TW544643B - Method for cropping area of license plate from vehicle image - Google Patents

Method for cropping area of license plate from vehicle image Download PDF

Info

Publication number
TW544643B
TW544643B TW90110883A TW90110883A TW544643B TW 544643 B TW544643 B TW 544643B TW 90110883 A TW90110883 A TW 90110883A TW 90110883 A TW90110883 A TW 90110883A TW 544643 B TW544643 B TW 544643B
Authority
TW
Taiwan
Prior art keywords
license plate
range
vehicle image
pixel
gradient
Prior art date
Application number
TW90110883A
Other languages
Chinese (zh)
Inventor
Kuen-Rung Wu
Bo-Shuen Jeng
Heng-Sung Liou
Rung-Ming Chen
Shin-Liang Lai
Original Assignee
Chunghwa Telecom Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Chunghwa Telecom Co Ltd filed Critical Chunghwa Telecom Co Ltd
Priority to TW90110883A priority Critical patent/TW544643B/en
Application granted granted Critical
Publication of TW544643B publication Critical patent/TW544643B/en

Links

Landscapes

  • Image Processing (AREA)

Abstract

The present invention provides a method for cropping area of license plate from the vehicle image, which includes the following steps: using the CCD camera with associated lens and the image capturing card to take the vehicle image on the lane; using the vehicle image reading unit to read out the image captured by the image capturing card; next, using the logarithmic gray level operation unit for pixels to calculate the logarithmic gray levels for each pixel in the vehicle image; then, using the horizontal gradient operation unit for logarithmic gray levels to calculate the horizontal gradients of the logarithmic gray levels on each pixel; and, using the binary operation unit for horizontal gradients to convert the real number of the horizontal gradients into binary values, 0 or 1; finally, using the searching unit for the most dense area of binary horizontal gradient to search the area with the highest sum of binary horizontal gradients in the whole vehicle image according to the default length and width values of the license plate, and cropping out the area as the location of the license plate area.

Description

經濟部智慧財產局員工消費合泎社印:tEmployees' Consumption Cooperative Press of the Intellectual Property Bureau of the Ministry of Economic Affairs: t

15 20 10 【技術領域】 別是車靖之車牌區域切取方法,裝 5門禁的車牌區二吏 動辨識系統及交通監控、車輕 【先前技術】像切取的車輛影像車律區域方法。 辨識二像中之車牌區域的定位及切取是車牌自動 整體性能;此外,在#、心+ …車碑目動辨識的 管理人員中I〆 乂空上乃至於車輛門禁管理上, 機及顯示㉞看來往車輛的車牌號 二=目動f車牌區域的部嶋切取並顯示出來,將 像中恭2人貝晛看。因此’如何自動且有效的從車輛影 找出車牌的位置所在並予以切取出來,便是車牌自動 辨識减及交通監控、車輛門禁等㈣㈣要課題。 j於T輛衫像中之車牌區域的定位及切取,一般係直 接使用像素灰階值來運算獲得結果,但此方式容易因光線 化碰影使得灰階值的運算結果發生誤差而造成車牌區 域的定位錯誤。 彩像中母個像素的灰階俊係由明亮度及物體反射強度 等兩個成分相乘(Muwplication)而得(參考Rafad c_ Gonzalez 及 Richard R Woods 所合著的 Digitanmage pr〇⑽邮, 1993 );明亮度係由外在的光線強弱所決定,而物體反射 強度才真正是反映物體本身的影象特性。 由此可見,上述習用物品仍有諸多缺失,實非一良善 丨!!·裝·1 丨(請先閱讀背面之注意事項再—貪) J I J J . .參15 20 10 [Technical field] Especially the method of cutting out the license plate area of the car, equipped with a five-access control license plate area, a motion recognition system, traffic monitoring, and vehicle light. [Previous technology] A method of extracting a vehicle image from a car law area. The identification and extraction of the license plate area in the second image is the automatic overall performance of the license plate. In addition, among the management personnel who ## 心 +… It seems that the license plate number 2 of the vehicle is cut and displayed on the license plate area of the moving f license plate, and it will be seen by two people like Zhong Gong. Therefore, how to automatically and effectively find out the location of a license plate from a vehicle image and cut it out is an important issue such as automatic license plate identification and reduction, traffic monitoring, and vehicle access control. The positioning and cutting of the license plate area in the T-shirt image is generally directly obtained by using the pixel gray level value to calculate the result. However, this method is prone to the error caused by the grayscale value calculation result due to the light and shadow. Positioning error. The gray scale of the mother pixel in the color image is obtained by multiplying the two components (brightness and reflection intensity of the object) (Muwplication) (refer to Digitanmage pr0⑽, co-authored by Rafad c_ Gonzalez and Richard R Woods, 1993) ; Brightness is determined by the intensity of the external light, and the reflection intensity of the object really reflects the image characteristics of the object itself. It can be seen that there are still many shortcomings in the above-mentioned custom items, which is not a good one! !! · Equipment · 1 丨 (Please read the precautions on the back first—greedy) J I J J ..

If n n 94 δ— u n 9— ‘紙張尺度適周中國國家標準(CNS)A4規袼(210 x 297公f \ σ又扑者,而亟丨寺加以改良q 法所μ"人鑑於上述㈣車姉像之車牌區域切取方 :生的各項缺點’乃亟思加以改良創新,並經多车苦 後,终於成功研發完成本件車輛影像之車 【發明目的】 的自t㈣之㈣即在於提供—輯車«像之車牌區域 10 、:刀取方法’可應用於車牌自動辨識及交通監控、車 兩門禁之車輛影像的車牌區域切取方法。 【技術内容】 15 味可達成上職明目的之車輛影像㈣牌區域切取方 )要賴攝職拍取車輛的影像,並從料畫面中各 爾I之騎灰難的二錄水平梯度分布敎來搜尋定 牌區域在整張車輛影像中的位置,並將車牌區域切取 m I發明包括搭配鏡頭的CCD攝影機、影像捕 捉r及車辆影像讀取單元、像素之對數灰階值運算單元、 f數灰階值之水平梯度運算單元、水平梯度之二值化運算 早7L、二值化水平梯度最密區域搜尋單元等模組。 20 一影像中每個像素的灰階值係由明亮度及物體反射強度 等兩個成分相乘而得(參考C. G_I_ Rlchard E. W〇〇dS所合者的D袖aUm哪Processing, 1993 );明亮度係由 :在的光線強弱所決定’而物體反射強度才真正是反映物 體本身的影像特性。灰階渲經過對數運算所得到的對數灰 曰'直岛二十文文明兀度(Logarithmic illumination )與對數反射強 544643If nn 94 δ— un 9— 'The paper size is in accordance with Chinese National Standards (CNS) A4 Regulations (210 x 297 male f \ σ), and it is urgent to improve the Q Law Institute ’s view of the above The license plate area of the sister ’s image: the various shortcomings of the birth of 'is eager to improve and innovate, and after suffering from multiple cars, and finally successfully developed the car that completed the image of this vehicle [invention purpose] is to provide —Edition of the vehicle «image plate area 10 :: knife extraction method 'can be applied to the method of automatic license plate identification and traffic monitoring, and the vehicle license plate area extraction method of the two-door vehicle image. [Technical content] 15 We can achieve the purpose of the job The vehicle image of the license plate area is obtained from the camera. The position of the license plate area in the entire vehicle image is searched from the horizontal gradient distribution of the gray record of each rider. The invention of the license plate region is taken from the invention. The invention includes a CCD camera with a lens, an image capture r and a vehicle image reading unit, a logarithmic grayscale value calculation unit for pixels, a horizontal gradient operation unit for f-number grayscale values, and a horizontal gradient two Valued operations Early 7L, binarized horizontal gradient most dense area search unit and other modules. 20 The grayscale value of each pixel in an image is obtained by multiplying the two components such as brightness and reflection intensity of the object (refer to D. a.U.M. Processing, 1993, C. G_I_ Rlchard E. W〇〇dS ); Brightness is determined by the intensity of the light in the ', and the reflection intensity of the object really reflects the image characteristics of the object itself. Logarithmic gray obtained by logarithmic gray scale rendering Logarithmic illumination and logarithmic reflection are strong 544643

、發明說明 A7 B7 PA890608.AiVn , 5/; 經濟部智慧財產局員工消費合作社印製 度(Logarithmic reflectance) 之和。 叙設影像畫面的解析度(Resolution)為M*N㈣像素, 且在座標位置(X,y)的 &lt;象素之灰階值為g(x,y),0&lt;X&lt;M+卜0 &lt; y、Ν·τ1,g(x,y)的值可直接從CCD攝影機的輸出值來讀 取。若座標位置(x,y)的像素之明亮g%i(x,y)、反射強度為 r(x,y) ’ 則 g(x,yH(x,y)X r(x,y),i(x,y)的值及 r(x,y)的值無法從 CCD攝影機的輸出值來讀取。經過對數運算後,對數灰階 值為奶x,y)、對數明亮度為r(x,y)、對數反射強度為r,(x,y), 即 10 gf(x?y)= In g(x?y) =111 i(x,y) + In r(x,y) 二 i’(x,y) + r’(x,y),〇 &lt; x &lt; m+卜 〇〈 y〈 N+1。 我們以[g,(x+1,y) / 2 i故為座標位置(x,y)之像素 的對數灰階值水平梯度,而 b [g?(x+l?y) &lt; gHx-l^y)] / 2 [if(x+i?y) + Γϊχ+Ι,ν), iT{x.1?y) ^ f!(x. l,y)]/2;由於鄰近像素間明亮度的差異甚小,因此卩(奸 -iT(x-l,y)之值非常小而可忽略;所以座標位置(x,y)的像素的 對數灰階直水平梯度非常接近於吻彻^吻韓。如 此一來,光線變化或陰影對影像的影響便被消除。本發明 20係對像素灰階值施以對數運算,並利用對數灰階值來計算 定位出車牌的位置所在,所以不會因光線變化或陰影使得 灰階匱發生變化而造成車牌區域的定位錯誤。 (丁 本發明主要f系藉由搭配鏡頭的CCD攝影機及影像捕捉 下對車逗攝取車輛影像,並由車輛影像讀取單元將影像捕 參紙張尺芰適闬t國國家標準(CNS)XT^袼(2if^冬 釐) 1 -------- t - III!蓮! ti!----- I* (請先閱讀背面之注意事項再填寫本頁) 544643 A7 B7 PA890608.AM1 - 6/12 發明說明 10 經濟部智慧財產局員Η消費合作、社印製2. Description of invention A7 B7 PA890608.AiVn, 5 /; The sum of the Logarithmic reflectance of the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs. The resolution of the image screen is M * N㈣ pixels, and the gray value of the &lt; pixel at the coordinate position (X, y) is g (x, y), 0 &lt; X &lt; M + bu0 &lt; The values of y, N · τ1, g (x, y) can be read directly from the output values of the CCD camera. If the bright g% i (x, y) and the reflection intensity of the pixel at the coordinate position (x, y) are r (x, y) ', then g (x, yH (x, y) X r (x, y), The value of i (x, y) and r (x, y) cannot be read from the output value of the CCD camera. After the logarithmic operation, the logarithmic grayscale value is milk x, y), and the logarithmic brightness is r ( x, y), the logarithmic reflection intensity is r, (x, y), that is 10 gf (x? y) = In g (x? y) = 111 i (x, y) + In r (x, y) i '(x, y) + r' (x, y), 0 &lt; x &lt; m + bu 〈y <N + 1. We take [g, (x + 1, y) / 2 i as the horizontal gradient of the logarithmic grayscale value of the pixel at the coordinate position (x, y), and b [g? (X + l? Y) &lt; gHx- l ^ y)] / 2 [if (x + i? y) + Γϊχ + Ι, ν), iT (x.1? y) ^ f! (x. l, y)] / 2; The difference in brightness is very small, so the value of 奸 (i-iT (xl, y) is very small and can be ignored; therefore, the logarithmic gray-scale straight horizontal gradient of the pixel at the coordinate position (x, y) is very close to ^ kiss Han. In this way, the effect of light changes or shadows on the image is eliminated. The 20 series of the present invention applies a logarithmic operation to the pixel grayscale value and uses the logarithmic grayscale value to calculate the location of the license plate, so it will not The positioning of the license plate area is incorrect due to changes in gray levels due to light changes or shadows. (The main f of the present invention is to capture the vehicle image by the camera with a lens CCD camera and image capture, and read from the vehicle image. The unit adapts the image capture paper size to the national standard (CNS) XT ^ 袼 (2if ^ winter) 1 -------- t-III! Lotus! Ti! ----- I * (Please read the precautions on the back before filling this page) 544643 A7 B7 P A890608.AM1-6/12 Description of Invention 10 Member of Intellectual Property Bureau, Ministry of Economic Affairs

15 足卡^柄取的景&gt;(象讀取出I,接著由ί象素之對數灰階!直運 异:兀來對車輛影像中的各個像素計算岀其對數灰階值, =放灰1 3叙之水平梯度運算單元則接著對各像素計算其對 敛灰13直水平梯度,接著由水平梯度之二值化運算單元將 、=錢的貫數轉^或㈣二元值。最後由二值化水 平梯度最密區域搜尋單元依照預設的車牌長寬丨直來尋找整 ^車輛影像中㈣區域的二社水平梯度總和最高,則^ 邊區域切出為車牌區域的所在。 【圖式簡單說明】 ▲ %麥閱以下有關本發明一較佳實施例之詳細說明及其 4圖苻可進一步瞭解本發明之技術内容及其目的功 百關該實施例之附圖為: …方塊^為本發明車輛料之車牌區域峰方法之實施例 圖—為二值化水平梯度最密區域搜尋單元之粗押君 式示意圖;以及 又1方 化水平梯度最密區域搜尋單元之細搜尋方不\艺、圖。 【主要部分代表符號】 i CCD攝影機 2影像捕捉卡 3車輛影濛讀取單元 4溧素之對數灰階匱運算單元 5對數芡階匱之水平梯度運算單元 .度適用中國] (請先閱讀背面之注意事項再填寫本頁) t - 1 1 n i 3 —fl 1 n xi ^ 裝 SI i II-&quot;-δ、e n i n Φ n »1 —fl Is t— ϋ n15 The scene taken by the foot card ^ (The image reads out I, and then the logarithmic gray scale of the pixel! Direct transport difference: Wulai calculates the logarithmic gray scale value of each pixel in the vehicle image, = put The horizontal gradient computing unit of Gray 1 and 3 then calculates the horizontal gradient of the straight gray 13 for each pixel, and then the binary computing unit of the horizontal gradient converts the value of == money to ^ or ㈣ binary value. Finally According to the binarized horizontal gradient densest area search unit, according to the preset length and width of the license plate, the total sum of the horizontal gradients of the two companies in the ㈣ region of the vehicle image is the highest, and the ^ side area is cut out to the location of the license plate area. Brief description of the drawings] ▲% Mai read the following detailed description of a preferred embodiment of the present invention and its 4 drawings. You can further understand the technical content of the present invention and its purpose. The drawings of this embodiment are as follows: ^ This is an embodiment of the method for the peak area of the license plate region of the vehicle material of the present invention—a rough schematic diagram of the binarized horizontal gradient densest area search unit; and a detailed search method of the squared horizontal gradient densest area search unit不 \ 艺 、 图. [Main department Representative symbol] i CCD camera 2 image capture card 3 vehicle shadow reading unit 4 prime logarithmic gray level calculation unit 5 logarithmic level horizontal gradient operation unit. Degree applicable to China] (Please read the precautions on the back first Fill out this page again) t-1 1 ni 3 —fl 1 n xi ^ Install SI i II- &quot; -δ, enin Φ n »1 —fl Is t— ϋ n

i IB ϋ ί I ' *1 i n 6- 544643 A7i IB ϋ ί I '* 1 i n 6- 544643 A7

10 經濟部智慧財.暑『費合序社印製 2ϋ 6水平悌度之二值化運算單元 7二值化水平梯度最密區域搜尋單元 【較佳實施例】取太請參關―’本發明所提供之車輛影像之車牌區域切 =之實施例方塊圖’係由搭配鏡頭的ccd攝影機1、 =捕捉卡2及車輛影像讀取單元3、像素之對數灰階值運 -早兀4、對數灰階直之水平梯度運算單元5、水平梯产之 :值化運算單元6、二值化水平梯度最密區域搜尋單元7等 :=組成:,其中’該搭配鏡頭輸攝影機1及影像捕 旦像亚由車輛影像讀取單元3將 糊取的影像讀取出來,接著由像素之對數 呼窗㈣ 對放運异來對車柄影像中的各個像素 冲^rr出其對數灰階值纟#〖务I 噼备伽&quot; 的灰階值範圍為0〜255,為 &lt;免對0做對數運算的數學^^ «的灰_…使得灰二:=運算:= =:::r運算單元5則接著對各像素計算其= :’右座標位置㈣的像素的對數灰階值為 p,y ’則㈣素的對數灰階值水平梯度為 咖 l,y)] / 2 ;接著由皮单谜洚 ' ,」伽 的奋數$ /又〜一化運鼻單元6將水平悌度或1的二元這,轉換方… 的咖最大的百分之一與最小的百分之一,;最Γ由平梯度設為卜其餘像素之二值化水平梯度設的垂啤Μ信·;直化水平梯度最密區域搜尋單元7依照預設 瓦幻直來尋找整張車輪影緣令那値區域的二直化水 I ! I ! 背 面 i !II φ 填響I裝 頁! ί 訂 i10 Wisdom Wealth of the Ministry of Economic Affairs. Summer [printed by Fei Hexu Co. 2ϋ 6 binarization operation unit of horizontal degree 7 binarization level gradient most dense area search unit [preferred embodiment] Please refer to the question-'this The block diagram of the vehicle license plate area of the vehicle image provided by the invention = an embodiment block diagram 'is composed of a ccd camera with a lens 1, a capture card 2 and a vehicle image reading unit 3, and the logarithmic gray scale value of the pixel is moved- Logarithmic gray level straight horizontal gradient operation unit 5, horizontal ladder production: valued operation unit 6, binarized horizontal gradient most dense area search unit 7, etc .: = composition :, where 'this lens is used to input camera 1 and image capture The image is read by the vehicle image reading unit 3, and then the logarithm of the pixel is called. The logarithmic difference is used to punch each pixel in the image of the handle image. [The gray scale value range of the service I 噼 格 &quot; is from 0 to 255, which is &lt; the math that does not perform logarithmic operation on 0 ^^ «grey _... makes grey two: = operation: = = :::: r operation Unit 5 then calculates == 'the logarithmic gray value of the pixel at the right coordinate position ㈣ for each pixel is p, y' The horizontal gradient of the logarithmic gray scale value of the element is ca l, y)] / 2; Then, the skin's mystery 洚 ', "Gai's Fen number $ / again ~ the nasal transport unit 6 will be the horizontal degree or 1 two Here, the conversion side ... the largest one hundredth and the smallest one hundredth; the maximum Γ is set from the flat gradient to the binarized horizontal gradient of the remaining pixels. The vertical level The gradient densest area searching unit 7 searches for the entire wheel shadow edge to make the two straightened water of the whole area I! I! The back i! II φ fills the I page according to the preset tile magic straight! order

544643 ΡΑ89ϋ608 ΑΜ1 - 3/12 經濟部智慧財產局Ρ、工消費合作社印製 A7 B7 i五、發明說明(6) ϊ ! 平梯度總和隶[ij ’則將邊區域切出為車牌區域的所在。 ί | · | 二值化水平梯度最密區域搜尋單元7分為粗搜尋及細 Ϊ 搜尋兩個步驟。圖二所示的是二僮化水平悌度最密區域搜 1 | 尋單元之粗搜尋方式示意圖,以車牌長寬a各兩倍的範圍 ! 5 做為計算梯度總和的範圍窗框,在整張車輛影像中移勤計 算梯度總和的範圍窗框並計算每次範圍中的所有二值化水 丨平梯度的總和;計算梯度總和的範圍窗框每次水平移動一 : | 値車牌的長度或垂直移動一個車牌的寬度。從每次移動中 i 找出二值化水平梯度總和最大的一個範圍,此範圍即為二 | 10 值化水平梯度最密區域搜尋單元7的粗搜尋結果。 ί 圖三所示的是二值化水平梯度最密區域搜尋單元之細 搜尋方式示意圖。對二ί直化水平梯度最密區域搜尋單元的 粗搜尋結果所界定的範圍進行細搜尋:以車牌長寬ί直為範 圍做為計算悌度總和的範圍窗框,在粗搜尋結果所界定的 15 範圍中移動計算梯度總和的範圍窗框並計算每次範圍中的 所有二值化水平梯度的總和,計算梯度總和的範圍窗框每 | 次水平移動一個像素或垂直移動一個像素。從每次移動中 | 找出二值化水平梯度的總和最大的一觸範圍,此範圍即為 I 二值化水平梯度最密區域搜尋單元7的細搜尋結果,此範 I 20 圍也就是車牌區域。 ί | 【特點及功效】 | 本發明所提供之車輛影像之車牌區域切取方法,與其 I 他習用技術相互比較時,更具有下列之優點: j I 本發明係對像素灰階值施以對數運算(Logarithmic544643 ΡΑ89ϋ608 ΑΜ1-3/12 Intellectual Property Bureau of the Ministry of Economic Affairs P, printed by the Industrial and Consumer Cooperatives A7 B7 i 5. Description of the invention (6) ϊ! The flat gradient summator [ij ′ cuts out the edge area as the location of the license plate area. ί | · | The binarization horizontal gradient densest area search unit 7 is divided into two steps: coarse search and fine search. Figure 2 shows the rough search method of searching for the densest area of the second childhood level. The search method is based on the rough search method of the unit. The length and width a of the license plate are each twice the range! 5 As the range window for calculating the sum of the gradients, In the vehicle image, the mobile window calculates the range window frame of the sum of gradients and calculates the sum of all binarization levels in each range; the range window of the calculated gradient sum moves horizontally one at a time: | 値 The length of the license plate or Move the width of a license plate vertically. From each movement i find a range where the sum of the binarized horizontal gradients is the largest, and this range is the rough search result of the bin | 10 binarized horizontal gradient densest area search unit 7. ί Figure 3 shows the detailed search method of the binarized horizontal gradient densest area search unit. Perform a detailed search on the range defined by the coarse search result of the search unit of the second-horizontal horizontal gradient densest area: use the license plate length and width as the range window to calculate the sum of the degrees. 15 Move the range to calculate the range window frame of the sum of gradients and calculate the sum of all binarized horizontal gradients in each range. The range window to calculate the sum of gradients moves one pixel horizontally or one pixel vertically every time. From each move, find the one-touch range where the sum of the binarized horizontal gradients is the largest. This range is the fine search result of the search unit 7 for the densest region of the binarized horizontal gradients. The range of I 20 is the license plate. region. ί | [Features and effects] | The method for extracting the license plate area of the vehicle image provided by the present invention has the following advantages when compared with other conventional techniques: j I The present invention applies a logarithmic operation to the pixel grayscale value (Logarithmic

I ! -8- ί (請先1讀背面之注意事項再填窵本頁)I! -8- ί (Please read the precautions on the back before filling out this page)

用中國國家(2i〇~x^?公釐) ;~~~Use Chinese country (2i〇 ~ x ^? Mm); ~~~

D4404J A7 B7 五、發明說明(f PA890608 AMI _ 9/1 9D4404J A7 B7 V. Description of the invention (f PA890608 AMI _ 9/1 9

10 opemtion)以得到對數灰階匱,再 明亮度的影響並計m 4轉軸/錢随來云除 光線變t或陰影使歧階it發生織化’所以不會因 } 艾化而造成車牌區域的定 1立錯誤。 上列詳細說明係針對I發明之一可 J彳丁貫施例之具體說 明’惟該實施例並非用以限制本發明之專利範目,凡未脫 離本發明技藝精神所岛之等效實施或變更,均應包含於本 案之專利範圍中。 綜上所述,本案不但在技術思想上確屬創新,並能較 白'用物品增進上述多項功效,應已充分符合新穎性及進步 唑之法定發明專利要件,爰依法提出申請,懇請貴局核 准本泮㈣翻中請案,以勵發明,至感德便。 (請先閱讀背面之注意事項再填寫本頁) ·裝 訂-·10 opemtion) in order to get the logarithmic gray scale, and then the effect of brightness, and calculate the m 4 reels / money. Clouds in addition to the light changes to t or the shadow makes the heterogeneous it iterating 'so it will not cause the license plate area due to} The setting is wrong. The above detailed description is a specific description of one of the inventions that can be implemented by J. Ding, but the embodiment is not intended to limit the patent scope of the invention, and any equivalent implementation or implementation that does not depart from the technical spirit of the invention or Changes should be included in the patent scope of this case. To sum up, this case is not only technically innovative, but also can be used to enhance the above-mentioned multiple effects. It should have fully met the requirements of novelty and progressive azole's statutory invention patents, and applied in accordance with the law. To approve this case, to encourage invention, to the utmost convenience. (Please read the notes on the back before filling out this page) · Binding-·

ϋ ϋ 12 I 經濟部智慧財產局員工消費合作社印製ϋ ϋ 12 I Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs

Claims (1)

5 10 2. 15 申請專利範圍 1· 車牌㈣峰料,主㈣藉搭配鏡 4及料卡對輕攝取車輛影像, 影像讀取單元將影μ捉卡所錄的影像讀 接著由像素之對數灰㈤直運算單元來對車輛 衫像卞的各個像素計算出其對數灰階值 之水平梯度運算單元則接著對各像素計算其對數灰階 直水平悌度,接著由水平梯度之二這化運算單元將水 千梯度的實數值轉為二元值。最後由二值化水 平梯度最密區域搜尋單元依照預設的車牌長寬值來尋 2整張車輛影像”個區域的二值化水平悌度總和最 兩,則將該區域切出為車牌區域的所在。 如申請專利範圍苐i項所述之車輛影像之車牌區域切 取方去,其中該對數灰階值運算單元以對數運算來對 車輛影像中的各個像素計算出其對數灰階值,若像素 的灰值範圍為〇〜255,為避免對〇做對數運算的數學 、’曰溴則在做對數運异雨先將像素的灰階值加丨,陡 得灰階值範圍成為1〜256。 士申明專利範圍第1項所述之車辆影像之車牌區域切 取方法,其中該水平梯度運算單元中若座標位置(x,y) 的像素的對數灰階值為g,(x,y),則該像素的對數灰階 值水平悌度岛[g,(x+l,y) - g,(x-l,y)] / 2。 如申請專利範圍第1項所述之車輛影像之車牌區域切 取方法,其中該水平梯度之二值化運算單元將水平悌 -10- 中國國家標準(CNS)A4規格(21ϋ X 297公 (請先閱讀背面之注意事項再填寫本頁) I裝 訂&quot;------- 經濟部智慧財產aR員工消費合泎让印製 3. 20 4· 544643 經濟部智慧財產局員工消費合作社印製 Λ8 心 B8 J CS D8 _____________________________________________________________ PA^gnpop? ,^ρ,,ίΐΐ - 11/-1?f、申請專利範圍 度的實數值轉為0或i的二元值' 其轉換方法係將所有 濛素中水平悌度的實數僮最大的百分之一與最小的百 分之一,其二丨直化水平悌度設為1,其餘像素之二值 化水平梯度設為0。 5 5_如申請專利範圍第1項所述之車輛影像之車牌區域切 取方法,其中該二值化水平梯度最密區域搜尋單元分 岛粗搜尋及細搜尋兩個步驟。 6.如申請專利範圍第5項所述之車輛影像之車牌區域切 取方法,其中該二值化水平梯度最密區域搜尋單元的 10 粗搜尋以車牌長寬ί直各兩ί咅的範圍做為汁算梯度總和 的範圍窗框,在整張車輛影像中移動計算梯度總和的 範圍窗框並計算每次範圍中的所有二值化水平梯度的 總和,計算梯度總和的範圍窗框每次水平移動一個車 牌的長度或垂直移動一個車牌的寬度。從每次移動中 15 找出二僮化水平梯度的總和最大的一個範圍,此範圍 即為二(直化水平梯度最密區域搜尋單元的粗搜尋結 果。 7·如申請專利範圍第5項所述之車輛影像之車牌區域切 取方法,其中該二ί直化水平梯度最密區域搜尋單元f糸 20 對二值化水平梯度最密區域搜尋單元的粗搜尋結果所 界定的範圍進行細搜尋:以車牌長寬值岛範圍做為計 算梯度總和的範圍窗框,在粗搜尋結果所界定的範圍 中移動計算梯度總和的範圍窗框並計算每次範圍中的 所有二渣化水平悌度的總和,計算悌变總和的範圍窗 -11 - (請先閱讀背面之注意事項再填寫本頁) 一裝 t ti n 1 —- ____τ____ 1 1 I 1 舞 本紙張尺t過用办國國家丨票遑(CNS) A4規袼CIO :&lt;297 k f &gt; 544643 ^Α8906Ω8 ΑΜ1 - B8 CS D8 、申請專利範圍 框每次水平移動一個像素或垂直移動一個像素。從每 次移動中找出二值化水平梯度的總和最大的一個範 圍,此範圍即為二值化水平梯度最密區域搜尋單元的 細搜尋結果,此範圍也就是車牌區域。 ---1 (請先閱讀背面之注意事項再填寫本頁} T »4 fff n is — n 1-¾ 經濟部智慧財產咼員工消費合泎社印.t -12 - 払紙張已t適ΐ中國國家標KNS)A-i規格(21〇χ 297公f )5 10 2. 15 Scope of patent application 1. License plate and peak material. The main camera uses the lens 4 and the material card to lightly capture the vehicle image. The image reading unit reads the image recorded by the shadow capture card and then the pixel logarithmic gray. A straightening operation unit calculates a logarithmic gray level value for each pixel of a vehicle shirt image. A horizontal gradient calculation unit then calculates a logarithmic gray level straight horizontal degree for each pixel, and then calculates a horizontal gradient two operation unit. The real value of the water thousand gradient is converted to a binary value. Finally, the binarized horizontal gradient densest area search unit searches for the entire vehicle image according to the preset license plate length and width values. The total binarized horizontal degrees of the two regions are the two most, and then the region is cut out as the license plate region. The location of the license plate area of the vehicle image as described in item i of the patent application range, where the logarithmic grayscale value calculation unit calculates its logarithmic grayscale value for each pixel in the vehicle image. The gray value of the pixel ranges from 0 to 255. In order to avoid the mathematical operation of logarithmic operation to 0, 'Bromo' is added in the logarithmic operation. First, the gray value of the pixel is added, and the range of gray values becomes steep. The method for extracting the license plate area of a vehicle image as described in Item 1 of the patent scope, wherein the logarithmic grayscale value of the pixel at the coordinate position (x, y) in the horizontal gradient operation unit is g, (x, y) , Then the logarithmic gray level value of the pixel is a horizontal degree island [g, (x + l, y)-g, (xl, y)] / 2. The license plate area of the vehicle image as described in the first scope of the patent application A cutting method, in which the binarization operation unit of the horizontal gradient Ping 悌 -10- Chinese National Standard (CNS) A4 Specification (21ϋ X 297 male (please read the precautions on the back before filling out this page) I Binding &quot; ------- Intellectual Property aR Staff Consumption泎 Print 3.20 4 · 544643 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs Λ8 heart B8 J CS D8 _____________________________________________________________ PA ^ gnpop?, ^ Ρ ,, ίΐΐ-11 / -1? F. The real value is converted to a binary value of 0 or i '. The conversion method is to set the real number of horizontal children in all Mongolian to the largest one percent and the smallest one percent. Is 1, the binarized horizontal gradient of the remaining pixels is set to 0. 5 5_The method for extracting the license plate region of a vehicle image as described in item 1 of the scope of the patent application, wherein the binarized horizontal gradient is the densest area search unit for islands Coarse search and fine search are two steps. 6. The license plate region extraction method of the vehicle image as described in item 5 of the patent application scope, wherein the coarse search of the binarized horizontal gradient densest area search unit uses the license plate length and width. Straight two The range is used as the range window frame for calculating the gradient sum. The range window frame for calculating the gradient sum is moved in the entire vehicle image and the sum of all binarized horizontal gradients in each range is calculated. The range window frame for calculating the gradient sum is calculated. Move the length of a license plate horizontally or move the width of a license plate vertically each time. From 15 in each move, find the range where the sum of the two childish horizontal gradients is the largest. Coarse search results for the unit. 7. The method for extracting the license plate area of a vehicle image as described in item 5 of the scope of the patent application, wherein the two straightened horizontal gradient densest area search unit f 糸 20 is a rough search of the binarized horizontal gradient densest area search unit Perform a fine search on the range defined by the results: use the license plate length and width value island range as the range window to calculate the gradient sum, move the range window to calculate the gradient sum in the range defined by the coarse search result and calculate the Sum of all the di-slagging levels, calculate the range of the total change window -11-(Please read the precautions on the back before filling out this page) One pack t ti n 1 —- ____ τ ____ 1 1 I Dance paper ruler The country of use of the country of origin (CNS) A4 regulations CIO: &lt; 297 kf &gt; 544643 ^ Α8906Ω8 ΑΜ1-B8 CS D8, the patent application frame moves one pixel horizontally or one pixel vertically. From each move, find a range where the sum of the binarized horizontal gradients is the largest. This range is the detailed search result of the search unit of the binarized horizontal gradient's densest area. This range is also the license plate region. --- 1 (Please read the notes on the back before filling out this page} T »4 fff n is — n 1-¾ Intellectual Property of the Ministry of Economic Affairs 咼 Employee Consumption Cooperative Press. T -12-払 The paper has been t Chinese national standard KNS) Ai specification (21〇χ 297 Male f)
TW90110883A 2001-05-08 2001-05-08 Method for cropping area of license plate from vehicle image TW544643B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
TW90110883A TW544643B (en) 2001-05-08 2001-05-08 Method for cropping area of license plate from vehicle image

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
TW90110883A TW544643B (en) 2001-05-08 2001-05-08 Method for cropping area of license plate from vehicle image

Publications (1)

Publication Number Publication Date
TW544643B true TW544643B (en) 2003-08-01

Family

ID=29708142

Family Applications (1)

Application Number Title Priority Date Filing Date
TW90110883A TW544643B (en) 2001-05-08 2001-05-08 Method for cropping area of license plate from vehicle image

Country Status (1)

Country Link
TW (1) TW544643B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN100349175C (en) * 2004-05-14 2007-11-14 中华电信股份有限公司 Method for plucking license-plate region from vehicle image
CN103226812A (en) * 2013-03-19 2013-07-31 苏州橙果信息科技有限公司 Texture filtering method based on edge binary image
CN110569683A (en) * 2018-06-06 2019-12-13 伟诠电子股份有限公司 bar code detection method and bar code detection system

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN100349175C (en) * 2004-05-14 2007-11-14 中华电信股份有限公司 Method for plucking license-plate region from vehicle image
CN103226812A (en) * 2013-03-19 2013-07-31 苏州橙果信息科技有限公司 Texture filtering method based on edge binary image
CN110569683A (en) * 2018-06-06 2019-12-13 伟诠电子股份有限公司 bar code detection method and bar code detection system
CN110569683B (en) * 2018-06-06 2023-06-09 伟诠电子股份有限公司 Bar code detection method and bar code detection system

Similar Documents

Publication Publication Date Title
CN102930265B (en) A kind of many I.D.s scan method and device
CN104616275B (en) A kind of defect inspection method and device
TW201218129A (en) A vehicle registration-plate detecting method and system thereof
CN103488984B (en) Method and device for identifying second-generation identity cards based on intelligent mobile equipment
US7693304B2 (en) Method and system for image quality calculation
CA2642046A1 (en) Image analysis systems for grading of meat, predicting quality of meat and/or predicting meat yield of an animal carcass
CN105608456A (en) Multi-directional text detection method based on full convolution network
KR101246120B1 (en) A system for recognizing license plate using both images taken from front and back faces of vehicle
KR101597739B1 (en) Image processing apparatus, image processing method, and computer readable medium
WO2016101767A1 (en) Picture cropping method and device and image detecting method and device
CN102779264B (en) Method and device for realizing barcode recognition
KR101182173B1 (en) Method and system for recognizing vehicle plate
RU2008143205A (en) EFFICIENT CODING OF MANY SPECIES
CN101655981A (en) Method for detecting and adjusting inversion of certificate image
WO2015066984A1 (en) Complex background-oriented optical character recognition method and device
KR20150108118A (en) Remote automatic metering system based image recognition
TW544643B (en) Method for cropping area of license plate from vehicle image
CN107977648B (en) Identification card definition distinguishing method and system based on face recognition
CN108805060A (en) A kind of zebra line style crossing detection method
CN106485639A (en) The method and apparatus differentiating forged certificate picture
US20130208984A1 (en) Content scene determination device
CN105975903A (en) Automobile feature identification method
CN103337054A (en) Two-stage image haze removal method based on single images
CN110991434B (en) Self-service terminal certificate identification method and device
CN109146834A (en) Car damage identification method and device, computer readable storage medium, terminal

Legal Events

Date Code Title Description
GD4A Issue of patent certificate for granted invention patent
MM4A Annulment or lapse of patent due to non-payment of fees