基於影像之智慧路況發布系統

Abstract

近年來，隨著智慧城市的發展，智慧型運輸系統的應用成為了一個熱門的主題。隨著電腦視覺技術與行動通訊技術的提升，眾多基於視覺方法的智慧交通應用被廣泛用於城市治理，包含交通路況監控系統、車牌辨識系統、行人偵測系統與自動駕駛系統等。然而，影像處理仍存在許多問題與挑戰，例如對於影像來源的品質需求高、資料處理需要強大的電腦運算能力、以及在大量資料傳輸的同時需要維持低延遲的特性等。因此，本研究開發了一套基於影像之智慧路況發布系統，透過YOLOv5x與DeepSORT演算法進行車輛辨識與追蹤，用以獲取車輛種類、數量、經緯度、速度等道路交通資訊，未來可以更進一步分析出車輛違停、超速、蛇行及闖紅燈等交通安全訊息。最終將相關資訊轉換為SAE 所制定的 J2735國際通用格式，並透過4G-LTE通訊系統的模擬工具進行傳輸，最後於用戶端利用HTML網頁進行顯示。

In recent years, with the growing demands from smart city development, the research of intelligent transportation systems has become a popular topic. Due to the development of computer vision and mobile network technologies, plenty of vision-based applications in ITS are applied for city management, for example, traffic monitoring, license plate recognition, pedestrian detection, and autonomous vehicle. However, there are also some challenges associated with using vision-based methods, such as a need for the high-quality video source, the computational cost of computer vision algorithms, and the ability to transfer large amounts of data with low latency. To deal with the problem, we proposed a system that uses video data from government CCTV to obtainand broadcast basic safety messages (BSM) through the 4G-LTE eMBMS broadcasting technique. By detecting and tracking vehicles using YOLOv5x and DeepSORT, we can estimate each vehicle's speed, location, and heading to generate BSM in the SAE J2735 standard. Finally, to simulate the vehicle communication system, we utilize the srsLTE tool to transmit BSM between devices and display it on the HTML web page. This system has the potential to be widely used in ITS applications and improve smart city implementations.