傳統的事件偵測演算法(incident detection algorithms)，主要是依賴對事件發生後的交通特性的巨大變化，作為判斷事件發生之依據；必須等到事件發生之後才能得知。然而在某些車流狀況下，因為某種干擾因子，以致車流出現不種定現象，若此現象繼續擴大，就可能會發生車禍。因此，若能夠將車流不發定現象之擴大過程，使用某種方式去推演，極有可能事先預知事件之發生。本研究即依此觀念出發，將推演過程，以某種跟車模式代入，以便突破傳統之事後偵測方式，發展事前預測概念。除了可以預警事件之發生，進而達成預防之目的，亦可及早準備事件反應與管理措施，以便提高事件管理效率。由實際建立之演算法的模擬測試結果顯示，本研究所發展之方法確實可能達成預知之效果，亦即可以使偵知時間為負的，說明此研究方向可以繼續推動。若在未來加上良好的反應體系，可以事先知會駕駛員，甚至可能將此事件讓它不要發生。

Conventional incident detection methods can only detect the incident using the changed traffic patterns by incident afterwards. In this paper a new incident detection algorithm is developed with a prediction mechanism involving possible car-following procedure. A simulation program is developed to investigate the traffic parameter characteristic before an incident occurred A car-following model is set up to predict car reaction following the vehicles. The prediction mechanism is established using a new parameter with distance gap between the vehicles, which will be detected real time by using image processing sensor. According to the application results the new method can actually detect the incident in advance, i.e., with the negative detection time. Our method could be a useful one when combined with other conventional methods to enhance the detection performance in advance.