Automatic Craniofacial Structure Detection on
Download
Report
Transcript Automatic Craniofacial Structure Detection on
Ajay Kumar, Member, IEEE, and David Zhang, Senior
Member, IEEE
Introdution
This paper proposes a new bimodal biometric
system using feature-level fusion of hand shape and
palm texture.
Fingerprint
iris
palmprint
voice
feature subset
Proposed Work
Feature subset selection helps to identify and remove
much of the irrelevant and redundant features.
improving the performance of hand-shape
recognition by exploring new features
investigating the palmprint recognition in
frequency domain using popular discrete cosine
coefficients
Proposed Work
Evaluating the performance gain from the feature
subset selection and features combination.
Bayes
support vector machines(svm)
feed-forward neural networks(FFN)
K -nearest neighbor (K-NN)
decision-tree
Proposed Work
AUTOMATED EXTRACTION OF HAND-SHAPE AND
PALMPRINT IMAGES
extraction of these two images
1. a binary image depicting hand-shape
2.a gray-level image containing palmprint texture
AUTOMATED EXTRACTION OF HAND-SHAPE AND
PALMPRINT IMAGES
The magnitude of thresholding limit η is computed by
maximizing the object function Jop(η)
where the numbers of pixels in class 1 and 2 are
represented by P1(η)and P2(η) ,µ1(η) and µ2(η) are the
corresponding sample mean.
AUTOMATED EXTRACTION OF HAND-SHAPE AND
PALMPRINT IMAGES
The orientation of each of the binarized image P(x,y)
is estimated by the parameters of the best-fitting
ellipse is estimated by the parameters of the bestfitting ellipse .
The counterclockwise rotation of major axis relative to
the normal axis is used to approximate the orientation
θ
AUTOMATED EXTRACTION OF HAND-SHAPE AND
PALMPRINT IMAGES
ρ11 , ρ22 , and ρ12 are the normalized second-order moments of pixels
in the image P(x,y)
(cx,cy) denote the location of its centroid
AUTOMATED EXTRACTION OF HAND-SHAPE AND PALMPRINT
IMAGES
Remove isolated foreground blobs or holes by
morphological preprocessing
The distance transform of every pixel in the hand-shape
image is used to estimate the center of palmprint.
The location (u,v)of the pixel with highest magnitude of
distance transform is obtained.
All the gray-level pixels from the original hand image, in a
fixed-square region, centered at (u,v) and oriented along ,
are used as the palmprint image.
AUTOMATED EXTRACTION OF HAND-SHAPE AND
PALMPRINT IMAGES
Palmprint Features
The discrete cosine transform(DCT) that maps a Q × R
spatial image block Ω to its values in frequency
domain (fig4)
The feature vector from every palmprint image is
formed by computing standard deviation of these
significant DCT coefficients in each of these blocks.
Palmprint Features
Hand-Shape Features
We investigated seven such shape properties, i.e.,
perimeter (f1),solidity (f2) , extent (f3)
eccentricity (f4) , x – y position of centroid relative to
shape boundary (f5 - f6), and convex area (f7)to
improve the success of prior methods.
In addition, 16 geometrical features from the hand shape,
as proposed in prior work were also obtained; four finger
length (f8 – f11), eight finger width (f12 – f19), palm width
(f20), palm length (f21), hand area (f22), and hand length
(f23).
CLASSIFICATION SCHEMES
naive Bayes
normal- it traditionally makes the assumption that the feature values
are normally distributed
estimation - The distribution of features was also estimated using
nonparametric kernel density estimation
multinomial
K -nearest neighbor (k-NN) - minimum Euclidean
distance between the query feature vector and all the
prototype training data
support vector machine (SVM
CLASSIFICATION SCHEMES
The feed-forward neural network (FFN)
- a linear activation function for the last layer the
sigmoid activation function was employed for other
layers
The C4.5 decision tree
logistic model tree (LMT)
EXPERIMENTS AND RESULTS
EXPERIMENTS AND RESULTS
EXPERIMENTS AND RESULTS
EXPERIMENTS AND RESULTS
CONCLUSION
Hand-shape and palmprint image segmentation, and the combination
of features from these two images, has shown to be useful in achieving
higher performance.
Our experimental results in Section V suggested the usefulness of
shape properties (e.g., perimeter, extent, convex area) which can be
effectively used to enhance the performance in hand-shape recognition.
Although more work remains to be done, our results to date indicate
that the combination of hand-shape and palmprint features constitutes
a promising addition to the biometrics-based personal recognition
systems.
THE END
thanks for your listening