JP2000235618A - Character detecting device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a character detecting device capable of correctly detecting a character even in an image in which many edge points are generated. SOLUTION: This device is provided with an image inputting and storing means 1 which inputs an image and stores it as an input image, an edge image generating means 2 which generates an edge image from the input image, an edge line pair detecting means 3 which extracts two edge lines existing at a prescribed distance in the edge image as an edge line pair and generates an edge line pair distribution image and a character position detecting means 4 which connects adjacent edge line pairs from the edge line pair distribution image on the basis of the direction of each edge line pair, finds a connection area as a character area and outputs a character position. Thus, an edge occurring on a road surface, etc., is prevented from being detected mistakenly.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character detecting device for detecting a character in an image.

[0002]

2. Description of the Related Art A conventional character detection device will be described with reference to the drawings. FIG. 11 shows, for example, Japanese Patent Laid-Open No. 7-105.
FIG. 1 is a block diagram illustrating a configuration of a conventional display plate detection device disclosed in Japanese Patent Publication No. 340.

In FIG. 11, reference numeral 1 denotes an image input / storage means for digitally converting an image from a camera and storing it in a frame memory; 2, an edge image generation means for detecting a luminance change point in the image to obtain an edge image; A pair edge detecting means for detecting a pair of edges existing within a predetermined range from the edge image;
Plate position detection means for detecting a plate position from the edge pair obtained in step 001.

Next, the operation of the above-described conventional character detection device will be described with reference to the drawings. FIG. 12 is a diagram showing a vertical edge detection mask for creating an edge image in a conventional character detection device.

First, the image input / accumulation means 1 stores a captured image in a frame memory. Next, the edge image generating means 2 extracts a vertical edge from the accumulated image to create an edge image. In this edge image generation method, a target pixel is set in an input image, and a vertical edge detection mask 100 shown in FIG.
The corresponding pixel value is multiplied by 3, and the added value is stored in the edge image as a differential value of the pixel of interest, the pixel of interest is updated gradually, and the above operation is performed for all pixels.

Next, the pair edge detecting means 1001 scans the edge image in the horizontal direction to detect an up edge changing from a dark portion to a bright portion and a down edge changing from a bright portion to a dark portion. When the distance between the up edge and the down edge is within a predetermined range, the up edge and the down edge are registered as a pair edge. Then, in the plate position detecting means 1002, the pair
A partial area where an edge exists for a certain number or more is extracted as a license plate area.

[0007]

In the above-described conventional character detection device, a region where there are many pairs of adjacent edge points is defined as a region where a character exists. Therefore, a region where many edge points occur, for example, a road surface or the like. There is a problem that a portion where no character is present is erroneously detected as a character area.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problem, and has as its object to provide a character detecting device capable of correctly detecting characters even in an image having many edge points.

[0009]

According to the present invention, there is provided a character detecting apparatus for detecting a character in an image, comprising: an image inputting / accumulating means for inputting an image and storing the input image; Edge image generating means for creating an edge image; edge line pair detecting means for extracting two edge lines existing at a predetermined distance from the edge image as an edge line pair to create an edge line pair distribution image; A character position detecting means for connecting adjacent edge line pairs from the line pair distribution image based on the direction of each edge line pair, obtaining a connected region as a character region, and outputting a character position.

Further, the character detecting apparatus according to the present invention further comprises a testing means for testing whether or not each of the character areas obtained by the character position detecting means is a character.

Further, the character detecting device according to the present invention is
The edge line pair detecting means detects an edge line pair again in a partial region obtained by subdividing the edge line pair distribution image when an edge line pair cannot be obtained.

[0012]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiment 1 A character detection device according to Embodiment 1 of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing a configuration of a character detection device according to Embodiment 1 of the present invention. In the drawings, the same reference numerals indicate the same or corresponding parts.

In FIG. 1, reference numeral 1 denotes an image inputting / accumulating means for inputting an image and accumulating it as an input image, 2 an edge image generating means for generating an edge image from the input image, and 3 a character line from the edge image. An edge line pair detecting means for extracting two pairs of edge lines corresponding to the contour portion as an edge line pair and creating an edge line pair distribution image, and a character in the input image based on the edge line pair distribution image. Character position detecting means for detecting a character position and outputting a character position.

FIG. 2 is a diagram showing a configuration of the edge image generating means 2.
FIG. 3 is a diagram illustrating an example of a Laplacian filter composed of × 3 pixels. In the figure, 5 is a Laplacian filter, and 6 is a filter value corresponding to the pixel of interest.

FIG. 3 is a diagram for explaining the operation of the edge line pair detecting means 3. In the figure, reference numerals 101 to 104 denote an example of an edge line pair detection mask for detecting an edge line pair.
Is a horizontal line, 103 is a diagonal line that rises to the left, 104
Is an edge line pair detection mask for detecting two pairs of straight lines formed by oblique lines in the upward direction, 105 to 108 are examples of edge line pair detection mask values, 109 is a character image, 110 is an edge image of a character image, 111 is the distance L between the contours of the character line
It is.

Next, the operation of the character detecting apparatus according to the first embodiment will be described with reference to the drawings.
FIG. 4 is a flowchart showing the operation of the character detection device according to Embodiment 1 of the present invention. FIG. 5 is a flowchart for explaining the operation of the edge line pair detecting means 3. FIG. 6 is a diagram for explaining the operation of the character position detecting means 4.

In step S101 of FIG. 4, the image input / accumulation means 1 accumulates an image from a camera or the like as an input image in a frame memory or the like in multiple values. Next, in step S102, the edge image generation means 2 generates an edge image representing a change in luminance in the input image.

For example, the character image "1234" 10 in FIG.
When 9 is an input image, the edge image 110 is an image of an outline constituting a character. Edge image generation method,
For example, a differential image generation method using a Laplacian filter disclosed in “Introduction to Computer Image Processing” (supervised by Soken Publishing, Hideyuki Tamura, PP. 123-125). The target pixel is set in the input image, the target pixel and its eight neighboring pixels are multiplied by the corresponding pixel value of the Laplacian filter 5, and the added value is stored in the edge image as a differential value of the target pixel. This operation is performed for all the pixels by gradually updating the target pixel.

In step S103, the edge line pair detecting means 3 calculates a correlation value between the edge image and predetermined edge line pair detection mask values 105 to 108. For example, the edge line pair detection mask value is set to a value larger than 0 for a portion corresponding to the contour line, and is set to values other than the values indicated by reference numerals 105 to 108 in FIG.

The operation of the edge line pair detecting means 3 will be described with reference to the flowchart of FIG. Here, the edge line pair distribution image stores a correlation value for each edge line pair detection mask.

In step S201, a variable i representing a processing target edge line pair detection mask is initialized to 1.

Next, in step S202, the variable i
Is selected, and for all pixel values of the edge line pair detection mask and the edge image, a pixel of interest is set for the edge image in the same manner as in the edge image generation method. The pixel value of the corresponding edge image of the paired detection mask is multiplied by the edge line paired detection mask value, and the added value is stored as a correlation value at the pixel of interest in the edge line paired distribution image of the selected edge line paired detection mask.

In step S203, it is determined whether or not the processing has been completed for all pixels in order to update the pixel of interest gradually and perform the above operation for all pixels. If finished, S2
The process proceeds to step S05, and if not completed, the process proceeds to step S204 to update the target pixel and repeat the above operation.

Next, in step S205, it is determined whether or not all of the predetermined edge line pair detection masks 105 to 108 have been executed based on whether the variable i matches the predetermined character line pair detection mask number. If all the predetermined edge line pair detection masks have not been completed (i ≠ the number of predetermined character line pair detection masks), the process proceeds to step S206, in which the variable i is increased by 1, and the process proceeds to step S202. When all the predetermined edge line pair detection masks have been completed (i = the number of predetermined character line pair detection masks), the process proceeds to step S207.

In the edge line pair distribution image corresponding to each of the edge line pair detection masks 105 to 108, step S20 is performed.
At step 7, the target pixel is set.

Next, in step S208, the horizontal
The oblique line detection masks 106 to 108 refer to the pixel value of the distance L111 in the sub-scanning direction, and the vertical direction detection mask 105 refers to the pixel value of the distance L111 in the main scanning direction. Determine if it exists.

For example, it is determined whether or not the following equation 1 is satisfied.

| V (x, y) -V (x + a, y + b) | <VTh and V (x, y)> Vth1 and V (x + a, y + b)> Vth2 (Equation 1)

Here, V (x, y) is the pixel value of interest in the edge line pair distribution image, (x, y) is the coordinates of the pixel of interest,
V (x + a, y + b) is a horizontal / diagonal line detection mask 106
To 108, the pixel value of a = 0, b = L, the vertical direction detection mask 105, the pixel value of a = L, b = 0, VT
h, Vth1 and Vth2 are predetermined thresholds.

If the above expression 1 is satisfied, it is considered that there is an edge line pair in the target pixel, and the flow advances to step S210 to set the target pixel value in the edge line pair distribution image to 1. If the condition is not satisfied, the process proceeds to step S209, sets the target pixel value to 0, and proceeds to step S211.

In order to update the pixel of interest gradually and perform the above-described operation on all the pixels, it is determined in step S211 whether all the pixels have been set as the pixel of interest. When the above processing is completed for all the pixels, the processing of the edge line pair detecting means 3 is completed. If not, the target pixel is updated in step S212, and the process proceeds to step S208 to repeat the above operation.

In step S104, the character position detecting means 4 detects a character position based on the edge line pair distribution image. The operation of the character position detecting means 4 will be described with reference to FIG.

In the edge line pair distribution image, a region where the edge line pairs are smoothly connected to each other in the vicinity of eight adjacent pixels of interest is detected as a character.

In step S301, a target pixel is set in the edge line pair distribution image. At this time, if the target pixel has no label, a new label number is given to the target pixel.

Next, in step S302, the variable i
Is initialized with 1.

Next, in step S303, the adjacent 8
If any of the pixel values in the edge line pair distribution image detected by each of the edge line pair detection masks 105 to 108 corresponding to the i-th pixel in the vicinity is 1, it is checked whether the following expression 2 is satisfied. If it satisfies, in step S305, it is determined that there is an edge line pair that is smoothly connected, and the same label as the pixel of interest is given to the adjacent pixel, that is, the edge line pair distribution image whose value is 1.

| Θ attention edge line pair direction−θ adjacent edge line pair direction | <Th (Equation 2)

Here, the θ target edge line pair direction is the edge direction of the target point, the θ adjacent edge line pair direction is the edge direction of eight neighboring pixels adjacent to the target point, and Th is a predetermined angle.

In step S304, the variable i is updated.
Next, in step S306, it is determined based on the value of the variable i whether or not the above-described processing has been performed on eight neighborhoods of the target pixel. If the value of the variable i is 8, the pixel of interest is updated in step S307, and the process proceeds to step S308.

In step S307, it is checked whether a label is attached to the new pixel of interest. If no label is attached, a new label number is given to the pixel of interest. If the value of the variable i is other than 8, the process proceeds to step S303.

In step S308, if the above processing is executed for all pixels in the edge line pair distribution image as target pixels, an area consisting of pixels having the same label is output as a character area in step S309.
If not all pixels are set as the target pixel, step S
Go to 302.

As described above, according to the present embodiment, the continuity between an edge line pair corresponding to a contour portion forming a character and an adjacent edge line pair is determined based on the directional displacement of the edge line pair. By detecting as one character area,
For example, an edge region that frequently occurs on a road surface or the like is not erroneously detected as a character region.

In this embodiment, the edge line pair detection mask is made up of horizontal, vertical, left-up, and right-up diagonal lines, but a plurality of detection masks having other angles may be used. Although the straight line is detected in the edge line pair detection mask, a plurality of masks for detecting a curved portion may be used.

Embodiment 2 A character detection device according to a second embodiment of the present invention will be described with reference to the drawings. FIG. 7 is a block diagram showing a configuration of a character detection device according to Embodiment 2 of the present invention.

In FIG. 7, reference numeral 7 denotes a testing means for testing whether or not the result of the character position detecting means 4 is a character based on an edge line pair distribution image. The configuration other than the test means 7 is the same as that of the first embodiment.

Next, the operation of the character detecting device according to the second embodiment will be described with reference to the drawings.
FIG. 8 is a flowchart showing the operation of the character detecting device according to Embodiment 2 of the present invention. FIG. 9 is a flowchart for explaining the operation of the test means 7.

Steps S401 to S404 are the same operations as in the first embodiment.

Next, in step S405, the character position output from the character position detecting means 4 is verified by the verifying means 7.

The operation of the test means 7 will be described with reference to the flowchart of FIG. First, in step S501,
The variable i is initialized with 1.

Next, in step S502, it is checked whether the area, width, and height of the i-th region of the character position are within a predetermined range by using the following expressions 3 to 5.

| SS Standard | <Sth (Equation 3)

| W−W standard | <Wth (Equation 4)

| H−H standard | <Hth (Equation 5)

Where S is the area of the i-th character position area,
S standard is the standard character area area of the character to be detected, W is the i-th width of the character position, W standard is the standard width of the character to be detected, H is the i-th height of the character position, and H standard is the detection target. The standard heights of the characters, Sth, Wth, and Hth, are predetermined thresholds, which are obtained in advance by the following equations 6 to 8, for example.

Sth = α0 × L + β0 (Equation 6)

Wth = α1 × L + β1 (Formula 7)

Hth = α2 × L + β2 (Equation 8)

Here, α0, α1, α2, β0, β1,
β2 is a predetermined constant, L is an edge line pair detection mask 10
The edge line distance 111 is 5 to 108.

If all of the above expressions 3 to 5 are satisfied, the process proceeds to step S503; otherwise, the process proceeds to step S5.
In step 07, the character position is deleted, and the flow advances to step S506.

Next, in step S503, the area of the edge line pair distribution image corresponding to the character position area is normalized to a predetermined size. The normalization method is described in, for example, "Introduction to Computer Image Processing" (Supervised by Soken Publishing, Hideyuki Tamura,
P. 113-115).

Next, in step S504, the normalized edge line pair distribution image is compared with the edge line pair distribution image (dictionary) previously obtained for the character to be detected by the following equation 9 to determine the degree of difference. . However, the range of Σ is k
= 1 to the number of edge line pair detection masks.

Diff = Σ | Ek (x, y) −Dk (x, y) | (Equation 9)

Here, Diff is the ith difference degree at the character position, Ek (x, y) is the edge line pair distribution image pixel value detected by the kth edge line pair detection mask at the point of coordinates x and y, Dk (x, y) is a pixel value in the dictionary of the coordinates x and y detected in advance by the k-th edge line pair detection mask.

Next, in step S505, it is determined whether the difference is smaller than a predetermined threshold. If it is smaller than the predetermined threshold, the process proceeds to step S506. If it is equal to or greater than the predetermined threshold, the process proceeds to step S507, and the character position is deleted.

Next, in step S506, it is determined based on a variable i whether or not execution has been performed for all character positions. If all of them have not been completed, the process proceeds to step S508, and after updating the variable i, the process proceeds to step S502. When all the processes have been completed, the process of the test means 7 is completed.

As described above, according to the present embodiment, erroneous detection of a non-character area can be reduced by performing the test based on the size of the character area and the test based on the edge line pair distribution comparison in the test means 7. .

In this embodiment, the collation is performed on the edge line pair distribution image. However, the collation may be performed between the input image corresponding to the character position and the detection target character. The matching method between images may be matched by a normalized correlation method disclosed in, for example, "Introduction to Computer Image Processing" (published by Soken Publishing, Hideyuki Tamura, PP. 149-150).

Embodiment 3 In the first embodiment,
Although the edge line pairs are detected by the predetermined edge line pair detection mask, in the third embodiment, when the predetermined number of edge line pairs cannot be detected, the edge line detection mask is changed and detected again. It is.

A character detecting device according to Embodiment 3 of the present invention will be described with reference to the drawings. FIG. 10 is a flowchart showing the operation of the character detecting device according to Embodiment 3 of the present invention. The configuration of the third embodiment is the same as FIG. 1 showing the configuration of the first embodiment.

The operations in steps S601 and S602 are the same as those in the first embodiment.

Next, in step S603, the edge image is divided into partial areas. The division method is, for example, n division into rectangular regions (predetermined width W, predetermined height H).

Next, in step S604, a variable i representing a division area number is initialized to 1.

Next, in step S605, an edge line pair distribution image is created in the partial area i. The method of generating an edge line pair distribution image is the same as in the first embodiment.

Next, in step S606, it is determined whether the number of edge line pairs in the i-th partial area is equal to or smaller than a predetermined number. If the number of edge line pairs exceeds a predetermined number, step S60
Proceed to 7. If the number of edge line pairs is equal to or smaller than the predetermined number, the process proceeds to step S610, and the edge line pair detection mask is set to step S6.
The distance L111 between the edge lines used in step S03 is changed, and the process proceeds to step S605.

Next, in step S607, it is determined whether the process has been performed for all partial areas. If the variable i is equal to the total number n of partial areas, it is determined that the processing has been completed for all partial areas, and the process proceeds to step S608. If the variable i does not match the total partial area number n, it is determined that the processing has not been completed for all partial areas, and the process proceeds to step S605. Steps S608 and S609 perform the same operation as in the first embodiment.

As described above, according to the present embodiment, even when a character whose distance between edge lines does not match is included in the image, if the number of edge line pairs is less than the predetermined number in the partial area, the edge is re-established. Since the detection is performed by changing the distance between lines, the character area can be detected without fail.

In this embodiment, the distance L between the edge lines matching the character size is changed. However, the edge image is enlarged / reduced to obtain a correlation value with a predetermined edge line detection mask. You may.

[0078]

As described above, the character detecting device according to the present invention is an image inputting / accumulating means for inputting an image and storing it as an input image in the character detecting device for detecting characters in an image. Edge image generating means for creating an edge image from the image, two edge lines existing at a predetermined distance in the edge image are extracted as an edge line pair, and an edge line pair detecting means for creating an edge line pair distribution image; It is provided with character position detecting means for connecting adjacent edge line pairs from the edge line pair distribution image based on the direction of each edge line pair, obtaining a connected region as a character region, and outputting a character position. There is an effect that the generated edge is not erroneously detected.

Further, as described above, the character detecting apparatus according to the present invention further includes the verification means for verifying whether or not each character area obtained by the character position detection means is a character. And the test based on the edge line pair distribution matching, the erroneous detection of a non-character area can be reduced.

Further, the character detecting device according to the present invention
As described above, the edge line pair detecting means performs the edge line pair detection again on the partial region obtained by subdividing the edge line pair distribution image when the edge line pair cannot be obtained in the partial region. Even when characters that do not match the detection mask are included in the image, if the number of edge line pairs is less than the predetermined number in the partial area, the edge line pair detection mask is changed and detected again, and the character area is completely detected. It has the effect of being able to.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a character detection device according to Embodiment 1 of the present invention.

FIG. 2 is a diagram illustrating an example of a Laplacian filter composed of 3 × 3 pixels used in an edge image generation unit of the character detection device according to the first embodiment of the present invention.

FIG. 3 is a diagram illustrating an operation of an edge line pair detection unit of the character detection device according to the first embodiment of the present invention.

FIG. 4 is a flowchart showing an operation of the character detection device according to Embodiment 1 of the present invention.

FIG. 5 is a flowchart showing an operation of an edge line pair detecting means of the character detecting device according to the first embodiment of the present invention.

FIG. 6 is a flowchart showing an operation of a character position detection unit of the character detection device according to Embodiment 1 of the present invention.

FIG. 7 is a block diagram showing a configuration of a character detection device according to Embodiment 2 of the present invention.

FIG. 8 is a flowchart showing an operation of the character detection device according to Embodiment 2 of the present invention.

FIG. 9 is a flowchart illustrating an operation of a verification unit of the character detection device according to the second embodiment of the present invention.

FIG. 10 is a flowchart showing an operation of an edge line pair detection unit of the character detection device according to Embodiment 3 of the present invention.

FIG. 11 is a block diagram showing a configuration of a conventional character detection device.

FIG. 12 is a diagram illustrating an example of a mask for detecting a vertical edge in a conventional character detection device.

[Explanation of symbols]

1 image input / storage means, 2 edge image generation means, 3
Edge line pair detecting means, 4 character position detecting means, 7 testing means.

Claims (3)

[Claims] 1. A character detecting device for detecting characters in an image, an image inputting / accumulating unit for inputting an image and storing the image as an input image; an edge image generating unit for generating an edge image from the input image; Edge line pair detecting means for extracting two edge lines existing at a predetermined distance from the image as an edge line pair and creating an edge line pair distribution image; A character detection device comprising: character position detection means for connecting based on the direction of a line pair, determining a connected region as a character region, and outputting a character position. 2. The character detecting apparatus according to claim 1, further comprising a verification unit for verifying whether or not each of the character areas obtained by the character position detection unit is a character. 3. The edge line pair detecting means, when an edge line pair cannot be obtained in a partial region obtained by subdividing the edge line pair distribution image, performs edge line pair detection again on the partial region. The character detection device according to claim 1.