KR102598210B1 - Drawing information recognition method of engineering drawings, drawing information recognition system, computer program therefor - Google Patents

Abstract

The present invention relates to a drawing information recognition method of engineering drawings, a drawing information recognition system, and a computer program.
According to one aspect of the present invention, the digitalization of engineering drawings is achieved by automatically recognizing and managing drawing information such as symbols, lines, and text included in engineering drawings (e.g., P&ID) managed in the form of image files. A drawing information recognition method of engineering drawings, a drawing information recognition system, and a computer program that enable work to be easily processed are disclosed.

Description

Drawing information recognition method of engineering drawings, drawing information recognition system, computer program {Drawing information recognition method of engineering drawings, drawing information recognition system, computer program therefor}

The present invention relates to a drawing information recognition method of engineering drawings, a drawing information recognition system, and a computer program, which automatically recognize drawing information such as symbols, lines, and text contained in engineering drawings (e.g., P&ID) managed in the form of image files. It relates to a drawing information recognition method of engineering drawings, a drawing information recognition system, and a computer program that facilitates the digitization of engineering drawings by recognizing and managing the information.

Engineering drawings such as P&ID (Piping and instrumentation diagram) are generally created as digital drawings using CAD programs, but it is difficult for creators to directly provide original CAD files to external users due to concerns about technology leakage.

For this reason, the original CAD file is often converted to a PDF file or image file (e.g., JPG, PNG) and provided to external users.

Meanwhile, engineering drawings created manually before using CAD programs are generally managed in the form of paper drawings or blueprints, so there are cases where the original CAD file itself does not exist.

However, in order to effectively utilize engineering drawings, digitization of drawings is sometimes necessary, so there is a need to develop technology that can automatically recognize drawing information such as symbols, lines, and text contained in engineering drawings and manage the information. It was.

Republic of Korea Public Patent No. 10-2014-0046333 (2014.04.18) Republic of Korea registered patent 10-2117323 (2020.05.26) Republic of Korea registered patent 10-2454886 (2022.10.11)

The present invention was developed in consideration of the above problems, and is designed to automatically recognize and manage drawing information such as symbols, lines, and text included in engineering drawings (e.g., P&ID) managed in the form of image files. By doing so, the purpose is to provide a drawing information recognition method of engineering drawings, a drawing information recognition system, and a computer program that facilitate the digitization of engineering drawings.

According to one aspect of the present invention in consideration of the above purpose, a drawing information recognition method executed in a drawing information recognition system, 1) receiving a drawing image of an engineering drawing - the drawing image includes symbols, lines, and text as drawing information. Contains-; 2) recognizing symbols included in the drawing image and generating a symbol image for the recognized symbol; 3) generating a first modified drawing image by removing the recognized symbols from the drawing image, and recognizing drawing lines included in the first modified drawing image; 4) generating a first modified symbol image by removing the symbol line from the symbol image, and recognizing symbol text included in the first modified symbol image; and 5) generating a second revised drawing image by removing the recognized symbol and the recognized drawing line from the drawing image, and recognizing drawing text based on the second revised drawing image. An engineering drawing comprising a. A drawing information recognition method is disclosed.

According to another aspect of the present invention, a memory storing one or more instructions; and a processor executing the one or more commands stored in the memory, wherein the processor receives a drawing image of an engineering drawing, wherein the drawing image includes symbols, lines, and text as drawing information; Recognize symbols included in the drawing image and generate a symbol image for the recognized symbol; generating a first revised drawing image by removing the recognized symbols from the drawing image, and recognizing drawing lines included in the first revised drawing image; Generating a first modified symbol image by removing a symbol line from the symbol image, and recognizing symbol text included in the first modified symbol image; Drawing information of an engineering drawing, characterized in that: generating a second revised drawing image by removing the recognized symbol and the recognized drawing line from the drawing image, and recognizing drawing text based on the second revised drawing image; The recognition system is launched.

According to another aspect of the present invention, there is a computer program stored in a computer-readable medium to execute a drawing information recognition method in a drawing information recognition system, the drawing information recognition method comprising: 1) receiving a drawing image of an engineering drawing; - The drawing image includes symbols, lines, and text as drawing information; 2) recognizing symbols included in the drawing image and generating a symbol image for the recognized symbol; 3) generating a first modified drawing image by removing the recognized symbols from the drawing image, and recognizing drawing lines included in the first modified drawing image; 4) generating a first modified symbol image by removing the symbol line from the symbol image, and recognizing symbol text included in the first modified symbol image; and 5) generating a second revised drawing image by removing the recognized symbol and the recognized drawing line from the drawing image, and recognizing drawing text based on the second revised drawing image. A computer program is disclosed.

The present invention allows the digitalization of engineering drawings by automatically recognizing and managing drawing information such as symbols, lines, and text included in engineering drawings (e.g., P&ID) managed in the form of image files. It has the advantage of being easy to process.

In particular, the present invention stores and manages drawing information about recognized symbols, drawing lines, symbol text, and drawing text as files in a standardized data format (e.g., JSON), so you can work using drawing information recognized in a drawing authoring tool. There is an advantage in that it can be done.

1 is a configuration diagram of a drawing information recognition system according to an embodiment of the present invention.
Figure 2 is a configuration diagram from a hardware perspective of a drawing information recognition system according to an embodiment of the present invention.
3 to 4 are flowcharts of a drawing information recognition method according to an embodiment of the present invention.
5 to 15 are schematic diagrams for explaining a drawing information recognition method according to an embodiment of the present invention.

The present invention can be implemented in various other forms without departing from its technical spirit or main features. Accordingly, the embodiments of the present invention are merely examples in all respects and should not be construed as limited.

Terms such as first, second, etc. are used only for the purpose of distinguishing one component from another component. For example, a first component may be named a second component, and similarly, the second component may also be named a first component without departing from the scope of the present invention.

When a component is referred to as being “connected” or “connected” to another component, it may be directly connected to or connected to the other component, but other components may also exist in between.

As used in this application, singular expressions include plural expressions, unless the context clearly dictates otherwise. In this application, terms such as “comprise”, “provide”, “have”, etc. are intended to express the presence of the components described in the specification or a combination thereof, but do not indicate the possibility that other components or features may be present or added. It is not excluded in advance.

Hereinafter, preferred embodiments according to the present invention will be described in detail with reference to the attached drawings.

FIG. 1 is a configuration diagram of a drawing information recognition system according to an embodiment of the present invention, and FIG. 2 is a configuration diagram from a hardware perspective of a drawing information recognition system according to an embodiment of the present invention.

The drawing information recognition system 100 of this embodiment automatically recognizes drawing information such as symbols, lines, and text included in engineering drawings managed in the form of image files and enables information management.

For example, the engineering drawing of this embodiment may be a drawing that includes symbols, lines (piping lines, instrumentation lines), and text (identification characters for symbols or lines) related to mechanical equipment, piping, and electrical instrumentation, such as a P&ID.

As an example, the drawing image file input through the data input/output interface 6 of the drawing information recognition system 100 is an image file (e.g., by scanning an engineering drawing (e.g., P&ID) in paper or PDF format. It may be converted to JPG or PNG).

From a functional standpoint, the drawing information recognition system 100 of this embodiment includes a drawing image management module 101 that inputs, stores, and manages drawing images of engineering drawings; A symbol information management module 102 that recognizes symbols included in drawing images, extracts symbol images, and generates, stores, and manages symbol-related information; a drawing line information management module 103 that recognizes drawing lines included in a drawing image and stores and manages drawing line-related information; a text information management module 104 that recognizes text included in symbol images and drawing images and stores and manages text-related information; It includes an operation module 105 that provides input/output processing functions for various information related to the drawing information recognition process and provides overall system management functions, including an administrator mode.

As an example, the symbol information management module 102 may include an artificial neural network model to which a deep learning-based object recognition technique is applied for symbol recognition. As such an artificial neural network model, for example, the YOLO (You Only Look Once) model may be used, but is not limited thereto.

In addition, the drawing information recognition system 100 of this embodiment includes a drawing image storage 111 that stores and manages drawing images input through a data input/output interface; a symbol image storage 112 that stores and manages symbol images extracted from drawing images; It includes a drawing information storage 113 that stores and manages drawing information about symbols, lines, and text recognized from drawing images and/or symbol images.

Referring to FIG. 2, from a hardware perspective, the drawing information recognition system 100 of this embodiment includes a memory 2 that stores one or more instructions and a processor 4 that executes the one or more instructions stored in the memory 2. It is a computing device on which a computer program stored in a medium is executed to execute a drawing information recognition method of an engineering drawing. The drawing information recognition system 100 of this embodiment may include a data input/output interface 6, a communication interface 8, a data display means 3, and a data storage means 5.

3 to 4 are flowcharts of a drawing information recognition method according to an embodiment of the present invention, and FIGS. 5 to 15 are schematic diagrams for explaining a drawing information recognition method according to an embodiment of the present invention.

In step 1), the drawing information recognition system 100 receives a drawing image of an engineering drawing. The drawing image includes symbols, lines, and text as drawing information.

For example, the drawing image may be an engineering drawing (e.g., P&ID) in paper or PDF format converted into an image file (e.g., JPG, PNG) by scanning.

Various preprocessing (Normalization, Resize, Threshold, Blurr, Morphological Transformation, etc.) can be performed on the input drawing image to facilitate recognition of symbols, lines, and text.

In step 2), the drawing information recognition system 100 recognizes symbols included in the drawing image and generates a symbol image for the recognized symbol.

Preferably, in step 2), symbols included in the drawing image are recognized based on a deep learning-based object recognition technique.

As an example of a deep learning-based object recognition technique, an object recognition technique using the YOLO model can be used. YOLO is an algorithm that provides real-time object detection using artificial neural networks.

For example, the YOLO model can be learned in advance through the learning server of the drawing information recognition system 100 as follows. Figure 6 illustrates the process of model learning and execution of the learned model in this embodiment.

Drawing images of engineering drawings containing symbols are input to be used as learning data. Various preprocessing can be performed on the input drawing image to facilitate symbol recognition.

An annotation is assigned to each symbol included in the drawing image. The annotation consists of a class label and boundary-box for each symbol. Various well-known programs can be used for annotation assignment. Annotation-related information assigned to each symbol may be stored and managed in the form of an annotation document (e.g., annotation xml document).

The YOLO model is trained using each annotated symbol as training data. For example, a known deep learning framework (e.g., Darknet framework) may be used to learn and execute the YOLO model, but is not limited thereto.

Recognition of symbols included in drawing images can be performed using the YOLO model learned as above, and other known models may be used as long as they are artificial neural network models using deep learning-based object recognition techniques.

Preferably, unnecessary lines may be removed from the generated symbol image. Figure 8 illustrates a case of removing unnecessary lines from a symbol image. For example, removal of unnecessary lines can be accomplished by removing non-line parts constituting the symbol as noise using a morphological transformation technique in a manner similar to the drawing line recognition process described later.

In step 2), recognition of symbols included in the drawing image involves recognizing the type, location, and size of the symbol. Figure 7 illustrates a visualization state of the result of recognizing a symbol object in a drawing image.

For example, if it is necessary to recognize the direction angle of the symbol in step 2) above, for a symbol with symmetrical up-down and left-right directions, the direction angle of the symbol is recognized based on the horizontal and vertical length of the symbol, and the up-down or left-right For asymmetric symbols, the direction angle can be recognized based on the asymmetrical geometric features included in the symbol. For example, asymmetrical geometric features included in a symbol may be the number of asymmetrically arranged vertices of the symbol, the location and shape of polygons within the symbol, etc.

By recognizing symbols included in drawing images based on deep learning-based object recognition techniques, the type of symbol can be known. For each type of symbol, information about whether it is a symmetrical symbol or an asymmetrical symbol is stored in advance, the horizontal/vertical ratio is stored in advance for each type of symmetrical symbol, and the vertices are asymmetrically arranged for each type of asymmetrical symbol. If information on location, number of vertices, polygon location, polygon shape, etc. is stored in advance, the direction angle of the symbol can be determined based on the stored information.

In step 3), the drawing information recognition system 100 generates a first revised drawing image from which the recognized symbols in the drawing image are removed, and recognizes drawing lines included in the first revised drawing image.

Drawing lines refer to lines in a drawing rather than lines constituting symbols, and include, for example, piping lines, instrumentation lines, distance marking lines, etc.

In order to prevent the lines constituting the symbol from being recognized as drawing lines, in step 3), the symbols recognized in the symbol recognition process are removed from the drawing image and then the drawing line recognition process is performed. Figure 9 illustrates a process for generating a first revised drawing image by removing a recognized symbol from the original drawing image.

Preferably, in step 3), the drawing lines included in the first revised drawing image are recognized using a morphological transformation technique.

Images typically consist of raster images. Raster images contain a lot of noise, and since noise is a major factor that interferes with object detection, it is necessary to minimize or remove noise. For example, since drawing elements other than the object to be detected (e.g., drawing lines) can be viewed as noise, a morphological transformation technique is used to minimize or remove noise that interferes with the detection of drawing lines.

Morphological transformation is a technique that approaches images from a morphological perspective. In a binarized image, if pixels with the same value are adjacent, this can be considered a morphological region. Morphological transformation of an image is a method of changing a region using image filtering. It is widely used in the pre- or post-processing process of images, and is used to remove noise, define gradients, combine and separate elements, and detect intensity peaks. Morphological transformations include the basic operations of dilation and erosion. The dilation operation and the erosion operation are convolution operations between the image and the kernel, and complex and diverse morphological transformation techniques can be implemented based on these basic operations. The kernel is a small space in the image that includes the X and Y pixels and the surroundings of those pixels.

The dilation operation is an operation that replaces all pixel values within the kernel area with the local maximum inside the kernel. When the dilation operation is applied, the dark area decreases and the bright area increases. The dilation operation is mainly used when trying to restore the reduced size after noise removal.

The erosion operation is an operation that replaces the values of all pixels within the kernel area with the local minimum inside the kernel. When the erosion operation is applied, the bright area decreases and the dark area increases. The erosion operation is mainly used for noise removal.

By setting various shapes of the kernel using structural elements and applying various transformation operations based on the basic operations described above, drawing elements corresponding to noise are removed and drawing elements to be detected (e.g. drawing line) can be detected.

In this way, using the morphological transformation technique, noise in the first revised drawing image can be removed and drawing lines can be detected. For example, the outline of the detected drawing line can be extracted using Open CV's outline extraction library, and the coordinates of each vertex constituting the outline of the drawing line, the thickness of the line, etc. can be extracted using this. Figure 10 illustrates a visualization state of the result of recognizing a drawing line in a drawing image.

For example, the recognition of drawing lines included in the first modified drawing image is configured to first recognize one type of horizontal drawing line and vertical drawing line, and then recognize the other type. For example, when the horizon drawing line is first recognized, a morphological transformation technique can be used to remove drawing elements other than the horizon drawing line as noise and recognize the start point coordinates, end point coordinates, and thickness of the horizon drawing line. Afterwards, vertical drawing lines can be recognized in the same way.

For example, in step 3), the drawing line included in the first revised drawing image is recognized by recognizing the start point coordinates, end point coordinates, and thickness of the drawing line. The coordinates are the X, Y coordinates in the first modified drawing image.

Preferably, in step 3), the recognition of drawing lines included in the first modified drawing image is configured to recognize only lines having pixels greater than a preset reference value as drawing lines. For example, n pixels ( Yes, it can be configured to detect only lines of 50 pixels or longer.

Meanwhile, in step 3), the recognition of the drawing line included in the first revised drawing image detects a polygon in the area around the end point based on information on each end point of the recognized line, and determines whether it is an arrow or an arrow based on the polygon. It is configured to recognize whether it is a general straight line or not. For example, if the polygon in the area around the end point of the recognized line is a triangle, it can be recognized as an arrow, and if it is a square, it can be recognized as a regular straight line. For example, information on each end point of the outline surrounding a drawing line can be obtained by extracting the outline of the drawing line using Open CV's outline extraction library.

In step 4), the drawing information recognition system 100 generates a first modified symbol image by removing the symbol line from the symbol image and recognizes the symbol text included in the first modified symbol image. Figure 11 illustrates the process of recognizing symbol text within a symbol image.

Preferably, in step 4), the recognition of the symbol text is performed by detecting the text in the first modified symbol image from which the symbol line has been removed, and recognizing the text while rearranging the detected text into a preset format ( recognition).

In the case of symbols containing text within the symbol, the lines of the symbol may interfere with text recognition, so the lines of the symbol are first removed through preprocessing, and then detection of the symbol text is performed.

In order to prevent a decrease in recognition rate and recognition speed caused by irregular text arrangement within the symbol, the text within the symbol is rearranged into a standard format and then the symbol text is recognized. For example, formal formatting can be done by arranging text horizontally to facilitate OCR recognition of the text.

After text recognition is complete, the previous text location information and text content are collected to extract the text's location, size, content, and symbol affiliation information.

In step 5), the drawing information recognition system 100 generates a second revised drawing image by removing the recognized symbol and the recognized drawing line from the drawing image, and creates drawing text based on the second revised drawing image. recognize Figure 12 illustrates a process of generating a second revised drawing image by removing recognized symbols and recognized drawing lines from the drawing image.

Preferably, step 5) may have the following detailed configuration. Figure 13 is a diagram for understanding the detailed configuration.

In step 51), the drawing information recognition system 100 generates a second revised drawing image by removing the recognized symbol and the recognized drawing line from the drawing image. (a in Figure 13)

In step 52), the drawing information recognition system 100 detects character-by-character text in the second modified drawing image and generates a 2-1 modified drawing image from which text, lines, and symbols other than horizontal writing are removed. . (b in Figure 13)

In step 53), the drawing information recognition system 100 detects word-by-word text in the 2-1 revised drawing image. (c in Figure 13)

In step 54), the drawing information recognition system 100 removes the horizontal text detected as word-by-word text in step 53) from the second revised drawing image and generates a third revised drawing image rotated by 90 degrees.

In step 55), the drawing information recognition system 100 detects character-by-character text in the third modified drawing image and generates a 3-1 modified drawing image from which text, lines, and symbols other than horizontal writing are removed. In step 55), the horizontally written text corresponds to the vertically written text based on the original drawing image.

In step 56), the drawing information recognition system 100 detects word-by-word text in the 3-1 revised drawing image.

In step 57), the drawing information recognition system 100 recognizes the drawing text based on the word-unit text detected in steps 53) and 56), respectively. Recognition of drawing text can be performed using an OCR program.

For example, in step 57), the drawing text is recognized by rearranging the detected word-by-word text into a preset format and then recognizing the text. For example, text rearrangement can be accomplished by arranging text horizontally to facilitate OCR recognition.

For example, when recognizing drawing text, recognition is performed by extracting a text image corresponding to the location (X, Y, Width, Height) of the drawing text detected in the previous process. In addition, when recognizing drawing text, each text image is rearranged in a regular form to improve recognition rate and recognition speed, and text recognition is performed using an OCR engine, and then text information is extracted using the coordinate information of the original location. .

Figure 14 overall illustrates the process of recognizing symbol text and drawing text.

Through the above process, the following recognition results can be obtained.

Information such as X and Y coordinates, Width, Height, symbol type, and direction angle for the recognized symbol can be extracted. Information such as X and Y coordinates, text size, and text content about recognized text (symbol text, drawing text) can be extracted. You can extract line information such as start X and Y coordinates, end X and Y coordinates, thickness, and arrows of a recognized drawing line.

Additionally, the extracted information can be reconstructed to create a structured text file in JSON format.

Text files structured in JSON format can be converted to JSON files through a converter and stored. The JSON file can store and manage drawing information about symbols, drawing lines, symbol text, and drawing text recognized through the above process.

For example, in a drawing authoring tool, you can load a JSON file and modify/edit drawing information. Additionally, it is possible to create a database by batching multiple drawings using a drawing authoring tool.

Figure 15 illustrates a process for generating a structured drawing information file by extracting symbols, drawing lines, and text information from a drawing image.

Embodiments of the present invention include programs for execution on various computers and computer-readable recording media recording the programs. The computer-readable recording medium may include program instructions, data files, data structures, etc., singly or in combination. The media may be those specifically designed and constructed for the present invention or may be known and available to those skilled in the art of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical recording media such as CD-ROMs, DVDs, and USB drives, magneto-optical media such as floptical disks, and ROM, RAM, It includes hardware devices specifically configured to store and execute program instructions, such as flash memory. Examples of program instructions include machine language code, such as that produced by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc.

100: Drawing information recognition system

Claims (16)

As a drawing information recognition method implemented in a drawing information recognition system,
1) Step of receiving a drawing image of an engineering drawing - the drawing image includes symbols, lines, and text as drawing information -;
2) recognizing symbols included in the drawing image and generating a symbol image for the recognized symbol;
3) generating a first modified drawing image by removing the recognized symbols from the drawing image, and recognizing drawing lines included in the first modified drawing image;
4) Creating a first modified symbol image by removing the symbol line from the symbol image, and recognizing the symbol text included in the first modified symbol image - The symbol image for generating the first modified symbol image includes text within the symbol It is an image of a symbol containing -; and
5) generating a second revised drawing image by removing the recognized symbol and the recognized drawing line from the drawing image, and recognizing drawing text based on the second revised drawing image; of an engineering drawing configured including; How to recognize drawing information.
According to paragraph 1,
In step 2) above,
A drawing information recognition method of an engineering drawing, characterized in that symbols included in the drawing image are recognized based on a deep learning-based object recognition technique.
According to paragraph 1,
In step 2) above,
A drawing information recognition method of an engineering drawing, characterized in that recognition of symbols included in the drawing image recognizes the type, location, and size of the symbol.
According to paragraph 3,
If it is necessary to recognize the direction angle of the symbol in step 2) above,
For symbols that are symmetrical in the top-bottom and left-right directions, the direction angle of the symbol is recognized based on the horizontal and vertical length of the symbol.
A method for recognizing drawing information in engineering drawings, characterized in that the direction angle of a symbol that is asymmetrical up and down or left and right is recognized based on the asymmetrical geometric features included in the symbol.
According to paragraph 1,
In step 3) above,
A drawing information recognition method of an engineering drawing, characterized in that the drawing lines included in the first revised drawing image are recognized using a morphological transformation technique.
According to clause 5,
A drawing information recognition method of an engineering drawing, characterized in that the recognition of the drawing line included in the first revised drawing image is configured to first recognize one type of a horizontal drawing line and a vertical drawing line, and then recognize the other type. .
According to paragraph 1,
In step 3) above,
A drawing information recognition method of an engineering drawing, characterized in that the recognition of the drawing line included in the first revised drawing image recognizes the start point coordinates, end point coordinates, and thickness of the drawing line.
In clause 7,
In step 3) above,
A drawing information recognition method of an engineering drawing, characterized in that the recognition of drawing lines included in the first revised drawing image is configured to recognize only lines having pixels greater than a preset reference value as drawing lines.
In clause 7,
In step 3) above,
Recognition of the drawing line included in the first revised drawing image detects a polygon in the area around the end point based on information on each end point of the recognized line, and distinguishes whether it is an arrow or a general straight line based on the polygon. A drawing information recognition method of an engineering drawing, characterized in that:
According to paragraph 1,
In step 4) above,
Recognition of the symbol text is,
Detect text in the first modified symbol image from which the symbol line has been removed,
A drawing information recognition method for engineering drawings, characterized in that the text is recognized while the detected text is rearranged in a preset format.
According to paragraph 1,
In step 5) above,
51) generating a second revised drawing image by removing the recognized symbol and the recognized drawing line from the drawing image;
52) detecting character-by-character text in the second modified drawing image and generating a 2-1 modified drawing image by removing text, lines, and symbols other than horizontal writing;
53) detecting word-level text in the 2-1 modified drawing image;
54) removing the horizontally written text detected as word-by-word text in step 53) from the second revised drawing image and generating a third revised drawing image rotated by 90 degrees;
55) detecting character-by-character text in the third modified drawing image and generating a 3-1 modified drawing image by removing text, lines, and symbols other than horizontal writing;
56) detecting word-level text in the 3-1 modified drawing image; and
57) Recognizing drawing text based on word-unit text detected in steps 53) and 56), respectively. A drawing information recognition method of an engineering drawing, comprising:
According to clause 11,
In step 57) above,
Recognition of the drawing text is,
A drawing information recognition method for engineering drawings, characterized in that the text is recognized while the detected word-unit text is rearranged in a preset format.
According to paragraph 1,
A drawing information recognition method of an engineering drawing, characterized in that drawing information about the recognized symbol, recognized drawing line, recognized symbol text, and recognized drawing text are converted into a JSON file and stored and managed.
According to paragraph 1,
A method of recognizing drawing information of an engineering drawing, characterized in that the engineering drawing is a P&ID (Piping and instrumentation diagram).
Memory that stores one or more instructions; and
a processor executing the one or more instructions stored in the memory,
The processor,
Receives a drawing image of an engineering drawing - the drawing image includes symbols, lines, and text as drawing information -;
Recognize symbols included in the drawing image and generate a symbol image for the recognized symbol;
generating a first revised drawing image by removing the recognized symbols from the drawing image, and recognizing drawing lines included in the first revised drawing image;
Create a first modified symbol image by removing the symbol line from the symbol image, recognize the symbol text included in the first modified symbol image, and - the symbol image for generating the first modified symbol image includes text in the symbol It is an image of a symbol-;
Drawing information of an engineering drawing, characterized in that: generating a second revised drawing image by removing the recognized symbol and the recognized drawing line from the drawing image, and recognizing drawing text based on the second revised drawing image; Recognition system.
A computer program stored in a computer-readable medium to execute a drawing information recognition method in a drawing information recognition system,
The drawing information recognition method is,
1) Step of receiving a drawing image of an engineering drawing - the drawing image includes symbols, lines, and text as drawing information -;
2) recognizing symbols included in the drawing image and generating a symbol image for the recognized symbol;
3) generating a first modified drawing image by removing the recognized symbols from the drawing image, and recognizing drawing lines included in the first modified drawing image;
4) Creating a first modified symbol image by removing the symbol line from the symbol image, and recognizing the symbol text included in the first modified symbol image - The symbol image for generating the first modified symbol image includes text within the symbol It is an image of a symbol containing -; and
5) generating a second revised drawing image by removing the recognized symbol and the recognized drawing line from the drawing image, and recognizing drawing text based on the second revised drawing image; A computer program that does.