Exercise 1: Georeferencing of maps/toposheets/images using QGIS software

What is Raster data?

• Raster data is an image, or a grid of pixels, where each pixel has a unique value that
represents a data point on the ground
• It stores information about features in a cell-based manner.
• Raster data is any pixelated (or gridded) data where each pixel is associated with a
specific geographical location.
• The value of a pixel can be continuous (e.g. elevation) or categorical (e.g. land use).
• It is similar to representation of any digital image.
What is Georeferencing?
• Georeferencing is the process of assigning real-world coordinates to each pixel of the
raster image (a toposheet or an area map with defined coordinates).
• It is the process of aligning the raster data with a map coordinate system. This allows you
to view, query, and analyze the data with other geographic data.
• Most of the times these coordinates are obtained through field surveys using a GPS
device or from digitized scanned maps where coordinates can be obtained from the
markings on the map image itself.
• Using these sample coordinates or GCPs (Ground Control Points), the image is warped
and made to fit within the chosen coordinate system.
Common Georeferencing Tools: ArcGIS, QGIS, ERDAS IMAGINE, AutoCAD Map 3D,
GRASS GIS
QGIS software
• QGIS is a free and open-source geographic information system (GIS) software that
supports Windows, macOS, and Linux environment.
• The software supports viewing, editing, printing, and analysis of geospatial data in a
range of data formats.
• The software can be easily downloaded from https://www.qgis.org/resources/hub/ where
other information regarding the usage of software is also available.
Materials Required:
• GIS Software (QGIS)
• Scanned map/toposheet/image with coordinates in tiff or jpg format
Procedure:
• Download a map/toposheet/image with coordinates (latitude/longitude) of any area
(Uttarakhand)
• Open latest version of QGIS software
• Go to raster and then georeferencer
• Open the downloaded raster image in the georeferencer
• Add four ground control points (GCP) from the raster image
• After adding the points, go to settings and set the transformation type to linear and click
on save the GCP points and load in QGIS when done.
• The output modified raster image can be saved in the desired location using settings
option
• Now, start georeferencing the image using green play button on the top left corner
• After georeferencing the modified file can be viewed in QGIS
Exercise 2: To download remote sensing images of an interest area
What is Remote Sensing?
• Remote sensing is the process of detecting and monitoring the physical characteristics of
an area from a distance (eg. a satellite or aircraft).
• The physical characteristics of an object or area can be studied by measuring the reflected
and emitted radiation.
LANDSAT
• The Landsat Program is a series of Earth-observing satellite missions launched on July
23, 1972 jointly managed by NASA and the U.S. Geological Survey.
• It was launched as Earth Resources Technology Satellite (ERTS-1) was launched which
was later renamed as Landsat 1.
• The Landsat Program’s gives a continuous archive since 1972 to present representing the
world’s longest continuously acquired collection of remote sensing data.
• It has an optimal ground resolution and spectral bands which helps in keeping an efficient
track of land use changes due to climate change, urbanization, drought, wildfire, biomass
changes (carbon assessments), and a host of other natural and human-caused changes.
• Currently orbiting and active satellites are Landsat 8 and 9.
Material Required
• Account in USGS Earth Explorer
• Software used to view the image : ENVI 4.7
Procedure:
• Go to https://earthexplorer.usgs.gov/
• Create your login account
• After account creation login to view the page
• First enter the search criteria as address/place under geocoder
• After entering the address/place, latitude and longitude of the place is displayed
• Click on the displayed coordinates
• After adding the coordinates, enter the time period for which the images are required
under the data range
• Click on the datasets and select the satellite (Landsat 8) image results
• Further, download the required images in band 3 (Green), 4 (Red) and 5 (NIR)
• Open the images in ENVI software
Exercise 3: To layer stack the remotely sensed images of the interest area
What is layerstacking?
Layer stacking is a process of combining multiple separate bands in order to produce a new
multi band image. This type of multi band image is useful in visualizing and identifying the
available Land Use Land Cover (LULC) classes.
How many bands should be stacked?
In order to save the processing time, stacking is done for necessary bands only. For instance,
LULC extraction or agriculture acreage estimation require only blue, green red and NIR
bands.
Why layer stacking is important for an image?
• Each of the bands when displayed alone is a panchromatic image (gray-scaled), so to obtain
a beautiful natural image of the earth, compositing or “stacking” of multiple images is
required.
• Spectral bands used by our computer monitors, television screens and other electronic
displays in RGB (Red, Green and Blue) and their appropriate combination helps our eyes
to see the world in natural colours.
• For display of natural colour image on our software for our area of interest, a
composite/stacking of image bands 4-3-2 is done in which the Band 4 is displayed as red
and Band 3 is displayed as green.
• Usually, the image for clear view and analysis is displayed in a composite/layerstack of
image bands 5-4-3 in which the near infrared is mapped to red, red to green, and green to
blue.
Material Required
• Remote sensing images
• Software used to view and process the image: ENVI 5.3
Procedure:
• Go to https://earthexplorer.usgs.gov/
• Download the required images in band 3 (Green), 4 (Red) and 5 (NIR)
• Open the images in ENVI 5.3 software by selecting all the three images
• Now, search for layer stacking option
• Open layer stacking option and select the open bands
• Save the layer stacked file
• Open it on ENVI software
• View and analyse the different landscape features
Exercise 4: To select region of interest (ROI) in remote sensing images
What is ROI?
It is a specific area of an image or dataset that is of interest for analysis and manipulation.
Material Required
• Remote sensing images
• Software used to view and process the image: ENVI 5.3
Procedure:
• Open the layer stacked image as done in the last exercise
• Search for ROI tool on the menu bar
• After opening the tool, select the shape as polygon for marking the interest area
• Now click on your interest area and form a polygon
• After marking the area left click on finish polygon
• The ROI would be highlighted within the image
• Subset the region of interest and save the file
• Open the subset region with the marked interest area on ENVI 5.3
Exercise 5: Supervised and Unsupervised classification of region of interest
What is Classification?
Classification is the process of assigning spectral classes into information classes. Spectral
classes are groups of pixels that are uniform with respect to their brightness values in the
different spectral channels of data. Information classes are categories of interest that an analyst
attempts to identify in the image on the basis of his knowledge and experience about the area.
For example, a remote sensing image contains spectral signatures of several features present
on the ground in terms of pixels of different values. An interpreter or analyst identifies
homogeneous groups of pixels having similar values and labels the groups as information
classes such as water, agriculture, forest, etc. while generating a thematic map. When this
thematic information is extracted with the help of software, it is known as Digital Image
Classification.
Image Classification can be defined as a process of assigning all pixels in the image to
particular classes or themes based on spectral information represented by the digital numbers
(DNs).
Two General Approaches to Image Classification:
1. Supervised Classification: It is the process of identification of classes within a remote
sensing data. In this the inputs are directed by the user in the form of training data.
2. Unsupervised Classification: It is the process of automatic identification of natural groups
or structures within a remote sensing data. Both the classification approaches differ in the way
the classification is performed.
Material Required
• Remote sensing images
• Software used to view and process the image: ENVI 5.3
Procedure:
• Open the ROI subset image
• Open the ROI tool and assign homogenous spectral signatures under different classes
• Different classes like crop area, forest area, builtup land, water bodies, etc. are assigned as
per the knowledge and interest in the subset area
• Go to Supervised Classification in the toolbar
• Select Maximum Likelihood option under the maximum classification tool
• After selecting the option select the assigned classes
• Save the output file and apply the classification
• A colourful image with the assigned classes would be visible on the screen
• For unsupervised classification, go to unsupervised classification tool
• Under the tool select Isodata clustering
• After that select the minimum number of classes you would like to assign to your image
• Save the output file and apply the classification to see the results.
*Notes
Stages in the process of image classification:

Training Signature Decision Making

Evaluation

The image classification process consists of the following three stages:

1. Training: It is the process of generating spectral signature of each class. For example, a
forest class may be defined by minimum and maximum pixel values in different image
bands, thus defining a spectral envelope for it. This simple statistical description of the
spectral envelope is known as signature. Training can be carried out either by an image
analyst with guidance from his experience or knowledge (i.e. supervised training) or by
some statistical clustering techniques requiring little input from image analysts (i.e.
unsupervised training). There are no specific rules regarding the number of training sites
per class but it is advisable to take several training sites for each class to be mapped. If you
take very less number of training sites then it may be difficult to obtain a spectral signature
which truly represents that class and if you take large number of training sites then a
significant time may be getting wasted in collecting and evaluating signatures with
significantly improving the final signature.
2. Signature Evaluation: It is the checking of spectral signatures for their representativeness
of the class they attempt to describe and also to ensure a minimum of spectral overlap
between signatures of different classes.
3. Decision Making: It is the process of assigning all the image pixels into thematic classes
using evaluated signatures. It is achieved using algorithms, which are known as decision
rules. The decision rules set certain criteria. When signature of a candidate pixel passes the
criteria set for a particular class, it is assigned to that class. Pixels failing to satisfy any
criteria remain unclassified.
Difference between Supervised and Unsupervised Classification:
Aspect Supervised Classification Unsupervised Classification
Definition Classification guided by known training Classification without
samples predefined training data
Training Requires labelled training data for each No labelled training data
Data class required
Process Involves training the classifier using Uses clustering algorithms to
labelled samples group pixels
User Requires user to select training samples Minimal user intervention
Involvement during classification
Class Prior knowledge of class identities is Classes are discovered from
Information needed data patterns
Accuracy Often results in higher accuracy due to May have lower accuracy due
Assessment training to lack of training
Applicability Effective for identifying specific classes Suitable for exploratory data
analysis
Flexibility Less flexible, as predefined classes are More flexible, as classes are
used generated dynamically
Complexity Potentially more complex due to training Generally simpler as it relies
process on clustering