在交通動線繁亂的市區,駕駛人有可能因為特別注意某個物體,或者注意力渙散,而忽略行人的存在,導致碰撞事故的發生;在夜闌人靜、交通流量稀少的時候,駕駛人也會因注意力不集中、太過於鬆懈,或者視線不明,造成不必要的事端。有鑑於此,本論文以人車在交叉路口環境下,建構一套安全警示系統,以預防人車在交叉路口的碰撞產生。我們採用TI OMAP5912嵌入式平台建置一套碰撞風險系統,並整合GPS接收器(WBT-201)、觸控式面板等周邊,搭配全球定位系統。考慮在未來,人手一機且一機多用的情形將會是常態,無線網路無所不在,手機及裝設有車上單元(On Board Unit, OBU)的車輛,可以利用無線網路取得彼此定位資訊,再由OBU內建之模糊碰撞風險評估演算法計算出人車碰撞風險的程度。此系統將針對交叉路口進行碰撞風險評估,並依不同的風險程度顯示於車上單元,給予駕駛不同程度的警示,以達到降低交叉路口事故發生率的目的。 OBU使用Qt程式來設計應用程式作為風險程度顯示操作介面,應用程式結合模糊碰撞風險評估演算法,而模糊決策以相對距離、速度、行人路口的行進方向以及行人初始位置為歸屬函數,來建構碰撞風險評估的模糊規則庫,並計算出人車可能發生碰撞的風險程度,將其結果顯示至OMAP5912介面來提醒駕駛者。
In this thesis, an image-based embedded pedestrian detection system is presented. A Safety Warning System is developed by integrating the TI OMAP5912, GPS receiver(WBT-201), and touch panel. The system will calculate the collision risk of a particular vehicle, driving through the cross-road junction, and the results will be shown on the display of an On Board Unit (OBU) of that car. Specially, the position and relative distance between pedestrians and vehicles are determined with the aid of GPS and CCD cameras mounted on road side and inside the vehicle. Then the collision risk for a particular driver is calculated base on fuzzy inference scheme. Results will be given to the OMAP5912 in the case of high collision risk to warn the driver the dangerous situation he will be confronted with. Experimental results show that the applicability of the proposed approach.