본문 바로가기 주메뉴 바로가기
22대 대한민국 국회 국회정보길라잡이 국회의원검색
회원가입 로그인
HOME

정보검색

소장정보 검색

국회도서관 홈으로 정보검색 소장정보 검색
결과 내 검색 동의어 포함
결과 내 검색 동의어 포함

-

목차보기

표제지

목차

I. 서론 12

II. 지능형 휠체어의 보행자 인식 기술 15

2.1. 지능형 휠체어와 보행자 인식 기술 동향 15

2.2. 지능형 휠체어를 필요로 하는 사용자 특징 분석 19

III. 특징 추출 및 분류기 22

3.1. Haar-like특징 22

3.2. 기울기 히스토그램(HOG) 25

3.2.1. 기울기 값의 계산 25

3.2.2. 셀의 벡터화 26

3.1.3. 블록에 의한 정규화 27

3.3. Adaboost 알고리즘 29

3.4. Receiver Operating Characteristics(ROC) 곡선 31

3.5. 면적크기변화 Haar-like 특징 제안 34

IV. Haar-like와 HOG 연속기법 검증 36

4.1. 다양한 Haar-like와 HOG 연속기법 실험결과 40

4.2. 병원환경에서의 보행자 인식률 개선 79

V. 결론 92

참고문헌 94

Abstract 101

List of Tables

Table 1. A trend of study in intelligent wheelchair. 16

Table 2. The number of test Image using recognition rate. 36

Table 3. Experiments of Haar-like and HOG feature. 40

Table 4. Accuracy of single and cascade strong classifier of variety Haar-like and HOG using INRIA and MIT data. 73

Table 5. Experiment of Cascade Haar-like and HOG using hospital data. 79

Table 6. Accuracy of single and cascade strong classifier of variety Haar-like and HOG using hospital data. 86

List of Figures

Fig. 1. A field of Intelligent wheelchair. 15

Fig. 2. Prevalence of cerebral palsy at the age of 5 in South Korea. 19

Fig. 3. Types of Haar-like mask. 22

Fig. 4. The integral image evaluation. 24

Fig. 5. Examples of image(left) and integral image(right). 24

Fig. 6. Cell normalization using the histogram of oriented gradients; (a) input image, (b) HOG magnitude image (c) the cell of image (d) normalization by block. 26

Fig. 7. Feature extraction as it is done for each pixel (x,y) inside a cell/block: (a) computation of gradient orientation and magnitude (b) orientation interpolation and weighting (c) vote binning for the used directions. 27

Fig. 8. Confusion matrix and common performance metrics calculated from it. 31

Fig. 9. Examples of distribution for recognition data and ROC curve by change of threshold. 32

Fig. 10. Examples of ROC curve about 3 types distirbution for recognition data. 33

Fig. 11. Haar-like feature pixel samples : (a), (b), (c) - Horizontal Feature extraction, (d), (e), (f) - Vertical feature extraction. 34

Fig. 12. Feature extraction method using change of area size. 35

Fig. 13. The examples of pedestrian input image and non-pedestrian input image from INRIA and MIT. 37

Fig. 14. The examples of pedestrian input image and non-pedestrian input image from Hospital. 38

Fig. 15. Experiment method of Haar-like horizontal feature using INRIA data. 41

Fig. 16. Experiment Result of Haar-like horizontal feature using INRIA data(Pedestrian 500, Nonpedstrian 500). 42

Fig. 17. Experiment Result of Haar-like horizontal feature using INRIA data(Pedestrian 500, Nonpedstrian 1,000). 42

Fig. 18. Experiment method of Haar-like vertical feature using INRIA data. 43

Fig. 19. Experiment Result of Haar-like vertical feature using INRIA data(Pedestrian 500, Nonpedstrian 500). 44

Fig. 20. Experiment Result of Haar-like vertical feature using INRIA data(Pedestrian 500, Nonpedstrian 1,000). 44

Fig. 21. Experiment method of Haar-like horizontal & vertical feature using INRIA data. 45

Fig. 22. Experiment Result of Haar-like horizontal & vertical feature using INRIA data(Pedestrian 500, Nonpedstrian 500). 46

Fig. 23. Experiment Result of Haar-like horizontal & vertical feature using INRIA data(Pedestrian 500, Nonpedstrian 1,000). 46

Fig. 24. Experiment method of HOG feature using INRIA data. 47

Fig. 25. Experiment Result of HOG feature using INRIA data(Pedestrian 500, Nonpedstrian 500). 48

Fig. 26. Experiment Result of HOG feature using INRIA data(Pedestrian 500, Nonpedstrian 1,000). 48

Fig. 27. Experiment method of Cascade Haar-like horizontal and vertical feature using INRIA data. 49

Fig. 28. Experiment Result of Cascade Haar-like horizontal and vertical feature using INRIA data(Pedestrian 500, Nonpedstrian 500). 50

Fig. 29. Experiment Result of Cascade Haar-like horizontal and vertical feature using INRIA data(Pedestrian 500, Nonpedstrian 1,000). 50

Fig. 30. Experiment method of Cascade HOG and Haar-like horizontal feature using INRIA data. 51

Fig. 31. Experiment Result of Cascade HOG and Haar-like horizontal feature using INRIA data(Pedestrian 500, Nonpedstrian 500). 52

Fig. 32. Experiment Result of Cascade HOG and Haar-like horizontal feature using INRIA data(Pedestrian 500, Nonpedstrian 1,000). 52

Fig. 33. Experiment method of Cascade HOG and Haar-like vertical feature using INRIA data. 53

Fig. 34. Experiment Result of Cascade HOG and Haar-like vertical feature using INRIA data(Pedestrian 500, Nonpedstrian 500). 54

Fig. 35. Experiment Result of Cascade HOG and Haar-like vertical feature using INRIA data(Pedestrian 500, Nonpedstrian 1,000). 54

Fig. 36. Experiment method of variable area Haar-like feature using INRIA data. 55

Fig. 37. Experiment Result of variable area Haar-like feature using INRIA data(Pedestrian 500, Nonpedstrian 500). 56

Fig. 38. Experiment Result of variable area Haar-like feature using INRIA data(Pedestrian 500, Nonpedstrian 1,000). 56

Fig. 39. Experiment method of Haar-like horizontal feature using MIT data. 57

Fig. 40. Experiment Result of Haar-like horizontal feature using MIT data(Pedestrian 500, Nonpedstrian 500). 58

Fig. 41. Experiment Result of Haar-like horizontal feature using MIT data(Pedestrian 500, Nonpedstrian 1,000). 58

Fig. 42. Experiment method of Haar-like vertical feature using MIT data. 59

Fig. 43. Experiment Result of Haar-like vertical feature using MIT data(Pedestrian 500, Nonpedstrian 500). 60

Fig. 44. Experiment Result of Haar-like vertical feature using MIT data(Pedestrian 500, Nonpedstrian 1,000). 60

Fig. 45. Experiment method of Haar-like horizontal & vertical feature using MIT data. 61

Fig. 46. Experiment Result of Haar-like horizontal & vertical feature using MIT data(Pedestrian 500, Nonpedstrian 500). 62

Fig. 47. Experiment Result of Haar-like horizontal & vertical feature using MIT data(Pedestrian 500, Nonpedstrian 1,000). 62

Fig. 48. Experiment method of HOG feature using MIT data. 63

Fig. 49. Experiment Result of HOG feature using MIT data(Pedestrian 500, Nonpedstrian 500). 64

Fig. 50. Experiment Result of HOG feature using MIT data(Pedestrian 500, Nonpedstrian 1,000). 64

Fig. 51. Experiment method of Cascade Haar-like horizontal and vertical feature using MIT data. 65

Fig. 52. Experiment Result of Cascade Haar-like horizontal and vertical feature using MIT data(Pedestrian 500, Nonpedstrian 500). 66

Fig. 53. Experiment Result of Cascade Haar-like horizontal and vertical feature using MIT data(Pedestrian 500, Nonpedstrian 1,000). 66

Fig. 54. Experiment method of Cascade HOG and Haar-like horizontal feature using MIT data. 67

Fig. 55. Experiment Result of Cascade HOG and Haar-like horizontal feature using MIT data(Pedestrian 500, Nonpedstrian 500). 68

Fig. 56. Experiment Result of Cascade HOG and Haar-like horizontal feature using MIT data(Pedestrian 500, Nonpedstrian 1,000). 68

Fig. 57. Experiment method of Cascade HOG and Haar-like vertical feature using MIT data. 69

Fig. 58. Experiment Result of Cascade HOG and Haar-like vertical feature using MIT data(Pedestrian 500, Nonpedstrian 500). 70

Fig. 59. Experiment Result of Cascade HOG and Haar-like vertical feature using MIT data(Pedestrian 500, Nonpedstrian 1,000). 70

Fig. 60. Experiment method of variable area Haar-like feature using MIT data. 71

Fig. 61. Experiment Result of variable area Haar-like feature using MIT data(Pedestrian 500, Nonpedstrian 500). 72

Fig. 62. Experiment Result of variable area Haar-like feature using MIT data(Pedestrian 500, Nonpedstrian 1,000). 72

Fig. 63. ROC curve of INRIA data(Pedestrian 500, Nonpedstrian 500). 74

Fig. 64. ROC curve of INRIA data(Pedestrian 500, Nonpedstrian 1,000). 75

Fig. 65. ROC curve of MIT data(Pedestrian 500, Nonpedstrian 500). 76

Fig. 66. ROC curve of MIT data(Pedestrian 500, Nonpedstrian 1,000). 77

Fig. 67. Experiment Result of Cascade Haar-like horizontal and vertical feature using hospital data(Pedestrian 500, Nonpedstrian 500). 80

Fig. 68. Experiment Result of Cascade Haar-like horizontal and vertical feature using hospital data(Pedestrian 500, Nonpedstrian 1,000). 80

Fig. 69. Experiment Result of Cascade HOG and Haar-like horizontal feature using hospital data(Pedestrian 500, Nonpedstrian 500). 81

Fig. 70. Experiment Result of Cascade HOG and Haar-like horizontal feature using hospital data(Pedestrian 500, Nonpedstrian 1,000). 81

Fig. 71. Experiment Result of Cascade HOG and Haar-like vertical feature using hospital data(Pedestrian 500, Nonpedstrian 500). 82

Fig. 72. Experiment Result of Cascade HOG and Haar-like vertical feature using hospital data(Pedestrian 500, Nonpedstrian 1,000). 82

Fig. 73. Experiment Result of variable area Haar-like feature using hospital data(Pedestrian 500, Nonpedstrian 500). 83

Fig. 74. Experiment Result of variable area Haar-like feature using hospital data(Pedestrian 500, Nonpedstrian 1,000). 83

Fig. 75. Experiment Result of cascade HOG and variable area Haar-like feature using hospital data(Pedestrian 500, Nonpedstrian 500). 84

Fig. 76. Experiment Result of cascade HOG and variable area Haar-like feature using hospital data(Pedestrian 500, Nonpedstrian 1,000). 84

Fig. 77. Experiment Result of cascade variable area Haar-like and HOG feature using hospital data(Pedestrian 500, Nonpedstrian 500). 85

Fig. 78. Experiment Result of cascade variable area Haar-like and HOG feature using hospital data(Pedestrian 500, Nonpedstrian 1,000). 85

Fig. 79. Experiment result of cascade strong classifier using hospital data(Pedestrian 500, Nonpedstrian 500). 87

Fig. 80. Experiment result of cascade strong classifier using hospital data(Pedestrian 500, Nonpedstrian 1,000). 88

Fig. 81. Example of experiment result image about FalseNegative(FN). 89

Fig. 82. Example of experiment result image about FalsePositive(FP). 89

Fig. 83. 2step pedestrian recognition algorithm. 90

Fig. 84. Result image of string classifier I using variable Haar-like feature(Strong Classifier:500:1000, Th:25.0). 91

Fig. 85. Result image of string classifier II using HOG feature(Strong Classifier:500:1000, Th:25.0). 91

초록보기

In this paper, we suggest an advanced algorithm, to recognize pedestrian/non-pedestrian using 2 step cascade novel Haar-like and HOG feature, which apply Adaboost algorithm to make a strong classification from weak classifications. The novel Haar-like is a new Haar-like feature extraction method which include horizontal Haar-like feature and vertical Haar-like feature at once about each 2×2, 4×4, 8×8 pixel size. This feature extraction method shows a good result better than each horizontal Haar-like feature and vertical Haar-like feature as well as other cascade feature extraction methods by using INRIA images, MIT images and pedestrian images at hospital respectively.

And also based on our result of experiment, we proposed use 2 step pedestrian recognition algorithm. At first step, it recognize pedestrian through strong classifier I using novel Haar-like feature. The recognized data make pedestrian areas and it is tracked temporary pedestrian. Next step, data of pedestrian areas is recognized once more using strong classifier II. Finally, real pedestrian is detected and keep tracking by pedestrian.

This way to use 2 step cascade recognition method on hospital environment provides more reliable assessment of defect detection algorithms and then allows a better recongnition rate of pedestrian for autonomou driving of Intelligent wheelchair.

참고문헌 (52건) : 자료제공( 네이버학술정보 )

참고문헌 목록에 대한 테이블로 번호, 참고문헌, 국회도서관 소장유무로 구성되어 있습니다.
번호 참고문헌 국회도서관 소장유무
1 “Design and testing of a low-cost robotic wheelchair prototype”, Autonomous Robotics, vol. 2, pp.77-88, 1995. 미소장
2 “Adaptive shared control of a smart wheelchair operated by voice control”, Proc. IROS 97, vol. 2 , pp.622-626, 1997. 미소장
3 Pro-oxidant effects of δ-aminolevulinic acid (δ -ALA) on Chinese hamster ovary (CHO) cells 네이버 미소장
4 “Developing intelligent wheelchairs for the handicapped”, Assistive Technology and Artificial Intelligence, Lecture Notes in AI, vol. 1458, pp.150-178, 1998. 미소장
5 “Intelligent Wheelchair Moving among people based on their observations”, IEEE Journal. pp.1466–1471, 2002. 미소장
6 Monocular Pedestrian Detection: Survey and Experiments 네이버 미소장
7 "Multiple Pedestrians Detection and Tracking using Color Information from a Moving Camera," Journal of KIPS : Korea Information and Applications B, vol. 11-B, no. 3, pp.317-326, 2004. 미소장
8 "Shape-based Pedestrian Detection," Proc. of the IEEE Intelligent Vehicles Symposium 2000, pp.215-220, 2000. 미소장
9 "A multiple detector approach to low-resolution for pedestrian recognition," In Procs. IEEE Intelligent Vehicles Symposium 2005, pp.23-28, 2005. 미소장
10 "Pedestrian Detection from a Moving Vehicle," In Procs. of European Conference on Computer Vision, vol. 2, pp.37-49, 2000. 미소장
11 Walking pedestrian recognition 네이버 미소장
12 A Trainable System for Object Detection 네이버 미소장
13 "Boundary based feature selection," Technical report, KU. Leuven, 2002. 미소장
14 "Real-time object detection for "smart" vehicles," In IEEE International Conference on Computer Vision, pp.87-93, 1999. 미소장
15 Detecting Pedestrians Using Patterns of Motion and Appearance 네이버 미소장
16 A Face Detection Method Based on Adaboost Algorithm using New Free Rectangle Feature 소장
17 “Fast Fedestrian Detection Using a Cascade of Boosted Covariance Features”, IEEE Trans. Circuit and System for Video Technology, vol. 18, no. 8, pp.1140-1151, 2008. 미소장
18 Head gesture recognition for hands-free control of an intelligent wheelchair 네이버 미소장
19 “얼굴과 입 모양 인식을 이용한 지능형 휠체어 시스템”,정보과학회눈문지:소프트웨어 및 응용, 제36권, 제2호, pp.161-168, 2009. 미소장
20 study on the obstacle avoidance of the intelligent motorized wheelchair system 네이버 미소장
21 “휠체어 장착형 지능형 재활 로봇을 위한 칼라 비전 시스템”,전자공학회논문집, 제35권, 제11호, pp.75-87. 1998. 미소장
22 시선 인식을 이용한 자율 주행 휠체어 시스템 소장
23 User and social interfaces by observing human faces for intelligent wheelchairs 네이버 미소장
24 Building Local Safety Maps for a Wheelchair Robot using Vision and Lasers 네이버 미소장
25 An Omnidirectional Stereo Vision-Based Smart Wheelchair 네이버 미소장
26 “An Extended Set of Haar-like Features for Rapid Object Detection”,IEEEE ICIP 2002, vol. 1, pp.900-903, 2002. 미소장
27 Multi Features Combination for Pedestrian Detection 네이버 미소장
28 “Vision-based Bicycle Detection and Tracking using a Deformable Part Model and an EKF Algorithm”,IEEE Conference on Intelligent Transportation Systems, pp.19-22, 2010. 미소장
29 "Wheelchair Detection Using Cascaded Decision Tree," IEEE Transactions on Information Technology Biomedicine, vol. 14, no. 2, pp.292-300, 2010. 미소장
30 An introduction to ROC analysis 네이버 미소장
31 Editorial 네이버 미소장
32 Prevalence and lifetime healthcare cost of cerebral palsy in South Korea 네이버 미소장
33 A Research trend on the Motor Skills Training of the Children with Cerebral Palsy 네이버 미소장
34 Study on the Relationship between ctivity of Daily Living and Job Status of the Disabled from the Functional Limitation Perspectives 소장
35 Effect of Self Care Training(based on International Classification of Functioning, Disability and Health) on Functional Independence in the Young Children with Spastic Cerebral Palsy 소장
36 Verification of Validity on the Manual Ability Classification System in Children With Spastic Cerebral Palsy 소장
37 A study on the motor development delay of children with spastic cerebral palsy 소장
38 “학령기 중증 뇌성마비아동의 건강관련 삶의 질에 관한 연구”,코창능력개발지, 제9권, 제1호,pp.207-217, 2007. 미소장
39 Transactions : Classification of body types of male wheelchair users 소장
40 “휠체어사용 지체장애학생의 정적 인체치수 특성”,Korean Journal of Physical and Multiple Disablities, vol. 51, no. 1, pp.191-209, 2008. 미소장
41 Intravenous patient-controlled analgesia for postoperative pain management in patients with cerebral palsy 소장
42 뇌성마비아 놀이지도의 이론과 실제,국립특수교육원, 1996. 미소장
43 "Robust Real-Time Face Detection," Proc. 8th IEEE Int'l Conf. Computer Vision, Vol. 20, pp.1254-1259, July 2001. 미소장
44 “Haar-like Feature 변형을 이용한 기울어진 얼굴 검출”, 대한전자공학회 하계종합학술대회, 제31권 제1호, pp.987-998, 2008 미소장
45 “A Two-staged Approach to Vision-based Pedestrian Recognition Using Haar and HOG Features”, IEEE Intelligent Vehicles Symposium, pp. 554-559, 2008. 미소장
46 “A Practical Use of ROC Analysis to Assess the Performances of Defects Detection Algorithms”, Journal of Electronic Imaging, vol.17, no.3, pp. 1-15, 2008. 미소장
47 Receiver operating characteristic analysis for intelligent medical systems-a new approach for finding confidence intervals 네이버 미소장
48 "Pedestrian recognition using differential Haar-like feature based on Adaboost algorithm to apply intelligence wheelchair", Journal of Biomed. Eng., vol.31, pp. 480-485, 2010. 미소장
49 Fast Pedestrian Detection Using a Cascade of Boosted Covariance Features 네이버 미소장
50 Pedestrian Detection Using Boosted Features over Many Frames 네이버 미소장
51 An Experimental Study on Pedestrian Classification 네이버 미소장
52 Pedestrian detection by modeling local convex shape features 네이버 미소장

추천서가 (다양한 추천 자료를 만나보세요)

권호기사보기

권호기사 목록 테이블로 기사명, 저자명, 페이지, 원문, 기사목차 순으로 되어있습니다.
기사명 저자명 페이지 원문 기사목차