Object Based Image Analysis

International Centre for Integrated Mountain Development
Kathmandu, Nepal
Object-Based Image
Analysis (GEOBIA)
Kabir Uddin
GIS and remote sensing analyst
Email: Kabir.Uddin@icimod.org kabir.Uddin.bd@gmail.com

- The last few decades have
seen high levels of
deforestation and forest
degradation in the region.
- The changes in forest cover
due to growth dynamics,
management, harvest and
natural disturbances may
change the role and function of
forest ecosystems.
- Land use conversions from
forest to other land uses often
result in substantial loss of
carbon from the biomass pool.
What happening in the region

Development of baseline information
The use of satellites to monitor processes and trends at the global scale is
essential in the context of climate change

Land cover and land use change
detection using remote sensing and
geospatial data provides baseline
information for assessing the climate
change impacts on habitats and
biodiversity, as well as natural
resources, in the target areas.

• The process of sorting pixels into a number of data
categories based on their data file values
• The process of reducing images to information classes
Image Classification

Image classification techniques
There are different types of classification
procedures:
● Unsupervised
● Supervised
● Knowledge base
● Object base
● Others

Unsupervised classification
– The process of automatically segmenting an image
into spectral classes based on natural groupings found
in the data
– The process of identifying land cover classes and
naming them
Label
Bare
Agriculture
Forest
Grass
Water
ISODATA
Class 1
Class 2
Class 3
Class 4
Class 5
Class Names

Supervised classification
– the process of using samples of known identity (i.e., pixels
already assigned to information classes) to classify pixels of
unknown identity (i.e., all the other pixels in the image)

Object based image analysis
Before knowing more details about object
based image analysis lets use relatively
simple example

Part of the application
of GEOBIA lets extract
water from the high
resolution image.
When visually
inspecting data sets
looks all the blue pixel
should be water.
√
√
√
.
. .
.
.
.
.
. .
. . . .
.
.
.
.....
.
.
.
.

water from the high
resolution image.
When visually
should be water.
√
√
√
.
. .
.
.
.
.
. .
. . . .
.
.
.
.....
.
.
.
.
√
√

water from the high
resolution image.
When visually
should be water.
.
. .
.
.
.
.
. .
. . . .
.
.
.
.....
.
.
.
.

– Color Statistics
– Shape
– Texture
– Hierarchy
– Relations to...
...neighbor objects
...super-objects
...sub-objects

• Object-Based Image Analysis also called Geographic
Object-Based Image Analysis (GEOBIA) and it is a sub-
discipline of geoinformation science. Object – based
image analysis a technique used to analyze digital
imagery. OBIA developed relatively recently compared
to traditional pixel-based image analysis.
• Pixel-based image analysis is based on the information
in each pixel, object based image analysis is based on
information from a set of similar pixels called objects or
image objects.

Software for Object Based
Classification
eCognition/ Definiens
IDRISI
ERDAS Imagine
ENVI
MADCAT

IMAGINE Objective
Automated Feature Extraction
Reduce the labor, time and cost associated with intensive
manual digitization.
Repeatability
Reliably maintain geospatial content by re-using your
feature models.
Emulates Human Vision
True object processing taps into the human visual system
of image interpretation.

IMAGINE Objective
Image segmentation in IMAGINE Objective

IMAGINE Objective
Classified building in IMAGINE Objective

Segmentation and Segment-Based
Classification using IDRISI

Object based image analysis using
MadCat
MadCat (MApping Device - Change Analysis Tool) is
software mainly devoted to optimizing the production of
vector polygon based maps.
It is part of the GEOvis set of tools developed by FAO
The software also includes a module for change assessment
and analysis
MadCat version 3.1.0 Release (12.02.2009).
MadCat is FREE

Object based image analysis using
MadCat

eCognition/Definiens
• eCognition/Definiens software employs a flexible
approach to image analysis, solution creation and
adaption
• Definiens Cognition Network Technology® has been
developed by Nobel Laureate, Prof. Dr. Gerd Binnig
and his team
• In 2000, Definiens (eCognition) came in market
• In 2003 Definiens Developer along with Definiens
eCognition™ Server was introduced. Now,
Definiens Developer 8 with updated versions is
available

eCognition Developer
… is the development environment
for object-based image analysis.
eCognition Software Suite
eCognition Architect
… provides an easy-to-use front end
for non-technical professionals
allowing them to leverage eCognition
technology.
eCognition Server
… provides a processing
environment for the batch execution
of image analysis jobs.

eCognition Developer
• Develop rule sets
• Develop applications
• Combine, modify and
calibrate rule sets
• Process data
• Execute and monitor
analysis
• Review and edit results
Product Components:
 eCognition Developer client
 Quick Map Mode
 User Guide & Reference Book
 Guided Tours
 Software Development Kit (SDK)

eCognition Architect
• Combine, modify and
calibrate Applications
• Process data
• Execute and monitor
analysis
• Review and edit results
Product Components
 eCognition Architect client
 Quick Map Mode

eCognition Server
• Batch process data
• Dynamic load balancing
• Service oriented
architecture
• Highly scalable
Product components:
 eCognition Server
 HTML User Interface: Administrator
Console

Steps of land cover mapping
Legend development and classification
scheme
Data acquisition
Image rectification and enhancement
Field training information
Image segmentation
Generate image index
Assign rules
Draft land cover map
Validation and refining of land cover
Change assessment
Land cover map

Segmentation
• The first step of an eCognition image analysis is to cut the
image into pieces, which serve as building blocks for further
analysis – this step is called segmentation and there is a
choice of several algorithms to do this.
• The next step is to label these objects according to their
attributes, such as shape, color and relative position to other
objects.

Types of Segmentation
Chessboard segmentation
Chessboard segmentation is the
simplest segmentation available as it
just splits the image into square objects
with a size predefined by the user.

Quadtree based segmentation
Quadtree-based segmentation is similar to
chessboard segmentation, but creates
squares of differing sizes.
Quadtree-based segmentation, very
homogeneous regions typically produce
larger squares than heterogeneous
regions. Compared to multiresolution
segmentation,quadtree-based
segmentation is less heavy on resources.

Contrast split segmentation
Contrast split segmentation is similar to
the multi-threshold segmentation
approach.
The contrast split segments the scene
into dark and bright image objects
based on a threshold value that
maximizes the contrast between them.

Contrast split segmentation
Contrast split segmentation is similar to the
multi-threshold segmentation approach.
The contrast split segments the scene into
dark and bright image objects based on a
threshold value that maximizes the
contrast between them.

Spectral difference segmentation
Spectral difference segmentation lets you
merge neighboring image objects if the
difference between their layer mean
intensities is below the value given by the
maximum spectral difference. It is
designed to refine existing segmentation
results, by merging spectrally similar image
objects produced by previous
segmentations and therefore is a bottom-
up segmentation.

Multiresolution segmentation
Multiresolution Segmentation groups areas
of similar pixel values into objects.
Consequently homogeneous areas result
in larger objects, heterogeneous areas in
smaller ones.
The Multiresolution Segmentation
algorithm1 consecutively merges pixels or
existing image objects. Essentially, the
procedure identifies single image objects of
one pixel in size and merges them with
their neighbors, based on relative
homogeneity criteria.

Multiresolution
Segmentation, Parameters
Scale
• The value of the scale parameter
affects image segmentation by
determining the size of image
objects;
• Defines the minimum size of the
object through threshold value;
• The larger the scale parameter,
the more objects can be fused and
the larger the objects grow;

Image analysis assumptions
• Similar features will have similar spectral
responses.
• The spectral response of a feature is unique
with respect to all other features of interest.
• If we quantify the spectral response of a
known feature, we can use this information to
find all occurrences of that feature.

Generating arithmetic
Feature
 The Normalized Difference Vegetation Index (NDVI) is a standardized
index allowing to generate an image displaying greenness (relative
biomass)
 Index values can range from -1.0 to 1.0, but vegetation values typically
range between 0.1 and 0.7.
 NDVI is related to vegetation is that
healthy vegetation reflects very well in
the near infrared part of the spectrum.
 It can be seen from its
mathematical definition that the NDVI
of an area containing a dense
vegetation canopy will tend to positive
values (say 0.3 to 0.8) while clouds
and snow fields will be characterized
by negative values of this index.
NDVI = (NIR - red) / (NIR + red)

Land and Water Masks (LWM)
Index values can range from 0 to 255, but water
values typically range between 0 to 50
Water Mask = infra-red) / (green + .0001) * 100
(ETM+) Water Mask = Band 5) / (Band 2 + .0001) *
100

Comparing features using the
2D feature space plot

Investigation of classified land
cover

Object Based Image Analysis

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to Object Based Image Analysis (20)

Recently uploaded (20)

Object Based Image Analysis

Editor's Notes