与抽样相结合的县级作物遥感面积估算应用实例

为了探索遥感卫星影像和无人机调查在实际遥感面积估算中的调查效果和适用性,以广东省阳春市晚稻为例,采用卫星影像与无人机调查结果相结合,使用两种面积估算方法进行2013年晚稻种植面积估算。实验结果表明,比估计和回归估计方法的估算结果分别为22 501.1hm2和22 781.1hm2,二者的CV分别为8.84%和1.03%。研究结果表明:1无人机调查可获得高质量的面积测量信息;2优化后的样本能够支持面积估算过程,但其理论意义需要更进一步地讨论与证明;3卫星数据与无人机调查数据相结合的方法可以提供满足统计精度要求的面积测量结果,具有良好的应用前景。

A Case Study for Area Estimation of Crop in County Level based on Remote Sensing Data and Sampling Technology

Unmanned aerial vehicle (UAV) is an effective tool and an appreciate alternative mean to supply,even partially replace the traditional ground investigation,especially at large scale investigation,due to its greater availability and flexibility.However,there is little study on area estimation using UAV in domestic,and this investigation approach is not yet adopted operationally.Therefore,it is necessary to explore the applicability and flexibility,and point out the issues that should be noted at the area estimation processes with UAVs in practical cases.In this study,the area estimation of rice,main crop in south China was implemented in Yangchun,Guangdong Province.Satellite imageries were acquired at key growth period of late rice.Support Vector Machine (SVM) was used to generate the classification of arable land.This study constructed 300 m×300 m square grids covering the whole study area.Then,grids that contain the arable land in classification were reserved to form the population.After that,we proposed a sampling method to reduce the cost and made full use of UAVs’ intensive investigation ability.This method included four steps:firstly,stratified random sampling method was used to select investigation samples from 300 m×300 m square grids and the variance of those samples (V1) was calculated.Secondly,big rectangle grids were built up and covered the whole study area,and each big rectangle grid contained 5×6small square grids which may be from different strata.Thirdly,the variance of 5×6small square grids (V2) in each big rectangle grid was calculated and compared with the variance of selected samples in step1.If the V2 of a given rectangle grid is equal to or near to V1,the big rectangle grid can be used to replace the samples selected from step1.Fourthly,according to the required sample size and the comparison results in step 3,we selected three big rectangle grids as final samples for UAVs investigation.Ratio estimator and difference regression estimator were adopted to estimate the sowing area of late rice in Yangchun of Guangdong Province in 2013.Their area estimations were 22 501.1 hm² and 22 781.1 hm²,respectively,and their coefficient of variation were 8.84% and 1.03%,respectively.Above the results suggest that the UAVs investigation provide reliable ground truth.The proposed sample selection method in this study enhances the feasibility and operability of UAVs in crop area estimation.