Display wording inspection device and display wording inspection program

Abstract

To guarantee 100% accuracy for pass assessment in a pass / fail assessment and narrow the range visually confirmable by human to only a location having failed in assessment.SOLUTION: The present invention comprises: character cutting-out means for reading a character image from a character string image composed of regular character string images that are used for displaying a character string on the display screen of a display device, and cutting out the character image in character units; character image generation means for generating a character image in character units on the basis of the character codes and the font data of character strings prepared in advance for display on the display screen; character code acquisition means for acquiring a character code on the basis of the degree of similarity of the cut-out character image to the generated character image; and match / mismatch detection means for detecting a match and a mismatch between wording by characters in character string units represented by the character codes obtained by the character code acquisition means and wording by characters in the character string units represented by character codes of character strings prepared in advance for display on the display screen.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a display wording inspection device and a display wording inspection program. [Background technology]

[0002] For example, in-vehicle cockpits use products equipped with TFT panels to convey various warning information to the driver. Some warnings displayed on these TFT panels are related to driving safety, and the contents of the warnings must be displayed correctly, so high reliability is also required for the inspection of warning displays.

[0003] In conventional inspections, a human would visually check the product and judge it to be a pass, and the process of checking to ensure that no degradation has occurred has become more automated, allowing the inspection to be carried out efficiently and without human error. However, the current situation is that the initial inspection is still carried out visually by a human.

[0004] Conventionally, OCR has been considered as a technology related to automatic inspection of text display. However, there is a problem with OCR recognition that means 100% accuracy cannot be expected. For this reason, for example, the pass / fail judgment itself may contain errors, so in an inspection that requires 100% accuracy, it is not possible to adopt an inspection method using OCR, which may contain errors.

[0005] Patent Document 1 also shows an inspection device for characters displayed on a display device. In this device, a character image displayed on the display device is captured, and the character image of the captured image (hereinafter referred to as a captured image) is normalized and stored in a first RAM. Data of a second font is stored in a second storage unit. A character corresponding to a character to be inspected is a character corresponding to a character code read from a captured image including a character image of a first font. A drawing unit calls the second font from the second storage unit to draw a character corresponding to the character to be inspected, and passes an image related to the drawn character of the second font to a comparison unit.

[0006] The comparison unit acquires a character image of the first font from the first RAM, acquires a character image of the second font from the drawing unit, and performs a comparison process between the character image of the first font and the character image of the second font. The comparison of the character images is performed one by one for each character corresponding to each character code. In other words, the comparison unit performs a comparison process of all character images of the first font and all character images of the second font corresponding to each character code. The comparison unit calculates a difference ratio D from the difference between each pixel of both images and compares each character image using a predetermined formula. The judgment unit judges the degree of agreement between the character image of the first font and the character image of the second font based on the result of the comparison process by the comparison unit. This Patent Document 1 also discloses a method of directly acquiring data of the first font without relying on capture.

[0007] Patent Document 2 discloses an inspection device that properly inspects whether a document including text is correctly rendered by a browser. The inspection device provides a document to a browser and displays the document on a screen or renders the document in a virtual view of the browser, thereby acquiring an image of the rendered document, performs character recognition on the acquired image to acquire recognized text, extracts body text to be rendered from the document, and compares the recognized text with the body text to determine whether the document is correctly rendered by the browser. This inspection device makes it possible to properly inspect whether a document including text is correctly rendered by a browser.

[0008] The recognition unit performs character recognition on the image acquired by the acquisition unit, thereby acquiring the text drawn on the image as recognized text. The recognition unit, for example, uses morphological analysis to divide the image acquired by the acquisition unit into images representing each character, and extracts features (vectors indicating features) used for character recognition from the divided images. Character recognition is then performed by comparing the extracted features with features of characters previously stored in a storage device. The features used for character recognition are directional line element features utilizing the outline of the character, etc. The recognition unit also acquires position information indicating the position in the image from which each character obtained as a result of character recognition was recognized. When a defect is found in the drawing of the character, this position information is used to display a warning that there is a defect in the drawing of the character at or around the position where the defect is present on the screen of the display device.

[0009] Furthermore, in Patent Document 3, while the general operation of each Web specification is defined in the specification, when each Web specification is actually executed in a Web browser that displays a Web page, different operations (operational differences) may occur between Web browsers. In consideration of this, an operation difference information generating device is disclosed that accurately generates operation difference information between Web browsers for each Web specification. Examples of operation differences include display misalignment, garbled characters, and the presence or absence of an event. This operation difference information generating device includes a Web page decomposition unit, a single browser operation difference detection unit, a multi-browser operation difference detection unit, and an operation difference information storage unit. The Web page decomposition unit receives an input Web page as an input, and generates a basic Web page and candidate Web pages for evaluation based on the operation difference information (operational difference accumulation information) stored in the operation difference information storage unit.

[0010] The single browser operation difference detection unit receives a base web page and an evaluation web page candidate as input, and outputs the evaluation web page candidate as an evaluation web page when there is an operation difference between the two input web pages in at least one or more web browsers. Through the processing of this single browser operation difference detection unit, evaluation web page candidates that have operation differences caused by factors other than the web specifications are excluded from the evaluation web page candidates, and an evaluation web page suitable for checking the presence or absence of operation differences between web browsers caused only by the web specifications is generated.

[0011] The multi-browser operation difference detection unit has a function of checking the operation differences between two or more different web browsers using an evaluation web page as an input, and generating operation difference information that summarizes the checking results. Therefore, the operation difference information generated by the multi-browser operation difference detection unit indicates the operation differences between multiple types of web browsers for the web specification to be evaluated.

[0012] The behavioral difference information storage unit has a function of a repository that stores the behavioral difference information generated by the multi-browser behavioral difference detection unit. The behavioral difference information storage unit can update known behavioral difference information by adding and storing newly generated behavioral difference information to the behavioral difference accumulation information.

[0013] The behavioral difference information display unit has a function of outputting behavioral difference information (behavioral difference accumulated information) stored in the behavioral difference information storage unit. The behavioral difference information display unit reads the behavioral difference accumulated information stored in the behavioral difference information storage unit and displays and outputs the contents on a display device, and can also output the information to a memory, an external storage device, and a network communication device. Thus, in the behavioral difference information generating device, the functions of each unit (processing by each unit) are called in the order of the Web page decomposition unit, the single browser behavioral difference detection unit, and the multi-browser behavioral difference detection unit, and the behavioral difference accumulated information is displayed.

[0014] As described above, in the inspection device, the inspection quality precision of the inspection function is improved, but this is not intended to reduce the man-hours of the inspector who ultimately performs the inspection. [Prior art documents] [Patent documents]

[0015] [Patent Document 1] JP 2018-205673 A [Patent Document 2] Patent No. 6356924 [Patent Document 3] Patent No. 6993284 Summary of the Invention [Problem to be solved by the invention]

[0016] Even if an inspection device makes a mistake in the judgment result in a process where 100% accuracy cannot be expected, by devising a way to make an NG judgment in the next process where 100% accuracy can be expected, it is possible to ensure 100% accuracy in the pass / pass judgment in the sense that although a pass case may be judged as NG, an NG case will not be erroneously judged as a pass. By adopting an automatic inspection device such as the one described above, visual inspection can be limited to only those parts judged as NG by the automatic inspection.

[0017] The present invention is based on the above-mentioned idea and aims to provide a display wording inspection device and a display wording inspection program that can guarantee 100% accuracy in determining whether a product is pass or fail, and can narrow the area that humans visually check to only those areas that are determined to be fail, thereby reducing the possibility of human error and the amount of evaluation work required. [Means for solving the problem]

[0018] The display word inspection device according to the present invention includes a display device for displaying image data related to the inspection, a character extraction means for reading out a character string image from a VRAM in which a character string image, which is a message composed of a regular character string image that is a basis for displaying a character string on a display screen of a display device to be inspected, is stored, and extracting the character string image into character units, and The above character image generating means for generating character images for each character based on the character code and font data of the character string; character images extracted by said character extracting means as explanatory variables; character codes of the character images generated by said character image generating means; and font The system is created by machine learning with the objective variable of composition a character code acquiring means for acquiring a similarity between the character image extracted by the character extracting means and a character image of a corresponding character code; By percentage a character code acquisition means for acquiring a character code for acquiring a character string unit character string using the character code acquired by the character code acquisition means; The above A match or mismatch between the character code of the character string and the wording of the character string is detected, and the similarity of each character of the message and an overall similarity of the message are calculated using the similarity. By percentage Get, The result of the match or mismatch is displayed on the display in percentage, or the similarity of each character of the message is displayed on the display in percentage. The present invention is characterized by comprising a match / mismatch detection means.

[0019] In the display text inspection device according to the present invention, the character extraction means extracts the characters by character using AI.

[0020] In the display text inspection device of the present invention, the character extraction means extracts text on a character-by-character basis using a display area frame in which text on the screen is displayed, the distance to the outline of the characters displayed in this display area frame, and the font as explanatory variables.

[0021] In the display text inspection device according to the present invention, the match / mismatch detection means detects match / mismatch based on color information of characters displayed in the display area frame.

[0022] The display text inspection device according to the present invention is characterized in that it inspects character strings spanning multiple lines.

[0024] In the display text inspection device according to the present invention, the character code acquisition means obtains a similarity between the character image extracted by the character extraction means and the character image generated by the character image generation means.

[0025] The display word inspection program according to the present invention is configured by a computer, and includes a display for displaying image data related to the inspection, the computer being a display device for inspecting a display word, character extraction means for reading out a character string image from a VRAM in which a character string image is stored, the character string image being a message composed of a regular character string image that is a basis for displaying a character string on a display screen of a display device to be inspected, and character extraction means for extracting the character string image into character units, The above A character image generating means for generating a character image for each character based on the character code and font data of the character string, the character image extracted by the character extraction means being an explanatory variable, and the character code of the character image generated by the character image generating means being and font The system is created by machine learning with the objective variable of composition a character code acquiring means for acquiring a similarity between the character image extracted by the character extracting means and a character image of a corresponding character code; By percentage a character code acquisition means for acquiring a character code for acquiring the character code of the character string unit; The above A match or mismatch between the character code of the character string and the wording of the character string is detected, and the similarity is used to calculate the similarity of each character of the message and the overall similarity of the message. by percentage Get, The result of the match or mismatch is displayed on the display in percentage, or the similarity of each character of the message is displayed on the display in percentage.The present invention is characterized in that it functions as a match / mismatch detection means.

[0026] The display text inspection program according to the present invention is characterized in that it causes the computer to function as the character extraction means to extract text on a character-by-character basis using AI.

[0027] The display text inspection program of the present invention is characterized in that the computer is caused to function as the character extraction means to extract text made of characters on the screen on a character-by-character basis using a display area frame in which the text is displayed, the distance to the outline of the characters displayed in this display area frame, and the font as explanatory variables.

[0028] The display text inspection program of the present invention is characterized in that the computer is caused to function as the match / mismatch detection means to detect match / mismatch based on color information of the characters displayed in the display area frame.

[0029] The display text inspection program of the present invention is characterized in that it causes the computer to function as the character extraction means, the character image generation means, the character code acquisition means, and the match / mismatch detection means to process character strings spanning multiple lines.

[0031] The display text inspection program of the present invention is characterized in that the computer is caused to function as the character code acquisition means to obtain the similarity between the character image extracted by the character extraction means and the character image generated by the character image generation means. [Brief description of the drawings]

[0032] [Figure 1] 1 is a block diagram of a display text inspection device according to an embodiment of the present invention. [Diagram 2] FIG. 2 is a block diagram of a computer that realizes the display text inspection device according to the present embodiment. [Diagram 3]4 is a flowchart showing the operation of the display text inspection device according to the present embodiment configured by a computer. [Figure 4] 3A and 3B are diagrams showing a screen of the display device of the embodiment, and character strings and detection frames displayed on the screen. [Diagram 5] 4A and 4B are diagrams showing an inference rectangle and a circumscribing rectangle used for character segmentation by the display text inspection device according to the embodiment; [Figure 6] 3A to 3C are diagrams showing the frame lines of a detection frame, an inference rectangle, and a circumscribing frame used by the display text inspection device according to the embodiment. [Figure 7] 4 is a diagram showing determination information obtained by the display text inspection device according to the present embodiment in response to a character string of two rows and two columns. FIG. [Figure 8] 4A and 4B are diagrams showing judgment information around a character frame by the display text inspection device according to the embodiment; [Figure 9] 4A and 4B are diagrams showing judgment information around a character frame by the display text inspection device according to the embodiment; [Figure 10] 4A and 4B are diagrams showing judgment information around a character frame by the display text inspection device according to the embodiment; [Figure 11] 4A and 4B are diagrams showing judgment information around a character frame by the display text inspection device according to the embodiment; [Figure 12] 4A and 4B are diagrams showing judgment information around a character frame by the display text inspection device according to the embodiment; [Figure 13] 4A and 4B are diagrams showing judgment information around a character frame by the display text inspection device according to the embodiment; [Figure 14] 4A and 4B are diagrams showing judgment information around a character frame by the display text inspection device according to the embodiment; [Figure 15] 4A and 4B are diagrams showing judgment information around a character frame by the display text inspection device according to the embodiment; [Figure 16] 4A and 4B are diagrams showing judgment information around a character frame by the display text inspection device according to the embodiment; [Figure 17] 4A and 4B are diagrams showing judgment information around a character frame by the display text inspection device according to the embodiment; [Figure 18]4A and 4B are diagrams showing judgment information around a character frame by the display text inspection device according to the embodiment; [Figure 19] 4A and 4B are diagrams showing judgment information around a character frame by the display text inspection device according to the embodiment; [Figure 20] 4A and 4B are diagrams showing judgment information around a character frame by the display text inspection device according to the embodiment; [Figure 21] 4A and 4B are diagrams showing judgment information around a character frame by the display text inspection device according to the embodiment; [Figure 22] 4A and 4B are diagrams showing judgment information around a character frame by the display text inspection device according to the embodiment; [Figure 23] 4A and 4B are diagrams showing judgment information around a character frame by the display text inspection device according to the embodiment; [Figure 24] 3 is a diagram showing a main portion of data or information accumulated by the display text inspection device according to the embodiment; FIG. [Diagram 25] This figure shows the inspection text portion of messages that have been judged as OK and messages that have been judged as NG, the image to be inspected (the image read out from VRAM 22), the detection result image portion, the judgment result (NG or OK), the overall similarity portion, and the main parts of the judgment information, using the display text inspection device of this embodiment. [Figure 26] 13 is a diagram showing a judgment image for the message "RESOLVE R-1 NOT ADJUSTED", which is judged as OK by the display text inspection device according to the present embodiment. FIG. [Figure 27] 13 is a diagram showing an example of an inspection result image in which the inspection text that has been judged as NG by the display text inspection device according to the embodiment is "No Message." DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0033] Hereinafter, an embodiment of a display wording inspection device and a display wording inspection program according to the present invention will be described with reference to the accompanying drawings. In each drawing, the same components are given the same reference numerals and duplicated explanations will be omitted. Fig. 1 shows a block diagram of a display wording inspection device 100 according to this embodiment. The display wording inspection device 100 is connected to a display device 200 equipped with a display screen such as a TFT panel, and performs inspection.

[0034] The display device 200 is provided with a message image creation unit 21 that creates message images as various types of warning information to be conveyed to the driver. An input unit 20 is connected to the message image creation unit 21, which receives, for example, traffic information and sends it to the message image creation unit 21 as message information (code) to be displayed. Message information for inspection is sent from the input unit 20 when inspecting display wording.

[0035] Furthermore, a VRAM 22 is connected to the message image creation unit 21, and the message image created by the message image creation unit 21 is stored in the VRAM 22. The VRAM 22 is connected to the display unit 23, and the display unit 23 reads out the image stored in the VRAM 22 and displays it on the display screen.

[0036] The display text inspection device 100 is connected to the input unit 20 and the VRAM 22. When inspecting a display text, the display text inspection device 100 imports message information for inspection from the input unit 20 and performs inspection. In response to the import of this message information, the display text inspection device 100 imports a message image stored in the VRAM 22 and performs inspection. The message information is character code and font data of a character string prepared in advance to be displayed on the display screen, and is information used for inspection.

[0037] The display word inspection device 100 includes a character extraction means 11, a character image generating means 12, a character code acquiring means 13, and a match / mismatch detecting means 14. The character extraction means 11 reads out a character image from a character string image composed of a normal character string image that is a basis for displaying a character string on a display screen of a display device, and extracts the character image on a character basis. Here, the normal character string image is a character image displayed on a display device, and is image information without degradation that can perform match / mismatch detection in the present invention as described below. It may be a capture image, but does not include an image captured by a camera or the like of a displayed image. In this embodiment, it is assumed that a character string image composed of a normal character string image is stored in a VRAM 22. Therefore, the character extraction means 11 reads out the character image from the VRAM 22 of the display device 200 that displays characters on a display screen based on the character image of the character string stored in the VRAM 22, and extracts the character image on a character basis. This process of extracting the character image on a character basis is performed by AI (Artificial Intelligence). For example, a judgment model for obtaining correct extraction position information by machine learning can be created using some of the judgment information, such as character codes, such as unique codes for characters in some character strings, and font information, as explanatory variables, and extraction position information as a target variable, and this can be used as the AI-based character extraction means 11. In FIG. 4, the AI ​​in this case is shown as a "multilingual automatic recognition tool" or a "text learning image creation tool," and a display screen such as a TFT panel is shown by a black rectangular frame. The character string "There is no message" is shown to be surrounded by a detection frame. Judgment information such as the top and bottom coordinates, width, and height of the detection frame are specified (input) when the above tool is executed. The character extraction means 11 not only extracts characters that are separated one by one, such as the alphabet, kanji, and hiragana, but also extracts the part of the multiple characters of the combination as the extraction position in the case of a language that uses a combination of multiple characters, such as Arabic. The character extraction means 11 can extract characters when characters are consecutive, extract characters from a horizontally written character string, and extract characters from a vertically written character string.

[0038] The character image generating means 12 generates character images for each character based on the character code and font data of a character string prepared in advance to be displayed on the display screen. That is, when inspecting a displayed message, inspection message information provided from the input unit 20 is input and character images are generated.

[0039] The character code acquisition means 13 obtains a character code corresponding to the extracted character image based on the similarity between the character image extracted by the character extraction means 11 and the character image generated by the character image generation means, and can be performed by AI. That is, the character code acquisition means 13 obtains the similarity between the character image extracted by the character extraction means 11 and the character image generated by the character image generation means 12. For example, the character code acquisition means 13 can be configured as an AI for character image recognition by creating a judgment model by machine learning using learning data in which the character image extracted by the character extraction means 11 is an explanatory variable and information such as the character image generated by the character image generation means 12 (same character code and font) is an objective variable. In this case, the character code acquisition means 13 is configured to output the similarity. The match / mismatch detection means 14 detects match / mismatch between a wording by characters in a character string unit based on the character code obtained by the character code acquisition means 13 and a wording by characters in a character string unit based on the character code of a character string prepared in advance to be displayed on the display screen. The character code acquisition means 13 obtains (acquires) codes for each character, so at this stage it obtains the similarity (precision) for each character, which the match / mismatch detection means 14 uses to display the detected match / mismatch results for the entire character string as a percentage.

[0040] The display text inspection device 100 can be configured as a computer as shown in Fig. 2. That is, this computer operates as the display text inspection device 100 by controlling each part using a program stored in a main memory 31 by a CPU 30.

[0041] The CPU 30 is connected to an external memory interface 33, an input interface 34, a display interface 35, and data input interfaces 36 and 37 via a bus 32. The external memory interface 33 is connected to an external storage device 43. The external storage device 43 stores various data for operating the display wording inspection device 100, in addition to the above-mentioned character extraction means 11, character image generation means 12, character code acquisition means 13, match / mismatch detection means 14, and programs for implementing AI. The input interface 34 is connected to an input device 44 such as a keyboard or a touch panel and a pointing device 42 such as a mouse, and is configured to be able to input necessary data, commands, and the like. The display interface 35 is connected to a display device 45 equipped with a display device such as an LED or an LCD, and is able to display necessary images and data. The data input interface 36 is connected to the input unit 20 and is used to import message information for inspection provided from the input unit 20 during display wording inspection, and the data input interface 37 is connected to the VRAM 22 and is used to import message images stored in the VRAM 22.

[0042] The computer having the above configuration executes the program of the flowchart shown in Fig. 3 as the character extraction means 11, character image generation means 12, character code acquisition means 13, match / mismatch detection means 14, AI, etc., to function as a display text inspection device, so the operation will be described with reference to this flowchart. When the display text inspection starts, the image of the character string is read from the VRAM 22 and extracted into character units (S11). At this time, it is assumed that the judgment information has been input, as already explained with reference to Fig. 4.

[0043] When character segmentation is performed in step S11, a corresponding character code is obtained from the similarity between the character image segmented in character units and the character image generated by the character image generating means 12 (S12). At this time, the inference rectangle and the circumscribing rectangle used for segmentation are determined as shown in FIG. 5. Here, "inference" refers to the character image taken out from the VRAM 22. The detection frame in FIG. 5 is a thin frame compared to other frames as shown in FIG. 6. The inference rectangle in FIG. 5 is shown as a rectangular range obtained from inference information (information on the characters of the character string and the inferred area) as a thick frame in FIG. 6. The circumscribing frame in FIG. 5 is shown by a dashed line in FIG. 6 as a circumscribing rectangle for the characters in the inference frame. Here, "circumscribing" refers to the character image taken out from the VRAM 22. In the examples in FIG. 5 and FIG. 8 to FIG. 22, the displayed text is "There is no message."

[0044] Judgment information is obtained according to the detection frame, inference rectangle, and circumscribing rectangle as described above. Here, judgment information determined by the detection frame, inference rectangle, and circumscribing rectangle will be described as #1 to #16. In order to obtain the result of the display text inspection, judgment information is obtained in the process of steps S11 and S12 for each character string and each character therein. FIG. 7 shows the information obtained for a character string of two rows and two columns. The first column of the first row is related to "character 1", and judgment information such as similarity and font name is written in the part surrounded by a solid line frame. The first column of the second row is related to "character 2", and judgment information such as similarity and font name is written in the part surrounded by a two-dot chain line frame. The second column of the first row is related to "character 3", and judgment information such as similarity and font name is written in the part surrounded by a dashed line frame. The second column of the second row is related to "character 4", and judgment information such as similarity and font name is written in the part surrounded by a two-dot chain line frame.

[0045] As described above, there is a frame in which judgment information for "Character 1" to "Character 4" is written, and judgment information indicated by #1 to #16 is arranged around this frame. #1 is a numerical value indicating the distance between the left side of the detection frame and the left edge of the inferred image of the first character in the line, as shown in FIG. 8. #2 is a numerical value indicating the distance between the left side of the detection frame and the left edge of the circumscribing rectangle of the first character in the line, as shown in FIG. 9. #3 is a numerical value indicating the distance between the top side of the detection frame and the minimum coordinate of the top coordinate of the inferred rectangle of the first line, as shown in FIG. 10.

[0046] #4 is a value indicating the distance between the top of the detection frame and the minimum coordinate of the bounding rectangle of the first line, as shown in FIG. 11. #5 is a value indicating the distance between the bottom of the detection frame and the maximum coordinate of the estimated rectangle of the bottom line, as shown in FIG. 12. #6 is a value indicating the distance between the bottom of the detection frame and the maximum coordinate of the bounding rectangle of the bottom line, as shown in FIG. 13. #7 is a value indicating the distance between the right side of the detection frame and the right-end coordinate of the estimated rectangle of the last character of each line, as shown in FIG. 14. #8 is a value indicating the distance between the right side of the detection frame and the right-end coordinate of the bounding rectangle of the last character of each line, as shown in FIG. 15. #9 is a value indicating the distance between the bottom maximum coordinate of the inferred rectangle of the upper line and the top minimum coordinate of the inferred rectangle of the lower line, as shown in FIG. 16. #10 is a value indicating the distance between the bottom maximum coordinate of the bounding rectangle of the upper line and the top minimum coordinate of the bounding rectangle of the lower line, as shown in FIG. 17.

[0047] #11 is a value indicating the distance between the lower maximum coordinate of the inferred rectangle of the upper row and the upper minimum coordinate of the circumscribing rectangle of the lower row, as shown in FIG. 18. #12 is a value indicating the distance between the lower maximum coordinate of the inferred rectangle of the upper row and the upper minimum coordinate of the inferred rectangle of the lower row, as shown in FIG. 19. #13 is a value indicating the distance between the right coordinate of the inferred rectangle of the previous character and the left coordinate of the inferred rectangle of the following character, as shown in FIG. 20. #14 is a value indicating the distance between the right coordinate of the circumscribing rectangle of the previous character and the left coordinate of the circumscribing rectangle of the following character, as shown in FIG. 21. #15 is a value indicating the distance between the right coordinate of the inferred rectangle of the previous character and the left coordinate of the circumscribing rectangle of the following character, as shown in FIG. 22. #16 is a value indicating the distance between the right coordinate of the circumscribing rectangle of the previous character and the left coordinate of the inferred rectangle of the following character, as shown in FIG. 23.

[0048] The judgment information shown in each column in each of the boxes "Character 1" to "Character 4" in FIG. 7 is also obtained. The judgment information shown in each column in each of the boxes "Character 1" to "Character 4" in FIG. 7 will be explained below. In the first line of the box, from left to right, the font image, font image width, and font image height are set. In the second line of the box, the inferred image, inferred image width, and inferred image height are set. In the third line of the box, the circumscribed rectangle image, circumscribed rectangle width, and circumscribed rectangle height are set. In the fourth line of the box, the image type, font name, and similarity are set.

[0049] Returning to Fig. 3, the next step is to check whether the character code of a prepared character string matches the character code obtained from the similarity of the character image (S13), and to determine whether they match (S14). If they match, the text display of the character string is deemed to have passed (S15). If they do not match, the text display of the character string is deemed to have failed (S16).

[0050] The device of this embodiment accumulates data or information related to the processing result, including the above-mentioned judgment information, and this will be described. FIG. 24 shows an example of a main part of the accumulated data or information, which is shown in a table, but the format is not particularly limited. In the example of FIG. 24, each message is organized in a row direction, and information used by the system for management, such as a number, is set in the column direction from the left, the next row direction column is set with the message translated into Japanese, the next row direction column is set with the message translated into English, the next row direction column is set with the inspection text, which is the actual message, the next row direction column is set with the image of the inspection target (the image read out from the VRAM 22), the next row direction column is set with the detection result image, the next row direction column is set with the judgment result (NG or OK), the next row direction column is set with the overall similarity, and the next row direction column is set with the judgment information already described. The judgment information continues to the right side of the figure in accordance with the number of characters in the message text, but is omitted here.

[0051] FIG. 25 shows the inspection text of a message that was judged as OK and a message that was judged as NG, the image to be inspected (the image read from VRAM 22), the detection result image, the judgment result (NG or OK), the overall similarity, and the main part of the judgment information. FIG. 26 shows an image of the detection result of the inspection text "RESOLVE R-1 NOT ADJUSTED" of a message that was judged as OK. FIG. 27 shows an image of the detection result of the inspection text "No Message" of a message that was judged as NG. From FIG. 26 and FIG. 27, the similarity for character strings of an appropriate length is displayed.

[0052] Furthermore, as can be seen from Figure 25, in the test phrase "RESOLVE R-1 NOT ADJUSTED" of the message that was judged as OK, the similarity of each character is 100% and the overall similarity is also 100%, whereas in the test phrase "No Message" of the message that was judged as NG, the similarity of each character is not 100%, such as 55.73% and 61.46%, and the overall similarity is also not 100%, so it was judged as NG.

[0053] As described above, 100% accuracy can be guaranteed for pass / fail judgment of the character string of the displayed message. Therefore, in the device of this embodiment, the similarity and font name are displayed, so that the range that a human visually checks can be narrowed down to only the fail judgment points. Therefore, it is possible to provide a display wording inspection device and a display wording inspection program that can reduce the possibility of human error and the evaluation man-hours. [Explanation of symbols]

[0054] 11 Character cutting means 12 Character image generation means 13 How to obtain character code 14 Match / Mismatch Detection Method 20 Input section 21 Message Image Creation Section 22 VRAM 23 Display section 30 CPU 31 Main Memory 32 Bus 33 External memory interface 34 Input Interface 35 Display Interface 36 Data Entry Interface 37 Data Entry Interface 42 Pointing Device 43 External storage device 44 Input Devices 45 Display device 100 Display wording inspection device 200 Display device

Claims

1. a display for displaying images and data relating to the examination; a character extraction means for reading out a character string image from a VRAM in which a character string image is stored, the character string image being a message composed of a regular character string image serving as a basis for displaying a character string on a display screen of a display device to be inspected, and extracting the character string image on a character-by-character basis; character image generating means for generating character images on a character-by-character basis based on character codes and font data of the character strings prepared in advance to be displayed on the display screen; a character code acquisition means that is created by machine learning using a character image extracted by the character extraction means as an explanatory variable and a character code and a font of the character image generated by the character image generation means as objective variables, and that is configured to obtain a character code corresponding to the extracted character image, the character code acquisition means acquiring a similarity between the character image extracted by the character extraction means and the character image of the corresponding character code by a percentage; a match / mismatch detection means for detecting a match / mismatch between a wording of characters in a character string unit using the character code obtained by the character code obtaining means and a wording of characters in a character string unit using the character code of the character string previously prepared to be displayed on the display screen, and for obtaining a similarity of each character of the message and an overall similarity of the message in percentages using the similarity, and for displaying the result of the match / mismatch on the display device in percentages, or for displaying the similarity of each character of the message in percentages; A display wording inspection device comprising:

2. 2. The display text inspection device according to claim 1, wherein the character extraction means extracts the characters by character using AI.

3. The display text inspection device according to claim 2, characterized in that the character extraction means extracts text on a character-by-character basis using a display area frame in which text on the screen is displayed, the distance to the outline of the text displayed in this display area frame, and the font as explanatory variables.

4. 4. The display text inspection device according to claim 3, wherein the match / mismatch detection means detects a match / mismatch based on color information of characters displayed in the display area frame.

5. 3. The display text inspection device according to claim 2, wherein inspection is performed on a character string spanning multiple lines.

6. 2. The display text inspection device according to claim 1, wherein the character code acquisition means obtains a similarity between the character image extracted by the character extraction means and the character image generated by the character image generation means.

7. A display wording inspection device including a computer and a display device for displaying images and data related to the inspection, a character extraction means for reading out a character string image from a VRAM in which a character string image, which is a message composed of a regular character string image serving as a basis for displaying a character string on a display screen of a display device to be inspected, is stored, and extracting the character string image on a character-by-character basis; character image generating means for generating character images on a character-by-character basis based on character codes and font data of the character string previously prepared to be displayed on the display screen; a character code acquisition means that is created by machine learning using the character image extracted by the character extraction means as explanatory variables and the character code and font of the character image generated by the character image generation means as objective variables, and is configured to obtain a character code corresponding to the extracted character image, the character code acquisition means acquiring a similarity between the character image extracted by the character extraction means and the character image of the corresponding character code as a percentage; a match / mismatch detection means for detecting a match / mismatch between a wording of characters in the character string unit using the character code obtained by the character code acquisition means and a wording of characters in the character string unit using the character code of the character string previously prepared to be displayed on the display screen, and for obtaining a similarity of each character of the message and an overall similarity of the message in percentages using the similarity, and displaying the result of the match / mismatch on the display device in percentages, or for displaying the similarity of each character of the message in percentages on the display device. A display text checking program characterized by functioning as:

8. 8. The display text inspection program according to claim 7, wherein the computer is caused to function as the character extraction means by extracting characters by AI.

9. The display text inspection program according to claim 8, characterized in that the computer is caused to function as the character extraction means to extract text on a character-by-character basis using a display area frame in which text on the screen is displayed, and the distance to the outline of the text displayed in this display area frame and the font as explanatory variables.

10. The display text inspection program according to claim 9, characterized in that the computer is caused to function as the match / mismatch detection means to detect match / mismatch based on color information of characters displayed in the display area frame.

11. The display text inspection program according to claim 8, characterized in that the computer is caused to function as the character extraction means, the character image generation means, the character code acquisition means, and the match / mismatch detection means to process character strings spanning multiple lines.

12. The display text inspection program according to claim 7, characterized in that the computer is caused to function as the character code acquisition means to obtain a similarity between a character image extracted by the character extraction means and a character image generated by the character image generation means.

Images