Develop a system that can identify flags embedded in photos of natural scenes. Develop a system that can segment a flag portion automatically accurately. Reduce the identification time and produce a good result. Apply Support Vector Machine(SVM) to generate the correct Result.

1. Welcome
to the Thesis Presentation
Department of Information and Communication Technology
Mawlana Bhashani Science and technology University

2. Presented by
Shahidul Islam IT06008
Nasima Akter IT06031
Shameem Hossain IT06034
Tomal Kumar Mozumder IT06042

3. Supervisor
Md. Zamilur Rahman
Lecturer
Dept of ICT
Mawlana Bhashani Science and technology University

4. Title
Flag segmentation, feature extraction &
identification using support vector
machine(SVM)

5. Overview
Objective
Introduction
Design and Implementation
Experiment Results
Limitation
Future Work
Conclusion

6. Objective
Develop a system that can identify flags embedded in
photos of natural scenes.
Develop a system that can segment a flag portion
automatically accurately.
Reduce the identification time and produce a good result.
Apply Support Vector Machine(SVM) to generate the
correct Result.

7. Introduction
Acquire still image
Image segmentation
Feature extraction
Methodology
Conclusion

8. Flag
• Flags are everywhere. They are mainly associated with
geographical regions, countries and nations, but if we look
around we will find them as symbols of many other walks of
life. A flag is basically a piece of material that is flown from a
mast or pole, but once we start adding coloring, designs and
emblems to that piece of cloth we have a work of art.

9. Flag detection procedure
Acquire still image
Feature extraction
Segmentation
SVM input
Output

10. System Architecture
Plain Flag
Image
Acquisition
Flag Image
Database
Construction
Flags
embedded
natural
scenes
acquisition
Image
Preprocessing
Segmentation
Feature
Extraction
SVM Based
Classification
Output

11. Database Construction
There are three kinds of flag images: plain, synthetic, and
natural-scene
To generate the synthetic flag images
• cropped each plain image in five sections, four of them representing ¼ of
the image, and the fifth one a central area of 30% of the total image.
• Six more samples were generated by applying Wiener Deconvolution.We
used Wiener Decovolution to simulate the effect of camera movement or
out of focus.
• From this set of 12 samples we created 24 more samples by using the
Linear Conformal Transform.

12. Data generation for Bangladesh Flag

13. Image preprocessing
Input
Image
RGB
Image
Gray
Image
Noise
reduction

14. Segmentation
Segmentation is the partitioning of a digital
image into multiple regions according to a
given criterion.
Many important segmentation techniques
are
• Segmentation by thresholding
• Edge based segmentation
• Region based segmentation
• Watershed segmentation
• In our proposed system we use Edge based segmentation

15. Edge-based segmentations
• Edge-based segmentations rely on edges found in an image by edge
detecting operators these edges mark image locations of discontinuities in
gray level, color, texture, etc. But the image resulting from edge detection
cannot be used as a segmentation result. Supplementary processing steps
must follow to combine edges into edge chains that correspond better with
borders in the image. The final aim is to reach at least a partial
segmentation that is, to group local edges into an image where only edge
chains with a correspondence to existing objects or image parts are present.

16. Feature Extraction
In most of the actual photos , it is not possible to
determine the number of objects, symbols, colors (besides
red and yellow) or color distribution.
For this reason the feature vector for this preliminary
design of the system considers only colors and
percentages of color participation.
All colors are clustered into nine colors using HSV color
values.

17. Support Vector Machine (SVM)
• SVM are newly introduced two-class maximum margin
classifiers that have become very popular because they
perform well in high dimensional feature spaces, avoid over
fitting, and have very good generalization capability. Support
vector machines (SVMs),a rigorous theoretical foundation, are
a set of related supervised learning methods. It is a linear
classifier that finds a hyper plane to separate two classes of
data (positive & negative).A good candidate for those
classification problems with high dimensional input space.

18. Experimental result
We presented an interactive flag
recognition system that identifies
flags embedded in photos of natural
scenes.

19. Related Work
Several researches have been done in
the field of image processing. Such as
Interactive Flag Identification using
Image Retrieval Techniques
Interactive Flag Identification Using a
Fuzzy-Neural Technique

20. Future work
For future work, we
plan to make
improvements in three
main areas:
segmentation, data
generation, and
feature extraction
using neural network.
Try to improve more
stable feature
extractions.

21. Conclusion
We presented an interactive flag recognition system that
identifies flags embedded in photos of natural scenes.
Since obtaining a large volume of flag images is time-
consuming and difficult, we generated a large number of
synthetic flag images from plain flag images.
The proposed system is an interactive system because of
two reasons. First, auto select the region of interest by
cropping the perimeter of the flag area.
Second, the system does not automatically identify the
flag to its respective country but lists the countries based
on the color similarity.