現今的交通事故頻繁,事故頻繁的原因可能是駕駛人不專心沒有注意車輛周圍環境的狀況,才會發生悲劇,尤其是車輛周圍駕駛人看不見的盲點,更是發生事故的原因,所以本論文以電腦視覺為基礎開發行車輔助系統,系統主要的功能就是幫助駕駛人了解車輛周圍環境的相關資訊,尤其是在高速公路行進中的車輛,因為在高速公路上,車輛行進的速度每小時有100 公里以上,只要駕駛一不留神就很容易發生事故,在事故之前駕駛人可以反應的時間非常的短,所以利用電腦視覺技術開發前方車輛和後方車輛的偵測,在30~50 公尺的距離內可以先找出車輛,讓駕駛人能夠知道目前周圍環境車輛的資訊,增加駕駛人的反應時間,為了配合高速行駛的車輛,所以系統處理影像時的反應時間不可以超過0.98 秒,然而本論文所提出的的系統每秒可以處理20~40 畫格,也就是20ms~50ms 就能即時完成一張影像的處理,系統包含了道路偵測和車輛偵測,道路和車輛偵測的方法是利用基本影像處理方式,包含影像灰階、影像二值化、影像區域分群的方式處理,最後將原始影像轉換成俯視圖輔助駕駛人行車安全。
Traffic accidents are frequent nowadays. The high frequency of accidents may be because drivers do not pay enough attention to other vehicles or environmental conditions. In particular, the fact that a driver cannot see vehicles in the blind spot can cause accidents. Therefore, this thesis develops a driving assistance system based on computer vision. The system main function is to help the driver to identify the vehicles from the environment-related information. Vehicles move quickly on the highway, often approaching speeds of over 100km/h. Therefore, accidents can easily occur if drivers do not pay attention. The reaction time is often extremely short. The proposed system can identify vehicles at a distance of 30–50m, providing a driver with information about vehicles in the current environment, and enhancing the driver reaction time. Therefore, the system processing time must not exceed 0.98 seconds. However the system may process 20–40 frames in a second. A processing cycle can thus be completed in 20–50ms. The system includes both land detection and vehicle detection. The land and the vehicles detection use the basic image processing methods, which contains the gray scale, binary and region processing function. Finally, the original video image is mapped to be a vertical view displayed on the driver’s screen.