即時相機校正之前車偵測

Front Car Detection with Real-Time Camera Calibration

李濠欣(Hao-Hsin Li)

指導教授 ： 丁肇隆

共同指導教授 ： 張瑞益

國立臺灣大學/工學院/工程科學及海洋工程學研究所/碩士(2009年)

https://doi.org/10.6342/NTU.2009.00008

摘要

近年來世界各國漸漸的將智慧型運輸系統(Intelligent Transport System, ITS)視為重點發展項目。智慧型運輸系統的主要目的就是增進安全並且減少車輛、運輸時間以及燃料上的耗損。而車輛主動式安全系統 (Vehicle Active Safety System, VASS)是屬於智慧型運輸系統其中一環，利用架設在車輛上的感測器去收集週遭訊息並提醒駕駛人潛在的危險以減少意外事故之發生。本系統的目標是發展一套智慧型的行車輔助系統(Driver Assistance System, DAS)。藉由裝設在實驗車輛上的攝影機，配合電腦視覺及影像處理技術達成辨識車道線以及前方車輛的目的。在車道線偵測方面，我們採用了三個特徵，分別是高亮度、細長性及連續性特徵來設計演算法。而在前車偵測方面，由於白天及夜間的特性不同，我們分別使用車底陰影、車輛邊緣及車尾燈做為我們辨識的依據。除此之外，我們還利用消失點在影像上的特性，開發一套自動相機校正機制，進而提高距離估算的準確性。實驗結果顯示，車道偵測率接近99%，而車輛辨識率也可高達96%。這結果也顯示出我們所提出的演算法是可以有效地滿足實際上的需求。

關鍵字

行車輔助系統 ； 車道偵測 ； 車輛偵測 ； 相機校正

並列摘要

In recent years, Intelligent Transportation System (ITS) has drawn more and more attention in the world. The goal of ITS is to improve traffic safety and to reduce transportation times and fuel consumption of vehicles. Vehicle Active Safety System (VASS) is one subject of ITS. By different kinds of sensors mounted on vehicles, it can collect surrounding messages and notice drivers to avoid potential hazards. Our objective is to research and develop an intelligent Driver Assistance System (DAS), which is one kind of VASS. This system utilizes a monocular camera mounted on the experimental car, and applies computer vision and image processing techniques to detect lane markings and front vehicles. In the lane detection section, we combine three lane markings features: brightness, slenderness and continuity to design our algorithm. In the front car detection section, shadow beneath a car, vertical edges of car sides and taillight positions are used to recognize front car positions in daytime and nighttime respectively. In order to evaluate the relative distance of objects more exactly, we address an automatic camera calibration method based on the vanishing point location. The experimental results show that the recognition rate of lane detection approximate 99% and the recognition rate of front car detection is about 96%. It is concluded that the proposed recognition algorithm works effectively and very well.

並列關鍵字

DAS ； Lane Detection ； Car Detection ； Camera Calibration

參考文獻

[2] M. Bertozzi, and A. Broggi, “GOLD: a parallel real-time stereo vision system for generic obstacle and lane detection,” IEEE Trans. on Image Processing, Vol.7, No.1, pp.62-81, Jan. 1998.

[3] T. Bücher, “Measurement of distance and height in images based on easy attainable calibration parameters,” in Proc. IEEE Intelligent Vehicles Symp., pp.314-319, Oct. 2000.

[5] Y. Wang, E. K. Teoh, and D. Shen, “Lane detection and tracking using B-Snake,” Image and Vision Computing, vol. 22, pp. 269-280, 2004.

[7] H. Lin, S. Ko, S. Wang, Y. Kim, and H. Kim, “Lane departure identification on Highway with searching the region of interest on Hough space,” International Conference on Control, Automation and Systems, COEX, Seoul, Korea, Oct. 17-20, 2007.

[8] S. Zehang, G. Bebis, and R. Miller, “On-Road Vehicle Detection: A Review,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 28, no. 5, pp. 694-711, May, 2006.