.By pixel accuracy and regional accuracy,the paper analyzed the consistency of measuring results of winter wheat between MODIS and TM with different sample amount.By consistency check between different proportion of TM sample and measuring result,the study draws the conclusions that when the sample proportion reaches 5 percent of whole samples,the pixel accuracy is higher than 82 percent,and the region accuracy is about 90 percent.Moreover,when sample proportion reaches 40 percent,region accuracy in non-sample region stabilizes over 97 percent.The study result shows that the model in the paper can be regarded as the theoretical and experimental groundwork of the operational measurement of wheat plant area by remote sensing.In practical application,if high or moderate resolution image covering 40% of whole investigating region can be obtained,the accuracy can reach the criterion of operational crop area measurement using the method brought forward in this paper.