Study On The Method Of Accuracy Assessment Of Land Cover Remote Sensing Mapping Product

In order to meet the needs of global higher resolution land cover data,China has successfully developed the world's highest resolution land surface mapping product-GlobeLand30 in 2014.Accuracy evaluation is the premise of reasonable application of data,but the accuracy evaluation methods still have defects in sample size determination,sample layout and so on.In this paper,a more accurate method of accuracy evaluation is proposed to ensure the rationality of the accuracy of land cover products.Influenced by the spatial heterogeneity of land cover patches,the traditional sampling method has great limitations in sample size determination,inter layer sample allocation and sample layout.This study ameliorated the conventional method of accuracy assessment by studying landscape ecology.The spatial heterogeneity is quantified by calculating the shape index to reduce the influence of spatial heterogeneity in sample sampling.In this paper,a sample sampling method considering landscape shape is proposed to evaluate the accuracy of GlobeLand30 data in Jiangxi Province in 2010.At the same time,this study set up a comparative experiment and field data collection in the study area,which confirmed that the quality assessment method proposed in this study is practical and available.The main conclusions of this study are as follows:(1)At present,most of the accuracy assessment methods can not eliminate the influence of spatial heterogeneity on the accuracy evaluation process,and cannot accurately evaluate the accuracy of the high-resolution land cover products.This study shows that the landscape shape index,which makes the accuracy evaluation more authentic and reliable,can improve the sampling rate of the broken area,and reduce the influence of the spatial heterogeneity.(2)In this study,the accuracy evaluation method of traditional stratified sampling is used as a comparative experiment,and the accuracy evaluation results obtained by the two methods are compared and analyzed.The results indicate that OA of the two methods is similar,as well as Kappa coefficient.It shows that the accuracy evaluation method considering landscape shape is available.(3)The accuracy evaluation results are analyzed and evaluated by using field data acquisition,avoiding the misjudgment of some sample points caused by lack of experience,and ensuring the rationality of the accuracy evaluation results.