運用影像資料進行混合車流的安全評估方式

Abstract

台灣市區道路的交通狀況為混合車流，根據政府統計數據顯示，台灣近年來的私人運具使用率佔比超過70%，且由於機車擁有高效能與高機動性等特性，因此私人運具中，又以機車更受歡迎。儘管如此，機車對交通安全的影響著實為一項不小的隱憂。機車的車輛特性與駕駛行為和小汽車非常不同，也因為這些差異，導致台灣的機車事故率與肇事率是所有車種中最高的。

過去與交通安全評估相關的文獻大都以小汽車車流為主要研究對象，近年來，機車問題在世界各地機車使用率高的地方日趨嚴重，使得各地方越來越重視增進混合車流的安全與混合車流的安全評估。傳統的安全評估法存在其限制，促使新的評估法興起，也就是事故指標法。本研究即採用事故指標法，以time-to-collision (TTC)做為可行的事故指標法，並以交通衝擊為事故指標，探討TTC是否適合評估混合車流的安全。

本研究使用台北市忠孝東路四段的車流資料進行研究，車流資料中包含了每秒的車輛軌跡資料，並根據TTC的定義設計演算法，用以判斷車流資料中產生的交通衝擊數，也就是潛在的事故數目。此外，本研究亦進一步探討演算法結果的合理性與可靠性：利用空間分布熱圖可以辨識交通衝擊集中的區域是否符合一般印象中路段上的熱點，並且，本研究使用安全績效函數去分析混合車流中的機車比例與產生的交通衝擊數目兩者關係。

結果顯示本研究利用TTC法所得到的交通衝擊數目與發生地點均為合理，且交通衝擊的空間分布熱圖可以呈現既定印象中路段上的熱點，並且，透過安全績效函數亦得到混合車流中的機車比例確實與交通衝擊數有關的結果。

Taiwan’s traffic condition on urban arterials is mixed traffic. In Taiwan, the mode share of private transportation is over 70%, and scooters are popular in which. Despite of the high efficiency and mobility of scooters, traffic safety is a major concern. The attributes, maneuvering, and behavior of scooter riders are very different from those of cars drivers. The accident rate and fatality rate of scooters are very high among the mixed traffic in Taiwan. Previous studies focused on safety studies of automobiles, and there are recent interests in the safety evaluation of mixed traffic in which there is a high ratio of scooters. Since using crash surrogates on safety analysis is emerging in the past decade, this study aims to investigate the crash surrogate measures on safety of mixed traffic. We consider the time-to-collision (TTC) as the surrogate measure. A vehicle trajectory dataset collected in Taipei city is available for this study. The dataset contains the second-by-second positions of all vehicles in the traffic stream. An algorithm is developed to analyze the traffic conflicts, which is a potential of crashes, in two steps, (i) “Trajectories Projecting” and (ii) “Overlap Checking.” Then, we can derive all the traffic conflict cases for an assumed TTC value. Two further analyses are conducted to explore the rationality and reliability of the approach. The spatial distribution of traffic conflicts on the arterial are then investigated to check the rationality of the approach. Lastly, we propose to use a Safety Performance Function (SPF) to see if the frequencies of conflicts are related to the proportion of scooters in the mixed traffic. The results show that the conflicts identified with TTC are reasonable and a good representation of the observations. The spatial distribution can show the crash prone locations. Through the conflict-based SPF regression model, it is found that the percentage of scooters in mixed traffic is actually associated with conflict frequencies.