根據聯合國世界衛生組織統計全世界每年平均有120萬人死於交通意外並有兩千萬至五千萬人在交通意外中受到傷害,造成許許多多的生命及財產損失,預計2030年車禍將成為人類第五大死因。本論文的目的是使用式嵌入式手持裝置為基礎,發展出一套電腦視覺技術的行車安全輔助系統。由於近年來嵌入式手持裝置普及化,可以將多核心嵌入式手持裝置架設在車內擋風玻璃後方,並使用裝置的鏡頭來即時擷取影像並處理和分析車道線辨識、車燈辨識、追蹤駕駛者的眼睛以及事件紀錄之功能各個模組需要的資訊。 使用擷取的影像特徵來判斷出車道線、車頭燈及車尾燈並繪出其位置。並能在車道偏移及跟車過近時來提醒行車駕駛,來避免車禍的發生。在行車時系統會去追蹤及偵測駕駛者的眼睛,是否有離開前鏡頭的視線範圍;利用MPEG4壓縮技術來即時的紀錄行車影像以便事故不慎發生時來釐清肇事的責任歸屬。最後會整合並實現一套應用電腦視覺技術的嵌入視駕駛輔助系統。
According to the statistic report of the UN World Health Organization, 1.2 million people die from traffic accidents worldwide and twenty million to fifty million people injured, resulting in loss of many lives and property. Traffic accident is expected to be the fifth cause of death in 2030. The purpose of this paper is to develop a safety driving assistant system based on handheld embedded devices with the aid of the computer vision technology. In recent years, handheld embedded devices are becoming popular and powerful, such as smart phones. We can place a multicore embedded device behind the wind shield, and use the camera on the device to process and analyze lane mark, car light recognize; this system also can track driver’s eyes and record the traffic events in real-time. The system uses features from the image captured from the camera to detect the lane mark, car lights, and find the position of these items in each image. Thus the system can warn the driver when lane departure and tailgating happens, to prevent accident happen. The system can also monitor the driver’s eye to see if the driver does not watch the road. Besides, the system can record the traffic images by using the compression of MPEG4 technology. When the accident happens the investigator can observe the accident event. Finally, the several applications of computer vision technology are be integrated into the embedded system to assist the drivers.