基于相似性分析的中低分辨率复合水稻种植面积测量法

Monitoring of Paddy Rice Plant Area Based on Similar Index by Multi-Resolution Remote Sensing Data

【目的】利用遥感技术获取大范围水稻种植面积是遥感技术在农业领域的主要应用方向之一。本研究的目的是探索利用多尺度遥感数据复合测量水稻种植面积的方法。【方法】以SPOT5数据的水稻识别结果作为样本,构建图像相似性指数,通过支持向量机(SVM)混合像元分解模型,对MODIS-EVI时间序列数据进行水稻的种植面积测量。【结果】通过江苏省邳州市的试验研究得出:(1)在野外经验支持下,从MODIS-EVI时间序列数据中构建的水稻种植相似性指数可以有效反映水稻在整个研究区的空间分布情况;(2)利用图像相似性选取训练样本,能有效地提高MODIS-EVI数据的水稻种植面积的测量精度,当图像相似性指数越小,即图像相似性越高,提取的水稻种植面积也越准确;(3)通过与随机样本测量结果对比分析,基于相似样本的测量方法有着更高的稳定性;(4)该方法在不同种植结构分区内有着相似的总量精度与像元精度变化规律,均能获得较高的测量精度。【结论】基于相似样本的水稻种植面积测量方法,有助于发挥MODIS长时间序列优势,提高水稻种植面积遥感测量精度和稳定性,可以作为替代随机选取样本的方法之一。

[Objective] The acquisition of large area paddy rice acreage by remote sensing is one of the chief fields in the application of remote sensing in agriculture. And multi-resolution remotely sensed data method plays an important role. [ Method] In this study, paddy rice result, acquired from SPOT data, severs as high resolution samples, while MODIS-EVI time series data serve as low resolution data, and then using support vector machine （SVM） method to obtain the paddy rice coverage. [ Result]Some important advancements have been found from this study： （1） The similar index which is based on the MODIS-EVI time series RS-data and filed work samples can describe the paddy rice coverage obviously; （2） The method of choosing sample based on the image, similar index image, comparability has firmly theory foundation which means when the image comparability higher the accuracy better; （3） Compared to the result of random samples, the method based similar samples has better stability; （4） In different plant structure regions, the gross accuracy and pixel accuracy are similar and high. [Conclusion] So the method based on similar samples can help to utilize the advantage of MODIS data and improve the stability and accuracy of the plant area monitoring of paddy rice, which can be applicable to different regions as a substitution in acquisition of large area paddy rice coverage.