This study has suggested an image analysis system based on the Deep Learning for CCTV pedestrian detection and tracing improvement and did experiments for objective verification by designing study model and evaluation model. The study suggestion is that if someone’s face did not be recognized in crime scene CCTV footage, the same pedestrian would be traced and found in other image data from other CCTV by using Color Intensity Classification method for clothes colors as body features and body fragmentation technique into 7 parts (2 arms, 2 legs, 1 body, 1 head, and 1 total). If one of other CCTV footage has recorded its face, the identity of the person would be secured. It is not only detection but also search from stored bulk storage to prevent accidents or cope with them in advance by cost reduction of manpower and a fast response. Therefore, CIC7P(Color Intensity Classification 7 Part Base Model) had been suggested by learning device such as Machine Learning or Deep Learning to improve accuracy and speed for pedestrian detection and tracing. In addition, the study has proved that it is an advanced technique in the area of pedestrian detection through experimental proof.

요 약
 Abstract
 1. 서 론
 2. 관련연구
  2.1 보행자 검출에 관한 배경이론
  2.2 영상분석기반 기술동향과 배경이론
  2.3 딥러닝 기반 보행자 검출 및 추적
 3. CCTV 보행자 검출 및 추적 알고리즘모델 설계
 4. 명도차분을 이용한 보행자 추적 설계
  4.1 명도 차분 영상화 및 차분 객체 컬러 모델데이터의 메타 추출
  4.2 차분 객체 컬러 모델을 이용한 이동 중 동일 보행자 판별
 5. 딥러닝 기반 CCTV 보행자 추적 설계
  5.1 SVM Cluster data mining model 설계
  5.2 학습훈련 메타 데이터베이스 설계
 6. 실험 및 평가 모델
  6.1 RNN 학습기를 사용하는 보행자 검출 알고리즘 실험 모델
  6.2 CNN 학습기를 사용하는 보행자 검출 알고리즘 실험 모델
  6.3 C1N7 추적 알고리즘 실험의 검증 결과
 7. 결론
 참 고 문 헌

• 박재규(숭실대학교 IT정책경학과 박사) | Jae-kyu Park
• 한경석(숭실대학교 경영학부 교수) | Kyeong-Seok Han 교신저자