PowerPoint 簡報

專題研討---心得報告
Face Recognition System with Genetic
Algorithm and ANT Colony Optimization
International Journal of Innovation, Management and Technology,
Vol. 1, No. 5, December 2010
S.Venkatesan
and
Dr.S.Srinivasa Rao Madane
報告者:黃柏翎
指導老師:鄭朝榮、王文彥
課程指導老師:蘇德仁
2017/7/14
1
Outline (1/2)
• ABSTRACT
• INTRODUCTION
• HEAD POSE ESTIMATION
– Face image Acquisition
– Filtering and Clipping
• PROPOSED ANT COLONY OPTIMIZATION
GENETIC ALGORITHM
2017/7/14
2
Outline (2/2)
• ACOG ALGORITHM
• EXPERIMENTAL RESULTS
• CONCLUSION
2017/7/14
3
ABSTRACT
• A novel face recognition system to detection
faces in images.
• This system is caped with three steps:
– Initially preprocessing methods are applied on the
input images.
– Consequently face features are extracted from the
processed image by ANT Colony Optimization.
– Recognition by Genetic Algorithm.
2017/7/14
4
INTRODUCTION
• Face recognition is the process of automatically
detection whether two faces are the same person.
• Face recognizers, like our detectors, have been
trained using novel statistical learning methods, to
deal with these diverse factors and provide accurate
results on real-world data.
2017/7/14
5
HEAD POSE ESTIMATION(1/3)
• Their face detection technology not only locates
faces, but it also estimates the 3D head pose.
• Detect one set of landmarks in frontal and
semi-profile faces.
• Detect a second set of landmarks in full-profile
faces.
2017/7/14
6
HEAD POSE ESTIMATION(2/3)
• Face Image Acquisition:
– To collect the face images, a scanner has been used.
– Saved into various formats such as Bitmap, JPEG, GIF
and TIFF.
2017/7/14
7
HEAD POSE ESTIMATION(3/3)
• Filter and Clipping
– Filter has been used for fixing these problems.
– Clipped to obtain the necessary data.
2017/7/14
8
PROPOSED ACOG
• The ACO system contains two rules:
– Local pheromone update rule, which applied whilst
constructing solutions.
– Global pheromone update rule, which applied after all
ants constrict a solution.
2017/7/14
9
ACOG ALGORITHM (1/8)
• ACOG is differing from previous algorithm.
• It consists of two main sections:
– Initialization
– Main loop
(Genetic Programming is used in the second sections)
2017/7/14
10
ACOG ALGORITHM (2/8)
• Initialization:
–
–
–
–
–
–
–
2017/7/14
variable
states
function
input
output
input trajectory
output trajectory
11
ACOG ALGORITHM (3/8)
• While termination conditions not meet do
Construct Ant Solution:
– Apply Local Search
– Best Tour check:
• If there is an improvement, update it.
– Update Trails:
• Evaporate a fixed proportion of the pheromone on each read.
• For each ant perform the “ant-cycle” pheromone update.
2017/7/14
12
ACOG ALGORITHM (4/8)
• Initial Population:
– Generate randomly a new population of
chromosomes of size N: x1, x2….xn.
– Assign the crossover probability Pc and the mutation
probability Pm.
2017/7/14
13
ACOG ALGORITHM (5/8)
f (n)  1 
2017/7/14

f ( x, y )  f n ,t ( x, y )
( x , y )W
Bmax  xSize  ySize
14
ACOG ALGORITHM (6/8)
• Selection:
– Select a pair of chromosomes for mating use the
roulette wheel selection procedure.
– To select highly fit of chromosome for mating a
random number is generated in the interval[0, 100].
2017/7/14
15
ACOG ALGORITHM (7/8)
• Crossover:
– To chooses a crossover point where two parent
chromosomes break and then exchanges the
chromosomes parts after that point.
• Single point
• Two point
• uniform
2017/7/14
16
ACOG ALGORITHM (8/8)
• Mutation:
– To set of mutation rate Pm.
– Random number to flip value from 0 to 1 or 1 to 0.
2017/7/14
17
EXPERIMENTAL RESULTS (1/2)
• From this Table:
• The results in next page.
2017/7/14
18
EXPERIMENTAL RESULTS (2/2)
• Therefore the efficiency of the Face Recognition
System by using Genetic and Ant Colony
Optimization Algorithm is Best than other methods.
2017/7/14
19
CONCLUSION
• In this paper, this method is more robust suitable
for low resolution.
2017/7/14
20