复杂地理背景下耕地分布监测与形成机理研究

Study on spatial differentiation of farmland and its formation mechanism under complicated geographical background

针对复杂地理条件下,遥感影像中地物纹理复杂、光谱多样,耕地难以提取的问题,该文提出了一种基于动态时间弯曲算法的耕地提取方法。采取动态时间弯曲法计算NDVI时间序列的相似性,将耕地与林地、草地等其他有相似物候特征的地物区分开来,并解决不同作物的时间序列曲线的弯曲和平移问题。实验表明,此方法提高了耕地提取的精度,长株潭城市群2013年耕地信息提取的正确率达82.08%,完整率达81.63%。使用景观格局指数定量计算耕地空间分异信息,以经济社会数据作为潜在驱动因子,使用冗余分析方法对驱动因子进行约束性排序分析,结果表明人口因素、经济因素、农业因素是影响长株潭地区耕地分布的主要驱动因素。

According to the problem of extracting farmland information under a complicated geographical environment,in which land covers display in remote sensing images are characterized by complicated textures and multiple spectra,this paper proposes a new method that extracting the farmland distribution information based on dynamic time warping algorithm.This strategy presents a way of calculating the similarity between NDVI time series to distinguish farmland from other land use types,and to deal with the warp and translation of the time series curves caused by different crops.The experiment results show that this method can improve the precision of farmland extraction.The user accuracy of the extraction result of cultivated land in Chang-Zhu-Tan urban agglomeration in 2013is 82.08%,and the producer accuracy is 81.63%.The landscape pattern indices are used to quantitatively assess the farmland distribution,as well as the socioeconomic data are selected as potential driving factors.The redundancy analysis is applied to select and constrainedly ordinate the driving factors.The results indicate that the population,economy and agricultural factors are the main driving forces of the farmland distribution in Chang-Zhu-Tan area.