【正文】 19]))。)。 figure,imshow(bg3)。 bg2=imopen(bg3,strel(39。 title(39。通過(guò)程序處理后的圖 37 至310，可以清楚的看到最后經(jīng)過(guò)開操作濾波后的可能的車牌區(qū)域都清楚的保留了下來(lái)，是非常成功的。 （ 1） 首先對(duì)圖像每個(gè)區(qū)域進(jìn)行標(biāo)記，如圖 311 所示。這個(gè)章節(jié)是本次設(shè)計(jì)的主要內(nèi)容，在汽車牌照自動(dòng)識(shí)別過(guò)程中，字符分割有承前啟后的作用?，F(xiàn)采用以下程序求出車牌 二值圖像 的垂直投影圖和水平投影圖，如圖 41 所示： histcol1=sum(sbw1)。title(39。 title(39。 二次計(jì)算車牌水平投影 對(duì)矯正后的車牌二值子圖再次進(jìn)行水平投影計(jì)算，如圖 42： 圖 42 投影圖 從圖 42 和圖 41 中可以明顯看到，車牌的矯正是很成功的，圖 43 的水平投影已經(jīng)相當(dāng)?shù)那逦?，可以明顯看出車牌字符的高度、上邊框、下邊框位置。 minrow=min(histrow)。 for k=1:hight if histrow(k)=levelrow count1=count1+1。 石家莊鐵道大學(xué)四方學(xué)院畢業(yè)設(shè)計(jì) 21 end count1=0。 markrow(l)=hight。 %下降點(diǎn) markrow4(k)=markrow3(k)markrow(k)。 rowtop=markrow(findc)。 %字符高度 (rowbotrowtop+1) 計(jì)算車牌垂直投影 去水平（上下）邊框后，用如下 MATLAB 程序?qū)崿F(xiàn)垂直投影計(jì)算，投影如圖 44所示： histcol=sum(sbw2)。)。,int2str(maxhight)],39。)。另外由于我國(guó)車牌的標(biāo)準(zhǔn)化，車牌上的第二個(gè)字符與第三個(gè)字符的距離大于其他字符及邊框的距離，由此采用在垂直投影中找最大的峰中心距離的方法，可以確定車牌上的第二個(gè)字符的中心位置，而第二大的峰中心距離即可以確定為車牌字符的最大寬度，即定為車牌字符寬度。 count1=0。 %字符上升點(diǎn) markcol1(l)=count1。 %字符距離（上升點(diǎn)至下一個(gè)上升點(diǎn)） [m1,n1]=size(markcol2)。 markcol2(n1)=markcol(l)markcol(l1)。 %字符寬度（上升點(diǎn)至下降點(diǎn)） markcol5(k)=markcol3(k)double(uint16(markcol4(k)/2))。 markcol6(findmax)=0。 for k=findmax1:findmax+5 cleft=markcol5(k)maxwidth/2。 SegBw2 = imresize(SegBw1,[32 16])。,int2str(maxwidth)],39。)。分割出來(lái)的結(jié)果如圖 46 所示： 圖 46 識(shí)別結(jié)果 石家莊鐵道大學(xué)四方學(xué)院畢業(yè)設(shè)計(jì) 26 為了使下一步的字符識(shí)別順利進(jìn)行，識(shí)別前要對(duì)單個(gè)字符歸一化，以保證每個(gè)不同的字符識(shí)別時(shí)都有一致的標(biāo)準(zhǔn)輸入。 所謂大小歸一化，就是對(duì)不同大小的字符做變換，使之成為同一尺寸大小的字符。按識(shí)別對(duì)象，文字識(shí)別分為西文字符，阿拉伯叔祖和漢字識(shí)別等，它們由可分為印刷體和手寫體的識(shí)別，手寫體字符的字體、大小千變?nèi)f化因此此印刷體識(shí)別困難得多。 字符識(shí)別方法簡(jiǎn)介 （ 1） 模板匹配法 這個(gè)方法是把輸入的字符直接和標(biāo)準(zhǔn)的字符原型進(jìn)行比較，找到與之最匹配的模板。 （ 3） 神經(jīng)網(wǎng)絡(luò)識(shí)別方法 此方法是利用神經(jīng)網(wǎng)絡(luò)的機(jī)制來(lái)提取字符中存在的特征。 石家莊鐵道大學(xué)四方學(xué)院畢業(yè)設(shè)計(jì) 28 自動(dòng)識(shí)別 用模板庫(kù)中的字符圖像矩陣與待識(shí)字符圖像矩陣作減法，求最小誤差，即其相似度，將最相似的模板庫(kù)中的字符圖像矩陣的代碼作為識(shí)別結(jié)果，并顯示出來(lái)。首先對(duì)圖像進(jìn)行預(yù)處理，將彩色圖像轉(zhuǎn)化為灰度圖像，然后對(duì)灰度圖像采用基于灰度跳變的方法進(jìn)行二值化處理，提取車牌照邊緣信息，既保留了車牌區(qū)域的信息，又減少了噪聲的干擾 ，提高了后續(xù)處理的速度；然后運(yùn)用基于藍(lán)色象素點(diǎn)統(tǒng)計(jì)特性的方法對(duì)車牌是藍(lán)色的車牌進(jìn)行定位，結(jié)果表明準(zhǔn)確的定位了車牌照區(qū)域。 最后謝謝各位同學(xué)和朋友們?cè)诰帉懗绦蛏辖o予我的大力幫助！ 石家莊鐵道大學(xué)四方學(xué)院畢業(yè)設(shè)計(jì) 32 附 錄 附錄 A 外文資料 Car license plates recognition system Along with the increase in the number of countries around the world, city car traffic increasingly attention on how to effectively L traffic management, has bee more and more governments and the relevant departments have the focus. To solve this problem, we use the advanced science and technology, we have developed the developed various traffic surveillance, management system, these systems are usually include vehicle detection device, through these devices on passing vehicles conduct test on traffic data extraction, in order to achieve monitoring, management and directing traffic purpose car licence plate recognition technology is one of the important vehicle detection system in traffic link monitoring and control, play an important role, has a wide variety of applications, such as automatic charge system, no parking capture to expend, stolen vehicle searches, parking lot vehicle management, special department vehicle access control, etc. Meanwhile, car licence plate recognition method can also be used to other detection and identification field so car licence to identify the problem has bee a modern traffic engineering field research hotspot and focus of one problem. Car license plates recognition system is for specific goals licence plates special puter vision system, is the puter vision and pattern recognition technology application in intelligent transportation one of the major issues, it can be used widely in the traffic flow testing, traffic control and induction, airports, ports, the vehicle management, no parking fee, running red lights automatic monitoring peccancy vehicles and vehicle safety guard against theft and other fields, and has a broad application prospect. Car license plates recognition system is an important part in intelligent transportation system, is a hitech highway traffic monitoring management system one of the main functional modules. It in traditional traffic monitoring technology is introduced on the basis of digital camera, technology and puter information management technology, USES the advanced image processing, pattern recognition and artificial intelligence technology, through to the vehicle image collection and processing of the digital information, obtain 石家莊鐵道大學(xué)四方學(xué)院畢業(yè)設(shè)計(jì) 33 vehicles, so as to achieve higher intelligent management level. This system is the tsinghua university electrical engineering department, intelligent graphic information processing laboratory research and developed, it utilizes the only vehicle license plate is marking the identity to intelligent identification and ideological concepts, involving the statistical vehicle image capture, processing, understanding and recording techniques. 1. The system working principle: A plete license plate identification process generally includes the following three main steps: first, through the camera will contain car licence information car image filmed, converted to digital image, the deposit in the puter used to deal with, and some necessary pretreatment work。 (2