PCA 복원과 HOG 특징 기술자 기반의 효율적인보행자 인식 방법 = An Efficient Pedestrian Recognition Method based on PCA Reconstruction and HOG Feature Descriptor

다국어 초록 (Multilingual Abstract)

In recent years, the interests and needs of the Pedestrian Protection System (PPS), which is mounted on the vehicle for the purpose of traffic safety improvement is increasing. In this paper, we propose a pedestrian candidate window extraction and unit cell histogram based HOG descriptor calculation methods. At pedestrian detection candidate windows extraction stage, the bright ratio of pedestrian and its circumference region, vertical edge projection, edge factor, and PCA reconstruction image are used. Dalal's HOG requires pixel based histogram calculation by Gaussian weights and trilinear interpolation on overlapping blocks, But our method performs Gaussian down-weight and computes histogram on a per-cell basis, and then the histogram is combined with the adjacent cell, so our method can be calculated faster than Dalal’s method. Our PCA reconstruction error based pedestrian detection candidate window extraction method efficiently classifies background based on the difference between pedestrian's head and shoulder area. The proposed method improves detection speed compared to the conventional HOG just using image without any prior information from camera calibration or depth map obtained from stereo cameras.

국문 초록 (Abstract)

최근 보행자의 교통안전 개선을 위한 목적으로 차량에 장착되는 보행자 보호 시스템(PPS, Pedestrian Protection System) 에 대한 관심과 요구가 증가하고 있다. 본 연구에서는 보행자 검출 후보 윈도우 추출과 셀(cell) 단위 히스토그램 기반의 HOG 특징 계산 방법을 제안하였다. 보행자 검출 후보 윈도우 추출은 주변밝기 비율체크, 수직방향 에지투영, 에지펙터(edge factor), 그리고 PCA(Principal Component Analysis) 복원 영상을 이용하였다. Dalal 의 HOG 는 겹침 블록 상의 모든 픽셀에 대해 가우시안 가중치와 삼선형보간에 의한 히스토그램 계산이 필요한데 반하여 제안하는 방법은 단위 셀마다 가우시안 가중 및 히스토그램을 계산하고 그것들을 인접 셀과 결합하므로 연산 속도가 빠르다. 제안하는 PCA 복원 에러 기반의 보행자 검출 후보 윈도우 추출은 보행자의 머리와 어깨 영역과의 차이를 기준으로 배경을 효율적으로 분류한다. 제안하는 방법은 카메라 컬리브레이션이나 스테레오 카메라를 이용한 거리 정보 없이도 영상만으로 전통적인 HOG 에 비하여 연산속도가 크게 개선된다.

참고문헌 (Reference)

1 G. Monteiro, "Vision-based pedestrian detection using Haar-like features" 16-20, 2006

2 D. Gavrila, "Vision-Based Pedestrian Detection : The PROTECTOR System" 13-18, 2004

3 P. Dollar, "The Fastest Pedestrian Detector in the West" 2010

4 H. Kataoka, "Symmetrical Judgment and Improvement of CoHOG Feature Descriptor for Pedestrian Detection" 13-15, 2011

5 A. Mazoul, "Street crossing pedestrian detection based on edge curves motion" 41 : 570-575, 2007

6 A. Broggi, "Shape-based Pedestrian Detection" 215-220, 2000

7 "SVMlight"

8 G. Lie, "STUDY ON PEDESTRIAN DETECTION AHEAD OF VEHICLE BASED ON MACHINE VISION" 570-575, 2007

9 G. Xu, "Real-time Pedestrian Detection Based on Edge Factor and Histogram of Oriented Gradient" 384-389, 2011

10 R. Benenson, "Pedestrian detection at 100 frames per second" 2903-2910, 2012

11 D. Cheda, "Pedestrian candidates generation using monocular cues" 7-12, 2012

12 A. Cosma, "Part-based pedestrian detection using HoG features and vertical symmetry" 229-236, 2012

13 P. Dollar, "Integral channel features" 1-11, 2009

14 "INRIA Person Dataset"

15 Dalal, N, "Histograms of oriented gradients for human detection" 886-893, 2005

16 Sang-Hun Kim, "High Performance Pedestrian Detection System Using A Cascade Algorithm Structure" 91-94, 2011

17 W. Xing, "Fast Pedestrian Detection Based on Haar Pre-Detection" 1 : 207-209, 2012

18 T. Kancharla, "Edge based Segmentation for Pedestrian Detection using NIR Camera" 1-6, 2011

19 P. Viola, "Detecting Pedestrians Using Patterns of Motion and Appearance" 153-161, 2005

20 "CAVIAR Test Case Scenarios"

21 G. Lie, "Algorithm Study for Pedestrian Detection Based on Monocular Vision" 83-87, 2006

22 Guilherme V. Carvalho, "A weighted image reconstruction based on PCA for pedestrian Detection" 2005-2011, 2011

23 Y. Fang, "A Shape-independent method for pedestrian detection with far-infrared images" 1679-1697, 2004