台灣地區腹地狹小，私人交通工具使用率遠高於大眾運輸使用率，代表著私有車輛持有率極高，車流量和道路狀況在不同時間點會有不同的狀況，更突顯出交通路況資訊對於用路人的重要性。近年隨網際網路通訊技術及移動型電子設備發展迅速，可隨時輕易獲取大量資訊，且使用地點不受侷限。目前各地道路的相關資訊，Google Maps已將其大量資料統整並描繪出完整地圖，不僅可展示普通地圖，更擁有龐大的影像資料庫所建立的實際街景地圖，這些街道實景影像能讓使用者猶如置身其中，得以了解路線並抵達不熟悉的目的地。然而，因所需收集資料量非常龐大，部分地區街景影像時間久遠，導致與現況實景有所出入。本論文提出一套結合現有資源並因地制宜的系統，不同於Google Maps以顏色區分各路段之路況，本系統以圓圈大小展示各路段路況壅塞程度。本系統運用群眾外包 （Crowdsourcing） 的模式，收集資料提供者即時拍攝的路況影像，範圍可由主要道路涵蓋至巷道內，使道路資訊即時呈現且更加完整。此系統將提供查詢者數據訊息並包含影像資料，多項資料彙整至單一頁面，毋須分別點選觀看。除系統提供之主要資訊，其影像資料也蘊含許多背景資訊，如天氣、路邊停車狀況…等訊息，使用者也能透過影像觀看而獲得更多額外資訊，收集之資料庫也可做為後續研究的訓練資料。

Nowadays, private vehicles have been popularized in our life. Despite the fact that private vehicles have brought a lot of convenience to people, many problems have also emerged accordingly. Taiwan is one of the most densely populated countries in the world, traffic congestion has become a common daily occurrence. Obtaining real-time traffic volume estimates is essential in using the limited road space and traffic infrastructure. Google Maps can be used to show a location, but only provides old street-level imagery from the Street View's archives.

The proposed method has been collecting and accumulating traffic data such as GPS positions, traffic speed, number of vehicles and images from all the drivers. The system is implemented by combining crowdsourcing, which involves obtaining information from a large group of people who submit their data via the Internet and an android-based mobile application. For those submitted images, a vehicle detection is conducted based on YOLO v2 deep learning algorithm. A complete information is then extracted, including map, images, average speed and average number of vehicles. This information can be used as a training data in further research as well.

致謝 II
摘要 III
Abstract IV
目錄 V
圖目錄 VI
表目錄 VII
第1章 緒論 1
1.1 研究動機與目的 1
1.2 章節架構 2
第2章 文獻探討 3
2.1 手機相關應用程式 3
2.2 車輛辨識 4
第3章 研究方法 7
3.1 資料收集 9
3.1.1 資料來源平台及接收資料 9
3.1.2 道路車輛偵測與壅塞程度評量 12
3.1.3 資料庫建立 13
3.2 資訊查詢 16
3.2.1 路況及車流資訊網頁視覺化界面 16
3.2.2 鄰近資料篩選 17
第4章 行有餘利-玩樂東台灣 18
4.1 點數積分合作模式 18
第5章 實驗結果 20
5.1 資料收集與介面 20
5.2 資訊查詢與介面 21
第6章 結論 25
參考文獻 26

[1]https://cctv.taskinghouse.com/
[2]F. Kerouh and D. Ziou. "Real-Time Android Application for Traffic Density Estimation," 2018 IEEE Access, 2018.
[3]K. Fukushima, S. Miyake, and T. Ito. "Neocognitron: A neural network model for a mechanism of visual pattern recognition," IEEE transactions on systems, man, and cybernetics, （5）, 826-834, 1983.
[4]R. Girshick, J. Donahue, T. Darrell and J. Malik. "Rich feature hierarchies for accurate object detection and semantic segmentation," In Computer Vision and Pattern Recognition （CVPR）, 2014 IEEE Conference on, pages 580–587. IEEE, 2014.
[5]R. Girshick. Fast R-CNN. arXiv:1504.08083, 2015.
[6]S. Ren, K. He, R. Girshick, and J. Sun. "Faster R-CNN: Towards real-time object detection with region proposal networks," In NIPS, 2015.
[7]W. Liu, D. Anguelov, D. Erhan, C. Szegedy, and S. E. Reed. "SSD: single shot multibox detector, " CoRR, abs/1512.02325, 2015.
[8]J. Redmon, S. Divvala , R. Girshick, A. Farhadi. "You only look once: Unified, real-time object detection," In CVPR, 2016.
[9]J . Redmon and A. Farhadi. "Yolo9000: Better, faster, stronger, "In Computer Vision and Pattern Recognition （CVPR）, 2017IEEE Conference on, pages 6517–6525. IEEE, 2017.
[10]https://pjreddie.com/projects/