近年來，知名之觀光遊憩區於周休及連續假期時常湧入大量人潮造成壅塞；政府為了提升景區交通服務水準，因此擬導入先進的智慧型運輸系統(Intelligent Transportation System, ITS) 科技，使景區之交通壅塞能事前預警，防範於未然。本研究以日月潭國家風景區為例，利用現有車輛偵測器資訊提前預測景點壅塞狀況，並嘗試比較路網中不同之路段(節線)或景點(節點)資訊於景區壅塞預測之準確與可靠度績效，做為未來評估與選擇交通管制之參考。 以路段速率透過加權平滑推估進行預測，並以卡門濾波法演算進行停車場流量動態推估，並嘗試以時空數列方法對日月潭路網進行時空分析，最後以VISSIM軟體模擬日月潭交通路網。 結果顯示，以近景區之停車場評估景區壅塞較週邊路段更具優勢，且根據壅塞門檻以預測壅塞預警，可以提早60分鐘準備應便，準確率約為80%。 Due to the overwhelming tourists, congestion in some scenic area became a dead lock recently, especially in long vocation. To improve the quality of well-known scenic spot, authority introduces some advanced technologies of Intelligent Transport System (ITS) into this area, and hope that it could response the incident in advance before it gets worse. Sun Moon Lake National Scenic Area is on the top list of well-known spots cross the Taiwan Strait to be the first choice of this study area. With density layout vehicle detector (VD) data, we compare with types of link and node information in order to provide the effective performance index for traffic control in the near future. This study applies weighted average method and extended Kalman Filter for link and node performance assessments separately. Furthermore, to explore the spatial relationships of traffic congestion of upstream and downstream, STARIMA is also included based on a full data by a traffic simulation software, VISSIM. After discussion in details, the congestion of parking lot near major scenic spot is best for traffic control than those from links in surrounded area. Based on the standard threshold of congestion index, the warning call could make one hour in advance with accuracy of 80%.
