[논문]A Novel Road Segmentation Technique from Orthophotos Using Deep Convolutional Autoencoders

Sameen, Maher Ibrahim; Pradhan, Biswajeet

doi:10.7780/kjrs.2017.33.4.8

[국내논문] A Novel Road Segmentation Technique from Orthophotos Using Deep Convolutional Autoencoders 원문보기

대한원격탐사학회지 = Korean journal of remote sensing, v.33 no.4, 2017년, pp.423 - 436

Sameen, Maher Ibrahim (Department of Civil Engineering, University Putra Malaysia) , Pradhan, Biswajeet (Department of Civil Engineering, University Putra Malaysia)

Abstract ▼ AI-Helper

This paper presents a deep learning-based road segmentation framework from very high-resolution orthophotos. The proposed method uses Deep Convolutional Autoencoders for end-to-end mapping of orthophotos to road segmentations. In addition, a set of post-processing steps were applied to make the model outputs GIS-ready data that could be useful for various applications. The optimization of the model's parameters is explained which was conducted via grid search method. The model was trained and implemented in Keras, a high-level deep learning framework run on top of Tensorflow. The results show that the proposed model with the best-obtained hyperparameters could segment road objects from orthophotos at an average accuracy of 88.5%. The results of optimization revealed that the best optimization algorithm and activation function for the studied task are Stochastic Gradient Descent (SGD) and Exponential Linear Unit (ELU), respectively. In addition, the best numbers of convolutional filters were found to be 8 for the first and second layers and 128 for the third and fourth layers of the proposed network architecture. Moreover, the analysis on the time complexity of the model showed that the model could be trained in 4 hours and 50 minutes on 1024 high-resolution images of size $106{\times}106pixels$, and segment road objects from similar size and resolution images in around 14 minutes. The results show that the deep learning models such as Convolutional Autoencoders could be a best alternative to traditional machine learning models for road segmentation from aerial photographs.

Keyword

AI 본문요약
AI-Helper

* AI 자동 식별 결과로 적합하지 않은 문장이 있을 수 있으니, 이용에 유의하시기 바랍니다.

문제 정의

This paper proposes a deep convolutional Autoencoder network for segmenting road objects from very high resolution orthophotos. Three optimization methods including grid search, random search, and Bayesian method were studied to obtain the best network for road segmentation.

제안 방법

They held that prior data is not readily available in all cases and thus limits the use of road network extraction to only a few countries where required training data is available. They proposed a new system with three modules: road center detection through probabilistic method, road shape extraction through binary balloon algorithm, and road network formation through graph theory. Unlike the method proposed by Mnih and Hinton (2010) were even with the use of GPU the time required for road detection was of many hours, Unsalan and Sirmacek (2011) were able to not only detect road but also generate road network graphs through CPU in less than an hour.
The proposed model recognizes patterns and translates them into states so that their tracking behaviors can be captured. Although their experiments showed that the proposed model was more accurate than traditional techniques, their method required road point initialization. Saito et al.
The remaining of this paper is organized as follows. Section 3 presents the methodology of the study, which includes the overall flowchart of the proposed network, data preprocessing and preparation, hyperparameter optimization, overfitting mitigation, road network extraction and accuracy assessment method. Section 4 discusses the model performance, demonstrates the results of road segmentation, sensitivity analysis, and discusses the computing complexity of the model.
Before orthorectification process, the acquired photos and the generated DEM were coregistered through 25 points carefully selected at road intersections and building corners. Finally, the geometric distortions in the photos were removed by applying the rational polynomial coefficients in ArcGIS after creating them from manually selected points used for coregistration.
20GHz processor, 16 GB-RAM, NVIDIA GeForce 840M and using Windows 10 64-bits. The implementation of the proposed model is based on a deep learning framework called Keras, which is a high-level neural networks API capable of running on top of Tensorflow, CNTK, or Theano.
In this study, deep convolutional autoencoders were evaluated for road segmentation from very highresolution orthophotos with the focus on optimizing hyperparameters of the model. Several optimization algorithms such as Adam, Nadam, RMSprop, Adadelta, Adamax, and SGD were explored to find the best algorithm for road segmentation and to study the model sensitivity to these algorithms.
Several optimization algorithms such as Adam, Nadam, RMSprop, Adadelta, Adamax, and SGD were explored to find the best algorithm for road segmentation and to study the model sensitivity to these algorithms. In addition, activation functions such as ReLU, Sigmoid, Tanh, and ELU were studied so that their effect on the proposed model was explained. Other parameters including number of convolutional filters, learning rate, momentum, and weight decay were also fine-tuned and discussed.

대상 데이터

The datasets used in this research was acquired by airborne laser scanning system over Kinta Valley area, a district located in Perak, East of Malaysia. This area is approximately 200 km north of the capital city Kuala Lumpur and geographically is bounded between 103° 3′ 5″ E and 101° 5' 47'' E and 4° 39′ 11'' N latitudes and 4° 41' 53'' N longitudes.

이론/모형

Thus the tradeoff between the accuracy and the cost should be made by the analyst. In this study, the photos were orthorectified by using 20-m contour maps collected from the Department of Survey and Mapping Malaysia (JUPEM). The contour lines were interpolated by the Inverse Distance Weighting (IDW) method at 20 m spatial resolution.

성능/효과

To make the output of the model practical, the output segmentation maps were thinned and converted into vector GIS data. Overall, the proposed network architecture with the best parameters achieved an accuracy of 88.5% on the validation image sets.

참고문헌 (38)

Bergstra, J. and Y. Bengio, 2012. Random search for hyper-parameter optimization, Journal of Machine Learning Research, 13: 281-305
Colak, S., A. Lima, and M.C. Gonzalez, 2016. Understanding congested travel in urban areas, Nature communications, 7

상세보기
Dahl, G. E., D. Yu, L. Deng, and A. Acero, 2012. Context-dependent pre-trained deep neural networks for large-vocabulary speech recognition. IEEE Transactions on audio, speech, and language processing, 20(1): 30-42

상세보기
Devkota, K. C., A. D. Regmi, H. R. Pourghasemi, K. Yoshida, B. Pradhan, I. C. Ryu, M. R. Dhital, and O.F. Althuwaynee, 2013. Landslide susceptibility mapping using certainty factor, index of entropy and logistic regression models in GIS and their comparison at Mugling- Narayanghat road section in Nepal Himalaya, Natural hazards, 65(1): 135-165

상세보기
Jones, D. R., 2001. A taxonomy of global optimization methods based on response surfaces, Journal of global optimization, 21(4): 345-383
Kaddah, W., Y. Ouerhani, A. Alfalou, M. Desthieux, C. Brosseau, and C. Gutierrez, 2016. Roadmarking features extraction using theVIAPIX(R) system, Optics Communications,371: 117-127

상세보기
Krizhevsky, A., I. Sutskever, and G.E. Hinton, 2012. Imagenet classification with deep convolutional neural networks, Advances in neural information processing systems, 1097-1105
Kumar, P., 2012. Road features extraction using terrestrial mobile laser scanning system, National University of Ireland Maynooth, Ireland.
Kumar, P., C. P. McElhinney, P. Lewis, and T. McCarthy, 2014. Automated road markings extraction from mobile laser scanning data, International Journal of Applied Earth Observation and Geoinformation, 32: 125-137

상세보기
Kussul, N., M. Lavreniuk, S. Skakun, and A. Shelestov, 2017. Deep Learning Classification of Land Cover and Crop Types Using Remote Sensing Data, IEEE Geoscience and Remote Sensing Letters, 14(5): 778-782

상세보기
Latinopoulos, D. and K. Kechagia, 2015. A GIS-based multi-criteria evaluation for wind farm site selection, A regional scale application in Greece. Renewable Energy, 78: 550-560

상세보기
LeCun, Y., Y. Bengio, and G. Hinton, 2015. Deep learning, Nature, 521(7553): 436-444

상세보기
Li, M., A. Stein, W. Bijker, and Q. Zhan, 2016. Regionbased urban road extraction from VHR satellite images using binary partition tree, International Journal of Applied Earth Observation and Geoinformation, 44: 217-225

상세보기
Majumder, N., S. Poria, A. Gelbukh, and E. Cambria, 2017. Deep Learning-Based Document Modeling for Personality Detection from Text, IEEE Intelligent Systems, 32(2): 74-79

상세보기
Mnih, V. and G.E. Hinton. 2010. Learning to detect roads in high-resolution aerial images, Proc. of 11th European Conference on Computer Vision, Grete, Greece, Sep. 5-11, pp. 210-223
Mockus, J., 1975. On Bayesian methods for seeking the extremum, Proc. of Optimization Techniques IFIP Technical Conference Novosibirsk, Berlin Heidelberg, Germany, Jul. 1-7, pp. 400-404
Nareyek, A., 2003. Choosing search heuristics by nonstationary reinforcement learning, Applied Optimization, 86: 523-544

상세보기
Osborne, B. P., V. J. Osborne, and M. L. Kruger, 2012. Comparison of satellite surveying to traditional surveying methods for the resources industry, Journal of the British Interplanetary Society, 65(2): 98-104

상세보기
Raschka, S., 2015. Python machine learning, Packt Publishing Ltd., U.K
Rathore, M. M., A. Ahmad, A. Paul, and S. Rho, 2016. Urban planning and building smart cities based on the internet of things using big data analytics, Computer Networks, 101: 63-80

상세보기
Saito, S., T. Yamashita, and Y. Aoki, 2016. Multiple object extraction from aerial imagery with convolutional neural networks, Electronic Imaging, 2016(10): 1-9
Sameen, M. I. and B. Pradhan, 2016a. Assessment of the effects of expressway geometric design features on the frequency of accident crash rates using high-resolution laser scanning data and GIS, Geomatics, Natural Hazards and Risk, 1-15
Sameen, M. I. and B. Pradhan, 2016b. A Simplified Semi-Automatic Technique for Highway Extraction from High-Resolution Airborne LiDAR Data and Orthophotos, Journal of the Indian Society of Remote Sensing, 45(3): 1-11
Sameen, M. I. and B. Pradhan, 2017. Severity Prediction of Traffic Accidents with Recurrent Neural Networks, Applied Sciences, 7(6): 476
Sameen, M. I. and B. Pradhan, 2017a. Severity Prediction of Traffic Accidents with Recurrent Neural Networks, Applied Sciences, 7(6): 476

상세보기
Sameen, M. I. and B. Pradhan, 2017b. A Two-Stage Optimization Strategy for Fuzzy Object-Based Analysis Using Airborne LiDAR and High- Resolution Orthophotos for Urban Road Extraction, Journal of Sensors.
Saxe, A., P.W. Koh, Z. Chen, M. Bhand, B. Suresh, and A.Y Ng, 2011. On random weights and unsupervised feature learning, Proc. of the 28th international conference on machine learning (ICML-11), Bellevue, WA, Jun. 28-Jul. 2, pp. 1089-1096
Shi, W., Z. Miao, and J. Debayle, 2014. An integrated method for urban main-road centerline extraction from optical remotely sensed imagery, IEEE Transactions on Geoscience and Remote Sensing, 52(6): 3359-3372

상세보기
Sohail, M., D.A.C. Maunder, and S. Cavill, 2006. Effective regulation for sustainable public transport in developing countries, Transport Policy, 13(3): 177-190

상세보기
Unsalan, C. and B. Sirmacek, 2012. Road network detection using probabilistic and graph theoretical methods, IEEE Transactions on Geoscience and Remote Sensing, 50(11): 4441-4453

상세보기
Vitor, G.B., D.A. Lima, A.C. Victorino, and J.V. Ferreira, 2013. A 2D/3D vision based approach applied to road detection in urban environments, Proc. of 2013 IEEE Intelligent Vehicles Symposium (IV), Gold Coast City, Australia, Jun. 23-26, pp. 952-957.
Wang, J., J. Song, M. Chen, and Z. Yang, 2015. Road network extraction: A neural-dynamic framework based on deep learning and a finite state machine, International Journal of Remote Sensing, 36(12): 3144-3169

상세보기
Wegner, J. D., J.A. Montoya-Zegarra, and K. Schindler, 2013. A higher-order CRF model for road network extraction, Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, Portland, OR, Jun. 23-28, pp. 1698-1705
Weihs, C., K. Luebke, and I. Czogiel, 2006. Response surface methodology for optimizing hyper parameters, Universitat Dortmund, Dortmund, Germany
Xie, X., K.B.Y. Wong, H. Aghajan, P. Veelaert, and W. Philips, 2016. Road network inference through multiple track alignment, Transportation Research Part C: Emerging Technologies, 72: 93-108

상세보기
Yang, B., L. Fang, Q. Li, and J. Li, 2012. Automated extraction of road markings from mobile LiDAR point clouds, Photogrammetric Engineering & Remote Sensing, 78(4): 331-338

상세보기
Zeyer, A., P. Doetsch, P. Voigtlaender, R. Schluter, and H. Ney, 2017. A comprehensive study of deep bidirectional LSTM RNNs for acoustic modeling in speech recognition, Proc. of Acoustics, Speech and Signal Processing (ICASSP), 2017 IEEE International Conference, New Orleans, LA, Mar. 5-9, pp. 2462-2466
Zhan, C., 1993. A hybrid line thinning approach, Proc. of Autocarto-conference-, ASPRS American society for photogrammetry and remote sensing, Bethesda, MD, p. 396

활용도 분석정보

상세보기

다운로드

내보내기

활용도 Top5 논문

해당 논문의 주제분야에서 활용도가 높은 상위 5개 콘텐츠를 보여줍니다.
더보기 버튼을 클릭하시면 더 많은 관련자료를 살펴볼 수 있습니다.

표제어: PCR

동의어: Packet Collision Rate

용어 설명 출처 목록 (6)

용어 설명: PCR은 세균 특이성이 있는 primer를 이용하여 적은 수의 세균이 있을지라도 쉽게 검출할 수 있는 유용한 방법이며, 이를 이용하여 구강 내 치면세균막이나 타액에서 직접 세균을 검출할 수 있게 되었다[8].

내보내기 구분	파일저장 인쇄 메일전송
구성항목	기본정보 상세정보 관리번호, 논문명, 저널/프로시딩명, 저자 , 발행년, 권, 호, 시작페이지, 끝페이지, 발행기관 관리번호, 논문명, 대등논문명, 저자 , 저널/프로시딩명, 발행기관, 발행년, 발행언어, 권, 호, 시작페이지, 끝페이지, ISBN, ISSN, 주제분야, 키워드, 초록(한글), 초록(영문), 저자(소속기관)
저장형식	Text(ASCII format) Excel format RefWorks Direct Export RIS format (for Reference Manager, ProCite, EndNote), Scholar's Aids, Mendeley
메일정보	받는사람 (필수) @ 보내는사람 (선택) @ 제목 내용 KISTI 검색결과 이메일 서비스
안내	총 건의 자료가 검색되었습니다. 다운받으실 자료의 인덱스를 입력하세요. (1-10,000) 검색결과의 순서대로 최대 10,000건 까지 다운로드가 가능합니다. 데이타가 많을 경우 속도가 느려질 수 있습니다.(최대 2~3분 소요) 다운로드 파일은 UTF-8 형태로 저장됩니다. 파일의 내용이 제대로 보이지 않을실 때는 웹브라우저 상단의 보기 -> 인코딩 -> 자동선택 여부를 확인하십시오. ~ Text(ASCII format) Excel format

연합인증