Character recognition method and character recognition system

Abstract

To provide a character recognition device capable of detecting a trailer ID or a character string that could be a container ID from multiple character strings.SOLUTION: The character recognition device includes: a recognition unit that recognizes at least one character string in a captured image including a trailer picked up by an imaging device; and a selection unit that selects a character string that conforms to the specified notation format out of the character strings recognized by the recognition unit as a candidate of the trailer ID or container ID and selects one that has the similarity between the pre-stored trailer ID or container ID and the master data is equal to or greater than a predetermined master data out of the trailer ID or container ID candidates as the trailer ID or container ID.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] The present disclosure relates to a character recognition method and system. [Background technology]

[0002] Patent Document 1 discloses a license plate recognition device that detects multiple quadrangles of license plate area candidates from an input image and performs character recognition on character areas included in the license plate area candidates. The license plate recognition device selects a license plate area candidate to be output from the multiple detected license plate area candidates based on the character recognition result and quadrangle information of each license plate area candidate. The license plate recognition device outputs information about the selected license plate area candidate. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2008-217347 A Summary of the Invention [Problem to be solved by the invention]

[0004] In recent years, logistics using trailers has become popular, and there is an increasing need to efficiently manage the location of each trailer. Here, a trailer means a towed vehicle towed by a towing vehicle (also called a tractor). A trailer may be loaded with a container for storing various items. Generally, an identification (hereinafter referred to as "ID") related to the identification of the trailer or container is written on a trailer or container, and it is considered that the trailer or container can be efficiently managed by recognizing this ID from a captured image of a vehicle or the like. However, a captured image may also contain various characters, and a successfully recognized character string is not necessarily an ID related to the identification of the trailer or container. Therefore, when an imaged object such as a trailer or container contains multiple character strings of various sizes, colors, and designs, there is a need to detect an ID related to the identification of the trailer or container from the multiple character strings and identify the imaged object. Hereinafter, an ID related to the identification of a trailer is referred to as a "trailer ID" and an ID related to the identification of a container is referred to as a "container ID." In the following, when referring to the trailer ID and container ID without distinction, they may be referred to as "trailer ID / container ID."

[0005] In Patent Document 1, a quadrilateral that satisfies a predetermined condition is determined to be a quadrilateral that is a candidate for a license plate area. However, since there is currently no unified standard for trailer IDs, the position, size, number of characters, etc. are not fixed. Therefore, even if a technology similar to Patent Document 1 is applied, it is difficult to detect the trailer ID from among various character strings written at any position on the imaged object. In addition, there are multiple standards for container IDs, and it is difficult to detect them from among various character strings, just like trailer IDs.

[0006] The present disclosure has been devised in view of the above-mentioned conventional situation, and aims to detect a character string that is considered to be a trailer ID or a container ID from among a plurality of character strings. [Means for solving the problem]

[0007] The present disclosure provides a character recognition method in which a character recognition device recognizes at least one character string from an image captured by an imaging device, including a trailer, and the character recognition device selects, from the recognized character strings, a character string that conforms to a specified notation format as a candidate for a trailer ID / container ID.

[0008] The present disclosure also provides a character recognition method in which a character recognition device recognizes at least one character string from an image including a trailer captured by an imaging device, and the character recognition device selects, from the recognized character strings, a character string whose similarity to pre-stored master data of a trailer ID / container ID is equal to or greater than a predetermined threshold as a trailer ID / container ID.

[0009] The present disclosure also provides a character recognition system including at least one imaging device that photographs a trailer, a recognition unit that recognizes at least one character string from an image including the trailer photographed by the imaging device, and a selection unit that selects, from the character strings recognized by the recognition unit, a character string that conforms to a specified notation format as a candidate for a trailer ID / container ID.

[0010] The present disclosure also provides a character recognition system including at least one imaging device that photographs a trailer, a recognition unit that recognizes at least one character string from an image including the trailer photographed by the imaging device, and a selection unit that selects, from among the trailer ID / container ID candidates, one whose similarity to pre-stored trailer ID / container ID master data is equal to or greater than a predetermined threshold, as the trailer ID / container ID.

[0011] Any combination of the above components, and conversion of the present disclosure into a method, device, system, storage medium, computer program, etc. are also valid aspects of the present disclosure. Effect of the Invention

[0012] According to the present disclosure, it is possible to detect a character string that is considered to be a trailer ID or a container ID from among a plurality of character strings. [Brief description of the drawings]

[0013] [Figure 1] A schematic diagram showing an example of a character string included in a trailer [Diagram 2] A schematic diagram showing an example of a character string included in a trailer [Diagram 3] FIG. 1 is a block diagram showing an example of the configuration of a character recognition system according to an embodiment of the present invention. [Figure 4] A flowchart of a process for selecting a trailer ID / container ID according to the present embodiment. [Diagram 5] FIG. 1 is a table for explaining the selection logic of the trailer ID / container ID according to the present embodiment. [Figure 6] Flowchart of selection process by the master filter according to the present embodiment [Figure 7] FIG. 1 is a table for explaining similarities according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0014] Hereinafter, with reference to the drawings as appropriate, a detailed description will be given of an embodiment specifically disclosing a character recognition device, a character recognition method, and a character recognition system according to the present disclosure. However, more detailed description than necessary may be omitted. For example, detailed description of already well-known matters and duplicated description of substantially identical configurations may be omitted. This is to avoid the following description becoming unnecessarily redundant and to facilitate understanding by those skilled in the art. Note that the attached drawings and the following description are provided to enable those skilled in the art to fully understand the present disclosure, and are not intended to limit the subject matter described in the claims.

[0015] The character string included in the trailer TR will be described with reference to Figures 1 and 2. Figure 1 is a schematic diagram showing an example of the character string included in the trailer TR. Figure 1 is a schematic diagram of the trailer TR as viewed from the front right side.

[0016] The trailer TR is the entire object towed by the tractor LE. Generally, a vehicle towed by a tractor is called a trailer. The trailer may be configured as a set of a container that stores cargo but does not have a mechanism for movement such as wheels, and a dolly on which the container is loaded. The trailer may also be configured as a vehicle in which a portion for storing cargo and a mechanism for movement are integrated. When a tractor tows a set of a container and a dolly, only the dolly may be called a trailer. However, in this specification, for the sake of convenience, the entire object towed by the tractor, that is, both the set of the dolly and the container, and the vehicle equipped with a portion for storing cargo and a mechanism for movement are called a trailer.

[0017] In addition to the trailer ID, the trailer TR may include a container ID, a company name, a telephone number, a number indicating the height of the trailer TR, a uniform resource locator (hereinafter referred to as "URL"), a word, a text, or other character strings. Note that the character strings included in the trailer TR are not limited to the above examples.

[0018] The character strings ST1, ST2, ST3, and ST4 are character strings indicating company names. The character strings ST1, ST2, ST3, and ST4 are character strings that are decorated with italics.

[0019] The character string ST5 is a character string expressed in English words and is written below the character string ST2 indicating the company name.

[0020] The character strings ST6 and ST7 are character strings related to a number for identifying the tractor LE.

[0021] The character string ST8 is a character string indicating the trailer ID. The character string ST8 is written vertically on a corner of the trailer TR (that is, a boundary part of the surface forming the casing of the trailer TR, which is located on the side of the trailer TR and is long in the direction of gravity).

[0022] Fig. 2 is a schematic diagram showing an example of a character string included in the trailer TR. Fig. 2 is a schematic diagram showing the trailer TR as seen from the rear.

[0023] The character string ST9 is a character string that indicates a trailer ID. The character string ST9 is the same as the character string ST8, but written horizontally.

[0024] The character string ST10 is a character string indicating a container ID. The character string ST10 is a character string indicated by English letters and numbers. In this specification, for the sake of simplicity, an example is shown in which both the trailer ID and the container ID are written in one trailer, but only one of the trailer ID and the container ID may be written in the trailer.

[0025] The character string ST11 is a character string indicating a telephone number. The character string ST11 is a character string indicated by numbers.

[0026] The character string ST12 is a character string indicating the license plate number.

[0027] In this way, the trailer TR contains multiple character strings.

[0028] In recent years, with the increase in logistics, there is an increasing need to efficiently manage a large number of trailers TR. For example, at a location where multiple trailers TR gather (e.g., a departure point, a relay point, or a destination point, etc.), it is necessary to identify each trailer TR. However, since a trailer TR has multiple character strings (e.g., character strings ST1, ST2, ST5, ST8), there is a problem that it is difficult to detect the trailer ID / container ID from the multiple character strings.

[0029] Next, a block diagram of the character recognition system 1 according to the present embodiment will be described with reference to Fig. 3. Fig. 3 is a block diagram showing an example of the configuration of the character recognition system 1 according to the present embodiment.

[0030] The character recognition system 1 includes at least an imaging device 10, a database 20, a display device 30, and a character recognition device 40.

[0031] The imaging device 10 is a device that photographs trailers. The imaging device 10 is installed, for example, at the entrance to an area (hereinafter also referred to as a "yard") where cargo transported by tractors and trailers is loaded and unloaded or reloaded, and where trailers pulled by tractors or containers mounted on trailers are changed, and photographs trailers and tractors entering and leaving the yard. The imaging device 10 is configured to have at least a lens (not shown) and an image sensor (not shown) as optical elements. The lens receives light reflected by an object within the angle of view of the area photographed by the imaging device 10, and forms an optical image of the object on the light receiving surface of the image sensor (in other words, the photographing surface). The image sensor is, for example, a Charged Coupled Device (hereinafter referred to as a "CCD") or a Complementary Metal Oxide The image sensor converts an optical image formed on a photographing surface through a lens into an electric signal at a predetermined time (for example, 1 / 30 (second)). For example, when the predetermined time is 1 / 30 (second), the frame rate of the image pickup device 10 is 30 fps. The image pickup device 10 may generate image data (video data) by performing a predetermined signal processing on the electric signal at the above-mentioned predetermined time. The image data is a still image, and the video data is a video. The image pickup device 10 outputs the image data (video data) to the character recognition device 40. Hereinafter, the image data and the video data are referred to as a captured image. Although the example of FIG. 3 shows one image pickup device 10, the character recognition system 1 may be configured to include a plurality of image pickup devices. When the character recognition system is configured to include a plurality of image pickup devices, it is preferable that the plurality of image pickup devices are arranged so as to be able to capture images of the trailer from a plurality of angles.

[0032] The database 20 is configured using a storage medium (for example, a flash memory, a hard disk drive (hereinafter referred to as an "HDD"), or a solid state drive (hereinafter referred to as an "SSD")). The database 20 holds trailer IDs / container IDs. The trailer IDs / container IDs held in the database 20 may be trailer IDs / container IDs written on trailers scheduled to arrive at a departure point, a relay point, a destination point, or the like. The trailer IDs / container IDs held in the database 20 may be trailer IDs / container IDs identified in the past. The trailer IDs / container IDs held in the database 20 may be registered in advance by a user, or trailer IDs / container IDs selected by the character recognition device 40 may be registered by the character recognition device 40. The trailer IDs / container IDs held in the database 20 are collectively treated as master data 21. Hereinafter, the trailer IDs / container IDs held as the master data 21 may be referred to as a correct ID. The database 20 may be incorporated in the character recognition device 40.

[0033] The display device 30 displays the result of the selection of the character string by the character recognition device 40. The display device 30 is a terminal equipped with a display unit, such as a personal computer (hereinafter referred to as "PC"), a tablet, or a mobile phone. The display device 30 may be configured integrally with the character recognition device 40. Note that there may be a plurality of display devices 30. When the character recognition system 1 includes a plurality of display devices 30, the form and user of the display devices 30 may be different. For example, the result of the selection of the character string may be displayed on a PC or the like placed in an office and a tablet or the like carried by a worker. In this case, each display device 30 may change the details of the information to be displayed depending on the performance of each display device 30 or the work required of the user of each display device 30. In addition, in this embodiment, the display device 30 may not be provided. This is because when the character recognition device 40 saves the selected trailer ID / container ID or transfers it to another system, it is not necessarily necessary for the user to visually check the result of the selection of the character string.

[0034] Character recognition device 40 includes at least processor 41, communication I / F 42, and memory 43. Character recognition device 40 is, for example, a computing device such as a PC. Note that, although an example in which character recognition device 40 is configured as one computing device will be described in this specification, character recognition device 40 may be realized as a plurality of devices working together to realize its functions, or may be realized as a cloud.

[0035] The processor 41 is configured using, for example, a central processing unit (hereinafter referred to as "CPU"), a digital signal processor (hereinafter referred to as "DSP"), or a field programmable gate array (hereinafter referred to as "FPGA"). The processor 41 functions as a controller that manages the overall operation of the character recognition device 40. The processor 41 performs control processing for managing the operation of each part of the character recognition device 40, input / output processing of data between each part of the character recognition device 40, calculation processing of data, and storage processing of data. The processor 41 operates according to a program stored in the memory 43. The processor 41 uses the memory 43 during operation, and temporarily stores data generated or acquired by the processor 41 in the memory 43. The processor 41 realizes the functions of the character string detection unit 41A, the character string recognition unit 41B, and the character string selection unit 41C by using the programs and data stored in the memory 43.

[0036] The character string detection unit 41A, which is an example of a detection unit, detects a character string from a captured image acquired from the imaging device 10. The character string detection unit 41A detects a character string using a known technique such as machine learning.

[0037] The character string recognition unit 41B, which is an example of a recognition unit, performs character recognition of the character string detected by the character string detection unit 41A. The character string is recognized by using a known technique such as Optical Character Recognition (hereinafter referred to as "OCR"). In the description of the present embodiment, character recognition is performed by OCR.

[0038] The character string selection unit 41C, which is an example of a selection unit, selects a trailer ID / container ID from one or more character strings recognized by the character string recognition unit 41B. The character string selection unit 41C selects a trailer ID / container ID by applying various processes and filters described below. Details of the various processes and filters used for selection will be described below. The character string selection unit 41C outputs the selected trailer ID / container ID to the communication I / F 42.

[0039] The communication I / F 42 is an interface circuit for the character recognition device 40 to communicate wirelessly or wired with the imaging device 10, the database 20, and the display device 30. Here, I / F stands for interface. The communication I / F 42 may communicate via a network. The communication method by the communication I / F 42 is, for example, a Wide Area Network (hereinafter referred to as "WAN"), a Local Area Network (hereinafter referred to as "LAN"), Long Term Evolution (hereinafter referred to as "LTE"), mobile communications such as 4G and 5G, power line communications, short-range wireless communications (e.g., Bluetooth (registered trademark) communications), or communications for mobile phones.

[0040] Memory 43 is configured using, for example, Random Access Memory (hereinafter referred to as "RAM") and Read Only Memory (hereinafter referred to as "ROM"), and temporarily stores programs necessary for the operation of character recognition device 40, as well as data generated during operation. The RAM is, for example, a work memory used during operation of character recognition device 40. The ROM stores and holds, for example, programs for controlling character recognition device 40 in advance.

[0041] Next, a flowchart of the process executed by the character recognition device 40 will be described with reference to Fig. 4. Fig. 4 is a flowchart of the process executed by the character recognition device 40. Each process of the flowchart in Fig. 4 is executed by the processor 41. At the start of this flowchart, the processor 41 has acquired a captured image from the imaging device 10. For ease of explanation, it is assumed that the captured image includes a plurality of character strings.

[0042] When the captured image acquired from the imaging device 10 includes an image that is likely to be a character string, the character string detection unit 41A detects at least one character string from the captured image, and the character string recognition unit 41B performs character recognition of the character string (step St100). For convenience of explanation, it is assumed that there are multiple recognized character strings. The character detection unit 41A may detect a trailer or container area from the captured image and detect a character string from that area. The captured image may include a character string unrelated to a trailer ID / container ID such as a sign, or a shadow that looks like a character string. Therefore, the character detection unit 41A can prevent a character string unrelated to a trailer ID / container ID and a shadow that looks like a character string from being detected as a candidate for a trailer ID / container ID by limiting the area in which the character string is detected to an area estimated to be a trailer or container. The function of detecting the trailer or container area from the captured image can be realized by pattern matching or by using a trained model that has been machine-learned to learn an image of a trailer or container.

[0043] The character string recognition unit 41B performs correction on the character string recognized in the process of step St100 (step St101). The content of the correction will be described later.

[0044] The character string selection unit 41C applies an exclusion filter to the character string corrected in the process of step St101 and the character string before the correction (step St102). As a result, character strings having specific characteristics are excluded (deleted) from the multiple character strings recognized in step St100. The contents of the exclusion filter will be described later. For convenience of explanation, it is assumed that there are multiple character strings after the exclusion filter is applied.

[0045] The character string selection unit 41C applies a priority filter to the character strings that are not excluded in the process of step St102 (step St103). As a result, the character string selection unit 41C selects, from among the recognized character strings, character strings that conform to a specified notation format as candidates for trailer IDs / container IDs. The contents of the priority filter will be described later. For the sake of convenience, it is assumed that there are multiple candidates for trailer IDs / container IDs.

[0046] The character string selection unit 41C performs mode selection on the character string selected in the process of step St103, and further selects candidates for the trailer ID / container ID (step St104). The content of the mode selection will be described later. For the sake of convenience of explanation, it is assumed that there are multiple candidates for the trailer ID / container ID after the mode selection.

[0047] The character string selection unit 41C performs selection by the master filter (step St105). The process of step St105 will be described later with reference to the flowchart shown in Fig. 6. For convenience of explanation, it is assumed that there are multiple candidates for the trailer ID / container ID after selection by the master filter.

[0048] The character string selection unit 41C executes remaining data selection for the character string selected in the process of step St105, and finally selects one trailer ID / container ID (step St106). As a result, one trailer ID / container ID related to the identification of the photographed trailer is selected. The contents of the remaining data selection will be described later.

[0049] The processor 41 outputs the result of step St106 to the display device 30 (step St107). Note that the process of step St107 may be omitted.

[0050] Next, each process and each filter that appeared in the flowchart of Fig. 4 will be described with reference to Fig. 5. Fig. 5 is a table diagram for explaining the selection logic of the trailer ID / container ID according to this embodiment.

[0051] The content of the correction process executed by the character string recognition unit 41B on the character string in step St101 of the flowchart in Fig. 4 will be described. The correction process is a process for correcting the character string recognized from the captured image into a form that is easy to handle in each subsequent process. As the correction process, for example, it is possible to exclude characters that are easily misrecognized, such as hyphens, and to replace characters that are easily misrecognized as each other, such as "1" and "I".

[0052] When a hyphen is included in the recognized character string by the correction process, the character string recognition unit 41B deletes the hyphen. A hyphen is a character that is easily misrecognized as dirt on a trailer. In addition, in a system such as a management system that uses a trailer ID / container ID, it is rare to distinguish between a trailer ID / container ID that includes a hyphen and a trailer ID / container ID that does not include a hyphen. Therefore, in this specification, it is assumed that the trailer ID / container ID is managed in a format that does not include a hyphen. However, in other systems that use the recognized trailer ID / container ID, for example, when a trailer ID / container ID in a format that includes a hyphen is used, it may be better not to delete the hyphen. Therefore, whether or not to execute the process of deleting the hyphen may be switchable by the user.

[0053] When the character string recognition unit 41B recognizes a character or a number at the edge of the captured image by the correction process, the character or the number is deleted. This is because characters may be cut off at the edge of the captured image, and an accurate recognition result may not be obtained. For example, if only the left half of "8" is included in the captured image, "8" may be erroneously recognized as "E". Also, if only the right half of "8" is included in the captured image, "8" may be erroneously recognized as "3". Note that the width (in other words, the area) of the edge of the captured image may be changeable according to a user's designation. Also, when a character or a number at the edge of the captured image is recognized, the entire character string including the character or the number may be deleted. As described above, characters recognized from the edge of the captured image may be erroneously recognized, and therefore, even if a character string including such characters is used as a candidate for a trailer ID / container ID, an accurate trailer ID / container ID may not be obtained.

[0054] When the recognized character string includes a character or a number that is easily misrecognized by the correction process, the character string recognition unit 41B converts the character or the number. For example, when the recognized character string includes a number "1", the character string recognition unit 41B determines that the number "1" is easily misrecognized as the English letter "I" and converts the number "1" to the English letter "I". The criterion for determining whether or not a character is easily misrecognized may be preset by, for example, a user. When a part of the recognized character string that is specified as an English letter part in a description format described later is a specified number, the character string recognition unit 41B may convert the number to a specified English letter. When a part of the recognized character string that is specified as a number part in a description format described later is a specified English letter, the character string recognition unit 41B may convert the English letter to a specified number. When the processor 41 performs the conversion process, the processor 41 leaves the character string before conversion as it is and adds the converted character string as a newly recognized character string. That is, the character string recognition unit 41B leaves both the character string before and after conversion as candidates for the trailer ID / container ID. As a result, the character string selection unit 41C selects the trailer ID / container ID from the character string after conversion and the character string before and after conversion. This is because, when the trailer ID / container ID is written in an unknown description format, the character string after conversion that is made to match the known description format may not be the correct trailer ID / container ID.

[0055] The contents of the exclusion filter applied to the character string by the character string selection unit 41C in step St102 of the flowchart in Fig. 4 will be described. The exclusion filter is a filter for excluding character strings having a pattern that is not assumed to be a trailer ID / container ID from candidates. The exclusion filter excludes, for example, character strings for which the recognition result by OCR itself is unreliable, character strings containing character strings unlikely to be included in a trailer ID / container ID, character strings with an extremely short number of characters, etc. An example of a character string unlikely to be included in a trailer ID / container ID is, for example, a character string defined by a different standard from that of a trailer ID / container ID.

[0056] The character string selection unit 41C applies an exclusion filter to exclude, from among the recognized character strings, character strings whose OCR reliability score is equal to or less than a predetermined threshold. The reliability score is a score that indicates the accuracy of a character string recognized by OCR. The threshold may be set in advance by, for example, a user. The character string selection unit 41C applies an exclusion filter to exclude, from among the recognized character strings, character strings that are all English characters from among the candidates for the trailer ID / container ID. The character string selection unit 41C applies an exclusion filter to exclude, from among the recognized character strings, character strings whose prefixes are specific English characters from among the candidates for the trailer ID / container ID. In this specification, a prefix refers to an English character string at the beginning of a character string. For example, the prefix of the character string "AAA1234" is "AAA". For example, if it is known in advance that a character string having a prefix such as "USDOT" or "KYU" cannot be a trailer ID / container ID, the user can set the character string having the prefix to be excluded. In this example, trailer ID / container ID candidates are excluded based on the prefix, because it is known that a character string whose prefix is ​​a specific English character is not a trailer ID / container ID. For example, a character string whose prefix is ​​"USDOT" is known to be a format for notating numbers for audit purposes by the United States Department of Transportation. However, the configuration of excluding trailer ID / container ID candidates based on the prefix is ​​only one example, and trailer ID / container ID candidates may be excluded based on other character strings. For example, if it is known that a specific number or a character string included in the latter half of a string is not used in trailer IDs / container IDs, an exclusion filter may be configured to exclude candidates in which such a number or character string appears.

[0057] The character string selection unit 41C applies an exclusion filter to exclude from the candidates for the trailer ID / container ID, character strings that do not have a predetermined number of characters among the recognized character strings. The predetermined number of characters can be, for example, three characters. This is because character strings that are less than three characters often indicate the dimensions of a container, and are often not suitable as candidates for the trailer ID / container ID. In addition, the character string selection unit 41C applies an exclusion filter to exclude from the candidates for the trailer ID / container ID, character strings that are not English letters or numbers but are erroneously recognized. This is because trailer IDs / container IDs are often composed of character strings that can be read without question in multiple countries, and it is unlikely that characters or figures that are used in limited areas, such as hiragana, will be used.

[0058] The contents of the priority filter that the character string selection unit 41C applies to the character string in step St103 of the flowchart in Fig. 4 will be described. The priority filter is a filter that selects a character string that is highly likely to be a trailer ID / container ID. For example, the priority filter selects a character string that matches a notation format known as a trailer ID / container ID.

[0059] The character string selection unit 41C applies a priority filter to select, from among the recognized character strings, a character string that matches a specified notation format as a candidate for the trailer ID / container ID. The notation format specifies, for example, the English character portion and the number portion of the character string. For example, the character string selection unit 41C applies a priority filter to select, as a candidate for the trailer ID / container ID, a character string whose first two characters are English characters and whose English characters are followed by numbers. For example, the character string selection unit 41C applies a priority filter to select, as a candidate for the trailer ID / container ID, a character string whose first four characters are English characters and whose English characters are followed by numbers. In this embodiment, it is assumed that the trailer ID and the container ID are managed without distinction, so the character string selection unit 41C selects, as a candidate for the trailer ID / container ID, both when the character string matches the notation format of the trailer ID and when the character string matches the notation format of the container ID. However, if it is necessary to distinguish between trailer IDs and container IDs, they may be treated as candidates for IDs that match the respective notation formats. For example, a character string whose first two characters are alphabetic characters followed by a number may be treated as a candidate for a trailer ID, and a character string whose first four characters are alphabetic characters followed by a number may be selected as a candidate for a container ID.

[0060] In addition, when trailer IDs and container IDs are handled separately, a discriminator created using machine learning may be used to distinguish whether a character string is a trailer ID or a container ID. The discriminator can be created, for example, by performing machine learning on a large amount of data by distinguishing between examples of character strings indicating trailer IDs and examples of character strings indicating container IDs. Since various machine learning methods are known, detailed explanations are omitted in this embodiment. Note that the character strings indicating trailer IDs and character strings indicating container IDs to be machine-learned may be the results manually input by a user as trailer IDs or container IDs into a system for managing trailers, etc., or may be the results of selection based on the above-mentioned notation format.

[0061] In addition, when the trailer ID and the container ID are handled separately, the estimation result of the type of the vehicle towed by the tractor may be used to distinguish whether the character string is a trailer ID or a container ID. For example, when the vehicle towed by the tractor is a combination of a container and a dolly, the detected character string is treated as a container ID. On the other hand, when the vehicle towed by the tractor is a vehicle in which the part for storing cargo and the wheels are integrated, the detected character string is treated as a trailer ID. For example, a classifier created using machine learning is used to estimate the type of the vehicle towed by the tractor. More specifically, it is considered to perform machine learning by distinguishing between an image of a container and an image of a vehicle in which the part for storing cargo and the wheels are integrated. Since various machine learning methods are known, detailed explanations are omitted in this embodiment. In this case, in order to narrow down the area to be processed by the classifier, a process of detecting the tractor from the image may be performed in advance, and the area of ​​the vehicle towed by the tractor may be estimated based on the detection result. The process of detecting the tractor from the image can be realized, for example, by using a trained model in which an image of a tractor is machine-learned. If the vehicle towed by the tractor is a combination of a container and a dolly, it may be further determined whether the character string is treated as a container ID or a trailer ID depending on whether the character string is written on the container or on the dolly. For example, if the character string is written on the container, it may be treated as a container ID, and if the character string is written on the dolly, it may be treated as a trailer ID.

[0062] As described in the explanation of the correction, the character string recognition unit 41B may execute the correction process based on the notation format. For example, when the notation format specifies that the first four characters of a character string are alphabetic characters and that the alphabetic characters are followed by numbers, the character string recognition unit 41B may execute the correction process based on the notation format. For example, when the character string recognition unit 41B recognizes a character string "HG1U21456", the character string recognition unit 41B converts the "1" in "HG1U" to "I" based on the rule that the first four characters of a character string are alphabetic characters. In addition, the character string recognition unit 41B does not convert the "1" in "21456" based on the rule that an alphabetic character is followed by a number. As a result, the character string "HGIU21456" is newly added to the recognized character string.

[0063] The character string selection unit 41C may select, as a trailer ID / container ID candidate, a character string having a character height higher than a predetermined height among the recognized character strings by applying the priority filter. Here, the character height means the vertical length of the character. Generally, the higher the character height, the higher the possibility that the character is larger, and the larger the character, the higher the possibility that the character is accurately recognized by OCR. Therefore, by selecting a character string having a high character height as a trailer ID / container ID candidate, it is possible to select a character string having a high possibility of being correctly recognized as a candidate. In addition, when a plurality of trailer ID / container ID candidates remain as a result of applying the priority filter, a character string having a high character height may be adopted as a candidate having a higher possibility of being a trailer ID / container ID than the other candidates. The reason for evaluating the character height rather than the character size is that the trailer ID / container ID is often written as a vertically long character string, or the captured image in this embodiment is captured from the front or rear of the trailer in the traveling direction, and therefore the character width may not be accurately detected. In addition, when the trailer is captured from another angle, and the character width can be accurately detected, the character size may be used instead of the character height.

[0064] Furthermore, the character string selection unit 41C may select vertically written character strings as candidates for trailer IDs / container IDs from among the recognized character strings by applying a priority filter. This is because, in general, character strings that are not trailer IDs / container IDs, such as telephone numbers and URLs, are likely to be written as horizontally written character strings, and vertically written character strings are likely to be candidates for trailer IDs / container IDs.

[0065] Further, the trailer ID / container ID candidate selected by the character string selection unit 41C through application of the priority filter is either one of the trailer ID candidate and the container ID candidate. This is because if either the trailer ID or the container ID can be identified, the trailer or container entering or leaving the yard can be uniquely managed, and also because in many cases only one of the trailer ID and the container ID is written on the trailer. The character string selection unit 41C may determine whether the character string to be finally selected is the trailer ID or the container ID at the time of application of the priority filter. In the process after application of the priority filter, i.e., in the process from step St104 onward in the flowchart of FIG. 4, either one of the trailer ID candidate and the container ID candidate may be the processing target.

[0066] The content of the mode selection performed by the character string selection unit 41C on the character strings in step St104 of the flowchart in FIG. 4 will be described. The mode selection is a process of selecting the character string that is most frequently recognized among the candidates for the trailer ID / container ID as the trailer ID / container ID. Since the imaging device 10 is set to an angle of view suitable for photographing the trailer ID / container ID, the correct trailer ID / container ID is highly likely to be recognized as the same character string from images captured at multiple imaging timings. However, if the character string is erroneously recognized, the recognition result is likely to be unstable, and therefore the same character string is less likely to be recognized from images captured at multiple imaging timings. In other words, if the same character string is recognized from images captured at multiple timings, the character string is highly likely to be the correct trailer ID / container ID. Therefore, the mode selection utilizes this property to select the correct trailer ID / container ID.

[0067] The character string recognized by the character string recognition unit 41B is included in the image captured by the imaging device 10. When the imaging device 10 captures an image of a moving trailer, the imaging device 10 captures images of the trailer from, for example, the front, the side, and the rear. At this time, the same character string may be included in each of the images captured by the imaging device 10 of the trailer from the front, the side, and the rear. Alternatively, the same character string may be included multiple times in one captured image. In this case, the character string recognition unit 41B recognizes the character string multiple times.

[0068] The character recognition system 1 may include multiple imaging devices including the imaging device 10. The multiple imaging devices may be arranged so as to be able to capture images of the trailer from multiple angles, and the character string recognition unit 41B may recognize a character string from multiple captured images of the trailer captured from multiple angles. In this case as well, the same character string may be included multiple times in one or multiple captured images. Therefore, the character string recognition unit 41B may recognize the character string multiple times.

[0069] The character string selection unit 41C selects a trailer ID / container ID from among the trailer ID / container ID candidates based on the number of times the trailer ID / container ID candidate is recognized from one or more captured images by mode selection. In other words, the character string selection unit 41C selects a trailer ID / container ID based on the frequency with which a character string that is a candidate for the trailer ID / container ID appears in one or more captured images captured by one or more imaging devices. Hereinafter, the number of times a character string is recognized by the character string recognition unit 41B may be referred to as the count number.

[0070] For example, the character string selection unit 41C selects a trailer ID / container ID candidate with the largest count as a trailer ID / container ID. For example, the character string selection unit 41C selects a trailer ID / container ID candidate with a count equal to or greater than a predetermined threshold as a trailer ID / container ID. The threshold may be preset by a user. For example, when there are multiple trailer ID / container ID candidates with the same count, the character string selection unit 41C may preferentially select a character string with a larger number of characters. In addition, when performing the mode selection, the character string selection unit 41C may calculate the count including a trailer ID / container ID candidate that was not selected by application of any of the criteria of the priority filter. This is because, for example, when only vertically written character strings are selected by application of the priority filter, horizontally written character strings may be excluded from the trailer ID / container ID candidates, and the correct trailer ID / container ID may not remain among the candidates.

[0071] The character string selection unit 41C can narrow down the trailer ID / container ID candidates to one by the mode selection. When the trailer ID / container ID cannot be narrowed down to one by the mode selection, the character string selection unit 41C executes the selection by the "master filter" and / or the "remaining data selection" described later. A specific example of the case where the trailer ID / container ID cannot be narrowed down to one by the mode selection is when there are multiple trailer ID / container ID candidates with the same count number. Also, when there are multiple trailer ID / container ID candidates with a small difference in the count number by the mode selection, the selection by the "master filter" and / or the "remaining data selection" described later may be executed. This is because when the difference in the count number is small, there is no clear superiority or inferiority between the trailer ID / container ID candidates from the viewpoint of the mode selection, and the result of narrowing down to one may be incorrect. Also, even if the count number of the most frequent value selection is small, the selection by the "master filter" and / or "remaining data selection" described later may be executed. This is because in this case, the information on which the most frequent value evaluation is based is scarce, and the result of the most frequent value selection itself is not sufficiently reliable.

[0072] In the above explanation, for simplicity, an example is given in which the candidate with the highest count is selected as the trailer ID / container ID. However, it is also possible to evaluate the frequency with which the trailer ID / container ID is recognized based on the count within a certain period of time, and select the most frequently recognized candidate as the trailer ID / container ID.

[0073] The content of the remaining data selection executed by the character string selection unit 41C for the character string in step St106 of the flowchart in Fig. 4 will be described. "Remaining data selection" is a process for selecting a trailer ID / container ID when the trailer ID / container ID has not been narrowed down even after the processes up to step St105 have been performed. In the remaining data selection, the trailer ID / container ID is selected based on, for example, the number of characters. This is because, since a large number of trailers or containers are handled, the trailer ID / container ID used to distinguish each trailer or container is likely to be a relatively long character string to avoid duplication.

[0074] The character string selection unit 41C selects the character string with the largest number of characters as the trailer ID / container ID from among the trailer ID / container ID candidates by the remaining data selection. The character string selection unit 41C executes the remaining data selection when the trailer ID / container ID cannot be narrowed down to one by other processes and selection by the filter. The remaining data selection has less grounds for trusting the result than the processes up to step St105. Therefore, the character string selection unit 41C may output, for example, information indicating that the trailer ID / container ID could not be confirmed as the remaining data process in the remaining data selection. In this case, the character string with the largest number of characters may be output as the trailer ID / container ID, or the trailer ID / container ID itself may not be output. In this way, it is possible to inform the user or an external system that a trailer ID / container ID with high reliability could not be selected, and therefore it is possible to encourage confirmation of the trailer ID / container ID by visual inspection, etc.

[0075] The contents of the master filter applied to the character string by the character string selection unit 41C in step St105 of the flowchart in FIG. 4 will be described. The master filter is a filter that determines whether a candidate for a trailer ID / container ID is a trailer ID / container ID based on whether the candidate resembles the master data 21 (correct ID) that is held as a character string known to be a trailer ID / container ID. As an example, the master data 21 is a character string that has been recorded as a trailer ID / container ID in the past. The master data 21 does not necessarily have to be recognized by OCR, and may be a character string manually input by an operator in a yard where recognition of trailer IDs / container IDs is not automated. In addition, when information on trailer IDs / container IDs scheduled to enter or leave the yard can be obtained from a higher-level system that manages the entire logistics, this information may be used as the master data 21. Note that the master data 21 does not necessarily have to include information that matches the trailer ID / container ID written on the trailer itself that enters or leaves the yard. This is because trailer IDs / container IDs tend to have similar notation formats to a certain extent, so by evaluating the similarity between a trailer ID / container ID candidate and the master data 21, it is possible to infer the possibility that the candidate is a trailer ID / container ID even if it does not completely match the master data 21. For example, if a trailer ID / container ID is composed of a combination of a transport company name and numbers, the transport company name may be the same even if the numbers are different between the master data 21 and the candidate. In this case, the similarity is evaluated to be high even if the trailer ID / container ID candidate and the master data 21 do not completely match, so it is possible to determine that the trailer ID / container ID candidate is highly likely to be a trailer ID / container ID.

[0076] The character string selection unit 41C selects, as a trailer ID / container ID, a trailer ID / container ID candidate whose similarity between the trailer ID / container ID candidate and the pre-stored trailer ID / container ID master data 21 is equal to or greater than a predetermined threshold value. Here, the similarity is a numerical value indicating how similar character strings (e.g., a trailer ID and a trailer ID in the master data 21) are. The similarity is calculated, for example, based on the Levenshtein distance between the character strings. However, the similarity of character strings may be evaluated based on other methods. For example, it may be evaluated based on the number of matching characters or the ratio of matching characters in the entire character string. Details of the selection process by the master filter will be described later with reference to FIG. 6.

[0077] The character string selection unit 41C can narrow down the trailer ID / container ID candidates to one by executing selection using the master filter. If the trailer ID / container ID cannot be narrowed down to one by the selection using the master filter, the character string selection unit 41C executes selection using the "remaining data selection" described above.

[0078] Next, the process of selecting trailer ID / container ID candidates by the master filter will be described with reference to Fig. 6. Fig. 6 is a flowchart of the selection process by the master filter according to this embodiment. Each process in the flowchart of Fig. 6 is executed by the processor 41. At the start of the flowchart of Fig. 6, the process of step St104 in the flowchart of Fig. 4 has been completed. At this time, the description will be given on the assumption that multiple trailer ID / container ID candidates remain.

[0079] The character string selection unit 41C acquires candidates for the trailer ID / container ID (step St200). Here, acquiring means focusing on the candidates for handling in each process described later.

[0080] The character string selection unit 41C searches the master data 21 to search for a trailer ID / container ID that matches any of the trailer ID / container ID candidates acquired in step St200 (step St201).

[0081] If a trailer ID / container ID that matches any of the trailer ID / container ID candidates acquired in step St200 is held as the master data 21 (step St202; YES), the character string selection unit 41C adopts (selects) the trailer ID / container ID (step St203). Then, the character string selection unit 41C ends this processing flow.

[0082] If there is no trailer ID / container ID held as master data 21 that matches any of the trailer ID / container ID candidates acquired in step St200 (step St202; NO), the string selection unit 41C acquires all trailer IDs / container IDs that are previously held as master data 21, which are called correct IDs (step St204).

[0083] The character string selection unit 41C focuses on one of the trailer ID / container ID candidates for which the calculation of the similarity with the correct ID has not been completed (step St205).

[0084] The character string selection unit 41C calculates the similarity between the trailer ID / container ID candidate focused on in step St205 and each of the correct IDs acquired in step St204 (step St206). For example, if there are 100 correct IDs acquired in step St204, the character string selection unit 41C calculates the similarity between the trailer ID / container ID candidate and each of the 100 correct IDs. In this case, the character string selection unit 41C calculates the 100 similarities.

[0085] The character string selection unit 41C retains the correct ID having the highest similarity to the trailer ID / container ID candidate calculated in step St206, among the correct IDs acquired in step St204, and the similarity (step St207). Retaining means that the character string selection unit 41C associates and stores the trailer ID / container ID candidate, the correct ID having the highest similarity to the trailer ID / container ID candidate, and the similarity.

[0086] If any trailer ID / container ID candidates remain for which the calculation of similarity with the correct ID has not been completed (step St208; YES), the processor 41 returns to step St205 and repeats the process. This causes the similarity of each of all trailer ID / container ID candidates to each of all correct IDs to be calculated. For example, if there are 10 trailer ID / container ID candidates, the character string selection unit 41C repeats the series of processes of steps St205, St206, and St207 10 times. At this time, the character string selection unit 41C holds the correct ID with the maximum similarity with respect to each of the 10 trailer ID / container ID candidates, and the similarity.

[0087] A specific example will be described with reference to FIG. 7. FIG. 7 is a table diagram for explaining the similarity according to the present embodiment. In the example of FIG. 7, the trailer ID / container ID candidates are six, namely, "UC15Y", "2346", "UC15YARD2", "7FHV188", "1D2346" and "5YARD". There may be, for example, 10 or 100 correct IDs. In the example of FIG. 7, the processor 41 calculates the similarity between the trailer ID / container ID candidate "UC15Y" and each of the correct IDs, and holds the correct ID "UC15B6" that has been calculated to have the highest similarity and the similarity of 0.67 at that time. Similarly, the character string selection unit 41C holds the correct ID that has the highest similarity to each of the other five trailer ID / container ID candidates and the similarity at that time.

[0088] If there are no trailer ID / container ID candidates whose similarity with the correct ID has not yet been calculated (step St208; NO), the character string selection unit 41C judges whether the maximum similarity among the similarities held in step St207 is equal to or greater than a predetermined threshold (step St209). For example, if there are 10 trailer ID / container ID candidates, 10 similarities are also held. At this time, the character string selection unit 41C judges whether the maximum similarity among the 10 similarities is equal to or greater than a predetermined threshold. The threshold may be set in advance by the user, for example.

[0089] In the example of FIG. 7, character string selection unit 41C determines whether or not 0.67, which is the maximum similarity among the six similarities held, is equal to or greater than a predetermined threshold value.

[0090] If the maximum similarity among the similarities held in step St207 is equal to or greater than a predetermined threshold (step St209; YES), the character string selection unit 41C adopts the trailer ID / container ID candidate associated with the maximum similarity (step St210). Then, this processing flow ends.

[0091] 7, the maximum similarity is 0.67. For example, if the predetermined threshold is 0.6, the character string selection unit 41C determines that the maximum similarity is equal to or greater than the predetermined threshold. In this case, the character string selection unit 41C adopts the trailer ID / container ID candidate "UC15Y" associated with the similarity (0.67).

[0092] If the maximum similarity among the similarities held in step St207 is not equal to or greater than a predetermined threshold (step St209; NO), the character string selection unit 41C does not select a trailer ID / container ID in this processing flow (step St211). Then, this processing flow is terminated.

[0093] 7, if the predetermined threshold is 0.7, the maximum similarity of 0.67 is not equal to or greater than the threshold. In this case, the character string selection unit 41C does not select a trailer ID / container ID by the master filter, and ends this processing flow.

[0094] (Other variations) In the above-described embodiment, the processor 41 of the character recognition device 40 executes each process in the flowchart of FIG. 4. However, the present invention is not limited to this. For example, each process in the flowchart of FIG. 4 may be distributed to multiple PCs. For example, a PC installed at the starting point, relay point, or destination point of the trailer may execute the process up to step St102 in the flowchart of FIG. 4, and the process from step St103 onward may be executed on a cloud in a server of a company or the like. For example, each process in the flowchart of FIG. 4 may be consolidated and executed on a cloud. Furthermore, the database 20 stored in the character recognition device 40 may be a database on a cloud. In particular, when the correct ID is collected from the recognition results in multiple yards, implementing the database 20 as a database on a cloud connected to multiple character recognition devices 40 makes it easier to manage data. Furthermore, the process of outputting the result to a display device (for example, the display device 30) may be replaced with a process of transmitting the selection result of the trailer ID / container ID to an external system, or with a process of outputting the result to the display device and the external system. This is because the selected trailer ID / container ID can also be used for managing trailers, containers, or cargo, and can therefore be used by systems that optimize logistics networks or operators of such systems.

[0095] In the above-described embodiment, an example has been shown in which the character recognition device 40 executes various processes such as character recognition from an image captured by the imaging device 10. However, this is not limited to this, and for example, an ID recognition processor having a character recognition function or the like may be incorporated in the imaging device 10, and the imaging device 10 may execute various processes such as character recognition and selection of a trailer ID / container ID.

[0096] In the above-described embodiment, the character string selection unit 41C determines whether the character string to be finally selected is a trailer ID or a container ID at the time of applying the priority filter. However, the present invention is not limited to this, and the character string selection unit 41C can obtain both the result when the trailer ID is selected and the result when the container ID is selected. That is, the selection of the trailer ID and the selection of the container ID may be performed in parallel at the same time. For example, for a user such as a trailer manager, it may be sufficient to select either the trailer ID or the container ID for the purpose of managing the trailer. However, even in such a case, the user can obtain the selection result of the trailer ID and the selection result of the container ID, respectively.

[0097] In the above-described embodiment, the character string selection unit 41C performs the selection by the priority filter, followed by the selection by the mode, the selection by the master filter, and the selection by the remaining data in this order. However, the present invention is not limited to this. If the trailer ID / container ID can be narrowed down to one by the selection by the mode, the character string selection unit 41C may end the selection process without performing the selection by the master filter and the selection by the remaining data. Similarly, if the trailer ID / container ID can be narrowed down to one by the selection by the master filter, the character string selection unit 41C may end the selection process without performing the selection by the remaining data. In addition, the character string selection unit 41C may, for example, change the order of the selection by the master filter, the selection by the mode, and the selection by the remaining data after the selection by the priority filter. In addition, the order of the processes after the detection of the character string may be arbitrarily changed. For example, it is conceivable to move step St101 to immediately before step St105, and apply the corrected character string only to the selection by the master filter. The execution order may be set in advance by the user, for example. For example, the character string selection unit 41C may select one trailer ID / container ID by selecting remaining data after the selection by the priority filter. Also, in the above-mentioned embodiment, any of the processes performed by the character string selection unit 41C may be omitted, or another process may be added.

[0098] In the above-described embodiment, the character string selection unit 41C applies a priority filter to the character string to select a trailer ID / container ID candidate. In this case, the number of trailer ID / container ID candidates selected by the character string selection unit 41C may be only one. In other words, the character string selection unit 41C may select one trailer ID / container ID by the priority filter.

[0099] In the above-described embodiment, an example has been shown in which the character string selection unit 41C executes a selection based on a notation format and a selection based on the master data 21. However, the present invention is not limited to this, and the character string selection unit 41C may execute either one of a selection based on a notation format and a selection based on the master data 21. For example, the character string selection unit 41C may select a trailer ID / container ID by executing a selection using a priority filter, a mode selection, and a remaining data selection after applying an exclusion filter to the character string.

[0100] For example, the processor 41 may select a trailer ID / container ID by executing a master filter selection, a mode selection, and a remaining data selection after applying an exclusion filter to the character string. The greater the number of trailer IDs / container IDs held as the master data 21, the higher the accuracy of the master filter selection. Therefore, depending on the number of trailer IDs / container IDs held as the master data 21, a trailer ID / container ID may be selected by the master filter selection without executing each process such as the mode selection and the remaining data selection. However, in consideration of the possibility that similar character strings (in other words, character strings for which the same similarity can be calculated) may be recognized, it is preferable to execute the mode selection.

[0101] In the above-described embodiment, the trailer ID / container ID selection result that is output is the result of selection by any one of the processes of mode selection, selection by a master filter, and remaining data selection. However, if the number of trailer ID / container ID candidates is narrowed down to one as a result of selection by an exclusion filter or a priority filter, that candidate may be selected and output as the trailer ID / container ID. However, since a trailer ID / container ID that has not undergone mode selection is likely to be a misrecognized trailer ID / container ID, the accuracy of the selected trailer ID / container ID is increased by undergoing mode selection.

[0102] In the above-described embodiment, an example has been described in which a trailer ID / container ID for identifying a trailer or container is recognized and notified. However, the concept of the above-described embodiment can also be applied to recognizing or notifying other IDs of moving objects entering and leaving a yard. For example, it can be applied to recognizing an ID for identifying a tractor, or to recognizing the IDs of other vehicles such as trucks and passenger cars. In addition, it may be possible to set which type of moving object ID is to be recognized when entering and leaving a yard, according to the user's request.

[0103] Summary of the Disclosure The above description of the embodiments discloses the following techniques.

[0104] <Technology 1> A character recognition device (e.g., character recognition device 40) according to this embodiment includes a recognition unit (e.g., character string recognition unit 41B) that recognizes at least one character string from an image including a trailer captured by an imaging device (e.g., imaging device 10), and a selection unit (e.g., character string selection unit 41C) that selects, from the character strings recognized by the recognition unit, a character string that conforms to a specified notation format as a candidate for a trailer ID or container ID, and selects, from the candidate trailer IDs or container IDs, one whose similarity with pre-stored trailer ID or container ID master data (e.g., master data 21) is equal to or greater than a predetermined threshold value as a trailer ID or container ID.

[0105] As a result, the character recognition device according to the present embodiment can detect a character string that is considered to be a trailer ID or a container ID based on a specified notation format from among multiple character strings written in the trailer. Furthermore, even if detection based on the notation format is difficult, the character recognition device can detect a character string that is considered to be a trailer ID or a container ID based on a known trailer ID or container ID. In other words, the character recognition device can detect a trailer ID or a container ID with high accuracy from among multiple character strings written in the trailer.

[0106] <Technology 2> In the character recognition device described in Technique 1, the notation format may define the English character portion and the numeric character portion in the string.

[0107] As a result, the character recognition device of this embodiment can detect the trailer ID / container ID based on a notation format that specifies the English letter and number portions of the string when the string of the trailer ID or container ID is composed of English letters and numbers.

[0108] <Technology 3> In the character recognition device described in Technology 1 or 2, the recognition unit may convert a part of the recognized character string that is specified as an English character part in the notation format into a specified number if the part is a specified English character, and may convert a part of the recognized character string that is specified as a number part in the notation format into a specified number if the part is a specified English character.

[0109] As a result, when a recognized character string contains a character that is easily misrecognized, the character recognition device according to the present embodiment can convert the character based on the notation format, thereby converting only the necessary parts of the character string and preventing misrecognition of the character string.

[0110] <Technology 4> In the character recognition device according to the third aspect, the selection unit may select the converted character string and the unconverted character string as selection targets.

[0111] As a result, the character recognition device according to the present embodiment can select not only a converted character string conforming to a prescribed notation format as a trailer ID or a container ID, but also a character string before conversion as a trailer ID or a container ID. For example, even if a character string that does not conform to a prescribed notation format is written in the trailer as a trailer ID or a container ID, the character recognition device can select the character string as a trailer ID or a container ID.

[0112] <Technology 5> In a character recognition device described in any one of Techniques 1 to 4, a selection unit may select a trailer ID or a container ID from the trailer ID or container ID candidates based on the number of times the trailer ID or container ID candidate is recognized by the recognition unit from one or more captured images.

[0113] As a result, when there are multiple candidates for the trailer ID or container ID, the character recognition device of this embodiment can select the character string that appears more frequently in the captured image and is more frequently recognized as the trailer ID or container ID.

[0114] <Technology 6> In the character recognition device according to any one of the first to fifth aspects, the selection unit may exclude, from among the character strings recognized by the recognition unit, a character string that is entirely made up of English characters, from candidates for the trailer ID or the container ID.

[0115] This allows the character recognition device according to the present embodiment to exclude character strings that are all English characters and that are considered to be neither trailer IDs nor container IDs from candidates for trailer IDs or container IDs.

[0116] <Technology 7> In the character recognition device according to any one of the first to sixth aspects, the selection unit may exclude, from among the character strings recognized by the recognition unit, character strings that do not include a predetermined number of characters, from candidates for the trailer ID or the container ID.

[0117] As a result, the character recognition device according to this embodiment can exclude character strings that are less than a specified number of characters and are considered to be neither trailer IDs nor container IDs from candidates for trailer IDs or container IDs.

[0118] <Technology 8> In the character recognition device according to any one of the first to seventh aspects, the selection unit may calculate the similarity based on a Levenshtein distance between the candidate trailer ID or container ID and the master data.

[0119] As a result, the character recognition device according to this embodiment can calculate the similarity based on the Levenshtein distance between a candidate trailer ID or container ID and the trailer ID or container ID held as master data.

[0120] <Technology 9> In the character recognition device according to any one of the first to eighth aspects, the selection unit may select, as the trailer ID or the container ID, a character string having the largest number of characters from among candidates for the trailer ID or the container ID.

[0121] As a result, the character recognition device according to this embodiment can select a character string with a large number of characters as the trailer ID / container ID when, for example, it is not possible to narrow down the trailer ID / container ID candidates to one.

[0122] Although the present embodiment has been described above with reference to the drawings, it goes without saying that the present disclosure is not limited to such examples. It is clear that a person skilled in the art can conceive of various modifications, corrections, substitutions, additions, deletions, and equivalents within the scope of the claims, and it is understood that these also naturally belong to the technical scope of the present disclosure. In addition, the components in the above-mentioned present embodiment may be arbitrarily combined within the scope of the invention. [Industrial Applicability]

[0123] The present disclosure is useful as a character recognition device, a character recognition method, and a character recognition system. [Explanation of symbols]

[0124] TR Trailer LE Tractor ST1,ST2,ST3,ST4,ST5,ST6,ST7,ST8,ST9,ST10,ST11,ST12 String 1. Character Recognition System 10 Imaging Device 20 Database 21 Master Data 30 Display Devices 40 Character recognition device 41 Processors 41A String detection section 41B Character string recognition section 41C Character string selection section 42 Communication I / F 43 Memory

Claims

1. a character recognition device that recognizes at least one character string from a captured image including a trailer captured by an imaging device; The character recognition device selects, from the recognized character strings, a character string that conforms to a specified notation format as a candidate for a trailer ID / container ID. Character recognition method.

2. a character recognition device that recognizes at least one character string from a captured image including a trailer captured by an imaging device; The character recognition device selects, from the recognized character strings, those having a similarity to pre-stored master data of trailer IDs / container IDs equal to or greater than a predetermined threshold as trailer IDs / container IDs. Character recognition method.

3. at least one imaging device for capturing an image of the trailer; a recognition unit that recognizes at least one character string from a captured image including the trailer captured by the imaging device; a selection unit that selects a character string that conforms to a specified notation format from the character strings recognized by the recognition unit as a candidate for a trailer ID / container ID. Character recognition system.

4. at least one imaging device for capturing an image of the trailer; a recognition unit that recognizes at least one character string from a captured image including the trailer captured by the imaging device; a selection unit that selects, from among the trailer ID / container ID candidates, a trailer ID / container ID having a similarity to pre-stored trailer ID / container ID master data equal to or greater than a predetermined threshold value as the trailer ID / container ID. Character recognition system.

Images