Method for extracting text information

Abstract

To improve the accuracy of extracting text information from drawings. [Solution] The information processing device performs the following steps: acquiring a drawing image (S100); dividing the acquired drawing image into a first area containing textual information about the part shape and a second area different from the first area (S110); extracting textual information from the first area using a first extraction method (S120); and extracting textual information from the second area using a second extraction method different from the first extraction method (S130).

Description

Technical Field

[0001] This invention relates to a method for extracting character information.

Background Art

[0002] In Patent Document 1, in order to restore an image with degradation to high precision, a learned model corresponding to each of a plurality of character types is prepared. Then, an image of character information is restored using the learned model corresponding to the specified character type.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] By the way, special symbols are used in the drawings, or characters are written in directions other than the horizontal direction. Therefore, when extracting character information from the image of the drawing, even if a learning model corresponding to the character type is used, there is a possibility that the character information cannot be accurately extracted from the drawing.

Means for Solving the Problems

[0005] A method for extracting character information for solving the above problems includes a step of acquiring a drawing image, a step of dividing the acquired drawing image into a first area in which character information related to the component shape is described and a second area that is an area different from the first area, a step of extracting character information from the first area by a first extraction method, and a step of extracting character information from the second area by a second extraction method different from the first extraction method.

[0006] Examples of first and second extraction methods include using OCR (Optical Character Recognition) with different learning models, or inputting different prompts when extracting text information using artificial intelligence. [Effects of the Invention]

[0007] According to the above method for extracting text information, the accuracy of extracting text information from drawings is improved. [Brief explanation of the drawing]

[0008] [Figure 1] Figure 1 is a schematic diagram showing the configuration of a character information extraction system. [Figure 2] Figure 2 is a schematic diagram showing the drawing from which text information is to be extracted. [Figure 3] Figure 3 is a flowchart showing the processing flow executed by the processing circuit of the information processing device. [Modes for carrying out the invention]

[0009] Below, one embodiment of the method for extracting text information will be described with reference to Figures 1 to 3. <Configuration of the Information Processing System> Figure 1 shows the configuration of the character information extraction system. The character information extraction system includes a camera 200, an information processing device 100, a display device 300, a terminal 400, etc.

[0010] The camera 200 is equipped with an image sensor and captures a drawing 500 containing text information, transmitting the drawing image to the information processing device 100. In this embodiment, the camera 200 is directly connected to the information processing device 100, but the image captured by the camera 200 may also be imported into the information processing device 100 via a network. Alternatively, the image captured by the camera 200 may be saved to a server or the like, and then imported into the information processing device 100 via a network.

[0011] As shown in Figure 2, drawing 500 contains a first area A, which contains part shapes and textual information related to the part shapes, and a second area B, which is a different area from the first area A. More specifically, the second area B is the area in drawing 500 other than the first area A.

[0012] The textual information regarding the part shape described in Area 1A includes, for example, dimensions. Additionally, the textual information regarding the part shape described in Area 1A includes special symbols such as surface roughness. The orientation of this textual information varies depending on the part shape and the direction of dimension extension lines, and therefore it is not always written horizontally. The characteristics of the textual information described in Area 1A are that it may include special symbols and is not always written horizontally.

[0013] Area 2B contains textual information such as product numbers, product names, and designer names. A characteristic of the textual information in Area 2B is that it is generally written horizontally.

[0014] The information processing device 100 includes a processing circuit 110. The processing circuit 110 includes a CPU and ROM, and the CPU executes various processes by running programs stored in the ROM.

[0015] The display device 300 displays the processing results output from the information processing device 100, such as drawing images and text information extracted from the drawing images. Terminal 400 is equipped with input devices such as a keyboard and can give various instructions to the information processing device 100.

[0016] <Extracting text information> The processing circuit 110 of the information processing device 100 extracts character information from the drawing 500 by implementing the character information extraction method described below.

[0017] FIG. 3 shows the procedure of the character information extraction method executed by the processing circuit 110. In the following, the step numbers of each procedure are represented by numbers with "S" attached at the beginning. In this extraction method, the processing circuit 110 reads a drawing image that is an image of Drawing 500 (S100). The drawing image read at this time is an image taken by the camera 200 or an image generated from drawing data.

[0018] Next, the processing circuit 110 executes a process of dividing areas in the read drawing image (S110). In this S110, the processing circuit 110 divides the drawing image into the above-described first area A and second area B. This division of areas can be performed by any one of the following division methods 1 to 3, for example.

[0019] · Division method 1: The first area A and the second area B are predefined for each drawing paper size, and the processing circuit 110 divides the first area A and the second area B according to the paper size of the read drawing image. In this case, each area can be automatically divided.

[0020] · Division method 2: The first area A and the second area B are automatically divided using a pre-learned AI model. Also in this case, each area can be automatically divided. Further, even when the second area B is not concentrated in one place and is divided, the second area B can be appropriately divided.

[0021] · Division method 3: The operator manually divides the first area A and the second area B by operating the terminal 400 while checking the display device 300. When the process of S110 is executed, next, the processing circuit 110 extracts character information from the first area A using the first extraction method (S120). The first extraction method can be performed by any one of the following extraction methods 1 and 2, for example.

[0022] · Extraction method 1: Prepare a learning model optimized for the characteristics of the character information in the first area A, and perform extraction of character information by OCR using that learning model. • Extraction Method 2: Extract character information using multimodal AI, and input prompts considering the characteristics of the character information in Area 1A. An example of a prompt in this case would be "Read while rotating the characters." Such prompts may be automatically provided from pre-prepared sources, or they may be manually provided by the operator.

[0023] After executing the process in S120, the processing circuit 110 then extracts character information from the second area B using the second extraction method (S130). The second extraction method can be performed using, for example, one of the following extraction methods 3 and 4.

[0024] • Extraction Method 3: Prepare a learning model optimized for the characteristics of the character information in Area B 2, and use OCR with that learning model to extract the character information. • Extraction Method 4: Extract text information using multimodal AI, and input prompts considering the characteristics of the text information in Area 2B. An example of a prompt in this case would be "The text is written horizontally in a horizontal orientation." Such prompts may be automatically provided from pre-prepared sources, or they may be manually provided by the operator.

[0025] When the process in S130 is executed, the processing circuit 110 then combines the character information extracted from the first area A and the character information extracted from the second area B so that they can be treated as a single character information file (S140).

[0026] Then, after executing the process in S140, the processing circuit 110 terminates this process. <Operation and Effects of This Embodiment> The method for extracting text information in this embodiment includes the steps of acquiring a drawing image (S100) and dividing the acquired drawing image into a first area A containing text information about the part shape and a second area B which is an area other than the first area A (S110). Furthermore, the method for extracting text information in this embodiment includes the steps of extracting text information from the first area A using a first extraction method (S120) and extracting text information from the second area B using a second extraction method different from the first extraction method (S130).

[0027] Therefore, according to the character information extraction method of this embodiment, character information is extracted using different extraction methods for each area of ​​the drawing image, rather than by character type. As a result, it becomes possible to apply an extraction method to each area that corresponds to the characteristics of the character information described in that area. Consequently, the extraction accuracy when extracting character information from a drawing is improved.

[0028] <Example of changes> This embodiment can be implemented with the following modifications. This embodiment and the following modifications can be combined with each other to the extent that they do not contradict each other technically.

[0029] The area may be divided using methods other than those described in division methods 1-3 above. The first extraction method may be a method different from the extraction methods 1 and 2 described above. The second extraction method may be different from the extraction methods 3 and 4 described above. [Explanation of symbols]

[0030] 100... Information Processing Device 110… Processing circuit 200... Camera 300…Display device 400… Terminal 500... Drawing

Claims

[Claim 1] Steps to obtain drawing images, The acquired drawing image is divided into a first area containing textual information about the part shape and a second area which is different from the first area. The first extraction method involves extracting character information from the first area, The process includes the step of extracting character information from the second area using a second extraction method different from the first extraction method. Method for extracting text information.

Images