[논문]Aerial Dataset Integration For Vehicle Detection Based on YOLOv4

Omar, Wael; Oh, Youngon; Chung, Jinwoo; Lee, Impyeong

doi:10.7780/kjrs.2021.37.4.6

Aerial Dataset Integration For Vehicle Detection Based on YOLOv4 원문보기

대한원격탐사학회지 = Korean journal of remote sensing, v.37 no.4, 2021년, pp.747 - 761

Omar, Wael (Department of Geoinformatics, University of Seoul) , Oh, Youngon (Department of Geoinformatics, University of Seoul) , Chung, Jinwoo (Department of Geoinformatics, University of Seoul) , Lee, Impyeong (Department of Geoinformatics, University of Seoul)

Abstract ▼ AI-Helper

With the increasing application of UAVs in intelligent transportation systems, vehicle detection for aerial images has become an essential engineering technology and has academic research significance. In this paper, a vehicle detection method for aerial images based on the YOLOv4 deep learning algorithm is presented. At present, the most known datasets are VOC (The PASCAL Visual Object Classes Challenge), ImageNet, and COCO (Microsoft Common Objects in Context), which comply with the vehicle detection from UAV. An integrated dataset not only reflects its quantity and photo quality but also its diversity which affects the detection accuracy. The method integrates three public aerial image datasets VAID, UAVD, DOTA suitable for YOLOv4. The training model presents good test results especially for small objects, rotating objects, as well as compact and dense objects, and meets the real-time detection requirements. For future work, we will integrate one more aerial image dataset acquired by our lab to increase the number and diversity of training samples, at the same time, while meeting the real-time requirements.

주제어

표/그림 (18)

그림 Fig. 1. Datasets integration flowchart for datasets.
그림 Fig. 2. YOLO v1network structure.
그림 Fig. 3. Comparison of YOLOv4 and other state-of-the-art object detectors.
그림 Fig. 4. Examples of the manually annotated frames in the UAVDT benchmark. The three rows represent the DET, MOT, and SOT tasks, respectively. The shooting conditions of UAVs are presented in the lower right corner. The pink areas represent the ignored regions in the dataset. Different bounding box colors denote different classes of vehicles.
그림 Fig. 5. In the VAID dataset, the common vehicles are classified into 7 categories, namely (a) sedan, (b) minibus, (c) truck, (d) pickup truck, (e) bus, (f) cement truck, and (g) trailer. The sample images are shown in the figure from the left to the right accordingly.
그림 Fig. 6. Some aerial images in the VAID dataset were captured using a drone. It consists of different road types and traffic scenes.
그림 Fig. 7. Samples of annotated images in DOTA. It shows three samples per category, except six for large vehicle.
그림 Fig. 8. Samples of annotated images in LSM.
표 Table 1. The basic information of the three public aerial image datasets
표 Table 2. The number of each category vehicle in the UAVDT dataset
표 Table 3. The number of each category vehicle in the VAID dataset
표 Table 4. The number of each category vehicle in the VAID dataset
표 Table 5. The actual class numbers in the LSM dataset are used in the training process
표 Table 6. The processed dataset information
표 Table 7. A merged classes names accordingly to my research purpose
그림 Fig. 9. Shows the distribution number of each class in the LSM dataset.
그림 Fig. 10. (a) LSM testing image, (b) result on LSM images.
표 Table 8. 5The test result

AI 본문요약
AI-Helper

제안 방법

), mostly focus on vehicle detection but also include a limited number of annotated vehicles. To improve vehicle detection research, including vehicle detection, counting, and tracking, we provide three customized aerial image datasets for real-time vehicle detection.

성능/효과

The vehicles in Fig. 11 are mostly horizontal with some vertical and rotated vehicles, the test result shows that the model has a good performance on the detection of rotating objects, while the detection may miss one object in the far lower right corner, but the model has 79.77% of (mAP) mean average precision score on LSM testing dataset.
2) Perspective specificity, the perspectives of aerial images are generally high-altitude overlooking, while most of the common datasets are ground level perspectives.
3) A small object, the objects of aerial images are normally only a few dozen or even a few pixels, so, their amount of information is less also.
4) Multidirectional, aerial images are taken from a bird’s view, and the object’s directions are unconfident (while the object directly on the conventional dataset tends to have confidently, such as pedestrians are generally upright).
Other data sets such as VEDAI, COWO, DLR3K MunichVehicle, and UCASAOD, mostly focus on vehicle detection but also include a limited number of annotated vehicles. To improve vehicle detection research, including vehicle detection, counting, and tracking, we provide three customized aerial image datasets for real-time vehicle detection.
This method integrates three public aerial image datasets suitable for YOLOv4. The training model presents good test results especially for small objects, rotating objects, as well as compact and dense objects, and meets the real-time requirements, and reached 79.77% of the mean Average Precision (mAP) score. Next, we will integrate one more aerial image dataset acquired by our lab to increase the number and diversity of training samples, at the same time.

후속연구

77% of the mean Average Precision (mAP) score. Next, we will integrate one more aerial image dataset acquired by our lab to increase the number and diversity of training samples, at the same time. While integrating more datasets, we will optimize the latest version of the YOLO algorithm to further improvement to the detection accuracy.
Next, we will integrate one more aerial image dataset acquired by our lab to increase the number and diversity of training samples, at the same time. While integrating more datasets, we will optimize the latest version of the YOLO algorithm to further improvement to the detection accuracy.

참고문헌 (29)

Ajay, A., V. Sowmya and K.P. Soman, 2017. Vehicle detection in aerial imagery using eigen features, Proc. of 2017 International Conference on Communication and Signal Processing ICCSP, Chennai, IN, Apr. 6-8, pp. 1620-1624.
Ammour, N., H. Alhichri, Y. Bazi, B. Benjdira, N. Alajlan, and M. Zuair, 2017. Deep Learning Approach for Car Detection in UAV Imagery, Remote Sens, 9(4): 312.

상세보기
Azevedo, C.L., J.L. Cardoso, M. Ben-Akiva, J.P. Costeira, and M. Marques, 2014. Automatic VehicleTrajectory Extraction by Aerial Remote Sensing, Procedia - Social and Behavioral Sciences, 111: 849-858.

상세보기
Bochkovskiy, A., C.Y.Wang, and H.Y.M Liao, 2020. YOLOv4: Optimalspeed and accuracy of object detection, arXiv print, arXiv: 2004.10934v1.
Chen, X., S. Xiang, C.-L. Liu, and C.-H. Pan, 2014. Vehicle Detection in Satellite Images by Hybrid Deep Convolutional Neural Networks, IEEE Geoscience and Remote Sensing Letters, 11(10): 1797-1801.

상세보기
Cheng, P. Z., 2009, Detecting and Counting Vehicles from SMALL LOW-COST UAV IMAGES, Proc. of ASPRS 2009 Annual Conference, Baltimore, MD, Mar. 9-13, pp. 1-7.
Deng,J.,W. Dong,R. Socher, L.J. Li, K. Li, and L. Fei-Fei, 2009. ImageNet:Alarge-scale hierarchical image database, Proc of 2009 IEEE Conference on Computer Vision and Pattern Recognition, Miami, FL, US, Jun. 20-25, pp. 248-255.
Everingham, M., S.M.A. Eslami, L.V. Gool, C.K.I. Williams, J. Winn, and A. Zisserman, 2014. The Pascal visual object Classes challenge: A retrospective, International Journal of Computer Vision, 111: 98-136.
Everingham, M.L.V. Gool, C.K.I. Williams, J. Winn, and A. Zisserman 2010. The pascal visual object classes (voc) challenge, International Journal of Computer Vision, 88(2): 303-338.

상세보기
Kim, C.E., M.M.D. Oghaz, J. Fajtl, V. Argyriou, and P. Remagnino, 2018. A comparison of embedded deep learning methods for person detection, arXiv PrePrint, arXiv:1812.03451.
Lewandowski, M., B. Placzek, M. Bernas, and P. Szymala, 2018. Road traffic monitoring system based on mobile devices and bluetooth low energy beacons, Wireless Communications and Mobile Computing, 2018: 1-12.

상세보기
Lin H.-Y., K.-C. Tu, and C.-Y. Li, 2020. VAID: An Aerial Image Dataset for Vehicle Detection and Classification, IEEE Access, 8: 212209-212219.

상세보기
Lin, T.-Y., M. Maire, S. Belongie, J. Hays, P. Perona, D. Ramanan, P. Dollar C, and L. Zitnick, 2014. Microsoft COCO: Common objectsin context, ECCV.
Liu, K. and G. Mattyus, 2015. Fast multiclass vehicle detection on aerial images, IEEE Geoscience and Remote Sensing Letters, 12(9): 1938-1942.

상세보기
Liu, W., D. Anguelov, D. Erhan, C. Szegedy, S. Reed, C.Y. Fu, and A.C. Berg, 2016. SSD: Single shot multibox detector, Proc. of European Conference on Computer Vision 2016, Cham, CH, Oct. 8-16, pp. 21-37.
Lu, J., C. Ma, L. Li, X. Xing, Y. Zhang, Z. Wang, and J. Xu, 2018. A Vehicle Detection Method for Aerial Image ased on YOLO. Journal of Computer and Communications, 6(11): 98-107.
Qiu, Y., (2014). Video-Based Vehicle Detection in Intelligent Transportation System, Master Thesis, Jilin University, Chang Chun, CN.
Razakarivony, S. and F. Jurie, 2016. Vehicle detection in aerial imagery: A small target detection benchmark, Journal of Visual Communication and Image Representation, 34: 187-203.

상세보기
Redmon, J. and A. Farhadi, 2017. YOLO9000: Better, Faster, Stronger. Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, HI, US, Jul. 21-26, Vol. 1, pp. 6517-6525.
Redmon, J. and A. Farhadi, 2018. YOLO v3: An Incremental Improvement, arXiv preprint, arxiv: 1804.02767.
Redmon, J., S. Divvala, R. Girshick, and A. Farhadi, 2016. You Only Look Once:Unified, Real-Time Object Detection, Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, LAS VEGAS, NV, US, Jun. 27-30, Vol.1, pp. 779-788.
Simonyan, K. and A. Zisserman, 2015. A Very Deep Convolutional Networks for Large-Scale Image Recognition, arXiv preprint, arXiv: 1409.1556.
Sivaraman, S. and M.M. Trivedi, 2010. A General Active-Learning Framework for On-Road Vehicle Recognition and Tracking, IEEE Transactions on Intelligent Transportation Systems, 11(2): 267-276.

상세보기
Mundhenk, T.N., G. Konjevod, W.A. Sakla, and K. Boakye, 2016. A large contextual dataset for classification, detection and counting of cars with deep learning, Proc. of European Conference on Computer Vision 2014, Zurich, CH, Sep. 6-12, Vol. 4, pp. 740-755.
Tehrani Niknejad, H., A. Takeuchi, S. Mita, and D. McAllester, 2012. On-Road Multivehicle Tracking Using Deformable Object Model and Particle Filter with Improved Likelihood, IEEE Transactions on Intelligent Transportation Systems, 13(2): 748-758.

상세보기
Xi, X., Z. Yu, Z. Zhan, and C. Tian, 2019. Multi-task Cost-sensitive-Convolutional Neural Network for Car Detection, IEEE Access, 7: 98061-98068.

상세보기
Xia, G.-S. X. Bai,J. Ding, Z. Zhu, S. Belongie,J. Luo, M. Datcu, M. Pelillo, and L. Zhang, 2018. Dota: A large-scale dataset for object detection in aerial images, Proc. of IEEE Conference on Computer Vision and Pattern Recognition, Salt Lake City, UT, US,Jun. 19-21, pp. 3974-3983.
Yu, H., G. Li,W. Zhang, Q. Huang, D. Du, Q.Tian, and N . Sebe, 2019. The unmanned aerial vehicle Benchmark: Object Detection, tracking and baseline, International Journal of Computer Vision, 128(5): 1141-1159.

상세보기
Zhu, H., X. Chen, W. Dai, K. Fu, Q. Ye, and J. Jiao, 2015. Orientation robust object detection in aerial images using deep convolutional neural network, Proc. of 2015 IEEE International Conference on Image Processing(ICIP), Quebec, QC, CA, Sep. 27-30, pp. 3735-3739.

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