Please use this identifier to cite or link to this item:
|Title:||STUDY AND CLASSIFICATION OF LUNAR SURFACE USING MINI-SAR DATA|
|Keywords:||ELECTRONICS AND COMPUTER ENGINEERING|
|Abstract:||In present study an attempt has been made to classify the craters to check the possibility of finding planetary water-ice deposits For this purpose, three different approaches (i.e., polarimetric, pattern analysis and fractal) have been applied for understanding the scattering behaviour of lunar surface. The polarimetric approach, m-S decomposition is used to identify craters having high probability of planetary water-ice deposits and then based on pattern analysis of CPR pixels a method is described to classify craters into two categories type-I and type-II. Type-I craters are those craters which have high probability of having planetary water-ice deposits and type-II craters are those craters which have low probability of having planetary water-ice deposits. This classification is based on degree of polarization and relative phase between LH-LV receptions. Different polarimetric parameters (child parameters) are calculated and statistics of these parameters are calculated for different craters. Further in this study unsupervised classification of mini-SAR images based on the surface texture is carried out. Texture parameter is measured with the help of fractal dimension (D), which lies in the range 2.0 and 3.0. Based on fractal values, i.e., `D', various regions of lunar surface is clustered in different classes. Using moving window approach, the local fractal dimension `D' is estimated with Triangular Prism Surface Area Method (TPSAM) and Differential Box Counting (DBC) method for different window sizes and corresponding fractal maps from TPSAM method are used for classification purpose. Different window. sizes may give different results thus the window size is very important for classification and hence effect of window size on fractal dimension and on Moran'sl is discussed. The K-means classifier has been used for classification which clusters the pixels according to `D' values. Although fractal dimension is able to provide the texture information very efficiently but it cannot uniquely identify all the classes. In order to remove this discrepancy, analysis based on spatial autocorrelation has been performed. For this purpose local value of Moran's I is calculated by varying window sizes, Moran's I gives spatial autocorrelation of pixels. Spatial autocorrelation measures the correlation of a variable with itself through space. Spatial autocorrelation can be positive or negative. Positive spatial autocorrelation occurs when similar values occur near one another. Negative value of spatial autocorrelation occurs when dissimilar values occur near one another thus classification based on Moran's I and fractal iii dimension `D' can give classification results with enhanced accuracy. Classification results are studied to draw conclusions, further the classification results are compared against images generated by m-S decomposition of mini-SAR images.|
|Appears in Collections:||MASTERS' DISSERTATIONS (E & C)|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.