Research On Cultivated Land Monitoring In Henan Province Based On Google Earth Engine

Cultivated land is an important guarantee and prerequisite for food security.In order to improve the efficiency of remote sensing monitoring of cultivated land information and expand the scope of data application,the method of extracting cultivated land information based on the NDVI time series difference index is further studied.As an efficient monitoring method,remote sensing can extract farmland information on a large scale.However,due to sensor and climate restrictions,the traditional stand-alone remote sensing monitoring method is facing great challenges in the extraction of large-scale,long-term sequence of cultivated land information,the research process and the results produced are greatly limited.With the development of remote sensing cloud computing technology and the emergence of cloud platforms,the traditional remote sensing data processing and analysis mode has been completely changed,making large-scale rapid analysis and application possible.As a global-scale geospatial analysis cloud platform,Google Earth Engine makes up for the lack of time-consuming stand-alone computing and brings new opportunities for rapid remote sensing classification.Therefore,this article is based on the Google Earth Engine platform,with Henan Province as the research area,and multi-temporal Sentinel-2 L1 C images from 2018 to 2020 as the data source,extracting the time series NDVI curve,and performing data reconstruction and HANTS harmonic smoothing on the curve.The NDVI Time Series Difference Index(NTDI)is constructed according to the difference of different land types on the time series NDVI curve.Combined with the phenological information,the vegetation growth season participating in the index calculation is defined,and then the cultivated land information of Henan Province from 2018 to2020 is extracted through the OTSU threshold method,and the accuracy of the extraction results is verified.And compared with other high-precision land cover products and arable land area statistics.The main results of this study are:(1)GEE can quickly access massive remote sensing images and other data resources.And with its high-performance cloud computing capabilities,it can quickly complete the processing of cloud removal,mosaic,cropping and index construction covering the image data of Henan Province.The NTDI index method is based on phenological information and combined with GEE to form a rapid extraction framework for cultivated land information,which can be used to extract cultivated land information conveniently and quickly,and has obvious advantages over local processing.(2)The NTDI index method can well identify small map spots for buildings and distinguish cultivated land from confusing linear features such as roads and rivers.Not only can clearly distinguish the boundaries of cultivated land,but also better restore the spatial form of cultivated land and non-cultivated land.At the same time,the distribution of cultivated land and non-cultivated land in the study area has obvious spatial differentiation.Cultivated land is mainly distributed in the eastern,central plains and southwestern basins with more residential areas and more convenient transportation;As the northern,western and southern mountainous and hilly areas are not conducive to farming,the settlements are scattered and the cultivated land is relatively small.(3)In 2018,the accuracy of cultivated land information extraction by the NTDI index method in Henan Province was the highest in the three years from 2018 to 2020,reaching 84.39%;In 2019,the accuracy of arable land information extraction by the NTDI index method in Henan Province was reduced to 84.06%,but it was 4.06%higher than the accuracy of the CGLS-LC100 product in the same year;In 2020,the accuracy of arable land information extraction by the NTDI index method in Henan Province was 82.89%,which was 2.83% lower than the accuracy of Globe Land30V2020 in the same year;The spatial distribution of cultivated land in the three classification data is relatively consistent,and the NTDI index method is superior to the Globe Land30 V2020 data and the CGLS-LC100 product in the recognition of linear features,point residential areas and the division of cultivated land boundaries.The NTDI index method can meet the accuracy requirements in the extraction of large-scale cultivated land information,and the extraction process is faster and more convenient.(4)The NTDI index method is simple to operate,with high extraction efficiency,and the extraction results vary from time to time.It is uniquely objective.The extraction effect is better in the central,eastern and northeastern regions of Henan Province,and can achieve high-precision and high-efficiency farmland extraction.It is an effective method for extracting farmland information.At the same time,the accuracy of extracting farmland information in some parts of northern Henan Province is low,and further data processing can be used to make up for the deficiencies caused by cloud pollution.