딥러닝 신경망모형을 이용한 실시간 교통정보수집 연구
Acquiring Real Time Traffic Information Using Deep Learning Neural Networks
0
조회
0
다운로드
5
인용
0
참고문헌
키워드
초록
비디오 기반의 실시간 교통정보수집 시스템은 자율주행을 포함한 지능형교통체계의 주요 서브시스템중 하나이며, 우수한 예측(추정)력 때문에 핵심 알고리즘으로 인공신경망 모형이 주로 사용되고 있다. 특히, 최근 과학 학술지 네이처에 구글 딥마인드의 알파고가 공개되면서 딥러닝과 인공지능에 대한 전 세계적인 관심이 집중되고 있다. 최근까지 가장 널리 사용된 인공신경망모형은 역전파 모형이었으나 최근 컴퓨터 하드웨어 및 알고리즘의 새로운 개발과 더불어 딥러닝 알고리즘을 이용한 딥인공신경망 모형에 대한 관심이 높아지고 있다. 그러나 아직 딥러닝 인공신경망에 대한 연구는 초기단계에 있는 실정이며 적용분야도 극히 한정되어 있다. 본 연구에서는 실시간 도로교통정보 수집을 위한 영상처리분야에서 딥러닝 알고리즘을 활용하고 실험결과를 통하여 향후 교통정보수집‧분석 분야에서의 딥러닝 인공신경망 모형의 적용가능성을 살펴보고자 한다. 특히, 본 연구에서 도출한 딥러닝 인공신경망모형의 실험결과를 기존 역전파 모형에 의한 실험결과와 비교하여 모형의 우수성을 비교‧평가하고자 한다.
Abstract
Real-time automatic video-based traffic surveillance system is one of the main subsystems of Intelligent Transportation Systems (ITS) which are advanced applications in order to provide innovative services relating to traffic management and enable various users to be better informed and make safer. Real-time automatic video-based traffic surveillance system is one of the main subsystems of Intelligent Transportation Systems (ITS) which are advanced applications in order to provide innovative services relating to traffic management and enable various users to be better informed and make safer. Normally, artificial neural networks have been used in real-time applications for collecting traffic information because of their remarkable performance in terms of predictive accuracy and real-time computing. Especially, the recent publication of a paper in the journal Nature and the success of Google DeepMind’s world-beating Go program AlphaGo caused significant increases of the public interest in artificial intelligence. Even though various types of learning models have been proposed, the Backpropagation has been the most popular neural network model. However, the Deep learning neural network model has attracted much interest recently. In this study, the Deep leaning neural network will be applied in image processing techniques for the real-time traffic information collection to investigate the performance of Deep learning algorithm in traffic engineering fields. In addition, the experimental results from Deep learing algorithm will be compared with the results from Backpropagation model to evaluate and propose better performing model.
참고문헌 (13)
- G. E. Hinton, "Reducing the dimensionality of data with neural networks", Science, 313(5786), pp.504~507 (2006)
- A. Krizhevsky, "Imagenet classification with deep convolutional neural networks", Advances in Neural Information Processing, 25 (2012), MIT Press
- D. E. Rumelhart, "Parallel distributed processing, Vol.1", pp.45~76 (1986), MIT Press
- D. E. Rumelhart, "Parallel distributed processing, Vol.1", pp.318~364 (1986), MIT Press
- An empirical study of learning speed in backpropagation networks, "S. E. Fahlman", Canegie Mellon University, ~, pp.Technical Report CMU-CS-88-162 (1988)
- D. Kim, "Standard and advanced Backpropagation models for image processing application in traffic engineering", ITS Journal, 7(3-4), pp.199~211 (2003)
- R. B. Palm, "Prediction as a candidate for learning deep hierarchical models of data", MSC, pp.2012
- G. E. Hinton, "A fast learning algorithm for deep belief nets", Neural Comput, 18(7), pp.1527~1554 (2006)
- Experiments with a restricted Boltzmann machine, "T. Amaral", NNIG, ~, pp.NNIG Technical Report No. 1 (2012), 2012
- D. Kim, "Creating a new input vector to improve the prediction performance of SVM classification", Journal of Basic and Applied Research International, 7(4), pp.232~238 (2015)
- D. Kim, "Performance Improvement by Modification of Learning Algorithm of Multilayer Feed-forward Neural Network Model", INFORMATION-AN INTERNATIONAL INTERDISCIPLINARY JOURNAL, 1(A)(4), pp.481~494 (2013)
- D. Kim, "Pre-processing of inputs to a neural network model for better performance in traffic scene analysis", CIVIL ENGINEERING AND ENVIRONMENTAL SYSTEMS, 27(1), pp.23~31 (2010)