Call Girls Walvekar Nagar Call Me 7737669865 Budget Friendly No Advance Booking
OBJECT DECOMPOSITION BASED ON SKELETON ANALYSIS FOR ROAD EXTRATION
1. Presented by
Amit D. Chandak
Guided by
Prof. P. M. Pandit
Co-Guided by
Prof. A. S. Kale
JAWAHARLAL DARDA INSTITUTE OF ENGINEERING AND
TECHNOLOGY, YAVATMAL
DEPARTMENT OF ELECTRONICS AND TELECOMMUNICATION
ENGINEERING
A SEMINAR ON
2. • INTRODUCTION
• METHODOLOGY
• Data Preparation And Concept Definition
• Road Skeleton Model
• ANALYSIS
• Experiment With Complex Objects
• RESULTS
• ADVANTAGES AND DISADVANTAGES
• FUTURE WORK
• CONCLUSION
• REFERENCES
3. • One of the common challenges of information extraction from remote sensing
image is the “different objects with similar spectrum” phenomenon, which
often causes combined objects.
• For instance, the results of most of spectrum-based segmentation , different
ground true objects are combined into one image object, shown in the Figure
below.
4. • Given image object, its skeleton can be extracted by using Euclidean skeleton based
on a signed sequential Euclidean distance map.
• And then pruned with Bending Potential Ratio.
• All the vertexes of vector skeleton are named as Skeleton Points .
For example J, P & Q shown in the figure below
• The skeleton point having more than two adjacent point is called as Junction Point.
for example point J in above figure.
5. • The skeletionation of the given object image have to be done by keeping some
characteristics of the road in mind.
• There are four most important characteristics of road to be considered,
• SMOOTHNESS
• STABILITY
• HOMOGENITY
• NARROWNESS
6. • It not only means that the road centre line should be continuous , but also that
the Direction of the road should not change sharply , especially at the
junction points.
• As seen in the below figure since the branch angle α is greater than the branch
angle β. Hence branch B is more appropriate than branch A. Based on these
criterion the road will be constructed to be a smoothest way.
7. • It means that the width of road should not change sharply ,
otherwise its centre line should be segmented.
• Let the Si be the ith skeleton point, so the path can be presented
as a sequence,
S={Si /1<i<N} where,
N is the total number of the path skeleton points
8. • Let “r” be the max change of the path radius of the adjacent extrema.
• If
• Then,
9. • Stability index c means the max change rate in a region of support (ROC, the
region between the last extremum and next extremum).
• Stability index can be represented as,
• So it can represent the width change situation of a path.
• If c ≥T , the path should be segment at this position.
Where T is threshold and is always set to 1.5
10. • The radius sequence and its first order are shown in Figure (a) and (b)
respectively.
• The segmentation points calculated with the criterion mentioned above are
shown in Figure(c) with red hollow circle, and the corresponding points are
denoted with green circle, shown in Figure (b).
11. • It is another characteristic to reflect the stability of road skeleton.
• The variance of radius sequence of skeleton segment can
represent homogeneity well.
• With this criterion, skeleton segments are selected as road line,
as shown in Figure shown below.
12. • It is observed from all kinds of remote sensing images, there
are many strip objects, such as rivers and cultivated lands.
• These objects have similar shape characteristics and skeletons
with road objects.
• But the width of road is limited, always no more than 100ft.
• So we can calculated the average radius of radius sequence.
• With this criterion, some non-road skeleton segments will be
filtered out.
13. Experiment with Complex Objects:-
• To validate the effectiveness of the method, a complex object is taken. It has
more than ten holes and consists of roads, houses and some grass lands.
• The skeletons are also complex, shown in the left sub-figure of Figure
• The result of skeleton decomposition with the method mentioned earlier is
shown in the middle sub-figure, where road parts are marked with red line
With the road parts of skeletons and their average width, the road regions, is
shown in the right sub-figure.
14. • To check the above features we use following formula:
where,
Lmr denote the length of matched reference
Lr the length of reference
Lme length of matched extraction
Le length of extraction
Lur length of unmatched reference
15. Here are two examples of rural and suburban regions respectively are taken into
experiment
• Table shows the average performance
Rural Region Sub Urban Region
16. • ADVANTAGES:
THE CORRECTNESS, COMPLETENESS AND THE QUALITY OF ANY ROAD CAN BE
CALCULATED FROM ITS REMOTE IMAGE ONLY.
IT SAVES THE MONEY, TIME AND LABOR WHICH NEEDS TO BE APPLIED FOR THE
ROAD EXTRACTION PROCESS.
• DISADVANTAGES:
IF THE WEATHER CONDITION OVER THE AREA WHERE ROAD EXTRACTION IS TO BE
CARRIED OUT ARE NOT GOOD THEN WE WILL NOT GET AN ACCURATE IMAGE
RESULTING INTO POOR AND NOT SO ACCURATE RESULTS
17. Future work will focus on two aspects:
(1) Inference method to decompose true ground objects from
combined objects more completely and correctly.
– The first one means to differ a potential road segment based on
its context shape information.
(2) Road centre line smoothness and perceptual organization.
– The second is to get smoother and complete road centre lines.
18. • With the characteristics of road skeleton, namely smoothness,
stability, homogeneity and narrowness, road can be extracted from
combined image object.
• Since image objects are always produced by image segmentation
based on spectrum and texture characteristic, it is actually a method
dealing with shape information and the texture characteristic.
• As a consequence, the performance of road extraction is improved.
19. 1) W. Shen, et al., "Skeleton growing and pruning with bending potential
ratio," Pattern Recognition., vol. 44, pp. 196 209, 2011.
2) W.-P. Choi, et al., "Extraction of the Euclidean skeleton based on a
connectivity criterion," Pattern Recognition,vol. 36, pp. 721- 729, 2003
3) S. Krinidis and V. Chatzis, "A Skeleton Family Generator via Physics-
Based Deformable Models," Image Processing, IEEE Transactions on,
vol. 18, pp. 1-11, 2009.
4) D. Ward and G. Hamarneh, "The GroupWise Medial Axis Transform
for Fuzzy Skeletonization and Pruning," Pattern Analysis and Machine
Intelligence, IEEE Transactions on, vol. 32, pp. 1084-1096, 2010.
5) C.Wiedemann, et al., Eds., Empirical evaluation of automatically
extracted road axes (CVPR Workshop on Empirical Evaluation
Methods in Computer Vision. California