具有適應性的交通標誌號誌偵測與辨識; An adaptive traffic sign and signal detection and recognition

NCUIR > College of Electrical Engineering & Computer Science > Graduate Institute of Computer Science and Information Engineering > Electronic Thesis & Dissertation > Item 987654321/9742

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/9742

Title:	具有適應性的交通標誌號誌偵測與辨識;An adaptive traffic sign and signal detection and recognition
Authors:	林盈宏;Ying-Hong Lin
Contributors:	資訊工程研究所
Keywords:	號誌;交通標誌;traffic sign;traffic signal
Date:	2009-07-02
Issue Date:	2009-09-22 11:55:00 (UTC+8)
Publisher:	國立中央大學圖書館
Abstract:	交通標誌和號誌的設置目的是要提供所有駕駛人關於道路狀況的資訊，以期達到整體交通安全的目標。對於一般駕駛人而言，大約利用八成的精神專注在視覺上，二成的精神使用耳朵注意駕駛環境。為了分擔駕駛人在視覺上的負擔，若有一套交通標誌及號誌自動偵測與辨識系統，將可增加駕駛人行車的安全性及方便性。在本論文的研究中，我們將一部相機架設在車上，用來偵測交通標誌及號誌，並且辨識速限交通標誌的速度來幫助駕駛人避免超速危險的發生。我們以IHS色彩資料來做為偵測外框為紅色的交通標誌。為了增加偵測率，我們將偵測環境自動區分成白天與夜間兩大類，再分別採取不同的色彩定義進行顏色擷取。接著產生連結區塊，並以這些區塊的形狀、大小、及寬高比進行篩選。篩選完後，將區塊正規到固定大小並用樣板比對方式分成圓形、三角形及半圓形三類。針對限速標誌的內含資訊，採用影像的疊代逼近門檻值法來做二值化處理，並透過形態學的技術得到更正確的限速資訊當作放射狀基底函數網路的輸入來辨識其內容。針對高速公路看板偵測，我們對黃色、綠色及藍色看板在IHS色彩空間上分別定義其顏色，產生連結區塊後，以區塊形狀、大小、及寬高比進行篩選就可以確認其位置。由於看板顏色在高速公路上差異較為大，因此只要做完區塊篩選，就有很高的正確率。交通號誌的偵測同樣也是以IHS色彩資料來做偵測，我們分別針對紅燈、黃燈及綠燈定義出三組色彩。經由產生連結區塊並針對區塊的形狀、大小、及寬高比進行篩選。我們利用號誌燈的特性對剩餘的區塊進行確認動作。若同時偵測到兩種不同的號誌燈，則比較其面積大小，以面積大者為主。我們在不同環境下測試我們的系統。交通標誌的偵測在893個含有交通標誌的影像中成功的偵測795個標誌，正確率89%，偵測錯誤的主要原因是由於標誌距離過遠時加上車子行進間的震動無法擷取完整交通標誌外框。限速標誌的辨識在327個測試資料中正確辨識289個資料，正確率有88.3%。交通號誌偵測實驗在339張含有交通號誌的影像中正確偵測311張影像，正確率91.7%。 The traffic signs and signals are used to provide the traffic information for all drivers and pedestrians. Generally, drivers pay much visual attention to gain the traffic information. The drivers would be more released if there is an automatic traffic sign and signal detector and recognition. In this study, we propose a traffic sign and signal detection and recognition system to aid the driving for safety. The traffic signs were detected in the IHS color space. To increase the accuracy, the captured images were automatically divided into two cases: day and night. Secondly, the detected color regions were classified into circle, triangle, and semicircle based on the shape, size, and the rate of width and height. The characters in the speed limit sign were extracted and recognized by the radial basis function network. The traffic signals were also detected in the IHS color space. The extracted color regions were classified by the shape, size, and the rate of width and height to make sure the traffic signals. The proposed systems were evaluated in several variant environments. The detection rate of traffic signs is 89% with detected 795 signs out of 893 signs. The errors mostly come from the far-distance unclear signs. The recognition rate of speed characters is 88.3%. The detection rate of traffic signal is 91.7%.
Appears in Collections:	[Graduate Institute of Computer Science and Information Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Size	Format

社群 sharing

Loading...