標題: 實現於手持式裝置之即時影像處理技術於車道偏離警示和車牌辨識之應用
Real-Time Lane Departure Warning and Plate Recognition on Hand-Held Devices
作者: 林玟欣
Wen-Hsin Lin
吳炳飛
Bing-Fei Wu
電控工程研究所
關鍵字: 車道偏離警示;手持式裝置;車牌辨識;lane departute warning;PDA;plate recognition
公開日期: 2007
摘要: 隨著時代的演變,幾乎每個人都有手機這類結合相機的手持式裝置。因此本論文著眼於手機的便利性,發展出一套結合車道線偏離警示以及車牌辨識的系統期以讓使用者在不需花錢購買額外設備情況下享受最多的服務。 在本論文中,我們透過DirectShow發展出一套能快速獲取手機內建相機鏡頭當下捕捉資料的程式並結合車道線偏離警示演算法,在偵測車道線發生偏移時會立即給予駕駛者一個警示聲藉以輔助駕駛者安全抵達目的地,並經過市面上販售的三款PDA手機的實際測試皆有很好的效果。另外當偵測出車道線連續偏移超過一定值時,本系統判斷可能會引發危險而會自行啟動錄影機制,記錄當時發生的情況,而後能交由專業人員判定相關的法律問題作為極佳的行車紀錄器。而在車牌辨識部分,本論文考量值法人員查緝贓車時的移動性和拍攝影像時可能會引發的偏差角度,提出了一套結合相機控制與角度校正的即時車牌辨識系統,期以提升執法人員在查緝贓車時的便利度與彈性。 本論文提出的車道線偏離警示和車牌辨識系統建構於作業系統為Windows Mobile 6,CPU 520MHz的PDA手機,針對LDW處理速度為20FPS,對於車牌辨識這部分,在750張影像測試下,車牌辨識率達89.7%,處理速度小於450毫秒。
As time moves forward, almost everyone has such a hand-held device with a cell phone. Therefore, this paper emphasizes on the convenience of cell phones. We developed a set of system called M-LDWS(Mobile-Lane Departure Warning System)and a system of license plate recognition to let users enjoy more service without extra equipments. In this paper, the raw data of current frame are captured by the built in CMOS camera of the PDA and to integrate the algorithm of LDW via DirectShow to give the driver warning signals when the lane departure is detected to help the driver reach the destination safely. The work has been tested through three PDA cell phones which can be purchased on market and have very good results. Besides, when detecting the number of frames for lane departure continuously over a threshold, our system determines whether it will cause danger and to switch on the recording mechanism automatically. The recording mechanism will record the situation at the moment and to use as an excellent recorder for the experts to judge the related issues of law. For the part of license plate recognition, after considering the law enforcement officers may take the pictures of plates with a slant angle when invesgating the stolen cars, the paper propose a set of system which combining the camera control and license plate recognition with slant angle correction. We expect the system will enhance the convenience and elasticity for agents. The LDWS and License Plate Recognition System are constructed through Windows Mobile 6, the PDA cell phones of CPU 520MHz。The processing time of M-LDWS is 20FPS.Under the test of 750 images for license plate recognition, the identify rates is 89.7%, and its processing time is less than 450mS.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT009512507
http://hdl.handle.net/11536/38228
Appears in Collections:Thesis