臺灣博碩士論文加值系統

在本論文中,我們發展了一套車輛牌照辨識系統。這套系統的特點是:能自
動的定位出車牌在影像中的位置,並且利用數學形態學的方法來做字元的
辨識。在許多的字元辨識系統中所用的方法,不外是分段,細化,再經由所
取得線段的特性,如:線段交點,角度等,去決定出正確的結果。本論文所使
用的方法,是採用數學形態學的方法,利用數學形態學中所提供的一些基本
運算元,萃取出每個字元的特性,再經由一個根據字元特性所建構成的決定
樹,來作出最後的判斷。本系統可分為三個部份:先前處理、車牌定位及字
元辨識等三大部分。第一部份先前處理包括影像的取得、影像二元化、去
除雜訊等,主要目的是強化影像的品質;第二部份是利用連接元件的分析,
找出車牌在影像中的位置,第三部份則是字元辨識部份。我們以40張影像
為測試樣本做實驗,結果車牌辨識率為90%,無法辨識的比例有7.5%,另
有2.5%是誤判的情形。

In this thesis, we develop a license plate recognition system.
The system first automatically locate the location of license
plate ,and then use the mathematical morphology to do
recognition. Traditional character recognition systems use
segmentation, thinning, stroke detection and so on. From the
strokes relationship, they use a similarity measure to select
best matched character. In this thesis, we use the mathematical
morphology method to extract the properties of each character
by a sequence of morphology operations. Then we can determine
the character by a decision tree which is built based on the
properties of the numerals and alphabets. The license plate
recognition system can be devided into three parts :
preprocessing, license plate localization, and character
recognition. In the first part, it includes image acquirement,
thresholding, and noise removal. The second part is to find the
location of the license plate by the connected components
analysis. The last part is mainly a character recognition
system. In experiment, we take 40 images as samples. The
results show that the successful rate is 90%, the unrecognized
rate is 7.5% and the misjudged rate is 2.5%.