根據內政部警政署統計台灣於2020年發生485,260件交通事故，造成龐大的傷害及的社會成本損失。隨著各項智慧交通技術的實施，許多研究開發了先進駕駛輔助系統來提升交通安全，其透過藉由感測器（如攝影機、超音波、雷達和光達等）收集大量的即時交通資訊。然而，超音波、雷達及光達等主動式感測器價格昂貴，且在道路擁擠時可能會引發干擾的問題。深度學習方法已被廣泛作為分類和檢測的方法，研究人員提出了許多與自駕車相關的深度預測模型，如碰撞風險預測模型。多數碰撞預測研究以二元分類將結果分為有碰撞及無碰撞，卻無進一步研究碰撞型態。因此本研究使用相對便宜且廣泛被使用的前向鏡頭行車紀錄器做為收集資料的感測器，提出了基於視覺的碰撞型態預測模型，其中碰撞影像從Youtube多媒體收集。該模型由兩個部分組成：使用YOLOv4和DeepSORT進行物件檢測和追蹤，以及基於卷積神經網路 (CNN) 和門控循環單元 (GRU) 的碰撞型態分類。其中CNN用於擷取影像中每幀的空間特徵。GRU用於訓練動態特徵和空間特徵以擷取時間序列特徵。最後，探討了每個碰撞類型中各碰撞類型的機率。
本研究目的為提供更多有關車輛碰撞的即時訊息。期望未來可以作為ADAS中警示或煞車系統的一部份，為駕駛員或車輛提供更精確的即時安全警示，以便實施應對措施，防止交通事故及最大程度地減少事故成本。

According to the statistics from the National Police Agency, Ministry of the Interior, 485,260 traffic accidents occurred in Taiwan in 2020, resulting in major injuries and huge social cost losses. With the advance in intelligent transportation technologies, Advanced Driver Assistance Systems (ADAS) is developed to enhance traffic safety, extensive real-time traffic data became available through sensors like ultrasound, radar, and LiDAR. However, those active sensors are expensive and may cause interference problems. On the other hand, deep learning methods have been widely used for classification and detection. Researchers have proposed many deep prediction models for autonomous vehicles or ADAS such as collision risk prediction models. Most of those studies use binary classification to classify the results into collision and non-collision, but do not further study the collision type. Therefore, this research uses a relatively inexpensive and widely used front-facing dashboard camera as a sensor to collect data and proposes a vision-based collision type prediction model. Vehicle collision videos from the dashboard camera were collected from YouTube. The model consists of two parts: object detection and tracking with YOLOv4 and DeepSORT methods, and collision type classification based on Convolutional Neural Network (CNN) and Gated Recurrent Unit (GRU), where CNN is used to extract spatial features of the video frames, and GRU is used to train with dynamic features and spatial features to extract the temporal features. Finally, the probability of collision types in each collision type is explored.
The purpose of the study is to provide more real-time information about vehicle collisions. It can be used in warning systems or actuating systems as a part of the ADAS in the future, to provide drivers with accurate real-time alerts and automatic vehicle controls, so that drivers or vehicles can make timely reactions to prevent traffic accidents or minimize the cost of accidents.

Aksan, N., Sager, L., Hacker, S., Marini, R., Dawson, J., Anderson, S., & Rizzo, M. (2016). Forward Collision Warning: Clues to Optimal Timing of Advisory Warnings. SAE international journal of transportation safety, 4(1), 107-112. Retrieved from https://pubmed.ncbi.nlm.nih.gov/27648455
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5026383/. doi:10.4271/2016-01-1439
Aslani, S., & Mahdavi-Nasab, H. (2013). Optical Flow Based Moving Object Detection and Tracking for Traffic Surveillance. International Journal of Electrical and Computer Engineering, 7, 1252-1256.
Automotive Research & Testing Center. (2016). 自動緊急煞車與車道追隨整合系統. The 21st National Conference on Vehicle Engineering, Southern Taiwan University of Science and Technology. Retrieved from https://www.artc.org.tw/upfiles/ADUpload/knowledge/tw_knowledge_556488673.pdf.
Bao, J., Liu, P., & Ukkusuri, S. V. (2019). A spatiotemporal deep learning approach for citywide short-term crash risk prediction with multi-source data. Accident Analysis & Prevention, 122, 239-254. Retrieved from http://www.sciencedirect.com/science/article/pii/S0001457518303877. doi:https://doi.org/10.1016/j.aap.2018.10.015
Bengio, Y., Courville, A., & Vincent, P. (2013). Representation Learning: A Review and New Perspectives. IEEE Transactions on Pattern Analysis and Machine Intelligence, 35(8), 1798-1828. doi:10.1109/TPAMI.2013.50
Bengio, Y., Simard, P., & Frasconi, P. (1994). Learning long-term dependencies with gradient descent is difficult. IEEE Transactions on Neural Networks, 5(2), 157-166.
Bochkovskiy, A., Wang, C.-Y., & Liao, H.-y. (2020). YOLOv4: Optimal Speed and Accuracy of Object Detection.
Bucsuházy, K., Matuchová, E., Zůvala, R., Moravcová, P., Kostíková, M., & Mikulec, R. (2020). Human factors contributing to the road traffic accident occurrence. Transportation Research Procedia, 45, 555-561. Retrieved from http://www.sciencedirect.com/science/article/pii/S2352146520302192. doi:https://doi.org/10.1016/j.trpro.2020.03.057
Chen, C., Xiang, H., Qiu, T., Wang, C., Zhou, Y., & Chang, V. (2018). A rear-end collision prediction scheme based on deep learning in the Internet of Vehicles. Journal of Parallel and Distributed Computing, 117, 192-204. Retrieved from http://www.sciencedirect.com/science/article/pii/S0743731517302447. doi:https://doi.org/10.1016/j.jpdc.2017.08.014
Chen, G. (1998). The Analysis of the Cause of accidents in Different Collision Type Behaviors and Responsibility Identification. Paper presented at the Road Traffic Safety and Law Enforcement Seminar.
Cho, K., van Merriënboer, B., Gulcehre, C., Bougares, F., Schwenk, H., & Bengio, Y. (2014). Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation. doi:10.3115/v1/D14-1179
Comaniciu, D., & Meer, P. (2002). Mean shift: a robust approach toward feature space analysis. IEEE Transactions on Pattern Analysis and Machine Intelligence, 24(5), 603-619. doi:10.1109/34.1000236
Corcoran, G., & Clark, J. (2019, 29-31 May 2019). Traffic Risk Assessment: A Two-Stream Approach Using Dynamic-Attention. Paper presented at the 2019 16th Conference on Computer and Robot Vision (CRV).
Dagan, E., Mano, O., Stein, G. P., & Shashua, A. (2004, 14-17 June 2004). Forward collision warning with a single camera. Paper presented at the IEEE Intelligent Vehicles Symposium, 2004.
Dagra Graph Digitizer. (2021). Linear Interpolation. Retrieved from https://www.datadigitization.com/dagra-in-action/linear-interpolation-with-excel/
Dingus, T. A., Guo, F., Lee, S., Antin, J. F., Perez, M., Buchanan-King, M., & Hankey, J. (2016). Driver crash risk factors and prevalence evaluation using naturalistic driving data. Proceedings of the National Academy of Sciences of the United States of America, 113(10), 2636-2641. Retrieved from https://wwww.unboundmedicine.com/medline/citation/26903657/Driver_crash_risk_factors_and_prevalence_evaluation_using_naturalistic_driving_data_
http://www.pnas.org/cgi/pmidlookup?view=long&pmid=26903657.
Drews, P., Williams, G., Goldfain, B., Theodorou, E., & Rehg, J. (2017). Aggressive Deep Driving: Model Predictive Control with a CNN Cost Model.
Fagnant, D. J., & Kockelman, K. (2015). Preparing a nation for autonomous vehicles: opportunities, barriers and policy recommendations. Transportation Research Part A: Policy and Practice, 77, 167-181. Retrieved from http://www.sciencedirect.com/science/article/pii/S0965856415000804. doi:https://doi.org/10.1016/j.tra.2015.04.003
Formosa, N., Quddus, M., Ison, S., Abdel-Aty, M., & Yuan, J. (2020). Predicting real-time traffic conflicts using deep learning. Accident Analysis & Prevention, 136, 105429. Retrieved from http://www.sciencedirect.com/science/article/pii/S000145751930973X. doi:https://doi.org/10.1016/j.aap.2019.105429
Girshick, R. (2015, 7-13 Dec. 2015). Fast R-CNN. Paper presented at the 2015 IEEE International Conference on Computer Vision (ICCV).
Girshick, R., Donahue, J., Darrell, T., & Malik, J. (2014, 23-28 June 2014). Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation. Paper presented at the 2014 IEEE Conference on Computer Vision and Pattern Recognition.
Hayward, J. C., Pennsylvania, T., Traffic Safety, C., & National Research, C. (1972). Near miss determination through use of a scale of danger. University Park, Pa.: Pennsylvania Transportation and Traffic Safety Center, The Pennsylvania State University.
Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural Comput, 9(8), 1735-1780. doi:10.1162/neco.1997.9.8.1735
Hsieh, C.-Y. (2020). Risk Prediction of Vehicle Collision Based on A Combined Neural Network of CNN and LSTM. National Cheng Kung University, Retrieved from http://dx.doi.org/10.6844/NCKU202001627
Huang, T., Wang, S., & Sharma, A. (2020). Highway crash detection and risk estimation using deep learning. Accident Analysis & Prevention, 135, 105392. Retrieved from http://www.sciencedirect.com/science/article/pii/S000145751930555X. doi:https://doi.org/10.1016/j.aap.2019.105392
Institute of Transportation, M. o. T. a. C. (2019). 道路交通事故成本推估之研究. Retrieved from file:///C:/Users/USER/Downloads/%E9%81%93%E8%B7%AF%E4%BA%A4%E9%80%9A%E4%BA%8B%E6%95%85%E6%88%90%E6%9C%AC%E6%8E%A8%E4%BC%B0%E4%B9%8B%E7%A0%94%E7%A9%B6.pdf
Jiménez, F., Naranjo, J. E., & García, F. (2012). An Improved Method to Calculate the Time-to-Collision of Two Vehicles. International Journal of Intelligent Transportation Systems Research, 11(1), 34-42. doi:10.1007/s13177-012-0054-4
Kale, K., Pawar, S., & Dhulekar, P. (2015, 2-4 Sept. 2015). Moving object tracking using optical flow and motion vector estimation. Paper presented at the 2015 4th International Conference on Reliability, Infocom Technologies and Optimization (ICRITO) (Trends and Future Directions).
Kalman, R. E. (1960). A New Approach to Linear Filtering and Prediction Problems. Journal of Basic Engineering, 82(1), 35-45. Retrieved from https://doi.org/10.1115/1.3662552. doi:10.1115/1.3662552
Khan, A., Sohail, A., Zahoora, U., & Qureshi, A. S. (2020). A survey of the recent architectures of deep convolutional neural networks. Artificial Intelligence Review, 53(8), 5455-5516. Retrieved from https://doi.org/10.1007/s10462-020-09825-6. doi:10.1007/s10462-020-09825-6
Khan, M. N., & Ahmed, M. M. (2020). Trajectory-level fog detection based on in-vehicle video camera with TensorFlow deep learning utilizing SHRP2 naturalistic driving data. Accident Analysis & Prevention, 142, 105521. Retrieved from http://www.sciencedirect.com/science/article/pii/S0001457519316422. doi:https://doi.org/10.1016/j.aap.2020.105521
Kim, S., Lee, S., Doo, S., & Shim, B. (2018, 3-7 Sept. 2018). Moving Target Classification in Automotive Radar Systems Using Convolutional Recurrent Neural Networks. Paper presented at the 2018 26th European Signal Processing Conference (EUSIPCO).
Krizhevsky, A., Sutskever, I., & Hinton, G. (2012). ImageNet Classification with Deep Convolutional Neural Networks. Neural Information Processing Systems, 25. doi:10.1145/3065386
Kumar, S., Shaw, V., Maitra, J., & Karmakar, R. (2020, 14-15 Feb. 2020). FCW: A Forward Collision Warning System Using Convolutional Neural Network. Paper presented at the 2020 International Conference on Electrical and Electronics Engineering (ICE3).
Lambert, S. (2020). 5 top autonomous vehicle companies to watch in 2020. Retrieved from https://www.intelligent-mobility-xperience.com/5-top-autonomous-vehicle-companies-to-watch-in-2020-a-958065/
Li, P., Abdel-Aty, M., & Yuan, J. (2020). Real-time crash risk prediction on arterials based on LSTM-CNN. Accident Analysis & Prevention, 135, 105371. Retrieved from http://www.sciencedirect.com/science/article/pii/S0001457519311108. doi:https://doi.org/10.1016/j.aap.2019.105371
Li, Y., Wang, C., & Han, K. (2018). RFAmyloid: A Web Server for Predicting Amyloid Proteins. International journal of molecular sciences, 19. doi:10.3390/ijms19072071
Lim, Q., He, Y., & Tan, U. (2018, 12-15 Dec. 2018). Real-Time Forward Collision Warning System Using Nested Kalman Filter for Monocular Camera. Paper presented at the 2018 IEEE International Conference on Robotics and Biomimetics (ROBIO).
Liu, S., Liu, L., Tang, J., Yu, B., Wang, Y., & Shi, W. (2019). Edge Computing for Autonomous Driving: Opportunities and Challenges. Proceedings of the IEEE, 107(8), 1697-1716. doi:10.1109/JPROC.2019.2915983
MarketsandMarkets. (2020). ADAS Market by System (ACC, DMS, IPA, PDS, TJA, FCW, CTA, RSR, LDW, AEB, & BSD), Component (Radar, LiDAR, Ultrasonic, & Camera Unit), Vehicle (PC, LCV, Bus, & Truck), Level of Autonomy (L1, L2&3, L4, L5), Offering, EV, and Region - Global Forecast to 2030 (AT 2068). Retrieved from https://www.marketsandmarkets.com/Market-Reports/driver-assistance-systems-market-1201.html
Marshall, B. (2018). Industrial Sensing, Lidar, Radar & Digital Cameras: the Eyes of Autonomous Vehicles. Retrieved from https://www.rs-online.com/designspark/lidar-radar-digital-cameras-the-eyes-of-autonomous-vehicles
Masumi, N., Raksincharoensak, P., & Nagai, M. (2008, 14-17 Oct. 2008). Study on forward collision warning system adapted to driver characteristics and road environment. Paper presented at the 2008 International Conference on Control, Automation and Systems.
Mazuryk, T., & Gervautz, M. (1999). Virtual Reality - History, Applications, Technology and Future.
Ministry of Economic Affairs. (2020). 無人載具科技創新實驗計畫. Retrieved from https://www.moea.gov.tw/MNS/doit/content/Content.aspx?menu_id=32501
Ministry of Health and Welfare. (2020). 108年國人死因統計結果. Retrieved from https://www.mohw.gov.tw/cp-16-54482-1.html
Morando, M., Tian, Q., Truong, L., & Vu, H. (2018). Studying the Safety Impact of Autonomous Vehicles Using Simulation-Based Surrogate Safety Measures. Journal of advanced transportation, 2018. doi:10.1155/2018/6135183
Narkhede, S. (2018). Understanding AUC - ROC Curve. Retrieved from https://towardsdatascience.com/understanding-auc-roc-curve-68b2303cc9c5
Naseralavi, S., Nadimi, N., Saffarzadeh, M., & Mamdoohi, A. R. (2013). A general formulation for time-to-collision safety indicator. Proceedings of the ICE - Transport, 166, 294-304. doi:10.1680/tran.11.00031
National Police Agency. (2020). 事故概況統計. Retrieved from https://ba.npa.gov.tw/npa/stmain.jsp?sys=220&ym=10600&ymt=10800&kind=21&type=1&funid=q06010101&cycle=4&outmode=610&compmode=0&ohtml=q250x&outkind=1&fldlst=111&cod00=1&rdm=enlfeNWi
Nita Congress. (1994). The Automated Highway System. Federal Highway Administration Research and Technology, 58(1). Retrieved from https://www.fhwa.dot.gov/publications/publicroads/94summer/p94su1.cfm
Ong, L. Y., Lau, S. H., Koo, V. C., & Hun, L. (2014). An experimental study on vision-based multiple target tracking. International Journal of Microwave and Optical Technology, 9, 134-138.
Papadoulis, A., Quddus, M., & Imprialou, M. (2019). Evaluating the safety impact of connected and autonomous vehicles on motorways. Accident Analysis & Prevention, 124, 12-22. Retrieved from http://www.sciencedirect.com/science/article/pii/S0001457518306018. doi:https://doi.org/10.1016/j.aap.2018.12.019
Park, S., Seonwoo, Y., Kim, J., Kim, J., & Oh, A. (2020). Denoising Recurrent Neural Networks for Classifying Crash-Related Events. IEEE Transactions on Intelligent Transportation Systems, 21(7), 2906-2917. doi:10.1109/TITS.2019.2921722
Patterson, J., & Gibson, A. (2017). Deep Learning (Vol. 4): O'Reilly Media, Inc.
Priyanka, S., & Kumar, D. N. (2016). Noise Removal in Remote Sensing Image Using Kalman Filter Algorithm.
Qun, L., He, Y., & Tan, U. X. (2018). Real-Time Forward Collision Warning System Using Nested Kalman Filter for Monocular Camera.
Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016, 27-30 June 2016). You Only Look Once: Unified, Real-Time Object Detection. Paper presented at the 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).
Redmon, J., & Farhadi, A. (2017, 21-26 July 2017). YOLO9000: Better, Faster, Stronger. Paper presented at the 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).
Ren, S., He, K., Girshick, R., & Sun, J. (2017). Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks. IEEE Transactions on Pattern Analysis and Machine Intelligence, 39(6), 1137-1149. doi:10.1109/TPAMI.2016.2577031
SAE INTERNATIONAL. (2014). SAE International Technical Standard Provides Terminology for Motor Vehicle Automated Driving Systems. Retrieved from https://www.sae.org/news/press-room/2014/10/sae-international-technical-standard-provides-terminology-for-motor-vehicle-automated-driving-systems
Salari, E., & Ouyang, D. (2013, 4-7 Aug. 2013). Camera-based Forward Collision and lane departure warning systems using SVM. Paper presented at the 2013 IEEE 56th International Midwest Symposium on Circuits and Systems (MWSCAS).
Scherer, D., Müller, A., & Behnke, S. (2010). Evaluation of Pooling Operations in Convolutional Architectures for Object Recognition.
Schoettle, B., & Sivak, M. (2015). Preliminary Analysis of Real-World Crashes Involving Self-Driving Vehicles (UMTRI-2015-34). Retrieved from
Sears, D. (2018). The Sightless Visionary Who Invented Cruise Control. Smithsonian. Retrieved from https://www.smithsonianmag.com/innovation/sightless-visionary-who-invented-cruise-control-180968418/
Stefanie Weber, K. T., Antonio Ernstberger,, & Volker Labenski, K. B. (2018). Different Types of Distraction Causing Accidents. In DDI 2018 Book of abstracts (pp. 4-7).
STMicroelectronics. (2020). Automotive Camera for ADAS. Retrieved from https://www.st.com/en/applications/adas/automotive-cameras.html
Sultan, B., & McDonald, M. (2003). Assessing The Safety Benefit of Automatic Collision Avoidance Systems (During Emergency Braking Situations).
Supervise.ly. (2017). Evolution: from vanilla RNN to GRU & LSTMs. Retrieved from https://towardsdatascience.com/lecture-evolution-from-vanilla-rnn-to-gru-lstms-58688f1da83a
Szarvas, M., Sakai, U., & Jun, O. (2006, 13-15 June 2006). Real-time Pedestrian Detection Using LIDAR and Convolutional Neural Networks. Paper presented at the 2006 IEEE Intelligent Vehicles Symposium.
Theofilatos, A., Chen, C., & Antoniou, C. (2019). Comparing Machine Learning and Deep Learning Methods for Real-Time Crash Prediction. Transportation Research Record, 2673(8), 169-178. Retrieved from https://doi.org/10.1177/0361198119841571. doi:10.1177/0361198119841571
Troppmann, R. (2006). Tech Tutorial: Driver Assistance Systems, an introduction to Adaptive Cruise Control: Part 1. Retrieved from https://www.eetimes.com/tech-tutorial-driver-assistance-systems-an-introduction-to-adaptive-cruise-control-part-1/
Wang, Y., & Kato, J. (2017, 29 Nov.-1 Dec. 2017). Collision Risk Rating of Traffic Scene from Dashboard Cameras. Paper presented at the 2017 International Conference on Digital Image Computing: Techniques and Applications (DICTA).
Wevolver. (2020). 2020 Autonomous Vehicle Technology Report. Retrieved from https://www.wevolver.com/article/2020.autonomous.vehicle.technology.report#reference24
Wojke, N., Bewley, A., & Paulus, D. (2017). Simple online and realtime tracking with a deep association metric. Paper presented at the 2017 IEEE international conference on image processing (ICIP).
Yang, X., & Niu, Q. (2019, 22-24 Nov. 2019). Comparison of Long-short Term Memory against Gate Recurrent Unit on Semiconductor Die Temperature Fitting. Paper presented at the 2019 IEEE 2nd International Conference on Automation, Electronics and Electrical Engineering (AUTEEE).
Yuan, J., Abdel-Aty, M., Gong, Y., & Cai, Q. (2019). Real-Time Crash Risk Prediction using Long Short-Term Memory Recurrent Neural Network. Transportation Research Record: Journal of the Transportation Research Board, 2673(4), 314-326. doi:10.1177/0361198119840611
Yuting, Z., Xiaomeng, L., Xuedong, Y., & Qingwan, X. (2015, 25-28 June 2015). Effects of collision warning system under different warning timing on driving speed and distanc. Paper presented at the 2015 International Conference on Transportation Information and Safety (ICTIS).