Extraction And Monitoring Technique For Cropland And Urban Land Using SAR Image Time Series

Oriented to the monitoring demand for cropland and urban land and enabled with time series diming and spatial data diming, we study the pixel-level SAR image time series modeling method, develop the novel extraction technique of cropland and urban land using SAR image time series, and achive technology breakthrough in monitoring the cropland to urban land, which can serve the monitoring, analysis and evalution of cropland and urban land geographic national conditions.Remote sensing time series is characteristic of big data, multi-scale, spatio-temporal correlation, fuziziness and much noise. Tapping their potential has particular theoretical and practical value. In recent years, research has mianly focused on optical image time series, the key of which lies on the reconstruction of time series. Most of the former research targets the differential InSAR, which is based on phase information of SAR image data; while the use of the SAR pixel-level time series backscattering coefficient is quite rare. Traditional remote sensing feature recognition is based on the spectral or spatial domain, while recognition based on the time-dimension focuses on the time vector of images. Thus, many studies focus on the similarity characteristics of the curve by using pixel-level SAR time series curves forego identification of cropland and urban land, ignoring the spatial correlation among pixels. As a result, it is quite a fresh issue to make full use of the spatial-temporal characteristics of the pixel-level SAR image time series to identify the ground-object. Moreover, using the SAR image time series monitor the conversion of cropland to urban land is quite rare, the difficulty of which lies in the selection of basic amount, removing noise spots, extraction of change nodes and the processing of mixing pixels.Two different SAR images data set of150m low spatial resolution and30m middle spatial resolution were selected as raw data in this paper, chosing two study area of Shenzhen city and Xuzhou urban area correspondingly. A novel framework of extraction and monitoring technique for conversion of cropland to urban land using SAR image time series was proposed. The main contents and conclusions are listed as follows:(1) Pixel-level SAR image time series modeling method. The definition and classification of pixel-level SAR image time series are given in the begining. To achieve pixel-level SAR image time series, we propose a locally adaptive image matching technique for the high-precision geometric registration of SAR images, which comprises three key parts:local error point clustering, iterative binary partition for local registration and boundary effects. Based on this proposed matching technique, we propose a pixel-level SAR image time series modeling method and introduce the denoising method for the time series.(2) Cropland and urban land extraction technique. We give the definition of the spatio-temporal similarity which is derived from the mathematical formulas of time series similarities and spatial neighborhood similarities. Spatio-temporal similarity analysis based spatial context was presented to improve feature extraction, considering both the similarity of a feature’s pixel-level time series and its spatial correlation. Similarity measurement via improved Dynamic Time Warping (DTW), yielding a DTW value between the pure and mixed pixel as the threshold for similarity judgment and extracting the cropland and urban land from SAR image.(3) Monitoring technique for conversion of cropland to urban land using SAR image time series. The change characteristics of pixel-level SAR image time series are discussed first. A spatio-temporal change detection strategy considering mixed pixel and noise were presented to extract the change nodes and change pixels by spatio-temporal change tree. The characteristic of mixed pixel which has changed was focused on discussing to obtain the threshold during the monitoring. Morphological-structural characteristics of time series was used to determine the conversion of cropland to urban land and the minimum of the series of mean square errors model was calculated to ascertain change node. The mean difference of time series of before and after the change node was utilized to deal with mixed pixel.In conclusion, to exploit the application potential in monitoring the conversion of cropland to urban land of SAR image data, this study proposes a novel framework of extraction and monitoring technique for conversion of cropland to urban land based on pixel-level SAR image time series, obtaining good result and bearing high application value to a certain extent. The proposed pixel-level SAR image time series modeling method yields a time series curve; the proposed spatio-temporal similarity analysis method can identify cropland and urban land with high accuracy; the proposed spatio-temporal change detection strategy considering mixed pixel and noise can effectively extract the change nodes and change pixels, achiving the monitoring the conversion of cropland to urban land.