SDA: A NOVEL SKEWED-DEEP-ARCHITECTURE FOR VEHICLE MOTION DETECTION IN DRIVING VIDEOS

Tansu Temel; Mehmet Kılıçarslan; Yaşar Hoşcan

doi:10.17780/ksujes.1358512

EN TR

SDA: A NOVEL SKEWED-DEEP-ARCHITECTURE FOR VEHICLE MOTION DETECTION IN DRIVING VIDEOS

Abstract

Collision avoidance mechanisms are important topics for studies in the field of autonomous vehicles. We could obtain prior information about the collision from the movement angles of vehicles. Therefore, it is important issue to learn the movement angles of vehicles in motion. In the study, an architectural model is developed that learns the horizontal movement angles of vehicles to form a base for collision warning systems. YOLOv3 is modified and used on motion profiles. Thanks to the learned angle values, also the bounding boxes match the traces in the motion profiles smoothly. The results obtained have a mAP value of 79% and an operating speed of 36 FPS. These results are better than when trained on motion profiles of the YOLOv3 architecture. In addition, the use of the new architecture on motion profiles and factors such as noise and bad weather in the image do not adversely affect the results. With these features, a fundamental step has been taken for anti-collision systems.

Keywords

EDM: SÜRÜŞ VIDEOLARINDA ARAÇ HAREKET ALGILAMASI IÇIN YENI BIR EĞIK-DERIN-MIMARI

Öz

Çarpışma önleme mekanizmaları otonom araçlar alanındaki çalışmalar için önemli bir konudur. Araçların hareket açılarından çarpışma hakkında önceden bilgi alma imkanımız olmaktadır. Bu nedenle hareket halindeki araçların hareket açılarının öğrenilmesi önemli bir konudur. Çalışmada çarpışma uyarı sistemlerine temel oluşturmak amacıyla araçların yatay hareket açılarını öğrenen bir mimari model geliştirilmiştir. Başarılı derin öğrenme mimarilerinden biri olan YOLOv3 geliştirilerek elde edilen yeni mimari hareket profilleri üzerinde kullanıldı. Öğrenilen açı değerleri sayesinde sınırlayıcı kutular da, hareket profillerindeki izlerle tam olarak eşleşmektedir. Elde edilen sonuçlar %79 mAP değerine ve 36 FPS çalışma hızına sahiptir. Bu sonuçlar, saf YOLOv3 mimarisinin hareket profilleri üzerinde eğitildiklerinde elde edilen sonuçlardan daha iyidir. Yeni mimarinin hareket profillerinde kullanılması ile görüntüdeki gürültü, kötü hava gibi etkenler sonuçlarımızı olumsuz etkilememektedir. Bu özellikleri ile çarpışma önleyici sistemler için önemli bir adım atılmıştır.

Anahtar Kelimeler

Destekleyen Kurum

Scientific and Technological Research Council of Turkey (TÜBİTAK)

Proje Numarası

122E586

Teşekkür

This work was supported by the Scientific and Technological Research Council of Turkey (TÜBİTAK) with projects numbered 122E586.

Kaynakça

Behrendt, K., Novak, L., & Botros, R. (2017, May). A deep learning approach to traffic lights: Detection, tracking, and classification. In 2017 IEEE International Conference on Robotics and Automation (ICRA) (pp. 1370-1377). IEEE. https://doi.org/10.1109/icra.2017.7989163
Cadieu, C., & Olshausen, B. (2008). Learning transformational invariants from natural movies. Advances in neural information processing systems, 21.
Cao, Z., Simon, T., Wei, S. E., & Sheikh, Y. (2017). Realtime multi-person 2d pose estimation using part affinity fields. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 7291-7299). https://doi.org/10.1109/cvpr.2017.143
Caraffi, C., Vojíř, T., Trefný, J., Šochman, J., & Matas, J. (2012, September). A system for real-time detection and tracking of vehicles from a single car-mounted camera. In 2012 15th international IEEE conference on intelligent transportation systems (pp. 975-982). IEEE. https://doi.org/10.1109/itsc.2012.6338748
Chen, L., Peng, X., & Ren, M. (2018). Recurrent metric networks and batch multiple hypothesis for multi-object tracking. IEEE Access, 7, 3093-3105. https://doi.org/10.1109/access.2018.2889187
Gordon, D., Farhadi, A., & Fox, D. (2018). Re3: Real-time recurrent regression networks for visual tracking of generic objects. IEEE Robotics and Automation Letters, 3(2), 788-795. https://doi.org/10.1109/lra.2018.2792152
Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural computation, 9(8), 1735-1780. https://doi.org/10.1162/neco.1997.9.8.1735
Hui, J. (2018). Real-time object detection with yolo, yolov2, and now yolov3. Available online: medium. com/@ jonathan_hui/real-time-object-detection-with-YOLO-YOLOv2-28b1b93e2088 (accessed on 24 February 2019). https://doi.org/10.22214/ijraset.2021.39044

Jazayeri, A., Cai, H., Zheng, J. Y., & Tuceryan, M. (2011). Vehicle detection and tracking in-car video based on motion model. IEEE Transactions on Intelligent Transportation Systems, 12(2), 583-595. https://doi.org/10.1109/tits.2011.2113340
John, V., & Mita, S. (2019). Vehicle semantic understanding for automated driving in multiple-lane urban roads using deep vision-based features. In International Joint Conferences on Artificial Intelligence; Macao, China (pp. 1-7).
Kilicarslan, M., & Temel, T. (2022). Motion-aware vehicle detection in driving videos. Turkish Journal of Electrical Engineering and Computer Sciences, 30(1), 63-78. https://doi.org/10.3906/elk-2101-93
Kilicarslan, M., & Zheng, J. Y. (2018). Predict vehicle collision by TTC from motion using a single video camera. IEEE Transactions on Intelligent Transportation Systems, 20(2), 522-533. https://doi.org/10.1109/tits.2018.2819827
Li, L., Zhou, Z., Wang, B., Miao, L., & Zong, H. (2020). A novel CNN-based method for accurate ship detection in HR optical remote sensing images via rotated bounding box. IEEE Transactions on Geoscience and Remote Sensing, 59(1), 686-699. https://doi.org/10.1109/tgrs.2020.2995477
Liang, Y., & Zhou, Y. (2018, October). LSTM multiple object tracker combining multiple cues. In 2018 25th IEEE International Conference on Image Processing (ICIP) (pp. 2351-2355). IEEE. https://doi.org/10.1109/icip.2018.8451739
Liu, Y., Lu, Y., Shi, Q., & Ding, J. (2013, December). Optical flow-based urban road vehicle tracking. In 2013 ninth international conference on computational intelligence and security (pp. 391-395). IEEE. https://doi.org/10.1109/cis.2013.89
Muehlemann, A. (2019). TrainYourOwnYOLO: Building a Custom Object Detector from Scratch. Disponible on-line: https://github. com/AntonMu/TrainYourOwnYOLO (Accedido Diciembre 2020). https://doi.org/10.5281/zenodo.5112375
Wang, L., Pham, N. T., Ng, T. T., Wang, G., Chan, K. L., & Leman, K. (2014, October). Learning deep features for multiple object tracking by using a multi-task learning strategy. In 2014 IEEE International Conference on Image Processing (ICIP) (pp. 838-842). IEEE. https://doi.org/10.1109/icip.2014.7025168
Yun, W. J., Park, S., Kim, J., & Mohaisen, D. (2022). Self-Configurable Stabilized Real-Time Detection Learning for Autonomous Driving Applications. IEEE Transactions on Intelligent Transportation Systems. https://doi.org/10.1109/tits.2022.3211326
Zhang, D., Maei, H., Wang, X., & Wang, Y. F. (2017). Deep reinforcement learning for visual object tracking in videos. arXiv preprint arXiv:1701.08936. https://doi.org/10.48550/arXiv.1701.08936
Zhou, H., Ouyang, W., Cheng, J., Wang, X., & Li, H. (2018). Deep continuous conditional random fields with asymmetric inter-object constraints for online multi-object tracking. IEEE Transactions on Circuits and Systems for Video Technology, 29(4), 1011-1022. https://doi.org/10.1109/tcsvt.2018.2825679

Ayrıntılar

Birincil Dil

İngilizce

Konular

Bilgisayar Görüşü , Görüntü İşleme , Örüntü Tanıma , Derin Öğrenme

Bölüm

Araştırma Makalesi

Yazarlar

Tansu Temel ^*
0000-0002-8359-1146
Türkiye

Mehmet Kılıçarslan
0000-0002-7212-5262
Türkiye

Yaşar Hoşcan
0000-0003-0789-6025
Türkiye

Yayımlanma Tarihi

3 Mart 2024

Gönderilme Tarihi

12 Eylül 2023

Kabul Tarihi

24 Kasım 2023

Yayımlandığı Sayı

Yıl 2024 Cilt: 27 Sayı: 1

DOI

https://doi.org/10.17780/ksujes.1358512

IZ

https://izlik.org/JA69CF97SK

Kaynak Göster

RIS / Bibtex

APA

Temel, T., Kılıçarslan, M., & Hoşcan, Y. (2024). SDA: A NOVEL SKEWED-DEEP-ARCHITECTURE FOR VEHICLE MOTION DETECTION IN DRIVING VIDEOS. Kahramanmaraş Sütçü İmam Üniversitesi Mühendislik Bilimleri Dergisi, 27(1), 92-104. https://doi.org/10.17780/ksujes.1358512

Cited By

Two-Strea[M] YOLOV8: Object and Motion Detection in Driving Videos

IEEE Transactions on Intelligent Vehicles

https://doi.org/10.1109/TIV.2024.3448631

An end-to-end deep learning model for vehicle speed estimation using a single camera

Internet of Things

https://doi.org/10.1016/j.iot.2026.101903