Reading device, reading method, and reading program

Abstract

This enables accurate recognition of numerical values ​​from image data captured from meters and other devices. [Solution] The setting reception unit 21 receives the setting of the display area in image data of a meter having a display area that displays the value of each digit of a multi-digit number. The section identification unit 22 divides the display area indicated by the setting received by the setting reception unit 21 into a reading area, which is the area where the value of each digit is displayed, and a blank area between the two reading areas. The reading unit 25 reads the numerical value displayed by the reading model, which is a trained model, for each section of the reading area identified by the section identification unit 22.

Description

[Technical Field]

[0001] This disclosure relates to a technology for reading numerical values ​​displayed on a meter from image data captured by photographing the meter. [Background technology]

[0002] Attempts are being made to read numerical values ​​from image data of meters using deep learning-based image recognition models. In this case, accurately reading the numerical values ​​is crucial.

[0003] Patent Document 1 describes a technique for accurately reading meter readings by correcting an image taken with a fisheye lens, where the left and right edges are smaller, so that the image components match the same image. [Prior art documents] [Patent Documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2019-16027 [Overview of the project] [Problems that the invention aims to solve]

[0005] By specifying the region of the image data to be recognized for an image recognition model, the recognition accuracy can be improved. However, if the region of the image data to be recognized contains extraneous information, such as the units of the displayed numbers, false detections may occur, leading to a decrease in recognition accuracy. Even when using the technology described in Patent Document 1, it is difficult to sufficiently improve recognition accuracy if the region to be recognized is not properly identified. This disclosure aims to enable accurate recognition of numerical values ​​from image data of meters, etc. [Means for solving the problem]

[0006] The reading device relating to this disclosure is A setting reception unit that receives the setting of the display area in image data of a meter having a display area that displays the value of each digit of a multi-digit number, A section identification unit divides the display area indicated by the setting received by the setting reception unit into a reading area which is the area where the values ​​of each digit are displayed, and a blank area between the two reading areas. It is equipped with. [Effects of the Invention]

[0007] In this disclosure, the display area for displaying the value of each digit is divided into a reading area, which is the area where the value of each digit is displayed, and a blank area between the two reading areas. This allows for the recognition of not only the entire multi-digit number, but also the recognition of each individual digit. As a result, it is possible to accurately recognize numerical values ​​from image data captured from meters, etc. [Brief explanation of the drawing]

[0008] [Figure 1] A diagram showing the configuration of the reading device 10 according to Embodiment 1. [Figure 2] An explanatory diagram of the display area 31 of the meter according to Embodiment 1. [Figure 3] Flowchart of the setting process according to Embodiment 1. [Figure 4] Flowchart of the setting acceptance process according to Embodiment 1. [Figure 5] An explanatory diagram of the display area reception process according to Embodiment 1. [Figure 6] An explanatory diagram of the setting value p according to Embodiment 1. [Figure 7] Diagram illustrating the digit count acceptance process and the set value acceptance process according to Embodiment 1. [Figure 8] Flowchart of the interval identification process according to Embodiment 1. [Figure 9] An explanatory diagram of the region setting process according to Embodiment 1. [Figure 10] An explanatory diagram of Method 1 when the quadrilateral region according to Embodiment 1 is not a rectangle. [Figure 11]Flowchart of the reading process according to Embodiment 1. [Figure 12] Flowchart of the section identification process according to Embodiment 2. [Figure 13] Explanatory diagram of the reading area identification process according to Embodiment 2. [Figure 14] Explanatory diagram of the reading area identification process according to Embodiment 2. [Figure 15] Explanatory diagram of the effects according to Embodiment 2. **Modes for Carrying Out the Invention**

[0009] Embodiment 1. ***Explanation of the Configuration*** Referring to FIG. 1, the configuration of the reading device 10 according to Embodiment 1 will be described. The reading device 10 is a computer. The reading device 10 includes hardware such as a processor 11, a memory 12, a storage 13, and a communication interface 14. The processor 11 is connected to other hardware via signal lines and controls these other hardware.

[0010] The processor 11 is an IC that performs processing. IC is an abbreviation for Integrated Circuit. Specific examples of the processor 11 are a CPU, a DSP, and a GPU. CPU is an abbreviation for Central Processing Unit. DSP is an abbreviation for Digital Signal Processor. GPU is an abbreviation for Graphics Processing Unit.

[0011] The memory 12 is a storage device that temporarily stores data. Specific examples of the memory 12 are SRAM and DRAM. SRAM is an abbreviation for Static Random Access Memory. DRAM is an abbreviation for Dynamic Random Access Memory.

[0012] Storage 13 is a storage device for storing data. Specific examples of storage 13 include HDDs and SSDs. HDD stands for Hard Disk Drive. SSD stands for Solid State Drive. Storage 13 may also be a portable recording medium such as an SD® memory card, CompactFlash®, NAND flash, flexible disk, optical disk, compact disk, Blu-ray® disc, or DVD. SD stands for Secure Digital. DVD stands for Digital Versatile Disk.

[0013] Communication interface 14 is an interface for communicating with external devices. Specific examples of communication interface 14 include Ethernet®, USB, and HDMI® ports. USB stands for Universal Serial Bus. HDMI stands for High-Definition Multimedia Interface.

[0014] The reading device 10 comprises a setting reception unit 21, a section identification unit 22, a setting storage unit 23, an image acquisition unit 24, a reading unit 25, and a display unit 26 as functional components. The functions of each functional component of the reading device 10 are implemented by software. The storage device 13 stores programs that implement the functions of each functional component of the reading device 10. These programs are loaded into the memory 12 by the processor 11 and executed by the processor 11. This enables the functions of each functional component of the reading device 10 to be implemented.

[0015] In Figure 1, only one processor 11 was shown. However, there may be multiple processors 11, and multiple processors 11 may work together to execute programs that implement each function.

[0016] ***Explanation of operation*** The operation of the reading device 10 according to Embodiment 1 will be explained with reference to Figures 2 to 11. The operating procedure of the reading device 10 according to Embodiment 1 corresponds to the reading method according to Embodiment 1. Furthermore, the program that implements the operation of the reading device 10 according to Embodiment 1 corresponds to the reading program according to Embodiment 1.

[0017] The reading device 10 performs a setting process and a reading process. The setting process is the process of setting the parameters for reading the numerical value displayed in the meter's display area. The reading process is the process of reading the numerical value displayed in the meter's display area according to the settings made in the setting process.

[0018] Referring to Figure 2, the display area 31 of the meter according to Embodiment 1 will be described. The meter's display area 31 displays the value of each digit in a multi-digit number. In Figure 2, a four-digit number is displayed in the display area 31. The area where the value of each digit is displayed is the area to be read in the reading process described later, and is called the reading area 32. Between each reading area 32, there is a blank area 33 where no numerical value is displayed. The blank area 33 may be blank, or it may display information that should not be read, such as units. Here, we assume that the width of each reading area 32 is the same. Also, we assume that the width of each blank area 33 is the same. In the following explanation, the direction in which the values ​​of each digit are arranged is referred to as the digit direction 34. In Figure 2, the horizontal direction is the digit direction.

[0019] Referring to Figure 3, the setting process according to Embodiment 1 will be explained. (Step S11: Setting acceptance processing) The setting reception unit 21 accepts basic setting information. Specifically, the setting reception unit 21 accepts the information of the display area 31, the number of digits n, and the setting value p as basic setting information. The setting reception process will be explained in detail with reference to Figure 4.

[0020] (Step S111: Display area reception processing) The setting reception unit 21 receives a setting for a display area 31 that displays the values ​​of each of the multiple digits in the meter. In this case, the setting reception unit 21 receives a setting for a rectangular area surrounding the display area 31 in image data taken by the target camera of a meter having a display area 31 that displays the values ​​of each of the multiple digits. For example, the setting reception unit 21 displays image data of the meter and prompts the user to specify a rectangular area surrounding the display area 31. Specifically, if the rectangular area is a rectangle, the setting reception unit 21 prompts the user to specify two opposite corners of the rectangle, such as the top-left and bottom-right corners. Note that the image data is not necessarily obtained by photographing the meter from the front, so the rectangular area may not be a rectangle. If the rectangular area is not a rectangle, the setting reception unit 21 prompts the user to specify the four vertices of the rectangular area. Alternatively, the setting reception unit 21 may prompt the user to specify the four sides surrounding the display area 31.

[0021] As shown in Figure 5, if the image data is obtained by the meter from a slightly rightward angle, the four vertices A, B, C, and D of the rectangular region are specified. Here, the specified rectangular region is assumed to enclose the area from the largest digit value displayed on one end of the digit direction 34 to the smallest digit value displayed on the other end. For example, if the largest digit value is displayed on the left, the rectangular region is specified so that no extra space is included to the left of the area where the largest digit value is displayed. Similarly, if the smallest digit value is displayed on the right, the rectangular region is specified so that no extra space is included to the right of the area where the smallest digit value is displayed.

[0022] (Step S112: Digit count acceptance processing) The setting reception unit 21 accepts input for the number of digits n of the numerical value to be displayed in the display area 31. For example, the setting reception unit 21 displays an input field for the number of digits and prompts the user to enter the number of digits.

[0023] (Step S113: Setting value acceptance processing) The setting reception unit 21 accepts the input of a setting value p. The setting value p is a value used to determine the width of the margin area 33. Here, the setting value p is the ratio of the width of the margin area 33 to the width of the reading area 32. Here, the width of the reading area 32 is the length of one reading area 32 in the digit direction 34. Similarly, the width of the margin area 33 is the length of one margin area 33 in the digit direction 34. In other words, as shown in Figure 6, when the setting value p is 1, the width of the reading area 32 and the width of the margin area 33 are the same. When the setting value p is less than 1, the width of the reading area 32 is wider than the width of the margin area 33. When the setting value p is greater than 1, the width of the reading area 32 is narrower than the width of the margin area 33.

[0024] For example, as shown in Figure 7, the setting reception unit 21 prompts the user to specify a setting value p along with the number of digits n. At this time, the setting reception unit 21 displays n reading areas 32 on the rectangular area received in step S111, and displays blank areas 33 between each reading area 32. When the setting value p is specified using the slider bar, the process in step S12, described later, is executed according to the specified setting value p. In the process of step S12, the width of the reading areas 32 and the width of the blank areas 33 are determined. The setting reception unit 21 changes the display to match the determined width of the reading areas 32 and the width of the blank areas 33.

[0025] (Step S12: Interval identification process) The section identification unit 22 identifies a reading area 32 for each digit of the numerical value within the rectangular area based on the rectangular area, the set value p, and the number of digits n received in step S11. Here, the section identification unit 22 identifies the reading area 32 for each digit value by dividing the rectangular area into sections of reading area 32 for each digit value in the digit direction 34 and sections of blank area 33, under the following conditions: (1) The sections of reading area 32 for each digit value are of the same width. (2) At one end of the rectangular area in the digit direction 34, there is a section of reading area 32 for the largest digit value, and at the other end, there is a section of reading area 32 for the smallest digit value. (3) Between the reading areas 32 for each digit value, there are sections of blank area 33 of the same width.

[0026] Refer to Figure 8 for a detailed explanation of the interval identification process. Here, we will first assume that the quadrilateral region is a rectangle. Then, we will provide supplementary information regarding the case where the quadrilateral region is not a rectangle.

[0027] (Step S121: Section length calculation process) The section identification unit 22 calculates the section length d. As shown in Figure 9, the section length d is the width of each reading area 32. Specifically, the section identification unit 22 calculates the section length d by calculating equation 11. In equation 11, AB is the width of the quadrilateral region in the digit direction 34. That is, AB is the length of side AB extending in the digit direction 34 of the quadrilateral region ABCD. Note that, since the quadrilateral region is a rectangle, the lengths of sides AB and CD extending in the digit direction 34 are the same.

number

[0028] The interval identification unit 22, in the processing from step S122 to step S125, identifies the division points t for each k = 1, ..., n × 2 - 2. k Identify the division point t, as shown in Figure 9. k This point indicates the boundary between the reading area 32 and the margin area 33.

[0029] (Step S122: Initial value setting process) The interval identification unit 22 sets the variable k to an initial value of 1.

[0030] (Step S123: Calculation of division points) The section identification unit 22 determines the division point t based on the section length d and the set value p. k Calculate the position. Specifically, the section identification unit 22 calculates the division point t by calculating the number 12. k Calculate the position of the division point t. k The position is calculated as the distance from one end of the quadrilateral region. In equation 12, equation 13 represents the smallest integer greater than or equal to k / 2. Also, in equation 12, equation 14 represents the largest integer less than or equal to k / 2.

number

number

number

[0031] (Step S124: Variable setting process) The interval identification unit 22 adds 1 to the variable k.

[0032] (Step S125: Termination determination process) The interval identification unit 22 determines whether the variable k has reached n × 2 - 2. The interval identification unit 22 proceeds to step S125 if the variable k reaches n × 2 - 2. On the other hand, if the interval identification unit 22 does not reach n × 2 - 2, it returns to step S123.

[0033] (Step S126: Area setting process) As shown in Figure 9, the section identification unit 22 identifies the division points t on sides AB and CD, respectively, that extend in the vertex direction 34 of the quadrilateral region ABCD, for each k = 1, ..., n × 2 - 2. kBy connecting them with lines, the rectangular area ABCD is divided into a plurality of areas. Then, the section specifying unit 22 alternately sets each area generated by being divided by lines as a reading area 32, a margin area 33, a reading area 32,... in order from one end side in the digit direction 34.

[0034] Here, the description has been made assuming that the rectangular area is a rectangle. Next, when the rectangular area is not a rectangle, for example, in the case of a trapezoid or a rhombus, a supplement is made. As processing when the rectangular area is not a rectangle, the following two can be considered.

[0035] (Method 1) As shown in FIG. 10, when the rectangular area ABCD is not a rectangle, the section specifying unit 22 targets each of the two sides AB and side CD extending in the digit direction 34 of the rectangular area, and for k = 1,..., 2n - 2, the position of the division point t k is specified. That is, the section specifying unit 22 targets side AB and executes the processing from step S121 to step S125 to specify the position of the division point t k in the case of targeting side AB. Also, the section specifying unit 22 targets side CD and executes the processing from step S121 to step S125 to specify the position of the division point t k in the case of targeting side CD. Then, in step S126, the section specifying unit 22 connects the division points t k for the same value k on each of the two sides AB and side CD with lines to divide the rectangular area ABCD into a plurality of areas. After that, similar to the case where the rectangular area is a rectangle, the section specifying unit 22 alternately sets each area generated by being divided by lines as a reading area 32, a margin area 33, a reading area 32,... in order from one end side in the digit direction 34. The processing in step S126 is the same as in the case where the rectangular area is a rectangle.

[0036] (Method 2) The section identification unit 22 projects the image data onto image data of the meter taken from the front. In other words, the section identification unit 22 projects the image data so that the rectangular area becomes a rectangle. Then, using the method described above, the section identification unit 22 identifies the reading area 32 and the blank area 33 for each digit's value. Then, for the reading area 32 and blank area 33 identified within the rectangularized rectangular area, the reverse transformation of the transformation that made the rectangular area a rectangle is performed on both the rectangularized rectangular area and the rectangularized rectangular area.

[0037] (Step S13: Setting storage process) The setting memory unit 23 writes information indicating each read area 32 that was finally set in step S12 to the storage 13. For example, when the confirmation button is pressed on the screen shown in Figure 7, the setting of each read area 32 is considered complete, and the process proceeds to step S13, where information indicating each read area 32 is written to the storage 13. The information indicating each reading region 32 includes the coordinate values ​​in the image data of the four vertices of each reading region 32. For example, the coordinate values ​​in the image data of the four vertices of each reading region 32 can be determined from the coordinate values ​​in the image data of the reference position of the rectangular region.

[0038] The reading process according to Embodiment 1 will be described with reference to Figure 11. (Step S21: Image acquisition process) The image acquisition unit 24 acquires image data of the meter captured by the target camera. The image data acquired here is image data acquired by the same camera and at the same field of view as the image data acquired in the setting process.

[0039] (Step S22: Read execution process) The reading unit 25 reads information indicating each reading area 32 stored in the storage 13 during the setting process. The reading unit 25 identifies each reading area 32 in the image data acquired in step S21 as the target range. The reading unit 25 reads numerical values ​​for each target area in the image data. Specifically, the reading unit 25 processes each reading area 32 and extracts a partial image of the image data within the target area of ​​the reading area 32. The reading unit 25 then inputs the partial image and the instruction to read the value into a pre-trained reading model, causing the image recognition model to read the value. The reading model is a machine learning model that has been trained to read numerical values ​​from image data.

[0040] (Step S23: Display process) The display unit 26 sets the values ​​read in step S22 to each digit and displays a multi-digit number.

[0041] ***Effects of Embodiment 1*** As described above, when the reading device 10 according to Embodiment 1 is given a rectangular area on which a numerical value is displayed, a set value p, and the number of digits, it identifies the reading area 32, which is the area on which the value of each digit is displayed, and the blank area 33 between the two reading areas 32. This makes it possible to easily specify the area to be recognized for each digit of a numerical value without having to manually specify the area to be recognized for each digit. As a result, the accuracy of numerical recognition can be improved.

[0042] Many meters have reading areas 32 where the value of each digit is displayed, arranged at equal intervals and with the same width. Therefore, as in the reading device 10 according to Embodiment 1, it is effective to identify each reading area 32 by considering the width of each reading area 32 section as the same and the width of each blank area 33 section as the same. The ratio of the width of the reading area 32 to the width of the margin area 33 varies depending on the meter. The reading device 10 according to Embodiment 1, as shown in Figure 7, displays the reading area 32 and margin area 33 according to the set value p, and allows the user to set the set value p. In addition, the display of the width of the reading area 32 and the margin area 33 is changed in conjunction with changing the set value p using the slider bar, making it easy to set the set value p appropriately. This makes it possible to set an appropriate value p. In other words, it becomes possible to set an appropriate ratio between the width of the reading area 32 and the width of the margin area 33. As a result, each reading area 32 can be appropriately identified.

[0043] ***Other configurations*** <Example 1> The reading device 10 may also include an editing reception unit that accepts edits to the reading area 32 and the margin area 33 identified by the section identification unit 22. For example, the editing reception unit allows the user to manipulate the reading area 32 or margin area 33 displayed together with the rectangular area shown in Figure 7, and accepts edits to the reading area 32 and margin area 33 identified by the section identification unit 22. For example, if only a portion of the margin area 33 is narrow, the unit accepts edits to narrow that margin area 33. Alternatively, if only a portion of the margin area 33 is misaligned, the unit accepts edits to move that margin area 33. This process is executed, for example, after the process in step S126. When an edit is received that changes the margin area 33, the process returns to step S126, and the read area 32 and the margin area 33 are reset in sequence.

[0044] Embodiment 2. Embodiment 2 differs from Embodiment 1 in that it identifies the reading area 32 and the blank area 33 by inputting image data into an image recognition model to detect the numerical area 35 where numerical values ​​are displayed. Embodiment 2 will explain this difference, and the same points will not be explained.

[0045] ***Explanation of operation*** Referring to Figure 3, the setting process according to Embodiment 2 will be explained. (Step S11: Setting acceptance processing) The setting reception unit 21 accepts input of basic setting information. Specifically, the setting reception unit 21 accepts information for the display area 31 as basic setting information. Here, the information for the display area 31 is a rectangular area surrounding the display area 31, as in Embodiment 1. However, in Embodiment 2, the information for the display area 31 is not limited to a rectangular area surrounding the display area 31, but is sufficient if it indicates the area surrounding the display area 31.

[0046] (Step S12: Interval identification process) The section identification unit 22 identifies the reading area 32 for each digit of the numerical value within the rectangular area based on the rectangular area received in step S11. The section identification process will be explained in detail with reference to Figure 12.

[0047] (Step S121A: Image acquisition process) The section identification unit 22 acquires image data of the meter. The image data acquired here may be the image data used when inputting information for the display area 31 in step S11, or it may be newly acquired image data.

[0048] (Step S122A: Numerical range identification process) The section identification unit 22 inputs the image data acquired in step S121A into the reading model, which is a trained model. This causes the reading model to detect the numerical region 35 where the numerical value is displayed.

[0049] (Step S123A: Exclusion Request Processing) In step S122A, the system accepts exclusion requests to exclude the numerical area 35 detected in step S122A from the settings for the reading area 32.

[0050] (Step S124A: Read area identification process) The section identification unit 22 determines the reading area 32 and the blank area 33 based on the position of the remaining numerical area 35 after excluding the numerical area 35 that was designated for exclusion in step S123A from the numerical area 35 detected in step S122A.

[0051] Specifically, as shown in Figure 13, the section identification unit 22 clusters the numerical regions 35 by enclosing multiple numerical regions 35 whose regions in the digit direction 34 overlap by a certain percentage or more. The section identification unit 22 determines the reading region 32 based on the numerical regions 35 obtained through clustering. Then, the section identification unit 22 determines the blank region 33 based on the remaining region after determining the reading region 32. For example, the section identification unit 22 determines each numerical region 35 obtained through clustering as the reading region 32, and determines the remaining region on the display region 31 after determining the reading region 32 as the blank region 33. In Figure 13, the numerical area 35 of numerical values ​​1, 2, and 3 remains unchanged because there are no other numerical areas 35 whose areas in the digit direction 34 overlap by a certain percentage or more. The numerical area 35 of numerical values ​​4 and 5 overlaps by a certain percentage or more in the digit direction 34, so it is clustered into one numerical area 35. Therefore, the numerical areas 35 of numerical values ​​1, 2, and 3 are each determined as readable areas 32. The numerical area 35 of numerical values ​​4 and 5 is clustered and determined as one readable area 32. A total of four areas are determined as readable areas 32. Then, the area in the display area 31 other than the readable areas 32 is determined as the blank area 33.

[0052] The section identification unit 22 may not simply determine each numerical area 35 obtained by clustering as the reading area 32, but may also determine the range of each numerical area 35 in the digit direction 34 in the display area 31 as the reading area 32, as shown in Figure 14. In other words, the reading area 32 may be determined to match the vertical size of the display area 31.

[0053] ***Effects of Embodiment 2*** As described above, the reading device 10 according to Embodiment 2 identifies the reading area 32 and the blank area 33 by inputting image data into an image recognition model and detecting a numerical area 35 in which a numerical value is displayed. This makes it possible to identify the reading area 32 and the blank area 33 without requiring input of the number of digits n and the set value p. Furthermore, as shown in Figure 15, even when the reading area 32 is unevenly spaced, the reading area 32 and the margin area 33 can be appropriately identified.

[0054] ***Other configurations*** <Modification 2> In Embodiment 1, each functional component was implemented in software. However, in Modification 2, each functional component may be implemented in hardware. The differences between this Modification 2 and Embodiment 1 will be explained below.

[0055] When each functional component is implemented in hardware, the reader 10 includes electronic circuits instead of the processor 11, memory 12, and storage 13. The electronic circuits are dedicated circuits that implement the functions of each functional component, as well as the functions of the memory 12 and storage 13.

[0056] Electronic circuits can include single circuits, complex circuits, programmed processors, parallel programmed processors, logic ICs, GAs, ASICs, and FPGAs. GA stands for Gate Array. ASIC stands for Application Specific Integrated Circuit. FPGA stands for Field-Programmable Gate Array. Each functional component may be implemented in a single electronic circuit, or it may be implemented by distributing each functional component across multiple electronic circuits.

[0057] <Variation 3> As a third variation, some of the functional components may be implemented in hardware, while others may be implemented in software.

[0058] The processor 11, memory 12, storage 13, and electronic circuitry are collectively referred to as the processing circuit. In other words, the function of each functional component is realized by the processing circuit.

[0059] Furthermore, the term "part" in the above explanation may be replaced with "circuit," "process," "procedure," "processing," or "processing circuit."

[0060] The various aspects of this disclosure are summarized below as an appendix. (Note 1) A setting reception unit that receives the setting of the display area in image data of a meter having a display area that displays the value of each digit of a multi-digit number, A section identification unit divides the display area indicated by the setting received by the setting reception unit into a reading area which is the area where the values ​​of each digit are displayed, and a blank area between the two reading areas. A reading device equipped with [a specific feature]. (Note 2) The setting reception unit accepts the setting of a rectangular area as the display area, The section identification unit identifies a reading area, which is the area in which the values ​​of each digit are displayed in the digit direction, which is the direction in which the values ​​of each digit are arranged within the rectangular area, and a blank area, which is the area between the two reading areas. A reading device as described in Appendix 1, comprising the following: (Note 3) The section identification unit identifies the sections of the reading area for each digit and the sections of the margin area based on a set value p that represents the ratio of the width of the margin area to the width d of the reading area. The unit assumes that the sections of the reading area for each digit have the same width, that there are sections of the margin area with the same width between the reading areas for each digit, and that there are sections of reading area at both ends of the rectangular area in the digit direction. The reading device described in Appendix 2. (Note 4) The section identification unit defines the division point t between the section of the reading area and the section of the blank area, expressed by the distance from one end of the rectangular area in the digit direction, for k=1,...,2n-2, where AB is the length of the side extending in the digit direction of the rectangular area, and n is the number of digits in the multiple digits. k By identifying the position based on Equation 1, the reading area interval for each digit and the blank area interval are determined. The reading device described in Appendix 3. (Math 1) (Note 5) The section identification unit targets each of the two sides extending in the direction of the vertex of the quadrilateral region and identifies the division point t for k=1,...,2n-2. k Identify the position of the division point t for the same value k on each of the two sides. k By connecting these points, the reading area interval for each digit and the blank area interval are identified. The reading device described in Appendix 4. (Note 6) The section identification unit projects the image data onto image data of the meter taken from the front, and then identifies the reading area section for each digit's value and the blank area section. The reading device described in Appendix 3 or 4. (Note 7) The section identification unit inputs the image data into a reading model, which is a trained model, and causes the reading model to detect numerical regions where numerical values ​​are displayed. Based on the position of the detected numerical regions, the reading region and the blank region are determined. The reading device described in Appendix 1. (Note 8) The section identification unit determines the reading area by clustering numerical areas by enclosing multiple numerical areas where the digit-direction areas (the direction in which the values ​​of each digit are arranged) overlap by a certain percentage or more, and determines the blank area using the remaining area. The reading device described in Appendix 7. (Note 9) The section identification unit determines the reading area and the blank area based on the position of the remaining numerical area after excluding the numerical area designated for exclusion from the detected numerical area. The reading device described in Appendix 7 or 8. (Note 10) The reading device further includes, A reading unit reads the numerical values ​​displayed by a pre-trained reading model, targeting each of the sections of the reading area identified by the section identification unit within the aforementioned display area. A reading device as described in any one of the appendices 1 to 9, comprising: (Note 11) The computer receives the setting of the display area in image data of a meter having a display area that displays the value of each digit of a multi-digit number, A reading method in which a computer divides the display area indicated by the setting into a reading area, which is the area where the value of each digit is displayed, and a blank area between the two reading areas. (Note 12) A setting acceptance process that accepts the setting of the display area in image data of a meter having a display area that displays the value of each digit of a multi-digit number, A section identification process divides the display area indicated by the setting received by the setting acceptance process into a reading area which is the area where the values ​​of each digit are displayed, and a blank area between the two reading areas. A reading program that makes a computer function as a reading device to perform this task.

[0061] The embodiments and variations of this disclosure have been described above. Some of these embodiments and variations may be implemented in combination. Alternatively, some or all of them may be implemented in part. However, this disclosure is not limited to the embodiments and variations described above, and various modifications are possible as needed. [Explanation of Symbols]

[0062] 10 Reading device, 11 Processor, 12 Memory, 13 Storage, 14 Communication interface, 21 Setting reception unit, 22 Section identification unit, 23 Setting storage unit, 24 Image acquisition unit, 25 Reading unit, 26 Display unit, 31 Display area, 32 Reading area, 33 Margin area, 34 Digit direction, 35 Numeric area.

Claims

1. A setting reception unit that receives the setting of the display area in image data of a meter having a display area which is a rectangular area that displays the value of each digit of a multi-digit number, A section identification unit identifies a section of the display area that divides the display area into a reading area, which is the area in which the values ​​of each digit are displayed in the digit direction, which is the direction in which the values ​​of each digit are arranged, and a blank area between the two reading areas, by identifying the section of the reading area, which is the area in which the values ​​of each digit are displayed, and a blank area between the two reading areas. Equipped with, The section identification unit identifies the sections of the reading area for each digit and the sections of the margin area based on a set value p that indicates the ratio of the width of the margin area to the width d of the reading area. The section identification unit assumes that the sections of the reading area for each digit have the same width, that there are sections of the margin area of ​​the same width between the reading areas for each digit, and that there are sections of reading area at both ends of the rectangular area in the digit direction.

2. The section specification unit defines the length of the side extending in the digit direction of the rectangular region as AB, and the number of digits in the multiple digits as n, for k = 1, ..., 2n-2, the digit direction of the rectangular region By identifying the position of the division point tk between the reading area section and the blank area section, expressed by the distance from one end in the direction, based on Equation 1, the reading area section and the blank area section for each digit value are identified. The reading device according to claim 1. [Math 1]

3. The section identification unit identifies the position of the division point tk for k = 1, ..., 2n-2 for each of the two sides extending in the digit direction of the rectangular region, and identifies the reading region section and the blank region section for each of the two sides by connecting the division points tk for the same value k on each of the two sides. The reading device according to claim 2.

4. The section identification unit projects the image data onto image data of the meter taken from the front, and then identifies the reading area section for each digit's value and the blank area section. The reading device according to claim 1.

5. A setting reception unit that receives the setting of the display area in image data of a meter having a display area that displays the value of each digit of a multi-digit number, The image data is input to a read model, which is a trained model, and the read model detects numerical regions where numerical values ​​are displayed. The numerical regions are clustered by enclosing multiple numerical regions where the regions in the digit direction (the direction in which the values ​​of each digit are arranged) overlap by a certain percentage or more, and the read region, which is the region where the values ​​of each digit are displayed, is determined from the region obtained. The remaining region is used to determine the blank region between the two read regions, and the section identification unit divides the display region indicated by the setting received by the setting reception unit into the read region and the blank region. A reading device equipped with [a specific feature].

6. The section identification unit determines the reading area by clustering the remaining numerical areas after excluding the numerical areas designated for exclusion from the detected numerical areas. The reading device according to claim 5.

7. The reading device further includes, A reading unit reads the numerical values ​​displayed by a pre-trained reading model, targeting each of the sections of the reading area identified by the section identification unit within the aforementioned display area. The reading device according to claim 1 or 5, comprising:

8. The computer receives the setting of the display area in image data of a meter having a display area which is a rectangular area that displays the value of each digit of a multi-digit number. The computer divides the display area into a reading area, which is the area in which the values ​​of each digit are displayed in the digit direction, which is the direction in which the values ​​of each digit are arranged within the rectangular area, which is the display area indicated by the setting, and a blank area, which is the area between the two reading areas, thereby dividing the display area into a reading area, which is the area in which the values ​​of each digit are displayed, and a blank area, which is the area between the two reading areas. A reading method in which a computer identifies the reading area for each digit and the margin area based on a set value p that indicates the ratio of the width of the margin area to the width d of the reading area. The computer assumes that the reading area intervals for each digit are of the same width, that there are margin area intervals of the same width between the reading areas for each digit, and that there are reading area intervals at both ends of the rectangular area in the digit direction.

9. The computer receives the setting of the display area in image data of a meter having a display area that displays the value of each digit of a multi-digit number, A reading method comprising: a computer inputting the image data into a pre-trained reading model, causing the reading model to detect numerical regions where numerical values ​​are displayed; clustering the numerical regions by enclosing multiple numerical regions where the regions in the digit direction (the direction in which the values ​​of each digit are arranged) overlap by a certain percentage or more, thereby determining the reading region, which is the region where the values ​​of each digit are displayed; determining the blank region between the two reading regions using the remaining region, thereby dividing the display region indicated by the setting into the reading region and the blank region.

10. A setting acceptance process that accepts the setting of the display area in image data of a meter having a display area which is a rectangular area that displays the value of each digit of a multi-digit number, The section identification process divides the display area into a section of reading areas, which are areas in the rectangular area that is the display area indicated by the setting received by the setting acceptance process, where the values ​​of each digit are displayed in the digit direction, and a section of blank areas between the two reading areas. The computer functions as a reading device to perform this task. A reading program that identifies the reading area for each digit and the margin area based on a set value p that indicates the ratio of the width of the margin area to the width d of the reading area in the aforementioned interval identification process, assuming that the reading area intervals for each digit have the same width, that there are margin area intervals of the same width between the reading areas for each digit, and that there are reading area intervals at both ends of the rectangular area in the digit direction.

11. A setting acceptance process that accepts the setting of the display area in image data of a meter having a display area that displays the value of each digit of a multi-digit number, The image data is input to a read model, which is a trained model, and the read model detects numerical regions where numerical values ​​are displayed. The numerical regions are clustered by enclosing multiple numerical regions where the regions in the digit direction (the direction in which the values ​​of each digit are arranged) overlap by a certain percentage or more, and the read region, which is the region where the values ​​of each digit are displayed, is determined from the region obtained. The remaining region is used to determine the blank region between the two read regions, and the interval identification process divides the display region indicated by the setting received by the setting acceptance process into the read region and the blank region. A reading program that makes a computer function as a reading device to perform this task.

Images