臺灣博碩士論文加值系統

隨著5G時代的來臨和自動駕駛技術的快速發展，未來的交通將會變得更加智能化。而未來的智慧交通必須建立在對交通流量資料的更進一步利用上，通過對交通流量特性的全面認知去更合理地規劃城市的建設，道路的設計，紅綠燈的時間設計等，使得交通效率得到極大提高，並且能夠有效減少碳排放。雖然5G時代還未真正來臨，但是目前世界各地的政府部門普遍有在各個交通要道安裝車流量偵測裝置，期望蒐集交通流量的資訊以幫助解決城市交通的各種擁堵問題。
以臺北市為例，臺北市交通局將車流量偵測器遍佈在台北市大大小小的路口，每五分鐘會收集一筆資料，記錄著過去一小段時間的平均車速、通過路口的車輛數等交通狀況。但由於資料量的龐大，除了統計上的數字，交通方面的專家無法更進一步有效利用這些資料形成關於交通流量特性的洞見，發掘不了隱藏在這些資料背後的更大價值。而另一個問題在於，由於感測器故障等原因，有相當一部分資料是錯誤或異常的，而且由於資料量巨大無法用人力去一一修正。本論文提出了一種利用生成式對抗網絡偵測出錯誤異常車流量的方法，並且提出了一個視覺化介面幫助交通方面的專家觀察每一個偵測器的車流量，並對交通流量的異常狀況進行進一步的判斷。

With the advent of the 5G era and the rapid development of autonomous driving technology, the future traffic will become more intelligent. In the future, smart transportation must be built on the further utilization of traffic flow data. Through comprehensive understanding of traffic flow characteristics to more rationally plan urban construction, road design, time design of traffic lights, etc., the traffic efficiency is greatly improved. Carbon emissions can be effectively reduced. Although the 5G era has not really come yet, government agencies around the world are generally installing traffic detection devices on various traffic routes, and it is expected to collect traffic flow information to help solve various traffic congestion problems in urban traffic.
Take Taipei City as an example. The Taipei City Transportation Bureau has set up traffic detectors at all intersections in Taipei. Every five minutes, a piece of information is collected, recording the average speed of the past small period of time, the number of vehicles passing through the intersection, etc. traffic conditions. However, due to the huge amount of data, except for the statistical figures, traffic experts cannot use these materials more effectively to form insights about the characteristics of traffic flow, and discover the greater value hidden behind these materials. Another problem is that due to sensor failures and the like, a considerable amount of data is wrong or abnormal, and because of the huge amount of data, it is impossible to manually correct it. This paper proposes a method for detecting traffic flow with anomalies using a generative adversarial network, and proposes a visual interface to help traffic experts observe the traffic flow of each detector and further judge the abnormal situation.

摘要 i
ABSTRACT ii
致謝 iii
目錄 iv
圖目錄 v
表目錄 vii
一、序論 1
二、相關研究 3
2.1車輛偵測器及偵測方式 3
2.2車流量預測 4
2.3車流資料填補 5
2.4異常偵測 6
2.5對抗生成網路 7
三、模型及方法 8
3.1資料和前處理 9
3.2損失函數和模型訓練 10
3.3後續處理 13
四、視覺化界面 14
4.1地圖部分 14
4.2圖表部分 16
4.3使用者交互 17
五、結果探討 22
5.1結果分析 22
5.2專家訪談 24
六、討論與結語 30
6.1 不足與限制 30
6.2 結語 32
參考文獻 33

[1] A. J. Smola and B. Schölkopf, “A tutorial on support vector regression,” Stat. Comput., vol. 14, no. 3, pp. 199– 222, 2004.

[2] M. M. Hamed, H. R. Al-Masaeid, and Z. M. B. Said, “Short-term prediction of traffic volume in urban arterials,” J. Transp. Eng., vol. 121, no. 3, pp. 249–254, 1995.

[3] W. Huang, G. Song, H. Hong, and K. Xie, “Deep Architecture for Traffic Flow Prediction: Deep Belief Networks With Multitask Learning.,” IEEE Trans. Intell. Transp. Syst., vol. 15, no. 5, pp. 2191–2201, 2014.

[4] Y. Lv, Y. Duan, W. Kang, Z. Li, F.-Y. Wang, and others, “Traffic flow prediction with big data: A deep learning approach.,” IEEE Trans. Intell. Transp. Syst., vol. 16, no. 2, pp. 865–873, 2015.

[5] X. Ma, Z. Tao, Y. Wang, H. Yu, and Y. Wang, “Long short-term memory neural network for traffic speed prediction using remote microwave sensor data,” Transp. Res. Part C Emerg. Technol., vol. 54, pp. 187–197, 2015.

[6] R. Yu, Y. Li, C. Shahabi, U. Demiryurek, and Y. Liu, “Deep learning: A generic approach for extreme condition traffic forecasting,” in Proceedings of the 2017 SIAM International Conference on Data Mining, 2017, pp. 777– 785.

[7] W. Jin, Y. Lin, Z. Wu, and H. Wan, “Spatio-Temporal Recurrent Convolutional Networks for Citywide Shortterm Crowd Flows Prediction,” in Proceedings of the 2nd International Conference on Compute and Data Analysis, 2018, pp. 28–35.

[8] Z. Cui, K. Henrickson, R. Ke, and Y. Wang, “Traffic graph convolutional recurrent neural network: A deep learning framework for network-scale traffic learning and forecasting,” 2018.

[9] Qu, L., Li, L., Zhang, Y., Hu, J.: ‘PPCA-Based missing data imputation for traffic flow volume: a systematical approach’, IEEE Trans. Intell. Transp. Syst., 2009, 10, (3), pp. 512–522

[10] L. Li, Y. Li, and Z. Li, “Efficient missing data imputing for traffic flow by considering temporal and spatial dependence,” Transp. Res. C, Emerg. Technol., vol. 34, pp. 108–120, Sep. 2013.

[11] J. Chen and J. Shao, “Nearest neighbor imputation for survey data,” J. Statist., vol. 16, no. 2, pp. 113–132, 2000.

[12] P. Cai, Y. Wang, G. Lu, P. Chen, C. Ding, and J. Sun, “A spatiotemporal correlative k-nearest neighbor model for short-term traffic multistep forecasting,” Transp. Res. C, Emerg. Technol., vol. 62, pp. 21–34, Jan. 2016.

[13] G. Chang, Y. Zhang, and D. Yao, “Missing data imputation for traffic flow based on improved local least squares,” Tsinghua Sci. Technol., vol. 17, no. 3, pp. 304–309, Jun. 2012.

[14] L. Qu, L. Li, Y. Zhang, and J. Hu, “PPCA-based missing data imputation for traffic flow volume: A systematical approach,” IEEE Trans. Intell. Transp. Syst., vol. 10, no. 3, pp. 512–522, Sep. 2009.

[15] Jin, Xuexiang, et al. “Robust PCA-Based Abnormal Traffic Flow Pattern Isolation and Loop Detector Fault Detection.” Tsinghua Science and Technology, vol. 13, no. 6, 2008, pp. 829–835., doi:10.1016/s1007-0214(08)72208-9.

[16] Martin Renqiang, et al. “Deep autoencoding gaussian mixture model for unsupervised anomaly detection.” International Conference on Learning Representations, 2018.

[17] H. Zenati, et al. “Adversarially Learned Anomaly Detection.” CoRR, (abs/1812.02288), dec 2018.

[18] Thomas Schlegl, et al. “Unsupervised anomaly detection with generative adversarial networks to guide marker discovery.” In International Conference on Information Processing in Medical Imaging, pp. 146–157. Springer, 2017.

[19] Goodfellow, Ian J., et al. “Generative adversarial nets.” In NIPS, 2014.

[20] M. Mirza and S. Osindero. “Conditional generative adversarial nets.” arXiv preprint arXiv:1411.1784, 2014.

[21] Phillip Isola, et al. “Image-To-Image Translation With Conditional Adversarial Networks” The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017, pp. 1125-1134.

[22] R. W. Schafer, “What is a Savitzky-Golay filter[lecture notes],” Signal Processing Magazine, IEEE 28, 111–117 (2011).

[23] T. Miyato, T. Kataoka, M. Koyama, and Y. Yoshida. “Spectral normalization for generative adversarial networks.” In ICLR, 2018.

[24] M. Bostock, V. Ogievetsky, J. Heer, "D3: Data-driven documents", IEEE Transactions on Visualization and Computer Graphics (InfoVis 11), vol. 17, no. 12, pp. 2301-2309, 2011.

[25]Modeling and Forecasting Vehicular Traffic Flow as a Seasonal ARIMA Process: Theoretical Basis and Empirical Results

[26] X. Jin, Y. Zhang, L. Li, and J. Hu, “Robust PCA based abnormal traffic flow pattern isolation and loop detector fault detection,” Tsinghua Sci. Technol., vol. 13, no. 6, pp. 829–835, Dec. 2008.