JP2561969B2 - Character recognition method for drawing reader - Google Patents
Character recognition method for drawing readerInfo
- Publication number
- JP2561969B2 JP2561969B2 JP2178177A JP17817790A JP2561969B2 JP 2561969 B2 JP2561969 B2 JP 2561969B2 JP 2178177 A JP2178177 A JP 2178177A JP 17817790 A JP17817790 A JP 17817790A JP 2561969 B2 JP2561969 B2 JP 2561969B2
- Authority
- JP
- Japan
- Prior art keywords
- character
- normalization
- character string
- recognition method
- deformation
- Prior art date
- Legal status (The legal status 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 status listed.)
- Expired - Lifetime
Links
Landscapes
- Character Discrimination (AREA)
- Character Input (AREA)
Description
【発明の詳細な説明】 ［概要］ 読取図面中に書かれた手書き文字列を認識する図面読
取装置の手書き文字認識方法に関し、 文字歪みを持った斜め文字列の認識率を向上すること
を目的とし、 読取図面中から抽出された文字列の斜め変形の有無を
検出し、斜め変形による歪みを除去して各文字を正規化
した後に各文字を認識するように構成する。また斜め変
形の有無は、正規化前と正規化後の文字列細線化ベクト
ルの総和の大小関係から求める。The present invention relates to a handwritten character recognition method of a drawing reading device for recognizing a handwritten character string written in a read drawing, and an object thereof is to improve a recognition rate of an oblique character string having a character distortion. Then, the presence or absence of diagonal deformation of the character string extracted from the read drawing is detected, the distortion due to the diagonal deformation is removed, each character is normalized, and then each character is recognized. The presence / absence of diagonal deformation is obtained from the magnitude relationship of the sum of the character string thinning vectors before and after normalization.
［産業上の利用分野］ 本発明は、読取図面中に書かれた手書き文字列を認識
する図面読取装置の手書き文字認識方法に関し、特に読
取図面のイメージデータからベクトルデータを生成して
文字列を認識する図面読取装置の手書き文字認識方法に
関する。BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a handwritten character recognition method of a drawing reading device for recognizing a handwritten character string written in a read drawing, and more particularly to generating vector data from image data of the read drawing to generate a character string. The present invention relates to a method for recognizing a handwritten character of a drawing reading device.
近年、プラント設計、電気設計などの分野ではCADシ
ステムへ手書きされた図面の読取データを自動入力する
システムが実用化されつつあるが、図面に記載された手
書き文字列の認識は、０度及び90度方向に並んだ文字列
のみに限定されており、０度、90度以外の傾いた文字列
に対しては認識率が低いため、文字認識の確認修正処理
もしくはCADシステムで再入力が必要であった。このた
め、傾きのある文字列を確実に認識する必要がある。In recent years, in the fields of plant design, electrical design, etc., a system for automatically inputting read data of a handwritten drawing into a CAD system is being put to practical use, but the handwritten character string described in the drawing is recognized at 0 degrees and 90 degrees. It is limited to character strings arranged in the direction of degree, and the recognition rate is low for inclined character strings other than 0 degrees and 90 degrees, so it is necessary to confirm and correct the character recognition or re-enter with the CAD system. there were. Therefore, it is necessary to surely recognize a character string having an inclination.
［従来の技術］ 従来、０度、90度以外の傾いた文字列の認識方法とし
ては、15度刻みの傾きのある文字列を対象とした認識方
法が知られている（「図面内の傾いた文字列抽出」情報
処理学会第37回（昭和63年度後期）全国大会講演論文集
2W−2,1988,pp1608−1609参照）。[Prior Art] Conventionally, as a method for recognizing a character string that is inclined other than 0 degrees and 90 degrees, a recognition method for a character string that has an inclination of 15 degrees is known. "Character string extraction" Proc. Of the 37th National Conference of IPSJ (Late 1988)
2W-2, 1988, pp 1608-1609).
［発明が解決しようとする課題］ しかしながら、このような従来の15度刻みの傾いた文
字列を対象とした文字認識方法にあっては、文字列を構
成する各文字が文字列の傾きに対して斜めの変形が無い
ものに対して有効であるが、文字列の傾きに対して正立
でなく斜めの変形があると、この斜め変形歪みにより正
しく文字認識できない。[Problems to be Solved by the Invention] However, in such a conventional character recognition method for a character string inclined by 15 degrees, each character forming the character string is However, when the character string is not upright and there is an oblique deformation, the character cannot be correctly recognized due to the oblique deformation distortion.
例えば、プラント設計のアイソメ図（配管図）を作成
する際には、一つの図面に対し複数の人間により文字列
が記入されるが、文字列を構成する文字の文字列の傾き
に対する斜め歪みがあるか否かは個人により異なり、個
人差を考慮して文字列を構成する文字の斜め歪みを検出
する必要があり、現在までのところ有効な解決策は見い
出されていない。For example, when creating an isometric view (piping diagram) of a plant design, a character string is entered by one or more people in one drawing, but the diagonal distortion of the character string forming the character string with respect to the inclination of the character string is Whether it exists or not depends on the individual, and it is necessary to detect the diagonal distortion of the characters that form the character string in consideration of individual differences, and so far no effective solution has been found.
本発明は、このような従来の問題点に鑑みてなされた
もので、文字歪みを持った斜めに傾いた文字列を確実に
認識して認識率を向上できる図面読取装置の文字認識方
法を提供することを目的とする。The present invention has been made in view of such conventional problems, and provides a character recognition method for a drawing reading device capable of reliably recognizing an obliquely inclined character string having character distortion and improving the recognition rate. The purpose is to do.
［課題を解決するための手段］ 第１図は、本発明の原理説明図である。[Means for Solving the Problems] FIG. 1 is a diagram illustrating the principle of the present invention.
読取図面中から抽出された文字列の斜め変形が検出さ
れたとき、第１図（ａ）に示すように、その変形による
歪みを除去して各文字を正規化し、その正規化された文
字列あるいは斜め変形なしと判定された文字列の各文字
を認識する図面読取装置の文字認識方法を対象とする。When the diagonal deformation of the character string extracted from the read drawing is detected, the distortion due to the deformation is removed to normalize each character as shown in FIG. 1 (a), and the normalized character string Alternatively, the present invention is directed to a character recognition method of a drawing reading device that recognizes each character of a character string that is determined to have no diagonal deformation.
ここで文字列の斜め変形の検出方法としては、 文字列を構成する各文字の細線化ベクトル長を算出す
る第１過程と； 該第１過程で算出された細線化ベクトル長の総和を算
出する第２過程と； 斜め変形歪みを除去する所定の正規化変換式により各
文字の細線化ベクトルを座標変換し、該座標変換ベクト
ルでなる各文字の細線化ベクトル長を算出する第３過程
と； 該第３過程で算出された正規化文字の細線化ベクトル
長の総和を算出する第４過程と； 前記第２過程で算出された正規化前の総ベクトル長と
前記第４過程で得られた正規化後の総ベクトル長とを比
較し、 正規化により総ベクトル長が短くなった時には斜め変
形ありと判定し、 正規化によりベクトル長が長くなった時には斜め変形
なしと判定する 第５過程と； を備えたことを特徴とする。Here, as a method for detecting the diagonal deformation of the character string, a first step of calculating the thinned vector length of each character that constitutes the character string; and a sum of the thinned vector lengths calculated in the first step is calculated. A second step; and a third step of coordinate-converting the thinning vector of each character by a predetermined normalization conversion formula that removes the oblique deformation distortion, and calculating the thinning vector length of each character formed by the coordinate conversion vector; A fourth step of calculating the sum of the thinned vector lengths of the normalized characters calculated in the third step; and a total vector length before normalization calculated in the second step and obtained in the fourth step. Compared with the total vector length after normalization, when the total vector length is shortened by normalization, it is determined that there is diagonal deformation, and when the vector length is long by normalization, it is determined that there is no diagonal deformation. Characterized by having To.
この斜め変形検出方法の第３過程で使用する正規化変
換式は、文字列の傾き角度をα、文字列の開始点を（x0
y0）、正規化する任意のベクトル座標値を（x,y）とし
た時、 ｘ＝ｘ−cos（α）×ｈ×K1（α） ｙ＝ｙ−sin（α）×ｈ×K1（α） で与えられる。The normalization conversion formula used in the third step of this oblique deformation detection method is that the inclination angle of the character string is α and the start point of the character string is (x 0
y 0 ), when an arbitrary vector coordinate value to be normalized is (x, y), x = x−cos (α) × h × K 1 (α) y = y−sin (α) × h × K It is given by 1 (α).
また正規化変換式に使用する係数ｈは文字列の開始点
（x0,y0）を通る傾きαの直線への垂線の長さであり、 ｈ＝｜−sin（α）×（ｘ−x0）＋cos（α） ×（ｙ−y0）｜ で与えられ、且つ正規化変換式に使用する係数K1（α）
は、経験により設定した１〜０の範囲の値とする。The coefficient h used in the normalization conversion formula is the length of a perpendicular line to a straight line having a slope α passing through the start point (x 0 , y 0 ) of the character string, and h = | −sin (α) × (x− x 0 ) + cos (α) × (y−y 0 ) | and the coefficient K 1 (α) used in the normalization conversion formula
Is a value in the range of 1 to 0 set by experience.
更に文字列の斜め変形検出方法に使用する正規化変換
式の係数K1（α）は経験により、 ０＜｜α｜≦π/4の時、K1（α）＝1.0 π/4＜｜α｜≦３π/8の時、K1（α）＝0.7 ３π/8＜｜α｜≦π/2の時、K1（α）＝0.0 とする。Furthermore, the coefficient K 1 (α) of the normalization conversion formula used in the method for detecting the diagonal deformation of a character string is empirically obtained, and when 0 <| α | ≦ π / 4, K 1 (α) = 1.0 π / 4 <| When α | ≦ 3π / 8, K 1 (α) = 0.7. When 3π / 8 <| α | ≦ π / 2, K 1 (α) = 0.0.
更に又、文字列の斜め変形検出方法における第５過程
の斜め変形の判定として、正規化前の総ベクトル長に重
み係数K2（α）を乗算した値と正規化後の総ベクトル長
とを比較し、 正規化により総ベクトル長が等しいか短くなった時に
は斜め変形ありと判定し； 正規化により総ベクトル長が長くなった時には斜め変
形なしと判定する； ことを特徴とする。Furthermore, as the determination of the diagonal deformation in the fifth step in the method for detecting the diagonal deformation of a character string, the value obtained by multiplying the total vector length before normalization by the weighting coefficient K 2 (α) and the total vector length after normalization are used. The comparison is performed, and when the total vector length is equal to or shorter than the normalization, it is determined that the diagonal deformation is present; when the total vector length is long due to the normalization, it is determined that the diagonal deformation is not present.
この場合の重み係数k2（α）の値は、経験により、 ０＜｜α｜≦π/4の時、 k2（α）＝1.0＋cos（α/2） π/4＜｜α｜≦３π/8の時、 k2（α）＝0.7＋cos（π/8） ３π/8＜｜α｜≦π/2の時、 k2（α）＝0.0 とする。The value of the weighting coefficient k 2 (α) in this case is empirically determined to be: when 0 <| α | ≦ π / 4, k 2 (α) = 1.0 + cos (α / 2) π / 4 <| α | ≦ When 3π / 8, k 2 (α) = 0.7 + cos (π / 8) When 3π / 8 <| α | ≦ π / 2, k 2 (α) = 0.0.
［作用］ このような構成を備えた本発明による図面読取装置の
手書き文字認識方法によれば、第１図（ｂ）に示すよう
に、斜め変形のある文字列が検出された場合には、斜め
変形による文字の歪みを取り除く正規化変換式に従って
正規化された文字列を生成し、変形歪みのない正規化文
字列に対し認識処理が行われることで、斜め方向に傾い
て書かれた文字列の認識率を大幅に向上することができ
る。[Operation] According to the handwritten character recognition method of the drawing reading apparatus according to the present invention having such a configuration, as shown in FIG. 1 (b), when a diagonally deformed character string is detected, Characters that are written in a slanted direction are generated by generating a normalized character string according to a normalization conversion formula that removes distortion of characters due to diagonal deformation, and performing recognition processing on the normalized character string that does not have deformation distortion. The recognition rate of columns can be significantly improved.
その結果、プラント設計等で用いられるアイソメ図は
30度、60度の傾きに並んだ文字列を含むアイソメ図中の
文字列を正確に認識でき、CADシステムに対する自動入
力を実用化できるに十分な認識率を得ることができる。As a result, the isometric drawing used in plant design etc.
It is possible to accurately recognize the character strings in the isometric drawing including the character strings arranged at the inclinations of 30 degrees and 60 degrees, and it is possible to obtain a recognition rate sufficient for practical application of automatic input to the CAD system.
［実施例］ 第２図は本発明の手書き文字認識方法が実施される図
面読取装置の実施例構成図である。[Embodiment] FIG. 2 is a block diagram of an embodiment of a drawing reading apparatus in which the handwritten character recognition method of the present invention is implemented.
第２図において、18は図面読取部であり、例えばイメ
ージスキャナを使用してプラント設計等の手書きされた
アイソメ図を読取ってイメージデータに変換する。20は
文字列抽出部であり、図面読取部18で読取られた図面の
イメージデータの中から傾きを持って書かれた文字列を
切り出し、文字列の傾きα、平行四辺形として設定され
た切り出し領域、切り出された各文字に対する輪郭ベク
トル、各文字の輪郭ベクトルの細線化処理により得られ
た細線化ベクトルを出力する。22は斜め変形検出部であ
り、後の説明で明らかにする正規化前の細線化ベクトル
長の総和と正規化後の細線化ベクトル長との総和との大
小比較により斜め変形の有無を検出する。24は文字正規
化部であり、斜め変形検出部22で斜め変形有りと判定さ
れた文字列を対象として所定の正規化変換式を使用した
各文字の正規化変換、具体的には各文字を構成する細線
化ベクトル座標の正規化変換を行なう。26は文字認識部
であり辞書を参照して斜め変形の無い文字を対象に文字
認識を行なう。In FIG. 2, a drawing reading unit 18 reads a handwritten isometric drawing such as a plant design using an image scanner and converts it into image data. Reference numeral 20 is a character string extraction unit, which cuts out a character string written with an inclination from the image data of the drawing read by the drawing reading unit 18 and extracts the inclination α of the character string and a parallelogram. The region, the contour vector for each character cut out, and the thinning vector obtained by the thinning process of the contour vector of each character are output. Reference numeral 22 denotes an oblique deformation detection unit, which detects the presence or absence of oblique deformation by comparing the sum of the thinned vector lengths before normalization and the total of the thinned vector lengths after normalization, which will be clarified later. . Reference numeral 24 denotes a character normalization unit, which normalizes and converts each character using a predetermined normalization conversion formula for a character string determined by the oblique deformation detection unit 22 to have a diagonal deformation, specifically, each character. Performs a normalization conversion of the thinned vector coordinates that compose it. A character recognition unit 26 refers to a dictionary to perform character recognition on characters without oblique deformation.
第３図は第２図の実施例における文字認識処理フロー
図である。FIG. 3 is a flow chart of character recognition processing in the embodiment of FIG.
第３図において、まず読取図面のイメージデータを対
象としてS1で文字列抽出、即ち読取図面中に存在する文
字列の切出しを行ない、文字列の傾きα、平行四辺形の
切出し領域、各文字に対する輪郭ベクトルと細線化ベク
トルを生成する。In FIG. 3, first, a character string is extracted in S1 from the image data of the read drawing, that is, the character string existing in the read drawing is cut out, and the inclination α of the character string, the cutout area of the parallelogram, and each character Generate a contour vector and a thinning vector.
続いてS2でS1で抽出した文字列の傾きαが０゜、ある
いは90゜以外の傾きか否か判定する。文字列の傾きαが
０゜または90゜以外の場合にはS3に進み、斜め変形があ
るか否かの検出処理を行ない、斜め変形が判定された文
字列についてのみS4で斜め変形による歪みを除去するた
めの文字正規化処理を行ない、最終的にS5で斜め変形の
無い文字を対象とした文字認識を行なうようになる。Subsequently, in S2, it is determined whether the inclination α of the character string extracted in S1 is an inclination other than 0 ° or 90 °. If the inclination α of the character string is other than 0 ° or 90 °, the process proceeds to S3 to detect whether there is diagonal deformation, and only the character string for which diagonal deformation is determined is distorted due to diagonal deformation in S4. Character normalization processing is performed for removal, and finally character recognition for characters without oblique deformation is performed in S5.
次に第３図の文字認識処理フロー図におけるS3の斜め
変形の検出処理及びS4の文字正規化処理について詳細に
説明する。Next, the oblique deformation detection process of S3 and the character normalization process of S4 in the character recognition process flow chart of FIG. 3 will be described in detail.
第４図は本発明の斜め変形の検出原理図であり、文字
列「AA」を例にとって斜め変形の有る文字列28、斜め変
形の有る文字列に正規化を行なった文字列30、正規化さ
れた文字列30を更に正規化した文字列32を示している。
即ち、文字列28は斜め変形が有り、この文字列28に斜め
変形による歪みを除く正規化を施すと斜め変形の無い文
字列30が得られる。一方、読取図面から抽出された文字
列が文字列30に示すように斜め変形が無かった場合に
は、この斜め変形の無い文字列30に正規化を施すことで
斜め変形による歪みを受けた文字列30に変換されること
になる。ここで斜め変形の有る文字列28を正規化して斜
め変形の無い文字列30とした場合をケース、斜め変形
の無い文字列30を正規化して斜め変形の有る文字列32と
した場合をケースとする。FIG. 4 is a principle diagram of detection of diagonal deformation of the present invention. A character string "AA" is used as an example, a character string 28 having oblique deformation, a character string 30 obtained by normalizing a character string having oblique deformation, and a normalization. A character string 32 obtained by further normalizing the generated character string 30 is shown.
That is, the character string 28 has an oblique deformation, and if the character string 28 is normalized to remove the distortion due to the oblique deformation, the character string 30 without the oblique deformation is obtained. On the other hand, when the character string extracted from the read drawing does not have the diagonal deformation as shown in the character string 30, the character string 30 without the oblique deformation is subjected to the normalization to obtain the character that is distorted by the oblique deformation. Will be converted to column 30. Here, the case where the character string 28 with diagonal deformation is normalized to form the character string 30 without diagonal deformation, and the case where the character string 30 without diagonal deformation is normalized to form the character string 32 with diagonal deformation is referred to as the case. To do.
ケースにおける正規化前の文字列28と正規化後の文
字列30について、各文字を構成する細線化ベクトル長の
総和を求めて大小関係を比較すると、斜め変形の有る文
字列28の場合には正規化後の総ベクトル長が長くなる関
係にある。For the character string 28 before normalization and the character string 30 after normalization in the case, the sum of the lengths of the thinned vector composing each character is calculated and the size relationship is compared.In the case of the character string 28 with diagonal deformation, There is a relation that the total vector length after normalization becomes long.
一方、ケースの斜め変形の無い文字列30を正規化し
た場合には、正規化前の細線化ベクトル長の総和と正規
化後の細線化ベクトル長の総和との間に、図示のように
正規化後の総ベクトル長が長くなる関係がある。このよ
うな斜め変形の有る場合と斜め変形の無い場合の文字列
に対する正規化前と正規化後の文字列を構成する各文字
の細線化ベクトル長の総和の大小関係から、処理対象と
している文字列、即ち正規化前の文字列が斜め変形を持
つか否かを検出することができる。On the other hand, when the character string 30 with no diagonal deformation of the case is normalized, the normalization is performed between the sum of the thinned vector lengths before normalization and the sum of the thinned vector lengths after the normalization as shown in the figure. There is a relation that the total vector length after conversion becomes long. Characters to be processed based on the size relationship of the sum of the thinned vector lengths of each character that makes up the character strings before and after normalization for the character string with and without diagonal deformation It is possible to detect whether a string, that is, a character string before normalization has an oblique deformation.
第５図は第４図の斜め変形の検出原理に基づいた具体
的な斜め変形の判定処理フロー図である。FIG. 5 is a flow chart showing a concrete judgment process of oblique deformation based on the principle of detecting oblique deformation in FIG.
第５図において、まずS1で文字列を構成する各文字の
細線化ベクトル長lijを算出する。例えば第６図に示す
傾きαを持つ文字列「Ａ・・・Ｚ」を処理対象とした場
合、例えば文字「Ａ」は細線化ベクトルV11〜V15で構成
されている。In FIG. 5, first, in S1, the thinning vector length l ij of each character forming the character string is calculated. For example, when the character string “A ... Z” having the inclination α shown in FIG. 6 is the processing target, for example, the character “A” is composed of thinning vectors V 11 to V 15 .
文字の細線化ベクトル長lijは、 で算出される。ここでｉは文字列を構成する各文字の順
番を示す番号、ｊは１つの文字を構成する細線化ベクト
ルの数を示す番号である。The thinned vector length l ij of the character is Is calculated. Here, i is a number indicating the order of each character forming the character string, and j is a number indicating the number of thinning vectors forming one character.
第６図の文字「Ａ」の場合、例えば最初の細線化ベク
トルV11の細線化ベクトル長l11は始点座標を（a11,
b11）、終点座標を（c11,d11）として前記（１）式によ
り算出される。このような細線化ベクトル長の算出を最
初の文字「Ａ」の細線化ベクトルV11から最後の文字
「Ｚ」の細線化ベクトルVn1まで行なう。For character of FIG. 6, "A", for example, the first thinning vector length l 11 of the start point coordinates (a 11 thinning vector V 11,
b 11 ), and the end point coordinates are (c 11 , d 11 ), and are calculated by the equation (1). The calculation of the thinning vector length is performed from the thinning vector V 11 of the first character “A” to the thinning vector V n1 of the last character “Z”.
続いてS2に進み文字列「Ａ〜Ｚ」の総ベクトル長Ｌ
を、 により算出する。Then, the procedure proceeds to S2, where the total vector length L of the character strings "AZ" is L.
To It is calculated by:
次にS3に進み、次の正規化変換式により各文字の細線
化ベクトルVijを座標変換して細線化ベクトルijを求
める。Next, in S3, the thinned vector V ij of each character is subjected to coordinate conversion by the following normalization conversion formula to obtain the thinned vector ij .
この（３）式による座標変換は第６図に示すように、
文字列の傾き角度をα、文字列の開始点座標を（x0,
y0）として任意の座標値（x,y）に対し斜め変形による
歪み補正の座標変換を行なうものである。 The coordinate conversion by the equation (3) is as shown in FIG.
The inclination angle of the character string is α, the start point coordinate of the character string is (x 0 ,
As y 0 ), the coordinate conversion for distortion correction by oblique deformation is performed on an arbitrary coordinate value (x, y).
ここで前記（３）式における定数ｈは、第７図に示す
ように文字列の開始点（x0,y0）を通る傾きαの直線に
対する正規化前の（x,y）の垂線の長さであり、 ｈ＝｜−sin（α）×（ｘ−x0）＋cos（α） ×（ｙ−y0）｜ （４） として与えられる。また第７図は正規化後の（，）
を併せて示している。更に前記（３）式の変換式におけ
る係数K1（α）の値は文字列の傾き角度αに依存して決
まる値であり、経験的に設定することが望ましく、例え
ば、 として設定される。即ち、第８図に示すように、文字列
の傾き角度αがπ/2以下ではK1（α）＝1.0として補正
を行ない、傾き角度αがπ/2を超え、3/8π以下の範囲
ではK1（α）＝0.7として角度αの正弦及び余弦成分に
よる補正割合を抑え、更に3/8π以上ではK1（α）＝0.0
とし、歪み補正は行なわないようにしている。Here, the constant h in the equation (3) is the normal line of (x, y) before normalization with respect to the straight line of the inclination α passing through the starting point (x 0 , y 0 ) of the character string as shown in FIG. It is a length and is given as h = | −sin (α) × (x−x 0 ) + cos (α) × (y−y 0 ) | (4). Fig. 7 shows (,) after normalization.
Are also shown. Further, the value of the coefficient K 1 (α) in the conversion formula of the above formula (3) is a value determined depending on the inclination angle α of the character string, and it is desirable to set it empirically. Is set as That is, as shown in FIG. 8, when the inclination angle α of the character string is π / 2 or less, correction is performed by setting K 1 (α) = 1.0, and the inclination angle α exceeds π / 2 and falls within the range of 3 / 8π or less. Then K 1 (α) = 0.7 to suppress the correction rate due to the sine and cosine components of the angle α, and if 3 / 8π or more, K 1 (α) = 0.0
Therefore, the distortion correction is not performed.
再び第５図を参照するに、S3で前記（３）式によりベ
クトルVijを座標変換して正規化されたベクトルijを
求めたならばS4に進み、S1の場合と同様、変換ベクトル
ijで成る各文字の細線化ベクトル長ijを算出し、続
いてS5に進み、変換後の文字列を対象とした総ベクトル
長をS2の場合と同様にして算出する。Referring to FIG. 5 again, if the vector V ij is coordinate-converted by the equation (3) in S3 to obtain the normalized vector ij , the process proceeds to S4, and as in the case of S1, the converted vector
calculating a thinning vector length ij of each character comprising in ij, then the process proceeds to S5, calculated in the same way as the case of the total vector length intended for the converted string S2.
次にS6に進み、S2で算出された変換前（正規化前）の
総ベクトル長Ｌと、S5で算出された変換後（正規化後）
の総ベクトル長の大小関係を比較する。具体的には変
換後の総ベクトル長と変換前の総ベクトル長Ｌに文字
列の傾き角度αによって決まる係数K2（α）を掛け合わ
せた値との比較を行なう。そして、 の条件式に従って変換後の総ベクトル長がK2（α）×
Ｌより短ければS7に進んで斜め変形有りと判定し、長け
ればS8に進んで斜め変形無しと判定する。Next, in S6, the total vector length L before conversion (before normalization) calculated in S2 and after conversion (after normalization) calculated in S5.
Compare the magnitude relation of the total vector length of. Specifically, the total vector length after conversion and the total vector length L before conversion are compared with a value obtained by multiplying the coefficient K 2 (α) determined by the inclination angle α of the character string. And The total vector length after conversion is K 2 (α) ×
If it is shorter than L, the process proceeds to S7 and it is determined that there is diagonal deformation, and if it is longer, it proceeds to S8 and it is determined that there is no diagonal deformation.
このS6に使用する（６）式の係数K2（α）の値は、経
験的に設定すれば良く、例えば次のように定められる。The value of the coefficient K 2 (α) of the equation (6) used for S6 may be set empirically, and is determined as follows, for example.
この第５図に示す斜め変形の判定処理により斜め変形
有りと判定された場合には、第３図のS4に進んで文字列
の正規化が行なわれる。この文字列の正規化は斜め変形
の判定における前記（３）式により行なうことができ、
具体的には第５図のS3で既に細線化ベクトルVijの歪み
を除去する変換が行なわれていることから、その変換結
果を使用する。一方、第５図の斜め変形判定処理で斜め
変形無しと判定された場合には、第３図に示すようにS2
で判定された傾斜角度αが０゜あるいは90゜の場合の文
字列と同様にそのままS5に進み文字認識を行なう。 If it is determined by the oblique deformation determination process shown in FIG. 5 that there is oblique deformation, the process proceeds to S4 in FIG. 3 to normalize the character string. The normalization of this character string can be performed by the above equation (3) in the determination of diagonal deformation,
Specifically, since the conversion for removing the distortion of the thinned vector V ij has already been performed in S3 of FIG. 5, the conversion result is used. On the other hand, if it is determined in the diagonal deformation determination process of FIG. 5 that there is no diagonal deformation, as shown in FIG.
In the same way as the character string when the inclination angle α determined in step 0 is 0 ° or 90 °, the process proceeds directly to S5 and character recognition is performed.
尚、上記の実施例にあっては、係数K1（α）及びK
2（α）については経験測に従って、例えば（５）式、
（７）式のように値を決めているが、本発明はこれらの
経験値に限定されず必要に応じて適宜の係数値を使用す
るようにしても良い。In the above embodiment, the coefficients K 1 (α) and K 1
2 For (α), according to empirical measurement, for example, equation (5),
Although the value is determined as in the equation (7), the present invention is not limited to these empirical values, and an appropriate coefficient value may be used if necessary.
次に本発明の手書き文字認識方法による認識率を従来
方法と対比して説明する。Next, the recognition rate by the handwritten character recognition method of the present invention will be described in comparison with the conventional method.
第９図は文字認識率を検証するために使用した異なる
傾きを持つ入力図面の文字列を示すもので、この第９図
の入力文字列について15゜刻みの傾きのある文字列の認
識と対象とした従来方法の認識結果は第10図に示すもの
であった。第10図において○印を付した文字は認識成
功、×印を付した文字は認識失敗を示している。第10図
の場合、認識対象とした55文字のうち、正解は30文字、
誤りは25文字であり、認識率は約54.5％であった。FIG. 9 shows the character strings of the input drawings with different inclinations used to verify the character recognition rate. Regarding the input character strings of FIG. The recognition result of the conventional method is as shown in FIG. In FIG. 10, characters marked with a circle indicate recognition success, and characters marked with a cross indicate recognition failure. In the case of FIG. 10, out of the 55 characters to be recognized, the correct answer is 30 characters,
The error was 25 characters, and the recognition rate was about 54.5%.
第11図は第９図の入力文字列を対象とした本発明の手
書き文字認識方法による認識結果を示す。第11図の本発
明の場合、認識対象とした55文字中、正確は47文字、誤
りは８文字であり、約85.5％の認識率を達成することが
でき、従来方法に比べ大幅な認識率の向上が確認され
た。FIG. 11 shows the recognition result by the handwritten character recognition method of the present invention for the input character string of FIG. In the case of the present invention shown in FIG. 11, of the 55 characters to be recognized, 47 are accurate and 8 are error, and a recognition rate of about 85.5% can be achieved, which is a significant recognition rate compared with the conventional method. It was confirmed that
［発明の効果］ 以上説明してきたように本発明によれば、プラントア
イソメ図等のような30゜や60゜の傾いた手書き文字列を
含む図面の文字認識を確実に行なうことができ、文字認
識率の向上によりCADシステムに対し実用的なレベルで
図面データを自動入力できる図面読取装置を接続した図
面入力システムを構築することができる。[Effects of the Invention] As described above, according to the present invention, it is possible to reliably perform character recognition of a drawing including a handwritten character string inclined at 30 ° or 60 °, such as a plant isometric drawing, etc. By improving the recognition rate, it is possible to construct a drawing input system that is connected to a drawing reading device that can automatically input drawing data to a CAD system at a practical level.
第１図は本発明の原理説明図； 第２図は本発明の実施例構成図； 第３図は本発明の文字認識処理フロー図； 第４図は本発明の斜め変形検出原理図； 第５図は本発明の斜め変形判定処理フロー図； 第６図は本発明の斜め変形の検出説明図； 第７図は本発明の正規化変換に用いるパラメータ説明
図； 第８図は本発明の正規化変換に用いる係数K1（α）説明
図； 第９図は本発明の認識率検証に用いた入力文字列説明
図； 第10図は従来方法の認識結果説明図； 第11図は本発明の認識結果説明図である。 図中、 18:図面読取部 20:文字列抽出部 22:斜め変形検出部 24:文字正規化部 26:文字認識部FIG. 1 is an explanatory view of the principle of the present invention; FIG. 2 is a block diagram of an embodiment of the present invention; FIG. 3 is a flow chart of character recognition processing of the present invention; FIG. 5 is a flow chart of the oblique deformation determination process of the present invention; FIG. 6 is an explanatory view of detection of the oblique deformation of the present invention; FIG. 7 is an explanatory view of parameters used in the normalization conversion of the present invention; FIG. 9 is an explanatory diagram of a coefficient K 1 (α) used for normalization conversion; FIG. 9 is an explanatory diagram of an input character string used for verification of the recognition rate of the present invention; FIG. 10 is an explanatory diagram of recognition results of a conventional method; It is a recognition result explanatory drawing of an invention. In the figure, 18: drawing reading unit 20: character string extraction unit 22: diagonal deformation detection unit 24: character normalization unit 26: character recognition unit
Claims (6)
形が検出されたとき、その変形による歪みを除去して各
文字を正規化し、その正規化された文字列あるいは斜め
変形なしと判定された文字列の各文字を認識する図面読
取装置の文字認識方法に於いて、 文字列を構成する各文字の細線化ベクトル長を算出する
第１過程と； 該第１過程で算出された細線化ベクトル長の総和を算出
する第２過程と； 斜め変形歪みを除去する所定の正規化変換式により各文
字の細線化ベクトルを座標変換し、該座標変換ベクトル
でなる各文字の細線化ベクトル長を算出する第３過程
と； 該第３過程で算出された正規化文字の細線化ベクトル長
の総和を算出する第４過程と； 前記第２過程で算出された正規化前の総ベクトル長と前
記第４過程で得られた正規化後の総ベクトル長とを比較
し、正規化により総ベクトル長が短くなったときには斜
め変形ありと判定し、正規化によりベクトル長が長くな
ったときには斜め変形なしと判定する第５過程と； を備えたことを特徴とする図面読取装置の文字認識方
法。1. When an oblique deformation of a character string extracted from a read drawing is detected, distortion due to the deformation is removed to normalize each character, and it is determined that the normalized character string or the oblique deformation is not present. A character recognition method of a drawing reading device for recognizing each character of a character string that has been generated; a first step of calculating a thinning vector length of each character that constitutes the character string; and a thin line calculated in the first step. A second step of calculating the sum of the generalized vector lengths; coordinate conversion of the thinned vector of each character by a predetermined normalization conversion formula that removes oblique deformation distortion, and the thinned vector length of each character formed by the coordinate conversion vector A fourth step of calculating the sum of the thinned vector lengths of the normalized characters calculated in the third step; and a total vector length before normalization calculated in the second step Normal obtained in the fourth step A fifth step of comparing the subsequent total vector length, determining that there is diagonal deformation when the total vector length is shortened by normalization, and determining that there is no diagonal deformation when the vector length is increased by normalization; A character recognition method for a drawing reading device, which is provided.
法に於いて、 前記第３過程で使用する正規化変換式は、文字列の傾き
角度をα、文字列の開始点を（x0,y0）、正規化する任
意のベクトル座標値を（x,y）、正規化後のベクトル座
標を、（，）とした時、 ＝ｘ−cos（α）×ｋ（ｘ）×k1（α） ＝ｙ−sin（α）×ｋ（ｘ）×k1（α） で与えられることを特徴とする図面読取装置の文字認識
方法。2. The character recognition method for a drawing reading device according to claim 1, wherein the normalization conversion formula used in the third step is that the inclination angle of the character string is α, and the start point of the character string is (x0 , y0), where (x, y) is an arbitrary vector coordinate value to be normalized and (,) is the vector coordinate after normalization, = x−cos (α) × k (x) × k1 (α ) = Y−sin (α) × k (x) × k1 (α), the character recognition method of the drawing reading device.
法に於いて、 前記正規化変換式に使用するｋ（ｘ）は文字列の開始点
（x0,y0）を通る傾きαの直線への垂線の長さであり、 ｋ（ｘ）＝｜−sin（α）×（ｘ−x0）＋cos（α） ×（ｙ−y0）｜ で与えられ、且前記正規化の変換式に使用する定数k1
（α）は経験により設定した１〜０の範囲の値としたこ
とを特徴とする図面読取装置の文字認識方法。3. The character recognition method for a drawing reading apparatus according to claim 2, wherein k (x) used in the normalization conversion equation is a straight line having an inclination α passing through a start point (x0, y0) of a character string. Is the length of the vertical line to k, and is given by k (x) = | −sin (α) × (x−x0) + cos (α) × (y−y0) | and used in the conversion formula for the normalization. Constant k1
(Α) is a value in the range of 1 to 0 set by experience.
法に於いて、 前記定数k1（α）を経験により、 ０＜｜α｜≦π/4の時、k1（α）＝1.0 π/4＜｜α｜≦３π/8の時、k1（α）＝0.7 ３π/8＜｜α｜≦π/2の時、k1（α）＝0.0 としたことを特徴とする図面読取装置の文字認識方法。4. The character recognition method for a drawing reading apparatus according to claim 3, wherein, when 0 <| α | ≦ π / 4, k1 (α) = 1.0 π by experience with the constant k1 (α). / 4 <| α | ≦ 3π / 8, k1 (α) = 0.7 3π / 8 <| α | ≦ π / 2, k1 (α) = 0.0 Character recognition method.
法に於いて、 前記第５過程の斜め変形の判定として、正規化前の総ベ
クトル長に重み定数k2（α）を乗算した値と正規化後の
総ベクトル長とを比較し、正規化により総ベクトル長が
等しいか短くなったときには斜め変形ありと判定し、正
規化により総ベクトル長が長くなったときには斜め変形
なしと判定することを特徴とする図面読取装置の文字認
識方法。5. The character recognition method for a drawing reading apparatus according to claim 1, wherein a value obtained by multiplying a total vector length before normalization by a weighting constant k2 (α) is used as the judgment of the oblique deformation in the fifth step. And the total vector length after normalization are compared, and when the total vector length is equal or shorter by normalization, it is determined that there is diagonal deformation, and when the total vector length is longer by normalization, it is determined that there is no diagonal transformation. A character recognition method for a drawing reading device, characterized in that.
法に於いて、 前記重み定数k2（α）の値は、経験により、 ０＜｜α｜≦π/4の時、k2（α）＝1.0＋cos（α/2） π/4＜｜α｜≦３π/8の時、k2（α）−0.7＋cos（π/
8） ３π/8＜｜α｜≦π/2の時、k2（α）＝0.0 と設定したことを特徴とする図面読取装置の文字認識方
法。6. The character recognition method for a drawing reading apparatus according to claim 5, wherein the value of said weighting constant k2 (α) is k2 (α) when 0 <| α | ≦ π / 4 by experience. ) = 1.0 + cos (α / 2) π / 4 <| α | ≦ 3π / 8, k2 (α) −0.7 + cos (π /
8) A character recognition method for a drawing reading device, wherein k2 (α) = 0.0 is set when 3π / 8 <| α | ≦ π / 2.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2178177A JP2561969B2 (en) | 1990-07-05 | 1990-07-05 | Character recognition method for drawing reader |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2178177A JP2561969B2 (en) | 1990-07-05 | 1990-07-05 | Character recognition method for drawing reader |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| JPH0465786A JPH0465786A (en) | 1992-03-02 |
| JP2561969B2 true JP2561969B2 (en) | 1996-12-11 |
Family
ID=16043956
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| JP2178177A Expired - Lifetime JP2561969B2 (en) | 1990-07-05 | 1990-07-05 | Character recognition method for drawing reader |
Country Status (1)
| Country | Link |
|---|---|
| JP (1) | JP2561969B2 (en) |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPS6024678A (en) * | 1983-07-21 | 1985-02-07 | Fujitsu Ltd | Picture reader |
| JPH0266690A (en) * | 1988-08-31 | 1990-03-06 | Mitsubishi Electric Corp | Image processor |
-
1990
- 1990-07-05 JP JP2178177A patent/JP2561969B2/en not_active Expired - Lifetime
Also Published As
| Publication number | Publication date |
|---|---|
| JPH0465786A (en) | 1992-03-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP3218004B2 (en) | Test signature verification method | |
| JP3335538B2 (en) | Method and apparatus for collating a handwritten character string | |
| JP5468332B2 (en) | Image feature point extraction method | |
| CN111626246B (en) | Face alignment method under mask shielding | |
| WO2018176514A1 (en) | Fingerprint registration method and device | |
| CN113378764A (en) | Video face acquisition method, device, equipment and medium based on clustering algorithm | |
| JP2561969B2 (en) | Character recognition method for drawing reader | |
| JPH10307889A (en) | Character recognition method, its device and recording medium recording character recognition program | |
| JP2833313B2 (en) | Authenticator for fingerprint feature points | |
| KR20090031084A (en) | Face recognition system, the method of face recognition registration and confirmation | |
| KR100564762B1 (en) | Authentication method and apparatus using fingerprint | |
| JPH01271884A (en) | Detecting system for center of fingerprint | |
| JP2001076142A (en) | Method and device for collating fingerprint image and recording medium with this method recorded thereon | |
| JP2008276758A (en) | Fingerprint singular point extractor and fingerprint singular points extraction method | |
| JPH07200735A (en) | Graphic recognition device | |
| JP2935331B2 (en) | Figure recognition device | |
| CN112016610B (en) | Image feature matching method and system | |
| JP3006823B2 (en) | Character and word recognition methods | |
| JP2856409B2 (en) | Character recognition apparatus and method | |
| JP3074691B2 (en) | Character recognition device | |
| JPH0235577A (en) | Individual certifying device | |
| CN110619295A (en) | Face recognition method and system | |
| JPH05205035A (en) | Fingerprint collating device | |
| JP2832035B2 (en) | Character recognition device | |
| TW200830197A (en) | Symbol recognition method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| FPAY | Renewal fee payment (event date is renewal date of database) |
Free format text: PAYMENT UNTIL: 20080919 Year of fee payment: 12 |
|
| FPAY | Renewal fee payment (event date is renewal date of database) |
Free format text: PAYMENT UNTIL: 20080919 Year of fee payment: 12 |
|
| FPAY | Renewal fee payment (event date is renewal date of database) |
Free format text: PAYMENT UNTIL: 20090919 Year of fee payment: 13 |
|
| FPAY | Renewal fee payment (event date is renewal date of database) |
Free format text: PAYMENT UNTIL: 20090919 Year of fee payment: 13 |
|
| FPAY | Renewal fee payment (event date is renewal date of database) |
Free format text: PAYMENT UNTIL: 20100919 Year of fee payment: 14 |
|
| EXPY | Cancellation because of completion of term |