JPH01266689A - Image recognition device - Google Patents

Abstract

PURPOSE:To obtain a device to recognize a document and a graphic on an original precisely by extracting the graphic from image information by a graphic recognizing means, and extracting and recognizing a character and a mark related to the graphic by a character string recognizing means. CONSTITUTION:The original 1 is read by an image reader 2, and the image information of the original is stored in a buffer memory 3. A circular graph recognition part 4 recognizes a circular graph in the image information developed on the memory 3. A character recognition part 5 recognizes a character pattern in the image information developed in the memory 3 based on information in a character font ROM 6 and the information obtained at the recognition part 4. The character pattern recognized at the recognition part 5 reads out a corresponding character pattern from the ROM 6, and performs a processing to write it on a VRAM 7 that is a RAM for display. Also, a various kinds of patterns stored in the VRAM 7 are displayed on a CRT 9 by a graphic control GC 8. A CPU 10 controls the whole of the device.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an image recognition device, and particularly to an image recognition device capable of recognizing characters from image information input from a document.

[Prior Art] Conventionally, there have been character recognition devices that recognize characters from patterns in image information input from a document reading device.

This character recognition device not only converts handwritten manuscripts into typeface, but also converts recognized patterns into codes (JIS codes in most cases).
For example, the data can be compressed into a small amount of data when stored in an external storage device.

[Problems to be Solved by the Invention] However, the document or image information to be recognized by this type of device is premised on the existence of only character symbol patterns. Therefore, in the case of a document containing only graphic patterns or a document containing a mixture of character/symbol patterns and figure/shape patterns, there is a drawback that the graphic pattern cannot be recognized.

Furthermore, even if a device can recognize graphic patterns, it cannot distinguish whether a character string near a graphic is part of the graphic or a text, making it unable to accurately recognize text and shapes. There are drawbacks.

The present invention eliminates the drawbacks of the conventional example and provides an image recognition device that accurately recognizes documents and figures on a manuscript.

Furthermore, when the figure is a pie chart, an image recognition device is provided that recognizes character strings near the pie chart by extracting and recognizing leader lines.

Furthermore, when the figure is a pie chart, by extracting the fan-shaped area surrounded by the partition line and arc and checking whether the entire character area or a part of the character area is included in the extended area, An image recognition device that recognizes character strings is provided.

[Means for Solving the Problem] In order to solve this problem, the image recognition device of the present invention has the following features:
an input means for inputting image information in which text and figures are mixed; a figure recognition means for extracting a figure from the image information inputted from the input means and recognizing characteristics of the figure; The apparatus further includes character string recognition means for recognizing the characters and symbols by extracting the characters and symbols associated with the graphics based on the recognition results.

The figure recognition means includes a circle extraction means for extracting a circle, and a partition line recognition means for extracting a straight line connecting the center and circumference of the circle and using it as a partition line.

The character string recognition means includes a cut-out line recognition means that extracts a straight line having at least one end outside the circle from among the straight lines other than the partition line and uses it as a lead line.

Further, the character string recognition means includes a character string extraction means for extracting characters and symbols in a fan-shaped extension area surrounded by the partition line and the arc.

[Operation] In this configuration, the figure recognition means extracts a figure from the image information input from the input means, and the character string recognition means extracts the characters and characters associated with the figure based on the recognition result of the figure recognition means. The characters and symbols are recognized by extracting the symbols.

The figure recognition means extracts a circle using a circle extraction means, recognizes a partition line using a partition line recognition means, and extracts and recognizes a pie chart from the inputted image information.

The character string recognition means uses the extraction line recognition means to extract a straight line having at least one end outside the inside from among the straight lines other than the partition line as a leader line, and extracts a character having a predetermined relationship with the leader line. and symbols are character strings indicating the ratio of the pie chart.

Further, the character string recognition means extracts characters and symbols in a sector-shaped extension area surrounded by the partition line and the arc, and converts the characters and symbols into a character string indicating the ratio of the pie chart. shall be.

[Embodiments] Hereinafter, embodiments of the image recognition apparatus of the present invention will be described in detail with reference to the accompanying drawings.

FIG. 1 is a block diagram of the image recognition apparatus according to the present embodiment. In FIG. 1, 1 is a document, and 2 is an image reading device that optically reads the document.

3 is a buffer memory for storing read image information; 4
5 is a pie chart recognition unit that recognizes a pie chart in the image information developed in the buffer memory 3, and 5 is a character font ROM.
This is a character recognition unit that recognizes the character symbol pattern in the image information developed in the buffer memory 3 based on the information in e and the information obtained by the pie chart recognition unit 4.

The character pattern recognized by the character recognition unit 5 is processed by reading the corresponding character pattern from the character font ROM 6 and writing it into the VRAM 7 which is a display RAM. Further, the various patterns stored in this VRAM 7 are displayed on the CRT 9 by the graphic controller GC8.

Furthermore, 10 is a CPtJ that controls the entire device, and an RO that stores a control program for controlling the entire device.
The RAM 10b is used as a work area for the CPU 10.

The basic processing procedure of this device with the above configuration is explained in the second section.
The explanation will be given according to the flowchart shown in the figure. First, in step S1, an image reading device 2 such as an image scanner is used.
Read original 1 from 0 and store it in buffer memory 2.
Next, in step S2, thinning processing is performed. Step S
In step 3, character patterns and graphic patterns are discriminated. The criteria for this determination is, for example, if a line segment is short, it is determined to be part of a character pattern, and if it is long, it is determined to be part of a graphic pattern.

If it is a character pattern, proceed to step S9, recognize the character in character extraction step SIO, and proceed to step Sll.
Display in .

On the other hand, if it is determined that it is part of a graphic pattern, the process proceeds to step S4, where the point sequence is divided into point sequences, linear approximation is performed for each point sequence, and a linearized vector is obtained. FIG. 3 shows the process of converting image information into vector information. Next, in step S5, the center point of the circle is calculated based on the vector obtained in step S4, and the radius is determined based on this center point.

The algorithm for calculating the center point of the circle will be explained in detail with reference to FIG. Vector is now VO.

■,... If there are vn pieces, the perpendicular bisector of the vector is 3. λ2. Flood...β, I can write a book.

Intersection of ℃1 and β, intersection of β and β3..., ρ, and 4
1 intersection, β2 and β8, β and β4..., β2 and βn
, f13 and 124, As and f2s..., ρ, and 12.
When the intersections of *a*11n-1 and 47 are calculated, a total of (n-1) 1 intersections are calculated. Next, the histogram of the distribution of the coordinate values (x, y) of the intersections. When , the intersection points are concentrated in a part (slanted part) as shown in FIG. Therefore, the intermediate value (Xwamx * Vw
Determine a□) as the center coordinates of the circle of the pie chart.

Next, the center coordinates (xs□+
From ym□), further vector ■l.

If we draw a median line on each of Vs, Vs...vn, a total of n median lines will be created. The angle between this median line and the vector is α
1. α2...α. Then, if an>07 (θ7 is a threshold), Vn is recognized as a circular vector forming a part of a circle, and if α7≦θd, v7 is recognized as a straight line vector forming a straight line.

In Figure 6, α. Since 〉θa, vI is a circular vector, and since α7〉0, V is a straight line element. In this way, all vectors are classified into circular elements, element elements, and straight line elements, and the length of the straight line connecting the circular element and the center point is made into a histogram, and the peak value is determined as the radius.

FIG. 7(a) is a diagram of this radius extraction algorithm, where the center point (Xmax, V□8) and the circular element γ1 are γ1. γ2. γ3...γ. and data can be created. When this data is plotted as a histogram, Figure 7 (
A histogram as shown in b) is obtained, and this peak value becomes the value of the radius of the circle.

In this way, the center coordinates and radius are determined, and a circle can be recognized.

Furthermore, when the radius of the circle is determined, among the circular elements, elements whose distance from the center point (xl, 8.y, . . . ) is different from the length of the radius are newly regarded as linear elements. In this way, from the extracted straight line elements,
Next, partition lines and leader lines are extracted.

In step S6, partition lines are first extracted. After recognizing a straight line element whose point coordinates at both ends are inside the circle as a partition line, in step S7, from among the remaining straight line elements, a straight line element whose at least one end is outside the circle is recognized as a leader line. are doing. FIG. 8 is a diagram in which all elements are classified into circle elements, partition lines, and leader lines. A circle element like this.

After extracting the partition line and leader line, in step S8, a character area such as a character representing a ratio is extracted based on the position of the outer end of the circle of the leader line.

As shown in FIG. 9, a character vector U+ that is closer to the tip P1 of the leader line is extracted. Next, a character vector U2 whose end is within a radius of 70 from the tip of this Ul is searched,
When similar processing is performed, =Ut, Us, Us・
. . U is determined, characters etc. guided by the lead line are extracted, and this can be recognized as a character string representing a ratio (percentage). After that, in step S9, the characters are cut out, and in step SIO, the characters are recognized.
The recognition result is displayed in step Sll.

[Second Example] In this example, a recognition procedure for a pie chart without a leader line as shown in FIG. 10 will be described. In this case, in the above-described embodiments, it is not possible to extract characters representing ratios. Therefore, the following processing is further performed in step S8.

When the center coordinates, radius, and partition line are recognized in the processing of steps 31 to S6, a fan-shaped area surrounded by the partition line and the circular arc is extracted, so in step S8, this fan-shaped area is first extracted. . As shown in FIG. 11(a),
The angle θ3 of the straight line, gl+122, ρ, with the X axis. θ2.
Examining the magnitude of θ3, we find that β, <βg < 12 s.

Also, if the radius of the pie chart is γ, then the equation of the circle is x”
+y" = γ2. When examining which area of the area A, B, or C in FIG. 11(b) the all-black pixel is included, we find that the coordinates (x, y) satisfy x2+y2<γ2. And the angle θ□ of the straight line connecting (0, O) and (x, y)
If it satisfies θ1 * θ8y * θ3, it is considered to be included in region A, and if θ2 * θ * θ3, it is included in region B, and if 0 * θ, ly < θ8, and θ3 * θ8y. is included in area C.

Next, by finding the X coordinate, the maximum value, and the minimum value of the X coordinate of the black pixel coordinate values included in each area of A, B, and C, the character area is extracted as shown in Figure 11 (C).・. in this way,
When the character area in the pie chart is extracted, since there is no character area in the fan-shaped area where there is no character string, a fan-shaped area where the character string exists and a blank fan-shaped area are obtained as shown in FIG.

Next, it is checked whether or not characters exist within the extension area of this blank area, and characters, etc. located outside the circle are extracted.
The figure shows a method for extracting the characters, etc. The shaded area is an area obtained by extending the blank fan-shaped area A of the pie chart to the outside of the circle by the length of the radius station. Only when at least one character vector exists in this diagonal area, a set of nearby character vectors (character area) is extracted.
The extracted character string is recognized as a character string indicating a ratio.

Thereafter, as in the above embodiment, characters are cut out in step S9, characters are recognized in step S10, and characters are recognized in step S1.
Display the recognition result with l.

As explained above, by recognizing figures such as pie charts, it has become possible to input and edit manuscripts consisting of text and figures using a scanner, increasing the amount of information input from manuscripts.

Moreover, it has become possible to recognize characters and symbols inside figures, such as characters representing percentages outside the circle of a pie chart, making it possible to recognize complex patterns.

In addition, in terms of recognition speed, it is possible to recognize and display pie charts, characters representing ratios, etc. using a simple processing method, which reduces processing time and enables high-speed character/figure recognition. .

Additionally, it has become possible to recognize leader lines of pie charts and characters representing ratios outside the circle, making it possible to recognize pie charts with complex patterns.

Furthermore, it has become possible to recognize characters that represent ratios that are part of a pie chart, making it possible to recognize complex patterns in pie charts.

In this example, the relationship is limited to the relationship between a pie chart and a character string indicating a ratio, but by extracting a character string according to the characteristics of the recognized shape, it is generally possible to associate a shape with characters and symbols related to the drawing. By the image recognition of the present invention, characters and character strings related to graphics can be clearly distinguished and recognized, so that it becomes possible to compress or edit images when they are stored.

[Effects of the Invention] According to the present invention, it is possible to provide an image recognition device that accurately recognizes documents and figures on a manuscript. Furthermore, by distinguishing between characters in a document and characters in graphics, it has become possible to edit manuscripts consisting of characters and graphics.

Furthermore, when the figure is a pie chart, by extracting and recognizing leader lines, it is possible to provide an image recognition device that recognizes character strings near the pie chart.

Furthermore, when the figure is a pie chart, by extracting the fan-shaped area surrounded by the partition line and arc and checking whether the entire character area or a part of the character area is included in the extended area, An image recognition device that recognizes character strings can be provided.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of the image recognition device of this embodiment, and FIG. 2 is a flowchart of the processing procedure of the image recognition device of this embodiment.
Charts, Figures 3 and 10 are diagrams showing the conversion of image information, Figure 4 is a diagram showing how to find the center of a circle, Figure 5 is a diagram showing the distribution of center points, Figure 6 is a diagram showing a circle Figure 7(a) is a diagram showing the radius of a circle. Figure 7(b) is a diagram showing a histogram of the radius of a circle. Figure 8 is a diagram showing the classification of elements. Figure 9 is a diagram showing the extraction of characters representing ratios, Figures 11 (a) to (c) are diagrams showing the method for extracting a fan-shaped area, Figure 12 is a diagram showing a fan-shaped blank area, FIG. 13 is a diagram showing a method for extracting character strings outside the circle. In the figure, 1... original, 2... image reading device, 3...
... Buffer memory, 4... Pie chart recognition unit, 5...
・Character recognition unit, 6...Character font ROM, 7...
VRAM, 8-G-C, 9-CRT, 10-...CP
U, 10a...ROM, 10b...RAM.

Claims (1)

[Claims] (1) An input means for inputting image information containing a mixture of text and figures, and a figure that extracts a figure from the image information inputted from the input means and recognizes the characteristics of the figure. Image recognition comprising: a recognition means; and a character string recognition means for recognizing the characters and symbols by extracting characters and symbols associated with the figure based on the recognition results of the figure recognition means. Device. (2) The figure recognition means includes a circle extraction means for extracting a circle, and a partition line recognition means for extracting a straight line connecting the center and the circumference of the circle and using it as a partition line, and the figure recognition means is provided with the input image. 2. The image recognition apparatus according to claim 1, wherein the image recognition apparatus extracts and recognizes a pie chart from the information. (3) The character string recognition means includes a leader line recognition means for extracting a straight line having at least one end outside the circle from among straight lines other than the partition line and using the extracted line as a leader line; 3. The image recognition apparatus according to claim 2, wherein the related characters and symbols are character strings indicating the ratio of the pie chart. (4) The character string recognition means includes a character string extraction means for extracting characters and symbols within a fan-shaped extension area surrounded by the partition line and the circular arc, and converts the extracted characters and symbols into the ratio of the pie chart. 3. The image recognition device according to claim 2, wherein the image recognition device is a character string indicating .