市區道路非重現性事件判定指標之建立

Abstract

市區道路非重現性事件發生，不僅可能對用路人生命造成極大的威脅，亦會造成交通壅塞，以致付出龐大社會成本，目前智慧型運輸系統(Intelligent Transportation System)為我國交通努力發展方向之一，透過路側所架設之CCTV(Closed Circuit Television)並搭配影像式事件偵測系統，即可達到交通管理與控制之效果，以減少交通壅塞與事故通報時間。本研究藉由CCTV所攝錄之事件發生影像過程，進行資料蒐集，並透過車流理論方法，分別以巨觀及微觀角度針對事件發生過程進行探討，以了解事件發生對於車流所造成之影響。本研究透過熵理論之概念，分別以橫向及縱向兩個維度進行探討，針對車輛於兩個方向維度之速度變化過程計算其熵值，當熵值越大，表示車輛速度變化情形越劇烈，所受干擾越大。當熵值越小，越接近0，則表示車輛速度變化極小，所受干擾越小。將計算所得熵值整理之後，更進一步建立事件判定指標，將事件判定分為確定有事件、確定無事件以及可能發生事件等三種情形，分別給予事件判定指標比例值，可以發現當熵值大於2.5時，將會被判定為有事件發生，而當熵值小於1時，將會被判定為無事件發生，至於熵值落於可能發生事件的區間內時，則可以參考事件判定指標比例值，研判事件發生的可能性。最後，透過實例應用說明本研究所建立之事件判定指標具有其實用性。

When the non-current incidents happened on urban road, it not only threatens with the drivers’ life but also increases traffic congestion. It leads to pay huge social cost for it. In Taiwan, the intelligent transportation system is one of goals that are why we’re working hard for developing it. Now, we can get the effects of traffic management and controlling by the CCTV and image incident detection system. So, we can decrease traffic congestion and reduce the time of accident announcing. This study collected the traffic flow data by taking a film from the CCTV on urban road. We discuss with the horizontal and vertical directions for using the information entropy. The entropy parameter within a calculation a vehicle speed variation on the horizontal and vertical directions. When the entropy value is higher, the vehicle speed has more violent variation. It shows that the vehicle runs into the conflict many times. When the entropy value is smaller and it closed to zero, the vehicle speed has less violent variation. It shows that the vehicle drives freely. We can develop the entropy indication parameter for non-current incidents after analyzing these calculated entropy values. We sort the incident indication by three situations: true incident, no incident and incident maybe happened. Then we can calculate the incident indication ratio value. This study find out the entropy value greater than 2.5, it is defined the “true incident.” If the entropy value is smaller than 1, it is defined the “no incident.” Other entropy values have to depend on the incident indication ratio value, and infer an entropy value representing the percent of an incident happened. Therefore, we can indicate an incident happening. Finally, this study explained the incident entropy indication parameter with a real case that was about a scooter accident. Therefore, we could apply it and proved its practicability.