JPH05242294A - Drawing reader - Google Patents

Abstract

PURPOSE:To recognize a character and a graphic by surely distinguishing the one from the other even when the graphic with same size as that of the character and the character exist on the same drawing in a drawing reader which extracts character data from binary picture element data in accordance with black/white obtained by raster-scanning a subject in which the character, a line segment, and a symbol are mixed. CONSTITUTION:A character candidate is found by performing contour vectorization processing based on the binary picture element data (step S1), and polygonal graphic detection processing is applied to the character candidate and it is classified to the character candidate and a character graphic candidate(step S2), and the character candidate is made into a character string, and also, the character graphic candidate not being made into the character string is set as the graphic(step S3), and the character is segmented from the character string by unifying adjacent plural character candidates in the character string(step S4), and character recognition is performed based on segmented character data(step S5).

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a drawing reading apparatus for detecting character data from a document or drawing in which characters, line segments, symbols and the like are mixed and recognizing the character.

[0002]

2. Description of the Related Art Conventionally, a drawing reading device for reading characters and figures is disclosed in Japanese Patent Application No. 1-253451. This apparatus performs contour vectorization processing on binarized pixel data corresponding to black and white obtained by raster scanning an object such as a document or drawing in which characters, line segments, and symbols are mixed, and circumscribes the contour vector. A contour vectorization processing unit that obtains circumscribed quadrangle data as character candidates,
A registration unit that registers the circumscribed quadrangle data of the character candidates in a tree structure, a character string candidate extraction unit that extracts the character string candidates from the registered character candidates, and a character string extraction that extracts the character strings from the character string candidates. And a character cutout unit having a function of integrating a plurality of character candidates in the extracted character string.

[0003]

The drawing reading device is
This is effective when there are no isolated figures other than characters in the drawing to be entered, which are of the same size as the characters, but a figure representing a building that has the same size as the characters, such as an urban map, is used. The following problems occur when inputting a drawing that exists.

That is, when the elements are separated, a graphic having a size smaller than the character size is selected as a character candidate, resulting in an incorrect recognition result. For example, when recognizing a map as shown in FIG. 19 (a), the correct recognition result is as shown in FIG. 19 (b). However, when it is recognized by the drawing reading device, a figure is converted into a character as shown in FIG. 20 (a). The character may be erroneously recognized, or the character may be erroneously recognized as a figure as shown in FIG.

The present invention has been made in view of the above points, and an object thereof is to reliably distinguish a character from a figure even when a figure having a size similar to that of a character exists in the same drawing. It is to provide a drawing reading device that can be recognized.

[0006]

SUMMARY OF THE INVENTION The present invention is directed to characters, line segments,
Drawing reading for recognizing characters by extracting character data from binary pixel data corresponding to black and white obtained by raster scanning an object with mixed symbols and referring to the extracted character data and a character dictionary for comparison In the apparatus, based on the binarized pixel data corresponding to black and white obtained by raster-scanning the subject, data of a contour vector formed by combining vectors connecting two black pixels adjacent to each other is obtained, and A contour vectorization processing unit that obtains circumscribed quadrangle data circumscribing the contour vector as a character candidate, and a polygonal figure is detected from the character candidate data obtained by the contour vectorization processing unit, and the detected polygonal figure A polygon figure detection processing unit that obtains data as character and figure candidates, and circumscribed quadrangle data of the character candidates obtained by the contour vectorization processing unit The registration range to be registered in the Lee structure, and the search range is determined from the core of the character string candidates set to a predetermined size, and among the character candidates registered in the registration unit, the central coordinates are characters within the search range. The character string candidate extraction unit that extracts a string as a character string candidate, and the height h _C of the character string candidate core and the adjacent character candidates in the character string candidate from the data extracted by the character string candidate extraction unit. The distance d is d
<H _C × α (α is a constant) A character string extraction unit for extracting data as a character string, and a function for integrating a plurality of adjacent character candidates in the character string extracted by the character string extraction unit And a character slicing unit for slicing a character from a character string, and performing character recognition based on the character data cut out by the character slicing unit, and a character and figure candidate obtained by the polygonal figure detection processing unit. The feature is that character figure candidates that are not converted into character strings are recognized as figures.

[0007]

When the circumscribed quadrangle is obtained by the vectorization processing unit, character candidate data can be obtained from a subject (document or drawing) in which characters, line segments, and symbols are mixed. The circumscribed quadrangle data of the character candidates is registered in the registration unit in a tree structure. The polygonal figure detection processing unit detects a polygonal figure from the character candidate data and obtains a character figure candidate.

The character string candidate extraction unit searches the data in the registration unit within a search range determined by the character string candidate kernel, and extracts a character string whose center coordinates are within the search range as a character string candidate.
Since the circumscribed quadrangle data is registered in a tree structure, the area quadrangle neighborhood search processing is reduced. Among the character string candidates, data in which the relationship of d <h _C × α (d is the distance between adjacent character candidates, h _C is the height of the character string candidate kernel, and α is a constant) is satisfied is the character string extraction unit. It is extracted as a character string by. That is, line segments and symbols mixed in the subject are excluded and only character string data is extracted.

Next, the character cutout unit cuts out a character from the character string. At this time, if the size of the plurality of adjacent character candidates in the character string and the distance between the character candidates are predetermined values,
The plurality of character candidates are integrated and the integrated character candidate is cut out as one character. As a result, the characters can be read as characters regardless of the size and format of the characters.

Further, among the character and figure candidates obtained by the polygonal figure detection processing section, character and character candidates which are not converted into character strings are recognized as figures. Therefore, even when a figure having a size similar to that of a character exists in the same drawing as in a city map, the character and the figure can be reliably distinguished and recognized.

[0011]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. First, the drawing reading apparatus according to the present invention includes a contour vectorization processing unit 1, a registration unit 2, a character string candidate extraction unit 3, a character string extraction unit 4, a character cutout unit 5 and a character recognition unit 6 as shown in FIG. The overall processing flow is shown in FIG. That is, first, in step S ₁ , the contour vectorization processing unit 1 performs preprocessing, in step S ₂ , polygon figure detection processing is performed, and in step S ₃ , the registration unit 2, the character string candidate extraction unit 3, and the character string extraction unit 4 are performed. There performs character string processing, character extraction section 5 performs a character segmentation process in step S _4, the character recognition unit 6 performs character recognition processing in step S _5.

The preprocessing performed by the contour vectorization processing unit 1 is represented by the flowchart shown in FIG. That performs image input processing in step S _1, whereby the border vectorized process in step S _2, the element separation process in step S _3, performs polygon graphic detection processing.

Of these, the image input processing and the contour vectorization processing are described, for example, in No. 4 of Japanese Patent Application No. 1-253451.
It is performed by the means described in FIGS. That is, based on the binarized pixel data corresponding to black and white obtained by raster-scanning a subject in which characters, line segments, and symbols are mixed, a contour vector formed by combining vectors connecting two black pixels adjacent to each other. While seeking the data of
The circumscribed quadrangle data circumscribing the contour vector is obtained (FIG. 9).

The polygon figure detection processing is performed according to the flowchart of FIG. First, in step S ₁ , it is determined whether or not contour vectorization is completed, and in step S ₂ , it is determined whether or not it is a character candidate. If it is a character candidate, polygon figure detection processing is performed in step S ₃ , and step S _{3 Four}
It is determined whether or not it is a polygon figure, and if it is a polygon figure, the character candidate is changed to a character figure candidate (step S).
₅ ). By performing the polygonal figure detection processing as described above, it is possible to classify into two, character candidates and character figure candidates.

The extraction of the character candidates (element separation processing) is performed as follows. First, FIG. 5 (Japanese Patent Application No. 1-25345)
The short vector is removed using the contour vector sequence table (identical to FIG. 9 of No. 1). FIG. 6 is a flow chart for explaining the short vector removing means. In step S ₁ , the contour vector circumscribed quadrangle (shown in FIG. 7) from the contour vector series table shown in FIG. 5 (x _min , y _min ),
(x _max , y _max ) is calculated. Next, in step S ₂ , dx =
x _man −x _min , dy = y _max −y _min and threshold value DX _th ,
Compare with DY _th . Step S ₃ is the comparison / judgment unit of step S ₂ , and in this step S ₃ (dx> DX _th ) ∪
If (dy> DY _th ) is determined, and if YES, information of a circumscribed quadrangle is created as a contour vector in step S ₄ .
An example of this information is shown in FIG. 8 as a circumscribed quadrangle information table. Remove the vector of the same series in Step S ₅ from the outline vector sequence table shown in FIG. 5, if NO to remove noise (noise) in the step S _3.

FIG. 7 is an explanatory diagram showing the circumscribed quadrangle of the contour vector. In this figure, x _max , x _min , y _max ,
y _min is as follows. x _max = max (..., x _i , ...) x _min = min (..., x _i , ...) y _max = max (..., y _i , ...) y _min = min (..., y _i , ...) dx = x _max −x _min dy = y _max −y _min (dx <DX _th ) ∩ (dy <DY _th ) where DX _th and DY _th are threshold values.

Character candidates are extracted under the following conditions based on the circumscribed quadrangle information (FIG. 9) obtained as described above. Condition 1 w _x <threshold, and w _y <threshold (w _x is the length of the circumscribing rectangle in the x direction, w _y is the length of the circumscribing rectangle in the y direction) Condition 2 Outer contour vector and circumscribing rectangle of character candidate The contour vectors of the outer circumference and the inner circumference which are completely included in are linked to the contour vectors which are the character candidates.

Next, a character string conversion process (step S _{3 in} FIG. 2)
Will be described. This process is performed according to the flow charts of FIGS. 10A and 10B after the polygon figure detection process of FIG. That is, it is determined in step S ₁ whether or not contour vectorization is completed, it is determined in step S ₂ whether or not it is a character candidate or a character / graphic candidate, and in step S ₃ , the character candidate is set as a tree structure in the registration unit of FIG. Register in 2. Further, in step S ₄ , it is determined whether or not the character string conversion target ends, in step S ₅ it is determined whether or not the character string is not yet converted, and in step S ₆ , it is determined whether or not it becomes a character string nucleus. If it is the character string nucleus, the character string candidate range is searched centering on the character string nucleus in step S ₇ . Then it is determined whether or not found in the range in step S _8, if found contain characters which uses該見the string (step S _9). Also step S
_{In step 10} , it is determined whether or not it is the end of the character string conversion target, in step S ₁₁ , it is determined whether the character string is not yet converted, and in step S ₁₂ , it is determined whether or not it is a character / graphic candidate. in step S ₁₃ if determining the graphic character candidate figure.

As described above, as shown in FIGS. 4 and 10 (a),
By performing the process of (b), the condition that the character string nucleus that starts the character string conversion is only the character candidate is added, and the character string composed only of the character and figure candidates disappears and is converted into the character string. The character / figure candidates that are not included are automatically determined as the figure candidates.

The character string conversion process of FIG. 10 is performed by the registration unit 2, the character string candidate extraction unit 3, and the character string extraction unit 4 of FIG. 1 as follows. First, as shown in FIG. 11, character candidates are subjected to character cutting processing in the horizontal direction (step S ₁ ), the vertical direction (step S ₂ ) and the diagonal direction (step S ₃ ). The processing in each direction has the same content. As shown in FIG. 12, after the character string candidate extraction processing (step S ₁ ) is performed, the character string extraction processing (step S ₂ ) is performed, and then the character cutout processing is performed. perform (step S _3).

In the character string candidate extraction processing, as shown in FIG. 13, the character candidate tree structuring processing (step S ₁ ) is performed, and then the character string candidate range search processing (step S ₂ ) is performed. Actually, the tree is registered in the registration unit (registration unit 2 in FIG. 1) as a tree structure in which the spatial division is repeated based on the center coordinates of the circumscribed quadrangle of the character candidate obtained as shown in FIG.
Then, among the character candidates, one having the same height as the height of the character string to be searched is used as the core of the character string candidates.
A character candidate having center coordinates in the search range as shown in (a) is searched from the character candidate tree of the registration unit 2.

This search is carried out by the character string candidate extraction unit 3 of FIG. 1 in the following procedure. (1) A central point of another character candidate is searched for within the search range as shown in FIG. 14B on the right side of the character string nucleus. (2) Among the center points found in (1) above, the one farthest from the central nucleus is set as the start point of the next search range. If the center point is not found in (1), the search ends to the right. (3) The leftward direction is performed in the same manner as (1) and (2) above.

What is found by the above search is set as a character string candidate. By setting the core size of the character string candidates in several stages from the largest size and performing the search, it is possible to deal with character strings of different sizes.

Next, the character string extracting unit 4 extracts, from the character string candidates, those satisfying the following conditions as a character string. That is, as shown in FIG. 15, when the height hc of the character string candidate kernel and the distance d between the character candidates have a relationship of d (i, j) <hc × constant ... I and j are character strings.

Next, character separation processing (step S _{4 in} FIG. 2)
Will be explained. The character cutout unit 5 cuts out characters from the extracted character string in units of character candidates. At this time, if the conditions shown in the following formula (2) are met, the height direction separated characters are integrated as in step S ₁ of FIG. 16, and the conditions shown in formulas (3) and (4) are satisfied. If it matches, the width direction separated character integration is performed as in step S ₂ of FIG. For example, when the condition of the following expression (2) is satisfied when the character candidates i and j are separated in the height direction as shown in FIG. 17, the character candidates i and j are integrated as shown.

Max (x _min (i), x _min (j)) ≤ (x _cen (i) o
r x _cen (j)) ≤ min (x _max (i), x _max (j))
(2) (x _min : minimum coordinate in the x direction of the circumscribing quadrangle, x _max : maximum coordinate in the x direction of the circumscribing quadrangle, x _cen : central coordinate in the x direction of the circumscribing quadrangle) Further, for example, as shown in FIG. , K are separated, if the following conditions (3) and (4) are satisfied, the character candidates j and k are integrated as shown. d (j, k) <d (average in character string) x constant ... (3) w (j) + w (k) + d (j, k) ≤ h x constant ... (4) (d: character candidate (Interval distance, w: character candidate width, h: character string height) In the process of the character cutting process (FIGS. 11 and 12) described above, a processed mark is added every time a character candidate is determined as a character string, and the number of processing targets is set. Is less. Although the horizontal direction has been described in each process of FIG. 12, the x and y directions are opposite in the vertical direction. A predetermined inclination angle is assumed for the diagonal direction, and the coordinate conversion processing at that angle is performed by the character string extraction processing and the character cutout processing (step S _{2 in} FIG. 12,
S ₃ ).

The character recognition process (step S _{5 in} FIG. 2) is performed as follows. The character recognition unit 6 refers to and compares the characters cut out by the character cutting unit 5 with a dictionary in which the characters are registered in advance, and recognizes each character.

[0028]

As described above, according to the present invention, a contour vectorization processing section, a polygon figure detection processing section, a registration section,
A character string candidate extraction unit, a character string extraction unit, and a character cutout unit are provided, and character recognition is performed based on the character data cut out by the character cutout unit, and the polygonal figure detection processing unit obtains the character recognition. Since the character / figure candidates that are not converted into character strings among the character / figure candidates are identified as a graphic, the following excellent effects can be obtained.

(1) Even when a figure having a size similar to that of a character exists in the same drawing as in a city map, the character and the figure can be surely distinguished and recognized. (2) When reading and recognizing characters from a document or drawing in which characters, line segments, and symbols are mixed, it is easy and accurate to use characters of any size and format without limiting the size and format. Can be read. (3) Since the data of character candidates is registered in a tree structure, the processing speed can be increased.

[Brief description of drawings]

FIG. 1 is a block diagram of an overall configuration showing an embodiment of the present invention.

FIG. 2 is an overall flowchart of an embodiment of the present invention.

FIG. 3 is a flowchart of a contour vectorization processing unit of the present invention.

FIG. 4 is a flowchart of polygonal figure detection processing according to the present invention.

FIG. 5 is an explanatory diagram showing a contour vector series table according to an embodiment of the present invention.

FIG. 6 is a flowchart of the short vector removing means according to the embodiment of the present invention.

FIG. 7 is an explanatory diagram of a circumscribed quadrangle according to an embodiment of the present invention.

FIG. 8 is an explanatory diagram of a circumscribed quadrangle information table according to the embodiment of this invention.

FIG. 9 is an explanatory diagram of contour vectorization processing according to an embodiment of the present invention.

FIG. 10 is a flowchart of a character string conversion process of the present invention.

FIG. 11 is a flowchart of character cutting processing according to an embodiment of the present invention.

FIG. 12 is a flowchart of character cutting processing according to an embodiment of the present invention.

FIG. 13 is a flowchart of a character string candidate extraction process according to an embodiment of the present invention.

FIG. 14 is an explanatory diagram of character string candidate extraction processing according to an embodiment of the present invention.

FIG. 15 is an explanatory diagram of character string extraction processing according to an embodiment of the present invention.

FIG. 16 is a flowchart of character cutting processing according to an embodiment of the present invention.

FIG. 17 is an explanatory diagram of height direction separated character integration processing according to an embodiment of the present invention.

FIG. 18 is an explanatory diagram of width-direction separated character integration processing according to an embodiment of the present invention.

FIG. 19A is an explanatory diagram showing an example of a drawing input to a drawing reading device, and FIG. 19B is an explanatory diagram showing an example of a correct recognition result.

20A is an explanatory diagram showing an example in which a figure is mistakenly recognized as a character by a conventional drawing reading apparatus, and FIG. 20B is an explanatory diagram showing an example in which a figure is mistakenly recognized as a figure by a conventional drawing reading apparatus. Fig.

[Explanation of symbols]

1 ... Contour vectorization processing unit, 2 ... Registration unit, 3 ... Character string candidate extraction unit, 4 ... Character string extraction unit, 5 ... Character cutout unit, 6
… Character recognition part.

Claims (1)

[Claims] 1. Character data is extracted from binarized pixel data corresponding to black and white obtained by raster-scanning a subject in which characters, line segments, and symbols are mixed, and the extracted character data and a character dictionary are referred to. In a drawing reading apparatus for recognizing characters by comparison, a vector connecting two black pixels adjacent to each other is combined based on binary pixel data corresponding to black and white obtained by raster scanning the object. The contour vector data obtained by the contour vector, and a contour vector conversion processing unit that determines the circumscribed quadrangle data circumscribing the contour vector as a character candidate; and a polygon figure from the character candidate data calculated by the contour vector conversion processing unit. A polygon figure detection processing unit which detects and detects the detected polygon figure data as a character figure candidate, and a sentence which is obtained by the contour vectorization processing unit. The search range is determined from the registration unit that registers the circumscribed quadrangle data of the character candidates in a tree structure, and the core of the character string candidates set to a predetermined size, and the center coordinates of the character candidates registered in the registration unit are A character string candidate extraction unit that extracts a character string existing within the search range as a character string candidate, and a height h _C of the character string candidate core and a character string candidate from the data extracted by the character string candidate extraction unit. A character string extraction unit that extracts data having a relationship that the distance d between adjacent character candidates in d is d <h _C × α (α is a constant), and the character extracted by the character string extraction unit. Having a function of integrating a plurality of adjacent character candidates in a column, a character cutout unit for cutting out a character from a character string, and performing character recognition based on the character data cut out by the character cutout unit, Polygon figure detection processing Drawing reading apparatus characterized by qualifying the graphic character graphic candidates that are not stringified of the graphic character candidate obtained in.