近年來，台灣交通事故頻發，其原因與車輛總數增加和駕駛行徑息息相關。駕駛行駛在路上時，如遇到精神狀態不佳，或是突發之生理反應情況下，可能出現不安全的用路行為，此時，如果能及時偵測駕駛行為之異狀，則能提早避免意外之發生。 在汽車安全的領域中，隨著科技的發展，各式各樣的安全配備逐漸普及到所有車中，包括安全帶、防鎖死煞車系統、安全氣囊、自動緊急煞車系統等，上述保護措施旨在發生緊急情況時，車輛能最大限度的保護駕駛與乘客。然而，交通事故的成因不只有車輛本身的故障，駕駛行為也會造成交通事故，但是針對從駕駛方面改善行車安全與預防事故發生，則缺少詳細的研究。 危險駕駛行為極易造成交通事故，要識別這些危險行為，則需收集大量相關危險行為資料，收集過程不僅對測試者與實驗人員本身具有一定風險，對其他用路人也是一項危害。本團隊過去研究中，試圖透過虛擬駕駛環境收集資料，並應用到真實環境的駕駛行為判斷上，但虛擬駕駛環境的資料與真實環境的資料有一定落差造成判斷效果不如預期。 本研究著重在駕駛行為的分析上，透過萃取出駕駛人在不同環境之駕駛行為，能對後續研究的分類模型提供更優良的資料，從而幫助分類模型來判斷駕駛本人的行為，如果分類模型能正確地判斷駕駛當前行為之安全性，就有機會為後續的警告與通知爭取時間，進而達到預防事故之發生。 ;In recent years, there were many traffic accidents happened. The reason behind those tragedy was related to the driver’s behavior. When driver driving on the road in poor mental health, it may cause unsafe driving behavior. In this case, if we can detect the strange driving behavior, we can avoid the accident happening. In car safety field, many safe equipment is working on the modern car. safe equipment like seat belt, anti-lock braking system, supplementary and autonomous emergency braking system have an important role to protect driver from danger. However, the reason that cause car accident is not only the car itself, but also the driver behavior. There wasn’t much detail research on preventing accident on driver’s perspective. Unsafe driving behavior will easily cause car accident. To detect dangerous behaviors, we need to collect many unsafe behavior data. The process of collecting unsafe behavior were dangerous to subjects and research team. In our previous research, we tried to collect data from simulation environment and use it to classify real environment’s data. But there is a difference between simulation environment data and real environment data. This research is focus on driving behavior analysis. We extract driver’s behavior from different environment and use on classifier model. If the classifier model can detect the safety of current driving behavior, we can have chance to prevent car accident from happening.
