Scribd
Upload
18 views3 pages

RGIS603 Assignment 2

Uploaded by

sara22.alhasan
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
18 views3 pages

RGIS603 Assignment 2

Uploaded by

sara22.alhasan
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 3

RGIS603: Digital Image Processing in RS Name: Sara Alhasan

Id: 201950031

Assignment 3: Image classification

Read the attached article (Salah 2017) and answer the following questions.

1) What are the different steps of the remote sensing classification


process?

The major steps of the remote sensing classification process to produce Land Use and
Land Cover (LULC) maps include:

1. Choice of a suitable classification system


2. Design image classes such as urban, agriculture, water areas, etc.
3. Conduct field surveys and collect ground information
4. Image preprocessing for the enhancement of geometric and radiometric qualities
of satellite imagery
5. Feature extraction and selection
6. Selection of training samples
7. Image classification

2) Remote sensing data varies in spatial, spectral, radiometric and


temporal resolution. According to this article, which parameter is the
most important and how does it influence the classification process?

Spatial resolution is identified as the most important factor that affects classification details and
influences the selection of a classification algorithm. The spatial resolution of remotely sensed
data determines the level of detail and clarity in the imagery, which directly impacts the ability to
distinguish between different land cover classes. Higher spatial resolution data, such as HR data
from sensors like IKONOS and SPOT 5, are more suitable for detailed classification tasks at a
local level. On the other hand, coarse spatial resolution data, like AVHRR, MODIS, and SPOT
Vegetation, are more appropriate for global-scale analyses. The choice of spatial resolution is
critical in determining the accuracy and effectiveness of the classification process
3) Classification algorithms can be categorized based on different criteria.
Name two of these criteria and give the categories of classification algorithm
that each one results in.

Two common criteria and their resulting categories of classification algorithms are:

1. Training Sample:
 Supervised: Uses training sets to classify pixels of unknown identity. Examples
include Maximum Likelihood (ML), Parallelepiped (PP), and Minimum Distance
(MD).
 Unsupervised: Divides pixels into a number of classes based on natural groupings
without prior knowledge. Examples include K-means and ISODATA.

2. Assumptions on Data Distribution:


 Parametric: Based on the assumption of a specific data distribution, such as Gaussian
distribution.
4) Improvements in remote sensing data resolution and in technology in
recent years, lead to the creation of ‘Advanced classification algorithms’.
Name one of the advanced algorithms listed in the article and describe
briefly (one paragraph).

One of the advanced classification algorithms mentioned in the article is the Support Vector
Machines (SVMs). SVMs are powerful machine learning algorithms that are widely applied in
image classification tasks. SVMs work by finding the optimal hyperplane that best separates
different classes in a high-dimensional feature space. They are effective in handling complex and
non-linear relationships in data, making them suitable for tasks where traditional classifiers may
struggle. SVMs have been shown to outperform conventional classifiers in various remote
sensing applications, especially when incorporating non-spectral data into the classification
process. Their ability to handle high-dimensional data and complex decision boundaries makes
SVMs a valuable tool for accurate and robust image classification

5) What are hybrid classifiers?

Hybrid classifiers are classification systems that combine multiple individual classifiers to
improve overall classification performance. These systems leverage the strengths of different
classifiers to compensate for each other's weaknesses and enhance the accuracy and robustness
of the classification process. Hybrid classifiers can be categorized into two main types:

1) Classifier Ensembles (CE): CE involves combining multiple instances of the same base
classifier, such as multiple decision trees or neural networks, to create a diverse set of
classifiers. The final classification decision is often made through voting or averaging the
outputs of individual classifiers.

2) Multiple Classifier Systems (MCS): MCS combine different types of classifiers, such as
neural networks, support vector machines, and decision trees, to form a more
comprehensive classification system. Each classifier in an MCS operates independently,
and their outputs are combined using various fusion strategies to make the final
classification decision.

Article reference:

Salah, M. (2017). A survey of modern classification techniques in remote sensing for


improved image classification. Journal of Geomatics, 11(1), 21.

You might also like