Exploring Advanced Image Analysis Techniques
Project Overview
The purpose of this project is to evaluate a variety of advanced image analysis methods using a Landsat 8 scene. The methods used in this project include the Principal Component Analysis (PCA), Tasseled Cap Transformation (TCT), and Normalized Difference Vegetation Index (NDVI). The theory behind these enhancements is discussed, and their applicability to the chosen scene assessed. Three different software packages are used to complete this project: PCI Geomatica for the PCA, Erdas Imagine for the TCT, and ArcMap for the NDVI.
The chosen area of interest is the southern portion of Bangladesh, imagery for which was downloaded from the USGS Earth Explorer website (see Figure 1 for global location). This area is quite interesting, as it hosts the very populated capital city of Dhaka at the intersection of the Padma and Meghna rivers. As the greater metropolitan area contains over 15 million people, the food requirements to host the growing population requires extensive agricultural land use. As such, the various enhancement methods are used to separate the various land cover types into easily visible classes, which is
helpful in determining the extensity of farms surrounding
the city.
Before any enhancements could be applied, a Top of
Atmosphere (TOA) correction was applied to the imagery. This
is necessary to convert DN values into spectral reflectance
(% of incident energy reflected) , and account for the angle
(azimuth) and elevation (zenith) of the sun at the time the image
was taken.
Figure 1: Imagery locator map
PCA
TCT
NDVI
Principal Component Analysis is the technique of breaking down a multispectral image from all available bands into fewer ‘new bands’ used for further analysis. These ‘new bands’ act to eliminate interband redundancy, while maintaining most of the information from the original image. This procedure uses an orthogonal transformation on the original data values to segment linearly correlated variables into linearly uncorrelated variables called ‘principal components’. The first principal component accounts for a majority of the variability within the data. Successive principal components then contain the greatest variance of data that is orthogonal to the previous principal component. As such, a majority of most multispectral image variability is found in PC1 (Principal Component 1).
Tasseled Cap Transformation is used in agricultural monitoring by transforming a weighted sum from the original bands into ‘vegetation components’. The majority of the original information is allocated into three main bands: Brightness, Greenness, and Wetness. The Brightness component is a weighted sum of all bands, and is defined in the direction of soil reflectance. The Greenness band is roughly orthogonal to Brightness, and is determined through the contrast between the visible bands and the near-infrared band. The result is strongly correlated to the amount of green vegetation in the image. Last is the Wetness band, which relates to soil and canopy moisture.
A Normalized Difference Vegetation Analysis (NDVI) can be very useful in detecting variability within fields, helping farmers to identify which areas need the most attention and resources (fertilizer, pesticides, and water). The NDVI directly compares the red and near-infrared band values to create a normalized index. As such, the final values are relative, and rest on an index from -1 (representing lack of vegetation) to 1 (representing areas of thriving vegetation). Features such as rock and soil have similar reflectance values in both bands, and will therefore exhibit values close to zero. The NDVI is commonly used to monitor crop health, as it compensates for a variety of external factors such as changing illumination conditions and surface slope.
Landsat Scenes with Applied Enhancements
Created by Sean Thibert
References:
World Map: Wikimedia Commons
NDVI,PCA: "Remote Sensing and Image Interpretation 7th Edition", Lillesand, Kiefer & Chipman, 2015
TCT: “The tasseled Cap – A Graphic Description of the Spectral-Temporal Development of
Agricultural Crops as seen by LANDSAT”, Kauth & Thomas, 1976.
Datum: WGS1984 UTM45N
Landsat Image: USGS Earth Explorer