딥러닝에 기반한 족적수사 체계의 고도화 방안
An Advancement Plan of Footprint Investigation System Based on Deep Learning
1
조회
0
다운로드
1
인용
0
참고문헌
키워드
초록
본 논문은 족적 감정의 객관성과 신속성을 확보하기 위해 감정기법에 딥러닝을 도입하여 새로운감정기법 개발을 목표로 하였다. 실험환경은 범죄현장과 유사하게 구성하고 족적을 남긴 후, 휴대전화카메라로 촬영하였다. 인공지능은 3단계를 거쳐 보완하였다. 1단계에서 2개의 신발 모델을 대상으로 인공지능의 도입 가능성을 확인하였다. 2단계에서 5개의 신발 모델을 이용하여 7개의 환경에서 족적을 수집하고 딥러닝 모델을 적용하였다. 3단계에서 모델의 정확도 향상 및 양질의 데이터 확보를 위해 전처리 작업을 추가한 후, 딥러닝 모델을 적용하였다. 그 결과, 실험 대상 모델에 대해 정확도 0.9346을 달성하였다. 그러나 더 많은 신발 모델에서도 높은 정확도를 얻을 수 있도록 제안된 방법론의 지속적인고도화가 필요하다. 제안한 딥러닝 기반 족적검색 기법과 CCTV(폐쇄회로 텔레비전)의 영상을 연계하여용의자를 특정하는 등 과학기술을 도입한 족적수사 체계에 대한 발전방안을 제시하였다.
Abstract
This paper proposes a deep-learning-based classification scheme for footprint analysis to achieve the objectivity and swiftness of footprint analysis. Experiment environments resembling a crime scene are set up with footprints, and their photos are taken by smartphones. Artificial Intelligence (AI) are complemented with three steps. In the first step, the possibility of the introduction of AI to forensic science was verified with two kinds of shoe models. In the second step, the footprints were collected in seven environments with five kinds of shoe models, and a deep learning model was applied to them. In the third step, a preprocess was used in order to improve the accuracy of a model and obtain data in terms of quantity and quality, and then a deep learning model was applied to the data. As a result, the accuracy for the model for the experiment objects was 0.9346. However, a continuous enhancement for our methodology is needed to achieve a high accuracy for more shoe models. Thus, this paper suggested a proposal to enhance the footprint investigation framework that adopted the science technology to specify a suspect by associating the proposed deep learning-based footprint searching scheme with the video from CCTV (closed-circuit television).
참고문헌 (32)
- "초등학생 ‘장난’으로 밝혀진 용인캣맘 사건의 전말" (2015)[URL]
- "휴대전화 흘리고·족적 남기고’ 어설픈 도둑…경찰 구속" (2019)[URL]
- "Wood-Ridge man’s shoes presented as evidence at murder trial" (2017)[URL]
- "경찰청_족적 검색 처리 현황" (2020)[URL]
- "KOLAS 공인시험기관"[URL]
- "경찰, 발자국 수사전문가 양성한다…감정관 자격제 도입" (2022)[URL]
- K Książek, "Blood stain classification with hyperspectral imaging and deep neural networks", Sensors, 20(22), pp.1~3 (2022)
- P Tiwari, "Detection of subtype blood cells using deep learning", Cognitive Systems Research, (52), pp.1036~1044 (2018)
- T Bergman, "Automatic Classification of Bloodstains with Deep Learning Methods", KI-Künstliche Intelligenz, 36(2), pp.135~141 (2022)
- G Jaiswal, "An Analytical Review of Machine Learning and Deep Learning Approaches for Document Forgery Detection", Proceedings of 3rd International Conference on Machine Learning, Advances in Computing, Renewable Energy and Communication: MARC 2021, pp.629~638 (2022)
- P Yang, "Document Image Forgery Detection Based on Deep Learning Models", 2022 International Symposium on Electrical, Electronics and Information Engineering (ISEEIE), pp.36~41 (2022)
- 장요한, "석고분말의 종류와 물의 함량에 따른 족적 본뜨기에 관한 연구", 석사, pp.2017
- S Wiesner, "A comparative research of two lifting methods:electrostatic lifter and gelatin lifter", J Forensic Sci, (56), pp.S58~S62 (2011)
- 형사 사건의 물리적 증거 해석 향상을 위한 형태학적 흔적 연구, "서영일", 국립과학수사연구원, 1~58 (2017)
- "한국의 과학수사, ⑫ 범인 다녀간현장…‘신발의 지문’이 남았다" (2016)[URL]
- 경찰청, "2021 경찰백서", pp.381~ (2021)
- 양호진, "이미지 기술자를 활용한 범죄 족적의 매칭", 한국데이터정보과학회지, 33(2), pp.223~236 (2022), 한국데이터정보과학회
- "제3회 과학치안 아이디어 공모전시상식 개최" (2018)[URL]
- "과기정통부, ‘제4회 과학치안 아이디어 공모전’ 시상식 개최" (2018)[URL]
- 방향벡터를 이용한 족적이미지 자동검색 시스템, "최성호", 한국멀티미디어학회, 21~23 (2009)
- Y Tang, "Similarity and clustering of footwear prints", 2010IEEE International Conference on Granular Computing, pp.459~464 (2010)
- A Kortylewski, "Unsupervised footwear impression analysis and retrieval from crime scene data", Computer VisionACCV 2014 Workshops, (12), pp.644~658 (2015)
- SY Park, "The effect of image descriptors on the performance of classifiers of footwear outsole image pairs", Forensic Sci. Int’l, 331(111126), pp.7~9 (2022)
- M Hassan, "Restorable-inpainting:A novel deep learning approach for shoeprint restoration", Information Sciences, (600), pp.22~42 (2022)
- N Richetelli, "Classification of footwear outsole patterns using Fourier transform and local interest points", Forensic Sci Int’l, (275), pp.102~109 (2017)
- Z Wen, "Shoeprint image retrieval and crime scene shoeprint image linking by using Convolutional Neural Network and Normalized Cross Correlation", Science & Justice, 43, pp.9~450 (2023)
- S Crabbe, "3D-forensics-mobile highresolution 3D-Scanner and 3D data analysis for forensic evidence", Proceedings of the 9th Future Security, Security Research Conference, pp.16~18 (2014)
- F Liang, "Review of target detection technology based on deep learning", Proceedings of the 5th International Conference on Control Engineering and Artificial Intelligence, pp.132~135 (2021)
- A Paszke, "Pytorch: An imperative style, high-performance deep learning library", Advances in neural information processing systems, 32 (2019)
- 김남호, "글로벌기업들의 머신 비전 솔루션", 한국통신학회지(정보와통신), 39(10), pp.32~36 (2022)
- MC Chirodea, "Comparison of tensorflow and pytorch in convolutional neural network-based applications", 2021 13th international conference on electronics, computers and artificial intelligence (ECAI), pp.1~6 (2021), IEEE
- OC Novac, "Analysis of the application efficiency of TensorFlow and PyTorch in Convolutional Neural Network", Sensors, 22(22), pp.8872~ (2022)