一般的影像系統都是建立在高運算執行速度的電腦平台上,然因成本、耗電量特性及體積上的侷限,本論文採用高速嵌入式DSP處理器開發智慧型車輛影像辨識系統。本文所開發的影像系統功能分為道路標線辨識、前車辨識及盲點車輛辨識,這些資訊在智慧型車輛系統中可以輔助駕駛做出行車安全的判斷,例如車道偏離警示、前車碰撞警示、輔助變換車道、前車跟隨及主動式巡航等。本文使用的嵌入式平台搭配所開發的演算法可以達到每一個畫面運算處理僅需25ms左右的處理時間,足以即時(real-time)辨識車道線位置以及前車特徵和盲點區域車輛辨識,完成一個可實際應用於智慧型車輛之嵌入式影像系統。
Image processing systems are normally built on the high performance computer platform, however, the system cost is subjected to the size and power consumption. Therefore, this research focuses mainly on the development of a low cost intelligent vehicle detection driving assistant system using high-speed embedded DSP image processor. The functions of this vision-based system include Lane Sign Detection (LSD) system, Front Car Detection (FCD) function, and Blind Spot Detection (BSD). With the aid of the developed techniques, this intelligent image processing system can provide immediate assistance and accurate driving information to the drivers, such as Lane Departure Warning (LDW), Front Collision Warning (FCW), Lane Change Assistant (LCA), and Adaptive Cruise Control (ACC). The real-time capability of this developed embedded and system algorithm requires only 25ms per frame to perform the lane mark tracking and front car positioning. The blind spot car detection can be perform individually will in 7 ms. Thus the development of this research can be deployed in the commercial embedded vision-based driving assistant system and future research of intelligent vehicle system.