Overview and details of the sessions and sub-session of this conference. Please select a date or session to show only sub-sessions at that day or location. Please select a single sub-session for detailed view (with abstracts and downloads if available).
F3: ID.10668 LU Change & Water Quality
11:30am - 12:30pm
Session Chair: Yifang Ban
Session Chair: Chuanrong Li
Workshop: Land & Environment
Location: Sun Moon Room -2, 5.5 Floor, Junyi Dynasty Hotel
Improving the Accuracy of Land Cover Classification over Textural Area Using Accurate InSAR Covariance Matrix Estimation
1Hohai Univeristy, China, People's Republic of; 2Southeast Univeristy, China, People's Republic of;
The global view of urban areas makes satellite missions a valid instrument for updating urban maps and carrying out the analysis of settlement dynamics. Optical remote sensing is a well-established tool for land cover mapping, but it suffers from atmospheric limitations, especially when unpredictable abnormally long periods of cloud cover affect usually clear-sky regions. The use of synthetic aperture radar (SAR) might become suitable when a systematic and timely survey is required. However, as yet relatively few studies have developed analyses that have used SAR data types over urban areas. One of the main difficulty is that the shapes of the structures in SAR image cannot be represented in detail and mixed pixels are likely to occur when conventional SAR parameter estimation method has been used. For example, de-speckling procedure reduces the speckle effect and simultaneously attenuates the resolution of the image, especially over rich texture areas. Moreover, the InSAR coherence cannot be estimated accurately and the features cannot be discriminated due to the observations are highly biased over the fast decorrelation areas. In this context, one objective of this study is to determine the effectiveness of advanced InSAR covariance estimation for land cover mapping in textural areas.
Based on optimized InSAR layers (as covariance matrix consists of SAR Intensity, InSAR coherence), the study on the combination of mutli-temporal SAR dataset for land cover mapping have been further explored over the urban area in Pearl River Delta, China. Different classifications are first carried out using Random Forest and SVM classifiers at the SAR datasets estimated by different parameter estimation methods. The results show that accurate InSAR parameter estimation is able to generate higher classification accuracies when mapping land cover including linear features, infrastructures, and point targets. Moreover, the combination of mutli-source data and accurate InSAR covariance estimation method can provided the best information to map complex textures. The methods and analyses suggested in this paper extend previous research into remote monitoring of urban environment and illustrate the opportunities for mapping areas with rich textures afforded through combinations of mutlitemporal high resolution SAR data and accurate InSAR parameter estimation.