JP2001236501A - Device and method for recognizing image - Google Patents

Abstract

PROBLEM TO BE SOLVED: To appropriately recognize a recognition object without preparing a large number of templates even when the object is photographed from a direction other than right in front of the object. SOLUTION: A template to be collated is retrieved from a template storage part 28 storing plural templates by photographing the partial image in the form of horizontal line of the recognition object with a line CCD camera 22 and horizontally slicing the pattern such as character or symbol of this image in a line form. Then, the pattern is recognized by comparing the order of ID of the collated templates with the order of ID of templates in respective patterns stored in an ID order storage part 32. Even in the case of image photographing the recognition object from a direction other than right in front of the object, by dividing the object into linear areas, the dimension difference between the divided image and the template is small. As a result, even when the recognition object is deformed or distorted, the object can be recognized by using the template of the pattern having no deformation or distortion.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image recognition apparatus and an image recognition method, and more particularly, to an image recognition apparatus for recognizing a recognition target as one of a plurality of preset patterns.

[0002]

2. Description of the Related Art Conventionally, as this type of image recognition device,
One that performs matching processing between a template image and a search image has been proposed (for example, Japanese Patent Application Laid-Open No. H11-31224).
Issue publication). In this apparatus, a calculation value obtained by calculating a search partial image is used for a dead area in a template image.

Further, there has been proposed a technique in which a target pattern and a template are divided into a plurality of blocks corresponding to each other to determine the target pattern (for example, Japanese Patent Laid-Open No. 7-311).
No. 816). In this apparatus, a correlation value is obtained for each block, a minimum value of the block correlation value is extracted, and a target pattern is determined based on a comparison between the minimum values of the block correlation values for a plurality of reference patterns.

[0004]

However, these devices may not be able to properly recognize a recognition target. When the recognition target is photographed from directly in front, the recognition target has no distortion or deformation, so that it can be appropriately identified by the matching process.However, when the recognition target is photographed from a position other than right in front, the captured recognition target is distorted or deformed. In some cases, it cannot be properly identified.
In addition, even if the recognition target is photographed from the front, if the recognition target is slightly deformed, it cannot be properly recognized.

In order to solve such a problem, it is conceivable to prepare a template for a recognition object in which a distortion or a deformation is added, and perform matching using the template. And the processing takes time.

An object of the image recognition device and the image recognition method of the present invention is to appropriately recognize a target to be recognized even when the target is photographed from a position other than the front. Another object of the image recognition apparatus and the image recognition method of the present invention is to recognize a recognition target by simple processing without preparing a large number of templates.

[0007]

Means for Solving the Problems and Their Functions and Effects The image recognition apparatus and the image recognition method of the present invention employ the following means in order to achieve at least a part of the above objects.

An image recognition apparatus according to the present invention is an image recognition apparatus for recognizing a recognition target as one of a plurality of preset patterns, wherein at least a part of an area obtained by dividing the pattern into a plurality of patterns is provided. Template storing means for storing a plurality of corresponding templates, photographing means for photographing at least a part of the recognition target as a collation target, and collation for collating the photographed collation target with a template stored in the template storage means And a determination unit configured to determine which of the plurality of patterns the recognition target is based on a positional relationship between the regions of the collated template.

In the image recognition apparatus according to the present invention, the collation means includes a collation target, which is at least a part of the recognition object photographed by the photographing means, of an area obtained by dividing the pattern stored by the template storage means into a plurality of areas. The template is compared with at least a part of the template, and the determining unit determines which of a plurality of preset patterns is to be recognized based on the positional relationship between the compared templates. The distortion or deformation of the recognition target that occurs when the recognition target is photographed from a position other than directly in front of the recognition target appears largely in the recognition target as a whole. A pattern having no distortion or deformation can be collated with a template including a part of the similarly divided region. Then, since the positional relationship of the collated template is the same as the positional relationship of the collation target in the recognition target, any one of a plurality of patterns in which the recognition target is preset based on this positional relationship Can be determined.

In the image recognition device of the present invention,
The template storage means is means for storing an area formed by horizontally slicing the pattern into a plurality of sections as a template, and the determination means is a means for determining based on the collated order of the collated templates. It can be. In this case, the collation can be performed based on the ratio of the contrast on the straight line between the collation target and the template. Even if the recognition image is distorted or deformed, the ratio of the contrast on a part of the straight line does not change, so that it is possible to accurately collate distortion and deformation.

In the image recognition apparatus according to the present invention, wherein the template is formed by horizontally slicing a pattern and dividing the pattern, the photographing means forms a horizontal linear part of the recognition target on or above the recognition target. It is a means for sequentially photographing as a collation object from below, and the collation means may be a means for sequentially collating the collation objects sequentially photographed.

Further, in the image recognition apparatus according to the present invention, wherein the template is obtained by horizontally slicing and dividing the pattern, the photographing means is means for photographing the whole of the recognition target, and the collating means is The recognition target photographed by the photographing means is divided into an area corresponding to the plurality of divided areas to be a plurality of collation targets and a means for sequentially collating the divided plurality of collation objects. Can also.

Further, in the image recognition apparatus according to the aspect of the present invention, wherein the template is obtained by horizontally slicing and dividing the pattern, the collating means collates the collation targets in the order of the templates constituting the pattern. It can also be a means. In this case, the matching can be performed more quickly than in the case of matching with all the patterns. Note that “matching in the order of templates” includes not only matching in the order of templates, but also matching in which one or more templates are omitted if the order of templates is matched. .

In the image recognition apparatus according to the aspect of the present invention, in which the collation is performed in the order of the templates, the collation means may determine the collation target as a first template and one or two of the first and subsequent templates. The means for starting the matching from the above-described matching with the template may be used. In this way, the recognition target can be recognized even when the portion corresponding to the first template of the recognition target is not set as the matching target due to aging or the like.

Further, in the image recognition apparatus according to the aspect of the present invention in which the collation is performed in the order of the template, the collation means may determine the collation target with the collated template and one or more subsequent orders of the collated template. It may be a means for matching with one or more templates. By doing so, it is possible to recognize the recognition target even if the way of division as the collation target is slightly different from the way of division of the template in the pattern,
The recognition target can be recognized even when a portion corresponding to the recognition target template is not set as the verification target due to aging or the like.

In the image recognition apparatus according to the aspect of the present invention, wherein the template is obtained by horizontally slicing a pattern and dividing the pattern, the collating means is means for collating using a correlation between the collation target and the template. It can also be. In this way, it is possible to cope with deformation such as blurring of the recognition target.

Further, in the image recognition apparatus according to the present invention, wherein the template is obtained by horizontally slicing the pattern and dividing the pattern, the determining means is means for determining based on the correlation of the collated templates. You can also. This makes it possible to cope with deformation such as blurring.

Further, in the image recognition apparatus according to the aspect of the present invention, wherein the template is obtained by horizontally slicing the pattern and dividing the pattern, the determination unit determines the template including a characteristic portion of the pattern in consideration of a weight. It can also be a means to perform. In this case, the recognition target can be more accurately recognized.

Further, in the image recognition device of the present invention, the image recognition device is mounted on a vehicle, and the recognition target may be a character or a symbol attached to a road surface. In this way, it is possible to recognize characters or symbols attached to the road surface. In the image recognition apparatus of the present invention of this aspect, the image recognition apparatus further includes rainy weather determination means for determining whether it is rainy weather, and wherein the collation means determines whether the verification target or the template is present when the rainy weather determination means determines that it is rainy weather. It is also possible to employ a means for inverting one of the colors to an inverted color and comparing the inverted one with the other.
In this way, characters or symbols attached to the road surface can be recognized even in rainy weather.

An image recognition method according to the present invention is an image recognition method for recognizing a recognition target as one of a plurality of preset patterns, wherein a photographing step of capturing at least a part of the recognition target as a collation target, A collation step of collating the photographed collation target with a template that forms at least a part of an area formed by dividing the pattern into a plurality of patterns; and a step of collating the plurality of recognition targets based on a positional relationship between the collated templates. And a determining step of determining which of the patterns is the main pattern.

According to the image recognition method of the present invention, a template which is composed of a part of an area obtained by similarly dividing a pattern having no distortion or deformation with respect to a part of the area to be recognized having distortion or deformation. Can be collated. This is because distortion and deformation of the recognition target that occur when the recognition target is photographed from a position other than right in front of the recognition target appear as large as the entire recognition target, but the influence is reduced in a divided area. In addition, since the positional relationship of the collated template is the same as the positional relationship of the collation target in the recognition target, any one of a plurality of patterns in which the recognition target is set in advance based on this positional relationship. Can be determined.

In the image recognition method of the present invention,
The collating step is a step of collating the collation target with a template obtained by slicing the pattern horizontally and dividing the pattern into a plurality, and the determining step is a step of determining based on the collated order of the collated template. May be used. In this case, the collation can be performed based on the ratio of the contrast on the straight line between the collation target and the template. Even if the recognition image is distorted or deformed, the ratio of the contrast on a part of the straight line does not change, so that it is possible to accurately collate distortion and deformation.

In the image recognition method according to the aspect of the present invention, wherein the template is obtained by horizontally slicing and dividing a pattern, the photographing step includes the step of forming a horizontal linear part of the recognition target on or above the recognition target. The step of photographing sequentially as a collation target from below may be a step in which the collation step is a step of sequentially collating the collation targets sequentially photographed.

Further, in the image recognition method according to the present invention, wherein the template is obtained by horizontally slicing and dividing the pattern, the photographing step is a step of photographing the entire recognition target, and the collating step is the photographing step. The recognition target photographed by the means may be divided into regions corresponding to the plurality of divided regions to be a plurality of collation targets, and the divided plurality of collation targets may be sequentially collated. it can.

Further, in the image recognition method according to the present invention, wherein the template is obtained by horizontally slicing and patterning the pattern, the collating step collates the collation targets in the order of the templates constituting the pattern. It can be a step. This way,
Matching can be performed more quickly than matching with all patterns. Note that “matching in the order of templates” includes not only matching in the order of templates, but also matching in which one or more templates are omitted if the order of templates is matched. .

In the image recognition method according to the aspect of the present invention, in which the matching is performed along the order of the templates, the matching step includes the step of matching the object to be checked with a first template and one or two of the first template and the following. This may be a step of starting the matching from the above-described matching with the template. In this way, the recognition target can be recognized even when the portion corresponding to the first template of the recognition target is not set as the matching target due to aging or the like.

In the image recognition method according to the aspect of the present invention, in which the collation is performed in the order of the template, the collation step includes the step of recognizing the collation target with the collated template and the next or subsequent order of the collated template. It can be a step of matching one or more templates. By doing so, it is possible to recognize the recognition target even if the way of division as the collation target is slightly different from the way of division of the template in the pattern, and the portion corresponding to the template to be recognized may change over time. The recognition target can be recognized even when it is not set as the verification target.

[0028]

Next, embodiments of the present invention will be described with reference to examples. FIG. 1 is a block diagram schematically showing the configuration of an image recognition device 20 mounted on a vehicle as an embodiment of the present invention by functional blocks, and FIG. 2 shows an example of a hardware configuration of the image recognition device 20 of the embodiment. FIG.
As shown in FIG. 1, an image recognition device 20 according to the embodiment includes a line CCD camera 22 that captures a linear image, an image processing circuit 24 that binarizes the image captured by the line CCD camera 22, A rainy weather determination unit 26 for determining whether or not it is rainy; a template storage unit 28 for storing a plurality of templates obtained by dividing patterns of characters and symbols into areas; Of the image that has been converted to rainy weather
6, a collation unit 30 for collating with the template stored in the template storage unit 28 using the result of the determination, an ID sequence storage unit 32 for storing the order of template IDs of patterns such as characters and symbols, and a collation unit A recognition determining unit 34 for recognizing patterns based on the order of template IDs collated by 30 and the order of template IDs of each pattern stored in the ID order storage unit 32.

As shown in FIG. 2, the hardware configuration of the image recognition apparatus 20 according to the embodiment is constituted by a microcomputer 40 having a CPU 42 as a center.
The microcomputer 40 includes, in addition to the CPU 42, a ROM 44 storing a processing program, a RAM 46 temporarily storing data, and an input / output circuit (I / O).
48. This microcomputer 40 includes:
An image signal from the line CCD camera 22, a switch signal from the wiper switch 50, and the like are input via the input / output circuit 48. Also, the microcomputer 40
Is also connected via an input / output circuit 48 to a hard disk drive 52 for storing a plurality of templates and the order of template IDs. Image recognition device 20 of the embodiment
Implements the above-described functional blocks by this hardware configuration and software. Hereinafter, the function of each functional block will be described together with the corresponding hardware configuration.

The line CCD camera 22 is mounted horizontally at the front center of the vehicle toward the lower front, and has a horizontal line-like pattern of characters and symbols painted on the road surface on which the vehicle is traveling. The part is photographed as an image. The image processing circuit 24 corresponds to a microcomputer 40 as a hardware configuration, and performs a binarization process of an image as one of graphic processes. Perform value processing. The rainy weather determination unit 26 corresponds to a microcomputer 40 that receives a switch signal from the wiper switch 50 as a hardware configuration, and determines whether or not it is rainy based on a switch signal from the wiper switch 50.

The template storage unit 28 corresponds to a hard disk device 52 as a hardware configuration, and stores a template obtained by horizontally dividing a pattern of characters and symbols painted on the road surface into a plurality of pieces, together with its ID.
For example, in the case of "50" as a character pattern painted on the road surface, as shown in FIG. 3, templates 50A to 50G obtained by dividing a character pattern horizontally into seven parts are used.
Is stored together with the ID as a binary black and white ratio. For example, in the case of the template 50F, the binarized black and white ratio is “black 6, white 2, black 3, white 2, black 4, white 2, black 2, white 2, black 5” from the left. Is encoded as follows.

The matching unit 30 corresponds to a microcomputer 40 as a hardware configuration, and stores a linear image binarized by the image processing circuit 24 in the template storage unit 28.
Search and collate a plurality of templates stored in the. At this time, when the rainy weather determination unit 26 determines that it is rainy, the collation unit 30 causes the image processing circuit 24 to
The collation is performed after inverting the digitized line-shaped image. The collation is performed by searching the template storage unit 28 for a template having the above-described black and white ratio.

The ID order storage unit 32 corresponds to the hard disk device 52 as a hardware configuration, and stores the order of the IDs of the templates constituting each pattern in association with the pattern. For example, in the case of “50” as a character pattern painted on the road surface, the IDs of the templates 50A to 50G are stored in this order in association with “50”.

The recognition judging section 34 corresponds to a microcomputer 40 as a hardware configuration. The order of template IDs collated by the collating section 30 and the order of template IDs of each pattern stored in the ID order storing section 32 are stored. And recognizes characters or symbols painted on the road surface based on the information.

Next, the operation of the image recognition apparatus 20 of the embodiment configured as described above will be described. FIG. 4 is a flowchart illustrating an example of a recognition processing routine executed by the microcomputer 40 of the image recognition device 20 according to the embodiment. This routine is repeatedly executed at predetermined time intervals (for example, every 4 msec) after the vehicle starts running.

When the recognition processing routine is executed, the CPU 42 of the microcomputer 40 first
A linear image captured by the D camera 22 is input (step S100), and the input linear image is binarized (step S102). Next, a switch signal from the wiper switch 50 is input (step S10).
4). It is determined whether it is rainy or not (step S106).
When the input switch signal is on, it is determined that it is rainy, and the binarized image is inverted (step S10).
8) When the input switch signal is off, it is determined that it is not rainy and the inversion process is not performed. Here, it is the line CCD that inverts the binarized image in rainy weather.
This is because the image captured by the camera 22 is in a state where black and white are reversed by water covering the road surface.

Next, when it is not raining, a line CCD is used.
A collation process is performed on a binarized image obtained by the camera 22 or an inverted image in rainy weather as a collation target (step S110). The collation processing is performed by searching the template storage area of the hard disk device 52 functioning as the template storage unit 28 for the same template as the ratio of black and white to be collated. In the embodiment, the correlation is used in the collation processing, and when a template that completely matches is not retrieved, a template having a large correlation is retrieved as a result.

After performing the collation processing, it is determined whether or not a template to be collated has been searched (step S11).
2). When the template to be compared is found, it is determined whether or not the ID of the template to be compared is the same as the previous one (step S113).
The ID is stored in a predetermined area of the RAM 46 (step S114), and this routine ends. If the ID of the template to be collated is the same as the previous one, the routine ends without storing the ID. This is for the following reason. Now
Consider a case where “50” is attached as a character pattern on the road surface. In the line CCD camera 22, first, an image having the same or a large correlation with the template 50G illustrated in FIG. 3 is taken, and the template 50G is searched as a template to be compared. The retrieved ID of the template 50G is stored in a predetermined area of the RAM 46. Immediately after that, the recognition processing routine of FIG. 4 is repeatedly executed. Therefore, depending on the repetition frequency and the vehicle speed of the recognition processing routine of FIG. In some cases, the line CCD camera 22 captures the same black and white image as the template 50G. In such a case, if the ID of the template to be collated is stored in a predetermined area of the RAM 46, the same ID may be stored plural times continuously.
In addition, it is necessary to keep a large predetermined area, and it is also necessary to perform a processing of deleting them at the time of pattern determination processing described later. In the embodiment, in order to keep the predetermined area of the RAM 46 small, when the ID of the template to be collated is the same as the previous time, it is not stored.

If no matching template is found in step S112, it is determined whether or not the matching template ID is stored in a predetermined area of the RAM 46 (step S116). This determination is made by determining whether or not the pattern of characters and symbols on the road surface has been photographed by the line CCD camera 22 and determining whether or not a pattern of characters and symbols has been added to the road surface. Will also serve as That is, if the ID of the template to be matched is R
When it is stored in a predetermined area of the AM 46, it means that a template to be collated was searched the last time the recognition processing routine was executed, and its ID is stored in the RAM 46.
Corresponds to the case where the data is stored in the predetermined area of the line CC.
This means that the photographing of the pattern such as characters and symbols on the road surface by the D camera 22 has been completed. When the ID of the template to be collated is not stored in the predetermined area of the RAM 46, this corresponds to the case where there is no template to be collated even when the recognition processing routine was executed last time, and a pattern such as a character or a symbol is attached to the road surface. Means when not. Therefore, when the ID of the template to be compared is not stored in the predetermined area of the RAM 46 in step S116, this routine ends without performing any processing.

When the ID of the template to be collated is stored in the predetermined area of the RAM 46 in step S116, a pattern determination process is performed based on the order of the template IDs stored in the predetermined area of the RAM 46 (step S118). This determination process is based on the order of the ID of the template stored in the predetermined area of the RAM 46 and the order of the ID of each pattern such as characters and symbols stored in the predetermined area of the hard disk device 52 functioning as the ID order storage unit 32. This is done by comparing. Now
Considering the case where "50" is given as a character pattern on the road surface, the IDs of the templates 50A to 50G illustrated in FIG. If the order of the IDs is reversed, the order of the ID of the template of “50” as the character pattern stored in the hard disk device 52 becomes the same as the order of the characters or symbols attached to the road surface. It can be determined that the pattern is “50”. Note that this determination process is performed in consideration of a case where a different template is searched for due to aging of patterns such as characters and symbols attached to the road surface, and if the order does not completely match, characters and symbols etc. The evaluation may be performed using a weighted pattern template, or the correlation of the order may be used.

When the pattern is determined in this way, the result of the determination is output (step S120), the ID of the template to be collated stored in a predetermined area of the RAM 46 is deleted (step S122), and this routine ends.
The output of the determination result in step S120 is described for the sake of convenience, and includes the case where the determination result is output to another device and the case where the determination result is written to a predetermined address of the RAM 46.

Next, a description will be given of a point that a pattern can be recognized even if deformation or distortion occurs in a recognition target by such recognition processing. FIG. 5 is an explanatory diagram showing an example of the appearance of a character pattern added to the road surface when the vehicle is traveling in the center of the traveling lane. FIG. 6 is a diagram in which the vehicle is traveling in the center of the traveling lane. FIG. 9 is an explanatory diagram of displaying a template which is sometimes photographed and searched by the line CCD camera 22; When the vehicle is traveling in the center of the driving lane,
The character pattern can be recognized by the recognition processing routine of FIG. 4 without having to add the above description.

FIG. 7 is an explanatory view showing an example of the appearance of a character pattern attached to the road surface when the vehicle is traveling on the left side of the traveling lane. FIG. 8 is a diagram showing the vehicle traveling on the left side of the traveling lane. FIG. 4 is an explanatory diagram showing a template that is captured and searched by the line CCD camera 22 when the camera is performing a search. FIG. 9 is an explanatory diagram showing an example of the appearance of a character pattern attached to the road surface when the vehicle is traveling on the right side of the traveling lane. FIG. FIG. 5 is an explanatory diagram of displaying a template that is shot and searched by the line CCD camera 22 when the camera is in the state. As can be seen by comparing FIG. 5, FIG. 7, and FIG. 9, when the vehicle travels on the left or right side of the traveling lane, the character pattern attached to the road surface appears to be deformed. Of course, the character pattern photographed by the line CCD camera 22 is also photographed as having undergone deformation. However, as can be understood by comparing FIG. 6 with FIG. 8 and FIG. 10, when the character pattern is sliced into a horizontal line to form a template and the ratio of black and white is considered, the vehicle is traveling anywhere in the traveling lane. Also, regardless of the deformation of the character pattern, the ratio of the template in black and white is substantially the same. As a result, the character pattern can be recognized using the templates 50A to 50G illustrated in FIG. 3 irrespective of where the vehicle is traveling in the traveling lane.

According to the image recognition apparatus 20 of the embodiment described above, the recognition target is deformed or distorted by photographing the recognition target from a position other than the front, by using the template obtained by slicing the pattern into horizontal lines. Even if it occurs, the pattern can be recognized. As a result,
There is no need to store a large number of templates, and retrieval and recognition can be performed in a short time. Further, according to the image recognition device 20 of the embodiment, it is determined whether it is rainy or not, and the image is inverted and collated in rainy weather. The pattern can be recognized.

Next, an image recognition apparatus 20B according to a second embodiment of the present invention will be described. FIG. 11 is a block diagram schematically illustrating the configuration of an image recognition device 20B according to the second embodiment by functional blocks. Image recognition device 20 of second embodiment
B has a point that a CCD camera 22B is provided instead of the line CCD camera 22 as shown in FIG.
The configuration is the same as that of the image recognition device 20 of the first embodiment except that an area dividing unit 25B is provided between the image recognition apparatus 20 and the matching unit 30. Therefore, the image recognition device 20B of the second embodiment
Among the configurations described above, the same configurations as those of the image recognition device 20 of the first embodiment are denoted by the same reference numerals, and description thereof will be omitted. The hardware configuration of the image recognition device 20B of the second embodiment is the same as that of the image recognition device 20 of the first embodiment illustrated in FIG.
Since only the line CCD camera 22 is replaced by the CCD camera 22B in the hardware configuration described above, its illustration is also omitted.

The CCD camera 22B, unlike the line CCD camera 22 included in the image recognition device 20 of the first embodiment, captures not a linear image but a rectangular image. For example, an image is captured including an object to be recognized as shown in FIG. 5, FIG. 7, and FIG. Image processing circuit 24
B performs binarization processing on a rectangular image captured by the CCD camera 22B, similarly to the image processing circuit 24 included in the image recognition device 20 of the first embodiment. The difference from the image processing circuit 24 of the first embodiment is that the image to be binarized is a linear image or a rectangular image. Area division unit 25B
In the binarized image, the whole image corresponding to the recognition target is horizontally sliced by the same number of divisions as the pattern template stored in the hard disk device 52 to divide the area, and each divided area is divided into areas. Is passed to the collating unit 30 as a collation target.

The recognition processing of the image recognition apparatus 20B of the second embodiment thus configured is performed by a recognition processing routine illustrated in FIG. The recognition processing by this recognition processing routine is different from the recognition processing of the image recognition device 20 of the first embodiment that executes the recognition processing routine illustrated in FIG. 4 except that the area division processing in step S203 is included. The same processing is performed. That is, in the image recognition device 20 of the first embodiment, the pattern is determined by sequentially collating the line-shaped images captured by the line CCD camera 22. In the device 20B, the entire recognition target photographed by the CCD camera 22B is divided into areas similar to the template, and the divided areas are collated to determine a pattern. Therefore, the description of the recognition processing routine of FIG. 12 will largely be the same as the description of the recognition processing routine of FIG. 4, and a description thereof will be omitted. FIG.
The reason why there is no processing for storing the ID of the template in the predetermined area of the RAM 46 in the recognition processing 2 is that the entire recognition image is divided into areas, and all the divided areas are collated. This is because they have all of the IDs.

The image recognition apparatus 2 of the second embodiment described above
The same effect as that of the image recognition device 20 of the first embodiment can be obtained by using 0B. That is, by using a template obtained by slicing a pattern into horizontal lines, the recognition target is photographed from a position other than directly in front, so that the pattern can be recognized even if the recognition target is deformed or distorted. It is possible to provide an effect that a pattern such as a character or a symbol appropriately attached to the road surface can be recognized.

In the above-described image recognition device 20 of the first embodiment and the image recognition device 20B of the second embodiment, all templates formed by horizontally slicing character patterns are stored in the hard disk device 52. Templates having a high correlation may be commonly used. Hereinafter, a case where a template having a high correlation is commonly used will be described.

FIGS. 13 to 16 show "30", "40", "50", and "6" as examples of character patterns added to the road surface when the vehicle is traveling in the center of the traveling lane.
It is explanatory drawing which illustrates the template which comprises "0".
Each template is configured as a horizontal slice of the entire pattern into ten pieces. Each template was given a serial number from below as a suffix to each pattern as a code. For example, in the case of the pattern “30”, FIG.
3 from the bottom, "30-01", "30-02" ...
・ "30-10". Of each of these templates,
For example, “30-02”, “50-02”, “50-0”
The templates “9”, “60-02”, and “60-09” are templates having similar black and white ratios. Now, a predetermined correlation value (for example, 80%) for each pattern
If the above are the same, the above-mentioned “30-
02 "and" 50-02 "can be replaced by the same template. FIG. 17 is an explanatory diagram exemplifying a common template when it is determined that a correlation value of 80% or more is the same for each of the templates in FIGS. For example, in FIG. 17, the second “SAME-2” from the bottom is
“30-02” which is the same as “50-02” and is regarded as the same
−02 ”,“ 50-09 ”,“ 60-02 ”,“ 60-
09 "can be replaced with each template. FIGS. 18 to 21 are obtained by replacing the templates illustrated in FIGS. 13 to 16 with the template of FIG. For example, FIG. 18 is configured by replacing the template of the pattern “30” of FIG. 13 that can be replaced with the template of FIG. In this case, all templates except the template “30-06” have been replaced. In the template of the pattern “40” in FIG. 14, “40-01” and “4
Seven templates except the three templates “0-08” and “40-10” are replaced.

As described above, if a template having a high correlation is commonly used, the number of templates stored in the hard disk drive 52 can be reduced. As a result,
The search time of the hard disk device 52 at the time of collation can be shortened, and the processing can be performed quickly.

Here, a template that could not be replaced with the common template of FIG. 17 (for example, template “30-06” of FIG. 18) can be said to be a characteristic template in the pattern. Therefore,
When the characteristic template is collated, the pattern can be recognized by giving a higher weight to the characteristic template than other templates in addition to the order of the templates. In this way, it is possible to more appropriately recognize.

The image recognition apparatus 20 of the first embodiment and the image recognition apparatus 20B of the second embodiment do not deal with the case where the pattern is partially erased due to blurring or the like. By using a template for recognizing a pattern in which the illustrated part has disappeared, the pattern can be recognized even if it is faint due to aging. In FIG. 22, the template "50B-
FIG. 23 shows a state where the right side of “0” in templates “60B-03” to “60B-07” has disappeared due to aging. By using such a template in which a part of the pattern has disappeared, the pattern can be properly recognized even if a part of the pattern painted on the road surface due to aging has disappeared.

In the image recognition device 20 of the first embodiment and the image recognition device 20B of the second embodiment, a line-shaped image taken by the line CCD camera 22 or an area dividing portion 25B taken by the CCD camera 22B is used. The hard disk drive 52 was searched for a line-shaped image area-divided in the above step, and matching was performed based on whether or not there was a template having the same ratio of black and white, and pattern recognition was performed based on the order of the compared templates. A plurality of patterns may be collated based on whether or not they can be collated with the first template or the next template among the templates constituting each pattern, and the patterns may be recognized in the order of the collated templates. . Hereinafter, the collation of the template and the recognition of the pattern in this modified example will be described in detail. Note that the hardware configuration of an image recognition device that performs template matching and pattern recognition according to the modified example is as follows.
Since the configuration is the same as that of the image recognition device 20 of the first embodiment, a description thereof will be omitted.

FIG. 24 is an explanatory view exemplifying a collation pattern when collating whether the pattern painted on the road surface is "30", and FIG. 25 is a diagram showing whether the pattern is "40". FIG. 26 is an explanatory diagram illustrating an example of a matching pattern when matching is performed. FIG. 26 is an explanatory diagram illustrating an example of a matching pattern when checking whether or not the pattern is “50”. FIG. It is explanatory drawing which illustrates the collation pattern at the time of collating whether it is ". Each collation pattern uses a common template illustrated in FIG. For example, in the case of the collation pattern in FIG. 24, the order from the top of the contents of each box is the same as the order from the bottom of the template (see FIG. 18) constituting the pattern “30”.
That is, when the line CCD camera 22 of the first embodiment is used, the patterns are photographed and collated in order from the bottom, and the collation pattern in FIG. 24 is in the order of collating the template in FIG. 18 from the bottom. 25 to 27
Are the corresponding patterns "40" to "60"
Are arranged from the bottom of the template (FIGS. 19 to 21).

The arrow in each collation pattern indicates a template that is allowed when collating the photographed image.
For example, in the case of the matching pattern of the pattern “30” illustrated in FIG. 24, the captured image is “SA” at the start of the matching.
ME-1 "or" SAME-
"SAM" and "SAM"
When the template of “E-1” is collated, the next captured image is collated with the same “SAME-1” template or with the next “SAME-2” template, or with the next “SAME”. -3 "is permitted. Here, the reason that the captured image is allowed to be compared with the same template is that when the vehicle speed is low, an image having the ratio of black and white belonging to the same template is captured many times. Also, it is permitted to match with the next (next) template when the vehicle speed is high, and when the image corresponding to the next template cannot be shot depending on the shooting timing. This is to allow. The determination as to whether or not to perform collation can be made based on whether or not the correlation value is equal to or greater than a predetermined value (for example, 60).

In the template collation of this modified example, the collation is advanced using the collation patterns created for each pattern with respect to the binarized collation target whose image is taken by the line CCD camera 22. As a specific example, "60" is a pane on the road surface, and this is "30",
Consider a case of recognizing which of “40”, “50”, and “60”. FIG. 28 shows "30" illustrated in FIG.
FIG. 29 is an explanatory diagram for explaining a state in which the matching is performed using the matching pattern of FIG. 29. FIG. 29 is an explanatory diagram for explaining a state in which the matching is performed using the matching pattern of “40” illustrated in FIG. FIG. 30 is an explanatory diagram illustrating a state in which matching is performed using a matching pattern of “50” illustrated in FIG. 26;
FIG. 31 is an explanatory diagram illustrating a state in which the matching is performed using the matching pattern “60” illustrated in FIG. 27. “Time” in each drawing indicates the timing at which the line CCD camera 22 captures an image. The numerical values in FIGS. 28 to 31 are correlation values. In this modification, the matching is determined when the correlation value is equal to or greater than the predetermined value (for example, 60) as described above. However, in the matching patterns illustrated in FIGS. In some cases, the correlation value for a plurality of templates is equal to or greater than a predetermined value. In this case, in the template matching of the modified example, the one with the larger correlation value is used as the template that has been matched. For example, when “time” in FIG. 28 is “03”, “SAM”
The template of “E-1” and the template of “SAME-2” are both collated with a correlation value of 60 or more, but the template of “SAME-1” having a large correlation value is the collated template. “*” On the right side of the collation template column in FIGS. 28 to 31 indicates a collated template used for pattern recognition. The case where the same template is successively collated is considered.

As a specific example, in the collation of “60” painted on the road surface, as shown in FIGS. 28 and 31, the collation pattern of “30” and the collation pattern of “60” are the boxes of “end”. 29 and FIG.
As shown in the figure, the matching pattern of “40” and the matching pattern of “50” do not reach the box of “end”. Therefore, the recognition target is recognized as one of the patterns “30” and “60”. Regarding which pattern to recognize, a recognition determination is performed based on the number of matched templates, a recognition determination is performed based on a correlation value of the matched templates, or a characteristic template is weighted. There are various determination methods such as performing recognition determination. For example, assuming that the recognition determination is performed based on the number of templates collated in the above specific case, the number of collated templates is 7 with respect to the total number 9 of templates constituting the pattern “30”. Since the number of templates collated with the total number of templates constituting the pattern "60" is 8, the comparison between 7/9 and 8/8 is made, and the pattern "60" is recognized. Also, recognition judgment based on the correlation value of the collated template, for example, as one of the methods, the correlation average value obtained by dividing the sum of the largest correlation value among the collated templates by the total number of templates constituting the pattern. As a result of the comparison, the correlation average of the pattern “30” is 60.74, and the pattern “6”
Since the correlation average value of “0” is 90.69, the pattern “60” is recognized. 28 to 3
1 shows the largest correlation value among the collated templates in a rectangular frame. The above-mentioned correlation average value is obtained by dividing the sum of those enclosed by the rectangular frame by the total number of patterns. Recognition judgment by weighting the characteristic template, for example, setting the normal template as point 1 and the characteristic template as point 2, and dividing the sum of the points of the matched template by the total number of templates constituting the pattern According to the comparison of the obtained point average values, the point average of the pattern “30” is 8/9 and the point average of the pattern “60” is 1
Since it is 1/8, the pattern "60" is recognized.

According to the template collation of the modified example described above, it is possible to collate the image photographed by the line CCD camera 22 based on the collation pattern of each pattern. Moreover, the matching by the matching pattern is
In addition to collating with the next template as the order, collation with the same template or the next template is allowed, so that it is possible to cope with a case where the vehicle speed is slow or fast. In addition, since the collation is performed using the correlation value, it is possible to cope with, for example, blurring due to aging of the pattern painted on the road surface.

According to the pattern recognition of the modified example, the number and the correlation value of the templates matched in the order of the templates constituting the pattern, the weight given to the characteristic template, and the like are taken into consideration, so that more appropriate This makes it possible to make an accurate recognition determination.

In the template collation of the modified example, the collation with the same template as the collated template, the next template, and the next template is permitted. May be allowed.

In the template collation of the modified example, the collation is performed using the collation pattern composed of the templates constituting the pattern. However, the collation is performed using the collation pattern composed of the template in which a part of the pattern has disappeared. It may be. Further, the matching may be performed using a matching pattern including a template constituting the pattern and a template in which a part of the pattern has disappeared.
FIG. 32 shows an example of the matching pattern. FIG. 32 is an explanatory diagram illustrating an example of a collation pattern formed by combining a template forming the pattern “60” illustrated in FIG. 21 and a template in which a part of the pattern “60” illustrated in FIG. 23 has disappeared. is there. The branch portion on the right side of the figure is a template in which a part of the pattern has disappeared. In the case of this matching pattern, the template "SA"
When “ME-8” is matched, “SAME-8” of the same template and “SAME-
2 "and the next template," SAME-9 "
Collation with the partially deleted template “60B-3” is allowed. At this time, if the template “60B-3” is collated, the next template “60B-3”
And the next template "60B-6" and the next template "60B-7" are allowed. If a template in which a part of the pattern has disappeared is provided as a branch part in this way, it is possible to reduce the number of collation patterns to start collation at the same time, and it is a pattern that has partially disappeared due to aging. Can also be matched. In the matching pattern shown in FIG. 32, when the process proceeds to the branch portion composed of the template in which a part has disappeared, the matching is performed in the order of the template of the branch portion. Alternatively, a transition from the trunk portion to the branch portion may be made halfway. A matching pattern in this case is illustrated in FIG.
In this matching pattern, for example, when the template “60B-3” of the branch portion is compared, as described above, “60B-3” of the same template, “60B-6” of the next template, and the next template "60
B-7 "is permitted, and further, the base part template" 60-06 "is permitted. Further, when the template “60-06” of the trunk portion is collated, the collation with the template “60B-7” of the branch portion is also permitted. As described above, by allowing mutual matching between the template of the branch portion and the template of the trunk portion, the pattern can be recognized even when the degree of blurring or disappearance is different.

Although the template collation and pattern recognition of the modified example have been described as the configuration of the image recognition device 20 of the first embodiment, the configuration can be similarly performed by the configuration of the image recognition device 20B of the second embodiment.

The result of recognition by the image recognition device 20 of the first embodiment, the image recognition device 20B of the second embodiment, and the image recognition device for performing template collation and pattern recognition of the modified example is to confirm the current position in the navigation system. For example, it can be used for various devices and systems, such as a device for issuing a warning by comparing with a vehicle speed of a vehicle, a device for giving information to a driver or an occupant, and the like.

In the image recognition device 20 of the first embodiment, the image recognition device 20B of the second embodiment, and the image recognition device of the modification,
Although the matching is performed based on the black and white ratio of the template, the matching may be performed on the template itself. Further, a part of the black-and-white ratio, particularly a characteristic part, may be given a weight, and the weight may be given a large specific gravity to calculate the correlation and collate. Furthermore, the image recognition device 20 of the first embodiment and the image recognition device 20B of the second embodiment
In the above, the binarization processing is used. However, the present invention is not limited to the binarization processing, and can be applied to a multi-level processing or the like.

In the image recognition device 20 of the first embodiment, the image recognition device 20B of the second embodiment, and the image recognition device of the modification,
The rainy weather determination unit 26 is provided to determine whether or not it is rainy, and the image is inverted in the rainy weather. However, although the recognition accuracy in the rainy weather is slightly reduced, the rainy weather determination unit 26 is not provided, and Regardless of whether or not, the same processing may be performed.

In the image recognition device 20 of the first embodiment, the image recognition device 20B of the second embodiment, and the image recognition device of the modification,
Although it was described as recognizing patterns such as characters and symbols attached to the road surface mounted on the vehicle, it is not limited to patterns attached to the road surface, but it can be used to recognize patterns attached to any objects. It may be applied. In this case, the template is not limited to one that is sliced in a line shape in the horizontal direction, may be one that is sliced in a line shape in the vertical direction, or may be one that is sliced in a line shape at a predetermined angle from the horizontal direction. Good.

The embodiments of the present invention have been described with reference to the embodiments. However, the present invention is not limited to these embodiments, and various embodiments may be made without departing from the spirit of the present invention. Of course, it can be carried out.

[Brief description of the drawings]

FIG. 1 is a block diagram schematically showing the configuration of an image recognition device 20 according to an embodiment of the present invention by functional blocks.

FIG. 2 is a configuration diagram illustrating an example of a hardware configuration of an image recognition device 20 according to the embodiment.

FIG. 3 is an explanatory diagram illustrating a template in the case of “50” as a character pattern painted on a road surface;

FIG. 4 is a flowchart illustrating an example of a recognition processing routine executed by the microcomputer 40 of the image recognition device 20 according to the embodiment.

FIG. 5 is an explanatory diagram illustrating an example of how a character pattern attached to a road surface looks when a vehicle is traveling in the center of a traveling lane.

FIG. 6 is an explanatory diagram showing a template that is captured and searched by the line CCD camera 22 when the vehicle is traveling in the center of the traveling lane.

FIG. 7 is an explanatory diagram illustrating an example of how a character pattern attached to a road surface looks when a vehicle is traveling on the left side of a traveling lane.

FIG. 8 is an explanatory diagram showing a template which is captured and searched by the line CCD camera 22 when the vehicle is traveling on the left side of the traveling lane.

FIG. 9 is an explanatory diagram showing an example of how a character pattern attached to a road surface looks when a vehicle is traveling on the right side of a traveling lane.

FIG. 10 is an explanatory view showing a template which is captured and searched by the line CCD camera 22 when the vehicle is traveling on the right side of the traveling lane.

FIG. 11 is a block diagram showing an outline of a configuration of an image recognition device 20B according to a second embodiment by functional blocks.

FIG. 12 is a flowchart illustrating an example of a recognition processing routine executed by the microcomputer 40 of the image recognition device 20B according to the second embodiment.

FIG. 13 is an explanatory diagram illustrating a template “30” as an example of a character pattern added to the road surface when the vehicle is traveling in the center of the traveling lane.

FIG. 14 is an explanatory diagram illustrating a template of “40” as an example of a character pattern added to the road surface when the vehicle is traveling in the center of the traveling lane.

FIG. 15 is an explanatory diagram illustrating a template of “50” as an example of a character pattern added to the road surface when the vehicle is traveling in the center of the traveling lane.

FIG. 16 is an explanatory diagram illustrating a template “60” as an example of a character pattern added to the road surface when the vehicle is traveling in the center of the traveling lane.

FIG. 17 is an explanatory diagram exemplifying a list of templates that can be used in common;

FIG. 18 is an explanatory diagram illustrating a state in which the template of pattern “30” in FIG. 13 is replaced with a template that can be used in common.

FIG. 19 is an explanatory diagram illustrating a state in which the template of the pattern “40” in FIG. 14 is replaced with a template that can be used in common.

20 is an explanatory diagram illustrating a state in which the template of the pattern “50” in FIG. 15 is replaced with a template that can be used in common.

FIG. 21 is an explanatory diagram illustrating a state in which the template of the pattern “60” in FIG. 16 is replaced with a template that can be used in common.

FIG. 22 is an explanatory diagram illustrating an example of a template in which a part of the pattern “50” has disappeared due to aging.

FIG. 23 is an explanatory diagram showing an example of a template in which a part of the pattern “60” has disappeared due to aging.

FIG. 24. The pattern painted on the road surface is “3”.
It is explanatory drawing which illustrates the collation pattern at the time of collating whether it is "0".

FIG. 25: The pattern painted on the road surface is “4”
It is explanatory drawing which illustrates the collation pattern at the time of collating whether it is "0".

FIG. 26: The pattern painted on the road surface is “5
It is explanatory drawing which illustrates the collation pattern at the time of collating whether it is "0".

FIG. 27: The pattern painted on the road surface is “6
It is explanatory drawing which illustrates the collation pattern at the time of collating whether it is "0".

FIG. 28 is an explanatory diagram illustrating a state in which matching is performed using a matching pattern of pattern “30” illustrated in FIG. 24;

FIG. 29 is an explanatory diagram illustrating a state in which matching is performed using a matching pattern of pattern “40” illustrated in FIG. 25;

FIG. 30 is an explanatory diagram illustrating a state in which matching is performed using a matching pattern of pattern “50” illustrated in FIG. 26;

FIG. 31 is an explanatory diagram illustrating a state in which matching is performed using a matching pattern of pattern “60” illustrated in FIG. 27;

FIG. 32 is an explanatory diagram showing an example of a collation pattern formed by combining a template constituting a pattern “60” and a template in which a part of the pattern “60” has disappeared.

FIG. 33 is an explanatory diagram illustrating an example of a matching pattern according to a modification.

[Explanation of symbols]

20, 20B image recognition device, 22 line CCD camera, 22B CCD camera, 24, 24B image processing circuit, 25B area division unit, 26 rainy weather judgment unit, 28
Template storage unit, 30 collation unit, 32 ID order storage unit, 34 recognition determination unit, 40 microcomputer, 42 CPU, 44 ROM, 46 RAM, 48
I / O circuit, 50 wiper switch, 52 hard disk drive.

Claims (19)

[Claims] 1. An image recognition apparatus for recognizing a recognition target as one of a plurality of preset patterns, wherein a plurality of templates corresponding to at least a part of an area obtained by dividing the pattern into a plurality of patterns are provided. Template storage means for storing; photographing means for photographing at least a part of the recognition target as a collation target; collation means for collating the photographed collation target with a template stored in the template storage means; A determination unit that determines which of the plurality of patterns the recognition target is based on a positional relationship between the regions of the template. 2. The image recognition apparatus according to claim 1, wherein the template storage unit is a unit that stores, as a template, an area obtained by slicing the pattern horizontally and dividing the pattern into a plurality of sections. Is an image recognition device that is a means for making a determination based on the collated order of the collated templates. 3. The image recognition apparatus according to claim 2, wherein the photographing unit photographs a horizontal linear part of the recognition target sequentially from above or below the recognition target as a verification target. The image recognition device, wherein the collation unit is a unit for sequentially collating the collation targets photographed sequentially. 4. The image recognition device according to claim 2, wherein the photographing unit is a unit for photographing the entirety of the recognition target, and the collation unit is configured to determine the recognition target photographed by the photographing unit. An image recognition device, which is a means for dividing into a plurality of regions to be compared corresponding to regions divided into a plurality of regions and for sequentially collating the plurality of regions to be collated. 5. The image recognition apparatus according to claim 2, wherein the matching unit matches the matching target in the order of the template forming the pattern. 6. The collating means is a means for starting collation of the object to be collated from the collation of a first template with one or more templates subsequent to the first template after the first template. An image recognition device according to claim 1. 7. The collating means for collating the collation target with a collated template and one or more templates subsequent to the collated template. An image recognition device according to claim 1. 8. The image recognition apparatus according to claim 2, wherein said matching means is means for performing matching using a correlation between said matching object and said template. 9. The image recognition apparatus according to claim 2, wherein the determination unit is a unit that determines based on a correlation between the collated templates. 10. The image recognition apparatus according to claim 2, wherein the determination unit is a unit that determines a template including a characteristic portion of the pattern in consideration of a weight. 11. The on-vehicle image recognition apparatus according to claim 1, wherein the recognition target is a character or a symbol attached to a road surface. 12. The image recognition device according to claim 11, further comprising rainy weather determination means for determining whether it is rainy weather, wherein said collation means determines that said rainy weather determination time is rainy weather, An image recognition device, which is means for inverting one of the object to be compared and the template to an inverted color and comparing the inverted one with the other. 13. An image recognition method for recognizing a recognition target as one of a plurality of patterns set in advance, comprising: a photographing step of photographing at least a part of the recognition target as a collation target; And a matching step of matching with a template forming at least a part of an area obtained by dividing the pattern into a plurality of patterns, and the recognition target is any of the plurality of patterns based on a positional relationship of the checked template. And a determining step of determining whether the image recognition is performed. 14. The image recognition method according to claim 13, wherein the collation step is a step of collating the collation target with a template obtained by horizontally slicing the pattern and dividing the pattern into a plurality of sections. The determining step is a step of determining based on the collated order of the collated templates. 15. The image recognition method according to claim 14, wherein, in the photographing step, a horizontal linear part of the recognition target is photographed sequentially from above or below the recognition target as a collation target. The image recognition method, wherein the collation step is a step of sequentially collating the collation targets photographed sequentially. 16. The image recognition method according to claim 14, wherein the photographing step is a step of photographing the entirety of the recognition target, and the collating step is a step of recognizing the recognition target photographed by the photographing means. An image recognition method comprising the steps of: dividing an area corresponding to a plurality of divided areas into a plurality of collation targets; and sequentially collating the divided plurality of collation objects. 17. The image recognition method according to claim 14, wherein the collating step is a step of collating the collation target in the order of templates constituting the pattern. 18. The method according to claim 17, wherein the collation step is a step of starting collation of the collation target with a first template and one or more templates subsequent to the first template after the first template. Image recognition method as described. 19. The collation step is a step of collating the collation target with a collated template and one or more templates subsequent to the collated template. Image recognition method as described.