닫기
Article Search
닫기

Original Article

Split Viewer

Int. J. Fuzzy Log. Intell. Syst. 2018; 18(2): 111-119

Published online June 25, 2018

https://doi.org/10.5391/IJFIS.2018.18.2.111

© The Korean Institute of Intelligent Systems

Building a HOG Descriptor Model of Pedestrian Images Using GA and GP Learning

Youngwan Cho, and Kisung Seo

Department of Computer Engineering, Seokyeong University, Seoul, Korea

Correspondence to :
Kisung Seo (ksseo@skuniv.ac.kr)

Received: May 17, 2018; Revised: June 9, 2018; Accepted: June 18, 2018

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

Go to

For detecting a pedestrian by using features of images, it is generally needed to establish a reference model that is used to match with input images. The support vector machine (SVM) or AdaBoost Cascade method have been generally used to train the reference pedestrian model in the approaches using the histogram of oriented gradients (HOG) as features of the pedestrian model. In this paper, we propose a new approach to match HOG features of input images with reference model and to learn the structure and parameters of the reference model. The Gaussian scoring method proposed in this paper evaluates the degree of feature coincidence with HOG maps divided with angle of the HOG vector. We also propose two approaches for leaning of the reference model: genetic algorithm (GA) based learning and genetic programming (GP) based learning. The GA and GP are used to search the best parameters of the gene and nonlinear function representing feature map of pedestrian model, respectively. We performed experiments to verify the performance of proposed method in terms of accuracy and processing time with INRIA person dataset.

Keywords: HOG model, Genetic algorithm, Genetic programming, Learning, Pedestrian detection

Conflict of Interest

Go to

No potential conflict of interest relevant to this article was reported.

Biographies

Go to

Youngwan Cho received the B.S., M.S., and Ph.D. degrees in electronic engineering from Yonsei University, Seoul, Korea, in 1991, 1993, and 1999, respectively. He worked as a Senior Research Engineer in the Control System Group at Samsung Electronics, Seoul, Korea, from 2000 to 2003. He was a visiting scholar at Department of Mechanical Engineering, Michigan State University from 2016 to 2017. He is currently working as an Associate Professor in the Department of Computer Engineering, Seokyeong University, Seoul, Korea. His research interests include fuzzy control theory and applications, intelligent control systems, machine learning, and robotics and automation.

E-mail: ywcho@skuniv.ac.kr

Kisung Seo received the B.S., M.S., and Ph.D. degrees in Electrical Engineering from Yonsei University, Seoul, Korea, in 1986, 1988, and 1993, respectively. He became Full-time Lecturer and Assistant Professor of Industrial Engineering in 1993 and 1995 at Seokyeong University, Seoul, Korea. He joined Genetic Algorithms Research and Applications Group (GARAGe) and Case Center for Computer-Aided Engineering & Manufacturing, Michigan State University from 1999 to 2002 as a Research Associate. He was also appointed Visiting Assistant Professor in Electrical & Computer Engineering, Michigan State University from 2002 to 2003. He was a Visiting Scholar at BEACON (Bio/computational Evolution in Action CONsortium) Center, Michigan State University from 2011 to 2012. He is currently Professor of Electronics Engineering, Seokyeong University. His research interests include deep learning, computer vision, evolutionary algorithm, genetic programming, evolutionary robotics, and evolutionary prediction for weather system.

E-mail: ksseo@skuniv.ac.kr

Article

Original Article

Int. J. Fuzzy Log. Intell. Syst. 2018; 18(2): 111-119

Published online June 25, 2018 https://doi.org/10.5391/IJFIS.2018.18.2.111

Copyright © The Korean Institute of Intelligent Systems.

Building a HOG Descriptor Model of Pedestrian Images Using GA and GP Learning

Youngwan Cho, and Kisung Seo

Department of Computer Engineering, Seokyeong University, Seoul, Korea

Correspondence to:Kisung Seo (ksseo@skuniv.ac.kr)

Received: May 17, 2018; Revised: June 9, 2018; Accepted: June 18, 2018

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

For detecting a pedestrian by using features of images, it is generally needed to establish a reference model that is used to match with input images. The support vector machine (SVM) or AdaBoost Cascade method have been generally used to train the reference pedestrian model in the approaches using the histogram of oriented gradients (HOG) as features of the pedestrian model. In this paper, we propose a new approach to match HOG features of input images with reference model and to learn the structure and parameters of the reference model. The Gaussian scoring method proposed in this paper evaluates the degree of feature coincidence with HOG maps divided with angle of the HOG vector. We also propose two approaches for leaning of the reference model: genetic algorithm (GA) based learning and genetic programming (GP) based learning. The GA and GP are used to search the best parameters of the gene and nonlinear function representing feature map of pedestrian model, respectively. We performed experiments to verify the performance of proposed method in terms of accuracy and processing time with INRIA person dataset.

Keywords: HOG model, Genetic algorithm, Genetic programming, Learning, Pedestrian detection

Fig 1.

Download
Figure 1.

Genetic configuration of the pedestrian HOG model.

The International Journal of Fuzzy Logic and Intelligent Systems 2018; 18: 111-119https://doi.org/10.5391/IJFIS.2018.18.2.111

Fig 2.

Download
Figure 2.

Calculation of the average score SP and SN.

The International Journal of Fuzzy Logic and Intelligent Systems 2018; 18: 111-119https://doi.org/10.5391/IJFIS.2018.18.2.111

Fig 3.

Download
Figure 3.

Effect of score variance to detect pedestrian.

The International Journal of Fuzzy Logic and Intelligent Systems 2018; 18: 111-119https://doi.org/10.5391/IJFIS.2018.18.2.111

Fig 4.

Download
Figure 4.

Angle score calculation process.

The International Journal of Fuzzy Logic and Intelligent Systems 2018; 18: 111-119https://doi.org/10.5391/IJFIS.2018.18.2.111

Fig 5.

Download
Figure 5.

GP score average calculation process.

The International Journal of Fuzzy Logic and Intelligent Systems 2018; 18: 111-119https://doi.org/10.5391/IJFIS.2018.18.2.111

Fig 6.

Download
Figure 6.

Threshold setting.

The International Journal of Fuzzy Logic and Intelligent Systems 2018; 18: 111-119https://doi.org/10.5391/IJFIS.2018.18.2.111

Fig 7.

Download
Figure 7.

Best GA model.

The International Journal of Fuzzy Logic and Intelligent Systems 2018; 18: 111-119https://doi.org/10.5391/IJFIS.2018.18.2.111

Fig 8.

Download
Figure 8.

Best GP model.

The International Journal of Fuzzy Logic and Intelligent Systems 2018; 18: 111-119https://doi.org/10.5391/IJFIS.2018.18.2.111

Table 1 . Configuration of GP terminal and function.

NodeArityDescription
R0Arbitrary values ranging from 0 to 1
angle score0~3400Score values of each angle
|Sin|1Absolute value of Sin function
|Cos|1Absolute value of Cos function
+2Sum of two channels
|−|2Absolute value of difference between two channels
*2Product of two channels
/2Division of two channels
Avg2Average of two channels
Wf12Weighted sum of two channels
Wf22Weighted sum of two channels
Max3Maximum value among three channels
Min3Minimum value among three channels

Table 2 . Results division table.

Predicted
YesNo
ActualYesTrue PositiveFalse Negative
NoFalse NegativeTrue Negative

Table 3 . GA parameter values.

ParameterValue
Population size50
Max generation100
Crossover rate0.9
Mutation rate0.1
Select methodTournament (size= 7)

Table 4 . GP parameter values.

ParameterValue
Population size1, 000
Max generation200
Crossover rate0.9
Mutation rate0.1
Select methodTournament (size= 7)
Initial depth6–8
Max depth17
Initial populationHalf and half

Table 5 . Experimental results of pedestrian detection performance.

Efficiency index
ACCPAGCSI
SVM0.8470.8010.724
GS0.7430.7520.592
GS + GA0.8900.8890.802
GS + GA + GP0.9270.9250.871

Table 6 . Comparison of execution time.

Execution time (FPS)
GS + GA44 ms
GS + GA + GP53 ms
SVM148 ms
GU + GA19 ms
GU + GA + GP28 ms

Article Tools

Metrics

Share this article on :

Related articles in IJFIS