JP2017138674A - License number plate recognition device, and license number plate recognition system as well as license number plate recognition method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce wrong recognition of a character string in a license number plate image of a photographed vehicle and the like.SOLUTION: A character string in a license number plate image is composed a plurality of characters in which an arrangement order from 1 to x (where x is an integer equal to or more than 2) in accordance with a position in the image or a relative position among characters is set. A processor 21 of a license number plate recognition device 3 is configured to: calculate a similarity degree between the character string in the license number plate image and a plurality of reference images, respectively; about a reference character included in each of the plurality of reference images corresponding to one character or y (where y is an integer smaller than x and equal to or more than 2) characters standing in the arrangement order serving as an evaluation object in the character string, calculate an evaluation value including a sum of the similarity degree for each reference image in which the reference character is identical; and determine a candidate of at least one character capable of constituting the character string in the license number plate image on the basis of the evaluation value.SELECTED DRAWING: Figure 1

Description

  The present disclosure relates to a license plate recognition device, a license plate recognition system, and a license plate recognition method for recognizing a character string in a license plate image of a photographed vehicle or the like.

  2. Description of the Related Art Conventionally, a technique for reading a license plate from an image of a vehicle photographed by a camera installed at a road or facility entrance / exit for the purpose of enhancing the security environment and managing customers has been widespread. For example, an automobile number automatic reading device installed on a main road or the like can read a license plate of a vehicle traveling on a road in a predetermined direction.

  On the other hand, even a device specialized for reading a license plate of a vehicle, such as the above-described auto number reading device, cannot obtain a clear captured image due to changes in shooting conditions such as bad weather or dirt on a camera lens. There is a case. Also, surveillance cameras installed in stores, streets, etc. may not always have a captured image suitable for license plate recognition, depending on the camera resolution and shooting conditions (number plate size, shooting angle, etc. in the captured image). It may not be obtained.

  Therefore, a technique for preventing erroneous recognition (reading mistake) of a character string in a license plate image has been developed even for such a captured image that is not suitable for license plate recognition. For example, a technique is known in which a correlation value between a character in a license plate image and a matching character pattern is calculated, and the character is recognized only when the correlation value is equal to or greater than a certain value (Patent Literature). 1).

  Also, the imaging pattern captured by the camera is compared with a plurality of dictionary patterns stored in advance, the similarity is calculated for each dictionary pattern, and the similarity is corrected based on the appearance frequency for each dictionary pattern stored in advance. A technique for calculating the corrected similarity and determining a dictionary pattern having the highest correction similarity is known (see Patent Document 2).

JP-A 64-74700 Japanese Patent Laid-Open No. 10-63786

  By the way, for example, in the automobile registration number mark used as a vehicle license plate in Japan, the fourth digit Arabic numeral and the middle dot “・” (the upper digit that is blank when the number is three digits or less) A series designation number consisting of symbols to be filled in is displayed. When such a series of designated numbers are recognized from a captured image according to the conventional techniques described in Patent Documents 1 and 2 above, the degree of similarity with a plurality of matching patterns made up of four-digit numbers prepared in advance And a pattern for matching having the maximum similarity is estimated as a series designated number.

  However, as a result of intensive studies, the inventors of the present application have recognized a character string (such as a multi-digit number or a series of characters) such as the above-described series designation number displayed on the license plate. Simply evaluating the similarity between the whole and the pattern for matching does not properly reflect the similarity between the individual numbers or characters that make up the string and the individual numbers or characters included in the pattern for matching. It has been found that the misrecognition of character strings displayed on the license plate can be effectively reduced by paying attention to such individual numbers and characters.

  The present disclosure has been devised in view of the problems of the prior art, and a license plate recognition device and a license plate recognition system capable of reducing erroneous recognition of a character string in a license plate image of a photographed vehicle or the like. The main object is to provide a license plate recognition method.

  The license plate recognition device of the present disclosure is a license plate recognition device including a processor that executes processing for recognizing a character string in a captured license plate image, and the character string is a position in the license plate image or It is composed of a plurality of characters set in an arrangement order from 1 to x (where x is an integer of 2 or more) according to the relative position between characters, and the processor includes the character string in the license plate image 1 or y (where y is an integer of 2 or more smaller than x) in the arrangement order to be evaluated in the character string by calculating the similarity to each of a plurality of reference images prepared in advance For the reference characters included in each of the plurality of standard images corresponding to the same, the sum of the similarities is calculated for each standard image in which the reference characters are the same. Calculating a zero-valued value, based on said evaluation value, and determining the candidate of the license plate of the at least one may constitute a character string in an image character.

  According to the present disclosure, it is possible to reduce erroneous recognition of a character string in a license plate image of a photographed vehicle or the like.

The block diagram which shows the outline of the license plate recognition system which concerns on 1st Embodiment The block diagram which shows the hardware constitutions of the license plate recognition apparatus shown in FIG. Functional block diagram of the license plate recognition device shown in FIG. The figure which shows an example of the calculation result by the similarity calculation part in FIG. The figure which shows the example of extraction of the top candidate of the correlation value by the similarity evaluation part in FIG. The figure which shows an example of the evaluation value calculation result of each digit by the similarity evaluation part in FIG. Explanatory drawing which shows the modification of the extraction method of the candidate number by the similarity evaluation part in FIG. Flow chart showing the flow of processing of the license plate recognition device shown in FIG. FIG. 8 is a flowchart showing details of step ST108 in FIG. The figure which shows an example (higher 2 digits) of the calculation result by the similarity calculation part of the license plate recognition apparatus which concerns on 2nd Embodiment. The figure which shows an example (lower 2 digits) of the calculation result by the similarity calculation part of the license plate recognition apparatus which concerns on 2nd Embodiment. The figure which shows an example of the evaluation value calculation result of each digit by the similarity evaluation part of the license plate recognition apparatus concerning 2nd Embodiment

  A first invention made to solve the above-described problem is a license plate recognition device including a processor for executing processing for recognizing a character string in a photographed license plate image, wherein the character string is the number It is composed of a plurality of characters set in an arrangement order from 1 to x (where x is an integer of 2 or more) according to the position in the plate image or the relative position between the characters, and the processor includes the license plate image The similarity between the character string in the image and a plurality of reference images prepared in advance is calculated, and 1 or y in the arrangement order to be evaluated in the character string (where y is 2 or more smaller than x) For each of the reference images having the same reference character, the reference characters respectively included in the plurality of reference images corresponding to the number of characters. Calculating an evaluation value comprising the sum of degrees, based on said evaluation value, and determining the candidate of the license plate of the at least one may constitute a character string in an image character.

  According to the license plate recognition apparatus according to the first aspect of the present invention, the reference images respectively included in the plurality of reference images corresponding to one or y characters in the arrangement order to be evaluated in the character string in the license plate image. For a character, an evaluation value including the sum of similarities is calculated for each reference image with the same reference character, and at least one character candidate that can form a character string in the license plate image is determined based on the evaluation value Therefore, it is possible to reduce erroneous recognition of character strings in a license plate image of a photographed vehicle or the like.

  According to a second aspect, in the first aspect, the processor calculates the evaluation value using only similarities from a larger one of the plurality of similarities to a number within a predetermined threshold. It is characterized by that.

  According to the license plate recognition apparatus according to the second aspect of the present invention, it is possible to obtain a more appropriate evaluation value by eliminating the reference image having a relatively low similarity in calculating the evaluation value.

  According to a third aspect of the present invention, in the first or second aspect of the invention, the y characters are in the arrangement order that is continuous with each other.

  According to the license plate recognition apparatus according to the third aspect of the present invention, the evaluation value is calculated by calculating the evaluation value including the sum of the similarities for each reference image in which a plurality of adjacent reference characters are the same. Processing becomes easy.

  According to a fourth aspect, in any one of the first to third aspects, the reference character is a single character.

  According to the license plate recognition apparatus according to the fourth aspect of the invention, by focusing on one character included in the character string, the similarity of each character is appropriately evaluated, and a more appropriate evaluation value is acquired. It becomes possible.

  In a fifth aspect based on the first aspect, the processor calculates an evaluation value including the similarity sum for each reference image having the same reference character for all the arrangement orders. It is characterized by doing.

  According to the license plate recognition apparatus according to the fifth aspect of the present invention, since the evaluation values are calculated for all the arrangement orders, it is possible to appropriately determine all the character candidates that can constitute the character string in the license plate image. It becomes possible.

  According to a sixth invention, in any one of the first to fifth inventions, the similarity is a correlation coefficient between the character string in the license plate image and the reference image. To do.

  According to the license plate recognition apparatus of the sixth aspect of the invention, it is possible to acquire an appropriate evaluation value by a simple process by using the correlation coefficient as the similarity.

  In a seventh aspect based on any one of the first to sixth aspects, the processor determines a character string candidate in the license plate image based on the similarity, and the at least one character And a display device for displaying the character string candidates.

  According to the license plate recognition apparatus of the seventh aspect of the invention, the user can easily search for a character string in the license plate image by referring to at least one character candidate based on the evaluation value and a character string candidate based on the similarity. Can be estimated.

  According to an eighth aspect of the present invention, there is provided a license plate recognition apparatus according to any one of the first to seventh aspects of the present invention, and an imaging apparatus that captures the number plate image.

  According to the license plate recognition system of the eighth aspect of the invention, it is possible to reduce erroneous recognition of character strings in a license plate image of a photographed vehicle or the like.

  The ninth invention is a license plate recognition method by a license plate recognition device for executing processing for recognizing a character string in a license plate image of a photographed vehicle, wherein the character string is included in the license plate image. It is composed of a plurality of characters set in an arrangement order from 1 to x (where x is an integer of 2 or more) according to the position or relative position between characters, The degree of similarity with the prepared reference image is calculated, and the plurality of characters corresponding to one or y characters (where y is an integer of 2 or more smaller than x) in the arrangement order to be evaluated in the character string For each reference character included in each of the reference images, an evaluation value that includes the sum of the similarities is calculated for each of the reference images that have the same reference character, and the evaluation value Based, and determining the candidate of the license plate image at least one may constitute a string of characters.

  According to the license plate recognition method of the ninth aspect, it is possible to reduce erroneous recognition of a character string in a license plate image of a photographed vehicle or the like.

  Hereinafter, embodiments of the present disclosure will be described with reference to the drawings.

(First embodiment)
FIG. 1 is a configuration diagram illustrating an outline of a license plate recognition system 1 according to the first embodiment of the present disclosure, and FIG. 2 is a block diagram illustrating a hardware configuration of the license plate recognition device 3 illustrated in FIG. 1. is there.

  The license plate recognition system 1 is a system for recognizing a character string in a license plate image obtained by photographing a license plate of a vehicle or the like. The license plate recognition system 1 includes an imaging device 2 that captures a vehicle or the like including a license plate, a license plate recognition device 3 that performs character string recognition processing on a license plate image acquired by the imaging device 2, and a license plate. An external storage device 4 that stores various data related to processing of the recognition device 3 and a display device 5 that displays a license plate image captured by the imaging device 2, a processing result of the license plate recognition device 3, and the like are provided.

  Here, as a license plate, an automobile registration number mark used in Japan will be described as an example. In the automobile registration number mark, a series of designated numbers is displayed as a character string to be recognized by the license plate recognition device 3. This series designation number is mainly composed of a 4-digit Arabic numeral in which a hyphen “-” is inserted between the upper 2 digits and the lower 2 digits. When the series designation number is a number of 3 digits or less, a middle point “·” is displayed in place of “0” (or a blank) in the upper digit (in this case, the middle point “·”). Is treated as a number.)

  The character string in the license plate image recognized by the license plate recognition system 1 is preferably composed of characters (particularly numbers), but is not strictly limited to characters (including numbers). It may include symbols and simple figures. Further, the number of characters in such a character string is not limited to four (here, four digits), and may be at least plural. Further, the characters in such a character string are not limited to those arranged in a single line such as a multi-digit number, and constitute at least a character group, relative to the position in the license plate image or between the characters. Any arrangement can be used as long as the arrangement order can be substantially set according to the position.

  The imaging device 2 is composed of a known digital still camera equipped with an image sensor such as a CCD or a digital video camera, and acquires a license plate image by imaging a vehicle or the like equipped with a license plate. The license plate image is a captured image including a license plate as a main subject, but usually includes a vehicle body other than the license plate, surrounding objects, and the like. As the imaging device 2, for example, a monitoring camera installed at a road or an entrance of a facility, an in-vehicle camera mounted on a vehicle, or the like can be used. The license plate images picked up by the image pickup device 2 are sequentially sent to the license plate recognition device 3, thereby being displayed on the display device 5 and stored in the external storage device 4. Further, the license plate image may be temporarily stored in the memory in the imaging device 2 without being sent directly to the license plate recognition device 3.

  The license plate recognition device 3 is a PC (Personal Computer) equipped with known hardware. As shown in FIG. 2, in the license plate recognition device 3, a processor 21 that comprehensively executes processing for recognizing a character string in a license plate image based on a predetermined control program, and a volatilization functioning as a work area of the processor 21. A random access memory (RAM) 22 and a read only memory (ROM) 23 that stores a control program executed by the processor 21 and data are connected to an input / output bus 25. In addition, the license plate recognition device 3 includes peripheral devices such as an input device 24 such as a keyboard and a mouse, a display device 5 including a monitor such as a liquid crystal, and a storage such as an HDD (Hard Disk Drive) and a flash memory. The external storage device 4 is connected. In the functional block diagram of FIG. 3 to be described later, the function of each part of the license plate recognition device 3 can be realized mainly by the processor 21 executing the control program.

  The connection form of the license plate recognition device 3, the imaging device 2, the external storage device 4, and the display device 5 is not particularly limited. For example, the license plate recognition device 3 may be directly connected by a known communication cable, or LAN ( A network connection form such as a local area network) via wired or wireless communication may be used. The license plate recognition device 3 is not limited to a PC, and can also be realized as a server having the same function, a digital camera, or the like.

  In the external storage device 4, together with the license plate image captured by the imaging device 2, image data of individual characters (hereinafter referred to as “reference characters”) that can form a character string of a license plate to be recognized. It is remembered. In the present embodiment, the reference image includes image data of all the Arabic numerals (here, “1” to “9”) and the midpoint “•” constituting the series designation number as the image data of the reference character. included.

  3 is a functional block diagram of the license plate recognition device 3 shown in FIG. 1, and FIG. 4 is a diagram showing an example of a calculation result by the similarity calculation unit 36 in FIG. 2, and FIG. 5 and FIG. FIG. 7 is a diagram showing an example of extraction of higher candidates for correlation values by the similarity evaluation unit 37 in FIG. 2 and a diagram showing an example of the evaluation value calculation result of each digit. FIG. It is explanatory drawing which shows the modification of an extraction method. In FIG. 7, for each digit, an evaluation value is shown on the vertical axis and a number is shown on the horizontal axis.

  In the license plate recognition device 3 shown in FIG. 3, the standard image generation unit 31 obtains an image of reference characters stored in the external storage device 4, and appropriately combines these reference character images to obtain recognition objects. A reference image of all character strings that can be displayed on the license plate is sequentially generated. In FIG. 3, a fourth digit character string “5775” (more precisely, a hyphen “-” between the upper two digits and the lower two digits is inserted. The same applies hereinafter) is generated as the reference image. In this case, in practice, 9999 reference images from “... 1” to “9999” constituting a series designation number (a character string to be recognized) are sequentially generated.

  Further, the projective transformation matrix acquisition unit 32 acquires the license plate image to be evaluated stored in the external storage device 4 and detects the corners (four corners) of the license plate in the image, whereby the coordinates of these four points are obtained. The plane projection transformation matrix is calculated from The projective conversion processing unit 33 sequentially performs projective conversion on the reference image generated by the reference image generating unit 31 using the planar projective conversion matrix. Thereby, a reference image (hereinafter referred to as “converted image”) that is deformed in the same manner as the license plate of the captured image is obtained. The image deterioration processing unit 34 performs simulated deterioration processing by known image processing such as down-sampling of the converted image and blurring processing (hereinafter referred to as “degraded image”). ) In sequence.

  The similarity calculation unit 36 acquires the license plate image to be evaluated stored in the external storage device 4, and sequentially calculates the similarity between each character string in the image and each deteriorated image. In the present embodiment, the following image correlation coefficient (hereinafter referred to as “correlation value”) is used as the similarity indicating the degree of similarity between images.

  As a result of the similarity calculation processing performed by the similarity calculation unit 36, as shown in FIG. 4, a correlation value for each of the series designation numbers “... 1” to “9999” is obtained. These correlation values are displayed on the display device 5 and stored in the external storage device 4.

  Based on the correlation value for each of the series designation numbers shown in FIG. 4, the similarity evaluation unit 37 is a series in which the order of the correlation values is within a threshold value (here, the top 80) as shown in FIG. Extract the specified number. FIG. 5 shows an example in which a series designation number “5025” in which the correlation value is 80th is extracted from a series designation number “5575” in which the correlation value is maximum. Note that, instead of extracting a predetermined number (here, 80) of serial designation numbers as described above, the similarity evaluation unit 37 extracts serial designation numbers whose correlation values are equal to or greater than a predetermined threshold value. It may be configured to.

  In addition, the similarity evaluation unit 37, for each digit (arrangement order) to be evaluated in the extracted series designation number, the similarity of the series designation number in the reference image in which the number (here, one number) is the same. Find the sum of degrees. Further, the similarity evaluation unit 37 obtains a value obtained by dividing the sum of the similarities by the number of extractions of the series designation number as an evaluation value. More specifically, the similarity evaluation unit 37 relates to each digit (here, the first digit—the fourth digit) for each number (10 numbers including “1”-“9” and “•”). The sum of the similarities of the series designation numbers (that is, the similarity of the fourth digit number) is obtained, and a value obtained by dividing the sum by 80 is set as the evaluation value.

  For example, when the similarity evaluation unit 37 obtains an evaluation value for the number “5” of the fourth digit (indicated by reference numeral 40 in FIG. 5) in the series designation number, the correlation value 0.3993 of the series designation number “5575” is obtained. , Correlation value 0.3575 of series designation number “5725”, correlation value 0.3537 of series designation number “5715”, correlation value 0.3534 of series designation number “5375”,..., Correlation value 0.3007 of series designation number “5175”, A value obtained by dividing the sum of the correlation value 0.3009 of the series designation number “5777” and the correlation value 0.2985 of the series designation number “5025” by 80 is used as the evaluation value.

  In this way, the similarity evaluation unit 37 can obtain evaluation values for the numbers “0” to “9” and “•” of the first digit to the fourth digit, respectively, as shown in FIG. Further, the similarity evaluation unit 37 extracts (determines) the number that takes the maximum value for the evaluation value of each digit in FIG. 6 as a character candidate (first candidate) that can form a character string in the license plate image. In the example of FIG. 6, the number “5” is the maximum evaluation value in the fourth digit, the number “7” is the maximum evaluation value in the third digit, and the number “7” is the maximum evaluation value in the second digit. In the first digit, the number “5” is the maximum evaluation value. Thereby, the similarity evaluation unit 37 determines the series designation number “5775” as the first candidate number of the character string in the license plate image, outputs it to the display device 5 and stores it in the external storage device 4. .

  In the present embodiment, an example in which only the first candidate number is obtained has been shown, but the similarity evaluation unit 37 considers the difference between the evaluation value of the first candidate number and another evaluation value, for example, A plurality of candidate numbers less than or equal to the second candidate number can be determined. Based on the evaluation value of each digit in each digit, the similarity evaluation unit 37 determines one number having the smallest difference from the evaluation value of each digit of the series designation number “5775” as a number constituting the second candidate number. can do. More specifically, in the example of FIG. 7, the evaluation value of “5” having the highest evaluation value (0.135) in the fourth digit and “9” having the second highest evaluation value (0.055). Difference (0.080) between the similar evaluation values in the other digits (for example, the difference between the evaluation values of “7” and “3” (0.085) in the third digit, “7” in the second digit). The difference between the evaluation values of “2” and “2” (0.088), and the difference between the evaluation values of “5” and “3” in the first digit (0.239)). Therefore, the similarity evaluation unit 37 can set “9775” obtained by replacing the fourth digit “9” with the corresponding fourth digit “5” in the series designation number “5775” as the second candidate number. . The similarity evaluation unit 37 may extract further candidate numbers (“5375”, “5725”, etc. in the example of FIG. 7) equal to or less than the third candidate number by the same method as described above. Further, in the present embodiment, the similarity evaluation unit 37 determines all candidates for four-digit numbers, but only a part of the candidates may be determined.

  In the present embodiment, the evaluation value for each digit (that is, one digit) is calculated, but at least a number smaller than the number of characters constituting the character string to be recognized (for example, two digits). It can be set as the structure which calculates an evaluation value with respect to.

  FIG. 8 is a flowchart showing the flow of processing of the license plate recognition device 3 shown in FIG. In the character string recognition processing by the license plate recognition device 3, first, a series designation number n is set to 1 (n = 1) (ST101), and a reference image (STEP) composed of reference characters corresponding to the series designation number ( That is, “... 1” or the like is generated (ST102).

  Thereafter, the reference image generated in step ST102 is projectively transformed to generate a converted image (ST103). Subsequently, the converted image generated in step ST103 is subjected to a degradation process, whereby a degraded image is obtained. Is generated (ST104). Further, the similarity (here, the correlation value) between the deteriorated image generated in step ST104 and the character string in the license plate image to be recognized is calculated (ST105).

  Therefore, it is determined whether the series designation number n is 9999 or less (n ≦ 9999) (that is, whether there is a reference image for which the similarity is to be calculated) (ST106), and the reference for which the similarity is to be calculated. If an image remains (Yes), 1 is added to n (n = n + 1) (ST107), and the process returns to step ST102 to execute the same processing as described above. When the series designation number n finally exceeds 9999 (ST106: No), based on the similarity calculated for each series designation number in step ST105, a similarity evaluation process described in detail later is executed (ST108). This completes the character string recognition process.

  FIG. 9 is a flowchart showing details of step ST108 in FIG. In the similarity evaluation process in step ST108, first, with respect to the similarity calculated in step ST105 in FIG. 9, a character string having a similarity from a larger one (maximum value) to a number within a predetermined threshold (here, , A series designated number) is extracted (ST201). Therefore, the character arrangement order d in the character string is set to 1 (d = 1) (ST202), and the character number s is set to 1 (s = 1) (ST203).

  Here, the character arrangement order d relating to step ST202 is set according to the character arrangement in the license plate image. Here, the digit of the serial designation number (any one of the first digit to the fourth digit). It corresponds to. Even when the character string to be recognized includes characters other than numbers, the arrangement order is determined according to the position of the character in the license plate image or the relative position between the characters. For example, when a character string is displayed in one line in the left-right direction in the license plate image, the arrangement number of the character located on the rightmost side is 1, and 2, 3,. , X is an integer) and an arrangement number can be set.

  The character number s related to step ST203 is a character type positioned in a predetermined character arrangement order, and corresponds to a digit of each digit here. When the character string to be recognized includes characters other than numbers, the character number s can be set based on a known character code, for example.

  Therefore, in the character arrangement order set in step ST202, the similarities of all character strings (characters extracted in ST201) including characters that match the character numbers set in step ST203 are added (ST204). The addition result is sequentially stored in the external storage device 4.

  Thereafter, it is determined whether or not the character number s is equal to or less than Ns (total number of characters) (that is, whether or not there is a character for which the addition of similarity has not been completed for each of all characters in a predetermined arrangement order). (ST206) If there is a character for which the addition of similarity is not completed (Yes), 1 is added to s (s = s + 1) (ST207), and step ST204 is performed again for the next character number. Execute. When the character number s finally exceeds Ns (ST206: No), it is determined whether or not the arrangement order d is 4 or less (d ≦ 4) (that is, the arrangement order d is less than or equal to its maximum value). (ST208).

  Therefore, when d ≦ 4 is satisfied (ST208: Yes), 1 is added to d (d = n + 1) (ST209), and the process returns to step ST203 again to perform the same processing as described above for the next digit (arrangement order). Execute. When the arrangement order d finally exceeds 4 (ST208: No), normalization processing is executed for each of the sums of similarities calculated in step ST204 (ST210), thereby completing the similarity evaluation processing. As the normalization process in step ST210, for example, an operation of dividing each of the sums of similarities calculated in step ST204 by the number of high-order candidates extracted in step ST201 can be executed.

  With such similarity evaluation processing, evaluation values as shown in FIG. 6 are calculated, and are output to the display device 5 and stored in the external storage device 4. Further, as described above, the license plate recognition device 3 can determine at least one character candidate that can form a character string in the license plate image based on the evaluation values, and outputs it to the display device 5. At the same time, it is stored in the external storage device 4. In addition, the license plate recognition device 3 outputs to the display device 5 the upper candidates for the character strings in the license plate image based on the similarity as shown in FIG. 5 together with the character candidates based on the evaluation value. Can do.

(Second Embodiment)
The license plate recognition system 1 according to the second embodiment of the present disclosure has the same configuration as the license plate recognition system 1 according to the first embodiment shown in FIGS. However, in the first embodiment described above, the similarity calculation unit 36 calculates the similarity of the entire character string (serial designation number), whereas in the second embodiment, the similarity calculation unit 36 calculates the character string. Similarity is calculated for a part of (but consisting of at least a plurality of characters). Regarding the second embodiment, matters not particularly mentioned below are the same as those in the first embodiment described above.

  10 and 11 are diagrams each illustrating an example (upper two digits and lower two digits) of a calculation result by the similarity calculation unit 36 of the license plate recognition device 3 according to the second embodiment of the present disclosure. These are figures which show an example of the evaluation value calculation result of each digit by the similarity evaluation part 37 of a license plate recognition apparatus.

  In the second embodiment, the similarity calculation unit 36, as shown in FIGS. 10 and 11, respectively, includes the upper 2 digits (up to “••” − “99”) and the lower 2 digits (“• 1”) of the serial designation number. “-99” and “00”-“09”), the correlation value is calculated.

  Thereafter, the similarity evaluation unit 37 extracts a series designation number (upper two digits) in which the order of the correlation value is within the threshold based on the correlation values shown in FIG. For each digit number to be evaluated in (digits), the sum of similarities is calculated for a series of specified numbers (the upper two digits) that contain the same number, and the value obtained by dividing the sum by the number of extracted series specified numbers is evaluated. Calculate as a value. In this way, the similarity evaluation unit 37 can obtain an evaluation value for each of the upper two digits as shown in FIG. Although the detailed description is omitted, similarly, the similarity evaluation unit 37 can obtain the evaluation value of each number of the lower two digits based on the correlation value shown in FIG.

  As mentioned above, although this indication was explained based on specific embodiments, these embodiments are illustrations to the last, and this indication is not limited by these embodiments. For example, the recognition target by the license plate recognition system is not limited to the license plate of the vehicle, but may be a license plate attached to another device or device. Note that the license plate recognition device, the license plate recognition system, and the license plate recognition method are not necessarily all necessary, and can be appropriately selected as long as they do not depart from the scope of the present disclosure.

  A license plate recognition device, a license plate recognition system, and a license plate recognition method according to the present disclosure are capable of reducing erroneous recognition of a character string in a license plate image of a photographed vehicle or the like, and in a license plate image of a photographed vehicle or the like. It is useful as a license plate recognition device, license plate recognition system, and license plate recognition method for recognizing character strings.

DESCRIPTION OF SYMBOLS 1 License plate recognition system 2 Imaging device 3 License plate recognition device 4 External storage device 5 Display device 21 Processor 24 Input device 25 Input / output bus 31 Reference image generation unit 32 Projection transformation matrix acquisition unit 33 Projection transformation processing unit 34 Image degradation processing unit 36 Similarity Calculation Unit 37 Similarity Evaluation Unit

Claims (9)

A license plate recognition device including a processor that executes processing for recognizing a character string in a photographed license plate image,
The character string is composed of a plurality of characters in which an arrangement order from 1 to x (x is an integer of 2 or more) is set according to a position in the license plate image or a relative position between characters,
The processor is
Calculating the similarity between the character string in the license plate image and a plurality of reference images prepared in advance,
For the reference characters included in each of the plurality of reference images corresponding to 1 or y (where y is an integer of 2 or more smaller than x) characters in the arrangement order to be evaluated in the character string, An evaluation value including the sum of the similarities is calculated for each of the standard images having the same reference character,
A license plate recognition apparatus that determines at least one character candidate that can constitute a character string in the license plate image based on the evaluation value.   2. The license plate recognition apparatus according to claim 1, wherein the processor calculates the evaluation value using only similarities from a larger one of the plurality of similarities to a number within a predetermined threshold. .   3. The license plate recognition apparatus according to claim 1, wherein the y characters are in the arrangement order of being consecutive to each other.   4. The license plate recognition apparatus according to claim 1, wherein the reference character is composed of a single character.   5. The license plate recognition according to claim 4, wherein the processor calculates an evaluation value including a sum of the similarities for each of the standard images having the same reference character with respect to all the arrangement orders. apparatus.   The license plate recognition apparatus according to claim 1, wherein the similarity is a correlation coefficient between the character string in the license plate image and the reference image. The processor determines the character string candidates in the license plate image based on the similarity,
The license plate recognition apparatus according to claim 1, further comprising a display device that displays the at least one character candidate and the character string candidate.   A license plate recognition system comprising: the license plate recognition device according to any one of claims 1 to 7; and an imaging device that captures the license plate image. A license plate recognition method by a license plate recognition device that executes processing for recognizing a character string in a license plate image of a photographed vehicle,
The character string is composed of a plurality of characters in which an arrangement order from 1 to x (x is an integer of 2 or more) is set according to a position in the license plate image or a relative position between characters,
Calculate the similarity between the character string in the license plate image and a reference image prepared in advance,
For the reference characters included in each of the plurality of reference images corresponding to 1 or y (where y is an integer of 2 or more smaller than x) characters in the arrangement order to be evaluated in the character string, An evaluation value including the sum of the similarities is calculated for each of the standard images having the same reference character,
A license plate recognition method, wherein at least one character candidate that can constitute a character string in the license plate image is determined based on the evaluation value.

Images