Detection Method Of Land Use Change Based On Landsat Image Time Series

The research on the spatio-temporal process of land use change has become the focus of global change research,and it is the premise and foundation to obtain the change information of the long time series of specific type or specific region.The spatial and temporal resolution of remote sensing images has been significantly improved.This provides abundant and reliable data support for detecting and charactering the temporal and spatial processes of regional land use change.Therefore,based on the time series of remote sensing images,it is one of the hotspots of land use/cover change to explore the time point,change position and change type of land use change.Landsat images record long time surface changes and have high spatial resolution,thus becoming the best image data for resource management.Therefore,exploring the method of land use change detection based on Landsat image time series can not only enrich the time series analysis method of remote sensing image,but also provide support for regional land use change analysis and planning management.The existing time series analysis of remote sensing images is mostly based on the analysis of the time series of single index,which can detect less land use change types,and the accuracy of land use change detection is not high.Therefore,more and more attention has been paid to the comprehensive identification of multiple indexes.The paper explores the construction methods of multi-exponent pixel-level image time series and multivariate time series similarity measurement methods,and forms the idea and implementation method of land-use change detection technology based on Landsat image time series.In this study,three indexes of NDVI,NDBI and MNDWI are selected to construct the ternary time series,and the similarity measurement method suitable for the ternary time series is constructed by improving the distance metric method.On this basis,the time points and change positions of land use change are detected,and the detailed land use change types are identified according to the land use data and remote sensing images.The main contents are as follows:(1)Construction of the ternary time series and analysis of its characteristics.Considering that NDVI,NDBI and MNDWI have significant effects on the detection of vegetation,buildings and water respectively,and the ranges are all between[-1,1],the three indexes of NDVI,NDBI and MNDWI are selected to construct the ternary time series.Combined with the land use data and Google Earth high resolution remote sensing images,the typical sample points of land use types that have not changed and changed are selected.When comparing the ternary time series of different types of land use,it was found that many time series of land use types are very similar and difficult to distinguish.The main reason is that the classification of land use types extracted from the land use data is fine.The spectral characteristics of the agricultural land and the subtypes of construction land in the remote sensing images are very similar,so the time series is also very similar.Based on the time series characteristic analysis of land use change types,the average time series of each change time point of the land use change sample pixels are used as the reference time series to detect the land use change of different points.(2)Research on similarity measure method of the ternary time series.In view of the characteristics of the ternary time series,the value of the ternary time series of each time point is mapped to the point coordinates of the three-dimensional space,and the ternary time series is the sequence of the point coordinate in the three dimensional space of different time.According to the distance of the point in the three-dimensional space,the similarity of the value of the ternary time series at each time point is quantified.The distance measurement method is improved,and the similarity calculation method of the ternary time series is proposed.The similarity of the ternary time series can be quantified by the sum of the distances in the three-dimensional space of all time points.(3)The detection of land use change.In view of the complex characteristics of time series and change types of land use change in Landsat image time series,this study proposes a method of land use change detection based on Landsat image time series:first detecting land use change at different time points,and then identifying the detailed types of land use change.The ternary time series of NDVI,NDBI and MNDWI are used as data sources.The land use change sample pixel average time series of 5 changing time points are used as the reference sequence.Four kinds of single time series similarity measure methods and improved ternary time series similarity measure methods are used to detect the land use change of different points.From the result of the accuracy evaluation,the overall accuracy of the detection results of improved ternary time series similarity measure methods is higher than 85%,the Kappa coefficient is higher than 0.8,compared with the single time series similarity measure methods.The results show that the similarity measurement of time series improves the accuracy of land use change detection.Based on the results of land use change detection,the detailed land use change types are identified by land use data and remote sensing images.In the experiment,52 types of land use change are identified,and more abundant information of land use change of changing time points and types of change are obtained.To some extent,this solves the problem of coarse classification of medium resolution images.The innovation of the research is:(1)Improved the similarity measurement method of the ternary time series,and improved the accuracy of land use change detection.(2)A method for detecting land use change is proposed,which first detects the change of time and location,and then identifies the change type.