應用共變異矩陣表示法之加強型交通號誌偵測與辨識系統

An Enhanced Traffic Sign Detection and Recognition System Using Covariance Matrix Descriptor

李政緯(Cheng-Wei Lee)

指導教授 ： 陳湘鳳

國立臺灣大學/工學院/機械工程學研究所/碩士(2011年)

https://doi.org/10.6342/NTU.2011.00167

摘要

隨著汽車數量與肇事數件遽增，智慧汽車系統逐漸受到重視。交通號誌偵測與辨識系統是智慧汽車系統中相當倚重的子系統，因為它能提醒與提供駕駛道路資訊。本系統含有四個模組:前處理模組、訓練模組、偵測模組和辨識模組。在偵測上，我們利用滑動視窗法在測試影像的任何位置來偵測不同大小的號誌，對於每個視窗採用共變數矩陣表示法作為特徵描述的方法，並判斷是否為是號誌。除此之外，採用Adaboost 演算法和分層偵測器 (cascade detector)的概念來縮減運算處理時間與降低假陽性率(false positive rate)；最後利用多類別支持向量機(multi-class Support Vector Machine)進行道路號誌辨識。提出的演算法分別在晴天狀況與四種干擾情形: 遮蔽、褪色、背光和模糊景色下進行測試。根據實驗結果，我們所提出的穩健系統可適用於任何環境且無論在偵測或是辨識交通標誌上，都具有高度準確率。

關鍵字

交通號誌偵測與辨識 ； 共變數矩陣表示法 ； Adaboost 演算法 ； 多類別支持向量機

並列摘要

Intelligent vehicle (IV) systems have gathered great importance in recent year. Many driver assistance systems have been developed to improve driving safety. Traffic sign recognition system is an important subsystem of driver assistance system because it can remind the drivers of the road sign information. The proposed system comprises four modules: preprocessing, training, detection and recognition. In detection phase, a sliding window is applied to the test image in different scales. For each sliding window, we compute covariance matrix descriptor for feature extraction, and determine whether it is a sign or not. Moreover, in order to reduce computational time and false positive rate, the detector is built by Adaboost algorithm and the cascaded decision. In recognition phase, we perform the sign identification by using multi- class Support Vector Machine (SVM). The proposed algorithms were tested in sunny conditions and four different noisy outdoor scenes: occluded, faded, backlight and blurred conditions. From the experimental results, the proposed system shown high performance to detect and recognize traffic signs.

並列關鍵字

traffic sign detection and recognition ； covariance matrix descriptor ； Adaboost algorithm ； multi-class SVM

參考文獻

[3] M. K. Hu, "Visual pattern recognition by moment invariants," IRE Transactions on Information Theory, vol. 8, pp. 179-187, 1962.

[4] K. H. Lim, L. M. Ang, and K. P. Seng, "New hybrid technique for traffic sign recognition," IEEE International Conference on Symposium on Intelligent Signal Processing and Communication Systems (ISPACS) , Swissotel Le Concorde, Bangkok, Thailand, pp. 1-4.,2008.

[5] J. P. Carrasco, A. de la Escalera, and J. M. Armingol, "Driving supervision through traffic sign analysis," IEEE International Conference on Vehicular Electronics and Safety (ICVES), pp. 42–247, 2008.

[6] Y. Y. Nguwi and A. Z. Kouzani, "Detection and classification of road signs in natural environments," Neural Computing & Applications, vol. 17, pp. 265-289, 2008.

[8] Y. Fatmehsan, A. Ghahari, and R. Zoroofi, "Gabor wavelet for road sign detection and recognition using a hybrid classifier," IEEE International Conference on Multimedia Computing and Information Technology (MCIT) , pp. 25-28, 2010.