Principal Components Analysis and Tasseled Cap Transformation used in Landsat ETM + images

Abstract

This paper describes how principal components analysis and Tasseled Cap Transformation were used in the Landsat ETM+ images of south-east London. The six-dimensional landsat dataset was reduced to three major components where the first component contributed for the maximum proportion of the variance of the original dataset, and subsequent orthogonal components accounted for the maximum proportion of the remaining variance. The first PCA component showed urban settings in the first component, vegetation in the second and water in the third. Similar to the PCA, the TC showed the urban settings in the brightness component, vegetation in greenness and water in the wetness component. As compared to the PCA, TC transformation has a more analytical basis as it combines a generalization from empirical observations. The PCA and TC components were compared to each other which showed that PC1 contained more brightness information than TC1. However, TC2 and TC3 included pertinent information than PC2 and PC3. The colour composites PCA123 and TC123 were visually compared which showed that for this study, TC123 produced better results.

Read entire article (It is password protected. Please contact me for liable access to this article.)

Read more

Unsupervised and Supervised Classification of Remotely Sensed Imagery

Introduction

Remote sensing has increasingly been used as a source of information for characterizing land use and land cover change at local, regional and global scales (Townshed and Justice, 2002; Lunetta and Lyons, 2003 in Jensen, 2005, p. 337). Land use/land cover classification based on statistical pattern recognition most often used methods of information extraction (Narumalani et al., 2002 in Jensen, 2005, p. 337). In this study, unsupervised and supervised classifications were carried out for land-cover mapping of a remotely sensed imagery of London.

Objectives of study

● To classify an image using both supervised and unsupervised techniques, including:
i) Evaluation of which Landsat bands to include in a classification
ii)Analysis of signature separability
iii)Class aggregation
iv)Inclusion of ancillary data
● To interpret the result of the classification using accuracy calculations.

Read entire article (It is password protected. Please contact me for liable access to this article.)

Read more

Report on Image calculations, Local window filters and Scaling image data

Abstract

This paper describes how band ratioing is useful in providing unique information not available in any single band that is useful for discriminating ground features. It also presents the significance of spatial filtering and rescaling in an image. In this study, four different band ratios were created which helped in extracting useful information regarding the ground features. The colour composite image of the three indices revealed some important information about the ground features and eliminated the effects of shadowing. The ratio with low correlation between the bands, i.e. ratio of red versus nir, was found to contain greater information than the ratios with high correlation between the bands. The NDVI ratio was rescaled by using both the standard scaling algorithm and raster calculator algorithm. A scatter plot was plotted to verify the results were same for both the rescaled NDVI channels. The panchromatic layer containing just band 4 was extracted and median low-pass filter and Sobel edge detector high-pass filter were applied to it. The median filter was found to smooth the image, remove noise and maintain the edges. The Sobel edge filter was found to enhance sharp edges.

Read entire article (It is password protected. Please contact me for liable access to this article.)  

Read more

Multi-scale information extraction of land cover from Landsat Imagery and DEM data

Introduction

The main purpose of this study is to perform a multi-scale information extraction of land cover and vegetation structure from Landsat TM imagery and DEM data, using classification and statistical modeling methods. This study also aims to evaluate the results of the analysis through accuracy statistics and meaningful map productions. The data available for this study is a TM Landsat image of the Hinton and Jasper, Alberta area with channels, 1 to 5 and 7. A channel containing a DEM of the area and three additional channels with tasseled cap (TCA) outputs are also provided. 

Additionally, three shape files containing land cover, leaf area index (LAI) and Crown closure are also available. These shape files are in the form of points and are basically used to create training sites and assess the accuracy of the results. The land cover shape file is a file containing 437 land cover calls made by field personnel observing a 90 x 90 meter area roughly equivalent to nine TM pixels. The leaf area index shape file is a file containing 37 estimates of LAI obtained by field personnel using an Accupar Ceptometer over a 30 x 30 meter ground plot roughly equivalent to one TM pixel. The crown closure shape file contains 73 estimates of crown closure measured by field personnel using spherical densiometers over a 30 x 30 meter ground plot roughly equivalent to one TM pixel.

The land cover shape file is used to create a classified image. The LAI and crown closure shape files are used to predict the LAI and Crown closure values for the entire Landsat scene. With the help of these datasets, it is possible to produce multiple maps that can be used to explore the relations hips between LAI, crown closure and the process that are taking place on the ground. The main objective of this study is to produce two maps that can demonstrate how these different sets of data can be used to generate meaningful information at different scales.

Read entire article (It is password protected. Please contact me for liable access to this article.)

Read more

Radar Polarimetry in Remote Sensing

Introduction

Radar polarimetry is defined as the science and techniques involved in measuring and analyzing the complex scattering matrix of pixels in a radar image (CCRS 2008). The microwave energy from a radar system includes wavelengths within approximate range of 1cm to 1m and is capable of  penetrating the atmosphere under virtually all conditions depending on the wavelengths involved (Lillesand et. al. 2004). The other advantages of radar over the optical sensors are that radar operates at user-specified times and provides unique information as it senses wavelengths outside the visible portion of the EMR spectrum (Geog 633 2008).

The radar transmits either horizontally polarized (H) or vertically polarized (V) microwave radiation which can then generate a back-scattered wave with a variety of polarizations. Any polarizations on either transmission or reception can be synthesized by using H and V components with a total of four combinations of transmit and receive polarizations (CCRS 2008). In this study we focus on dual-polarized and fully polarized data sets.

For the purpose of this assignment, we have two types of multi-polarized imagery of which one is dual-polarized and the other is polarimetric. The dual polarized data is from Germany taken by ENVISAT-ASAR sensor and is provided with a header file. The polarimetric data set is from an unknown area taken by Convair-580. The main objectives of this study are to explore the differences between multi-polarized imagery using PWS and PCI Geomatica software package, to examine how different polarizations can be used to understand the physical properties of the features, and to understand how the “polarization signature" of targets provides a convenient way of visualizing a target's scattering properties. 

Read entire article (It is password protected. Please contact me for liable access to this article.)

Read more