- 學位論文
植基於電腦視覺之道路標線辨識技術之研究與實作
A Study and Implementation of Lane Marks Recognition Technique Based on Computer Vision
摘要
交通事故經常因駕駛的疏忽而發生。在快速道路上，因為車速快，所以更容易發生重大交通事故。為了還原事故發生的情況，最直接的方法就是調閱攝影機的記錄畫面。然而，並不是每一個路段都有攝影機。因此，人們才會想在車輛上加裝攝影機，而這個車載攝影裝置就稱為行車記錄器。現在，已經有越來越多的駕駛安裝行車記錄器以記錄行車的情況。行車記錄器已漸漸成為車輛的標準裝置。 近年來，由於影像處理技術與硬體計算能力的進步，以電腦視覺技術為基礎應用於行車環境上的研究也大幅地增加。然而，在實際應用上還有很多困難需要解決。因此，相關的研究仍有很大的進步空間。然而現在的研究文獻中，多著重於日間、車輛較少和道路標線環境較單純的情況下之研究。在車輛較多和道路標線較複雜的行車環境情況下，由於行車環境的變異性大幅地增加。因此，相關的研究報告也較少。 本研究論文以道路標線之偵測及辨識為主要的研究重點。實驗上所使用的資料為行車記錄器的影片。本研究會先輸入原始影像作縮減取樣(down-sampling)來減少處理時間。然後再根據光線強度所造成的道路灰階值作處理，以取得重要辨識區域之分隔線，來減少複雜環境的干擾。因此，減少邊緣偵測和Otsu臨界值演算法的處理時間，以正確辨識車道標線，來提醒駕駛者注意車道標線的變換。本研究將會使用感興趣區域(ROI, Region Of Interest)、邊緣偵測、形態學(Morphology)和Otsu臨界值演算法來取得標線的紋理特徵，以讓電腦正確的去辨識。本研究論文所設計的演算法，能提高辨識的準確性，增加駕駛輔助的可行性。實驗結果顯示，本論文所提出的方法可以達到不錯的標線辨識準確性。
並列摘要
Traffic accident usually occurred is at the driver’s negligence. On the expressway, because the speed of vehicle is faster, it is easy to cause of the traffic accidents. In order to restore the situation of the traffic accident, the most direct way is to check the record screen of the camera. However, it is not that every the section of road has a camera. Therefore, people will want to install the camera on the vehicle, this camera is called event data recorder. Now, there have a lot of drivers to install the event data recorder to record traffic condition. The event data recorder has gradually become the standard equipment of the vehicle. In recent years, due to the image processing techniques and hardware computing power advances, based on computer vision technology is applied to the research of the driving environment also significantly increment. However, in practical application, there are many difficulties to be resolved. Therefore, the related research still has a lot of space for improvement. However, the current research literature, it the focus of study in the daytime, few vehicles, and lane marking pure driving environment. In the case of more vehicles and lane marking complex driving environment, it’s the variability of driving environment will significantly increased. Therefore, the related research papers also less. This research paper is the detection and recognition of the lane marking for major study focus. The experimental datum used of the movie of the event data recorder. First, this study will input the original image to down-sampling to reduce the processing time. Then, the road caused by the intensity of the light of grayscale value, in order to obtain the important recognition region dividing line, to reduce the interference of the complex environment. Therefore, it is by reducing the processing time of the edge detection and Otsu threshold algorithm, in order to correct recognition of lane markings of remind driver to pay attention to the transformation of the lane markings. This research will use the region of interest, edge detection, morphology, and Otsu threshold algorithm to obtain the texture features of the markings to let the computer correct recognition. In this research paper the design of algorithms, can be improve the recognition accuracy and increase the feasibility of driver assistance. The experimental results show that our proposed method can achieve good recognition accuracy of marking.
參考文獻
延伸閱讀
- Wu, Y. J. (2004). 電腦視覺為基礎之交通辨識技術應用於車載駕駛輔助與交通監控研究 [master's thesis, National Taiwan University]. Airiti Library. https://doi.org/10.6342/NTU.2004.01506
- 陳佶宏（2014）。應用電腦視覺於道路標線品質評鑑之研究〔碩士論文，義守大學〕。華藝線上圖書館。https://doi.org/10.6343/ISU.2014.00287
- 余文德、劉得廣、賴建宏（2015）。以知識管理為基礎之工程設計輔助系統研究—以道路工程設計為例。營建管理季刊，(100)，33-44。https://doi.org/10.6505/CMJ-2015-100-33-12
- 林庭慶（2019）。Case Study of an Automated Optical Inspection Based on Computer Vision〔碩士論文，朝陽科技大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0078-0510202013411415
- Chung, C. W. (1982). Computerized optimization of highway geometric alignment design [master's thesis, The University of Hong Kong]. Airiti Library. https://www.airitilibrary.com/Article/Detail?DocID=U0029-1812201200005495