| The agricultural plastic greenhouse is an important modern agricultural facility that can effectively improve the growing environment of crops and increase crop yields.However,the widespread use of agricultural plastic greenhouses has caused ecological and environmental problems such as water pollution,soil degradation,and white pollution within certain areas.Therefore,dynamic monitoring of the spatial distribution and usage of agricultural plastic greenhouses can provide decision reference for relevant departments in planning the use of agricultural plastic greenhouses,preventing non-point source pollution and protecting soil,which is of great significance for ecological environmental protection.However,the large-scale geographical environment’s complexity and the heterogeneity of the spectra of various types of agricultural plastic greenhouses make it difficult to use the plastic greenhouse spectral index to achieve rapid and high-precision extraction of agricultural plastic greenhouse in a large-scale complex environment.Therefore,studying method for extracting plastic greenhouse spectral index with strong generalization ability,higher level of automation,and recognition accuracy has important practical value.This paper takes the Dianchi Lake basin as the study area,and uses Sentinel-2optical image,based on geographical environmental background information and spectral index method,to carry out the research of agricultural plastic greenhouse mapping in large-scale and complex environments.Apply this method to extract agricultural plastic greenhouses in 30 provincial-level administrative regions within China.The specific research contents are as follows:1.Method of constructing spectral index of agricultural plastic greenhouse.This paper proposes an optimal index construction method with finite structure based on sample and feature matching(OICM).Firstly,match the samples and features,and obtain statistical information on the ground objects;Then,select features based on the standard deviation and Jeffries-Matusita Distance;Finally,the selected features are processed and combined to construct a Composite Plastic-Greenhouse Index(CPGI)suitable for extracting multiple types of agricultural plastic greenhouses.The experimental results show that:(1)The CPGI has high separability from background land cover,and agricultural plastic greenhouses can be extracted through single-index and single-threshold segmentation,effectively improving mapping efficiency;(2)The overall accuracy of the CPGI is 90.75%,and the Kappa coefficient is 0.81,and the classification accuracy is relatively stable at an annual scale.2.Extraction method of agricultural plastic greenhouse based on geographical environmental background information and spectral index.This paper proposes an agricultural plastic greenhouse extraction method based on geographical environmental background information and spectral index.Firstly,based on Simple Non-iterative Clustering(SNIC)clustering of segmented objects,the background similarity is calculated to divide geographic environments into urban and non-urban areas.Then,the OICM is used to construct plastic greenhouse indices suitable for different regions,and agricultural plastic greenhouses are extracted respectively.Finally,the results of classifying agricultural plastic greenhouses in different regions are integrated.The above process is simplified into a backgroundbasic composite plastic-greenhouse index(BCPGI).The experimental results show that:(1)Compared with the regular grid statistical units,the boundary division of the statistical units using SNIC segmented objects is more accurate and the calculation efficiency is higher;(2)Plastic greenhouse indices with regional characteristics have enhanced suppression capability for major interfering land cover in different regions,and the separability is improved;(3)The overall accuracy of the BCPGI is 92.43%,and the Kappa coefficient is 0.85,which shows a significant improvement in classification accuracy compared to the CPGI.3.Mapping agricultural plastic greenhouses in large-scale complex environments.This paper extraction of agricultural plastic greenhouses in 30 provincial administrative regions in China based on the Google Earth Engine platform and the BCPGI.The results show that the overall accuracy of the method is 81.32%,and the Kappa coefficient is 0.62.Compared with other plastic greenhouse indices,the BCPGI reduces commission errors of the agricultural plastic greenhouse by more than 9%,effectively reducing interference from background land cover.This method enables rapid mapping of agricultural plastic greenhouses in large-scale complex environments and provides support for the dynamic monitoring of the spatial distribution of agricultural plastic greenhouses. |
