JP7451373B2 - Drawing structuring system and drawing structuring method - Google Patents

Description

Embodiments of the present invention relate to drawing structuring techniques.

There are drawings that depict a large number of equipment or parts, such as plant system diagrams or circuit diagrams. Elements such as equipment or parts depicted in a drawing are called symbols. A large amount of such drawings are stored in raster format. 2. Description of the Related Art Conventionally, there is a known technique for creating structured data indicating connections between symbols by converting a raster format drawing into a vector format drawing.

Japanese Patent Application Publication No. 2018-206250

In the conventional technology, after converting a raster format drawing into a vector format drawing, symbols drawn in the vector format drawing are recognized (extracted). Therefore, there is a problem in that errors that occur when converting to a vector format drawing are added to the accuracy of symbol recognition, reducing the accuracy of structuring.

The embodiments of the present invention have been made in consideration of such circumstances, and an object thereof is to provide a drawing structuring technique that can improve the accuracy of recognizing symbols drawn on drawings.

A drawing structuring system according to an embodiment of the present invention includes a structured rule definition database that stores attribute information indicating attributes of symbols and rules for defining connection relationships between the symbols and line segments, and a raster format drawing. a symbol recognition function unit that recognizes the symbol drawn in the drawing, a vector conversion function unit that converts the raster format drawing into a vector format drawing, and a symbol recognition function unit that recognizes the symbol specified based on the recognition result of the symbol recognition function unit. A structure for structuring the vector format drawing by assigning the attribute information to the symbol using the attribute information and the rule placed in the vector format drawing and stored in the structuring rule definition database. and a conversion function section.

Embodiments of the present invention provide a drawing structuring technique that can improve the accuracy of recognizing symbols drawn in drawings.

1 is a block diagram showing a drawing structuring system. FIG. 3 is a block diagram showing a symbol recognition function unit. FIG. 3 is a block diagram showing a structuring function unit. An explanatory diagram showing a raster format drawing. An explanatory diagram showing symbols drawn on a raster format drawing. 1 is a flowchart showing processing performed by the drawing structuring system. 1 is a flowchart showing processing performed by the drawing structuring system. 1 is a flowchart showing processing performed by the drawing structuring system. 1 is a flowchart showing processing performed by the drawing structuring system. 1 is a flowchart showing processing performed by the drawing structuring system. 1 is a flowchart showing processing performed by the drawing structuring system.

Hereinafter, embodiments of a drawing structuring system and a drawing structuring method will be described in detail with reference to the drawings.

Reference numeral 1 in FIG. 1 is a drawing structuring system of this embodiment. This drawing structuring system 1 converts a raster format drawing 2 into a vector format drawing, and automatically generates a structured drawing 3 based on this vector format drawing. Note that the raster format drawing 2, vector format drawing, and structured drawing 3 will be described on the assumption that all of them are digitized electronic files (electronic data).

The raster format drawing 2 to be converted in this embodiment includes, for example, a drawing related to a plant. Here, the drawings related to the plant include the plant building, layout diagram, P&ID (piping and instrumentation diagram), ECWD (expanded connection diagram), IBD (interlock block diagram), single line diagram, soft logic diagram, etc. .

Many of these drawings are managed on paper or as raster format drawings 2 obtained by reading paper with a scanner. In such a raster format drawing 2, information indicating the mutual relationships between symbols, wiring, parts, etc. drawn in the drawing is lost, and the drawing is not structured. On the other hand, structured drawings 3 created using CAD software are created while adding information indicating the relationships between symbols, wiring, parts, etc., so they are structured drawings. There is.

If it is possible to convert the raster format drawing 2 into a structured drawing 3 created using CAD software, it is possible to use the structured drawing 3 to explore the circuit or system of a plant and simulate its behavior. Can be applied to simulation. For example, it becomes possible to construct a system for planning isolation work in a plant. Furthermore, even in areas other than plants, for example, by structuring raster-format drawings regarding water and sewage piping, it becomes possible to perform simulations to predict or predict the effects of water leakage due to pipe damage using the structured drawings. Therefore, the drawing structuring system 1 of this embodiment converts the raster format drawing 2 into a vector format drawing, and converts this into a structured drawing 3. In this way, the existing raster format drawing 2 can be effectively utilized.

Raster format drawing 2 (bitmap image) is a drawing in which equipment, valves, pumps, electronic elements, circuits, piping, wiring, tables, characters, etc. drawn on the drawing are all collections of dots (pixels). be. It is a paper drawing or an electronic file drawing such as TIF, PDF, or JPEG that is obtained by reading a paper drawing with a scanner. In the following description, components such as devices, valves, pumps, electronic elements, circuits, etc. depicted in the drawings are referred to as symbol 4 (FIG. 5). For example, when the symbol 4 is an electronic element, it includes a coil, a switch, a lamp, etc.

Note that the symbols 4 may include predetermined symbols, tables, characters, and the like. Furthermore, components connected to the symbol 4, such as piping and wiring, are referred to as line segments 5 (FIG. 5). Note that the line segment 5 may be a straight line, a curved line, or a bent line.

In the raster format drawing 2, attribute information indicating the attribute of the symbol 4 and connection information indicating the line segment 5 connected to the symbol 4 are not provided to each drawn symbol 4. In other words, it is an unstructured drawing.

The attribute information is information regarding the type of the symbol 4 (eg, type of coil, A contact, pump, etc.), the name of the symbol 4 (eg, device identification number), detailed specifications of the symbol 4, and the like. It also refers to information regarding the type of line segment (for example, type of wiring, piping, etc.), the name of the line segment (for example, the identification number of piping), the detailed specifications of the line segment, and the like. In addition, in the case where the line segment is a pipe, this is information regarding the type of fluid flowing inside the pipe, etc. It also refers to information regarding the board name of the board frame 8, etc.

The connection information is, for example, information indicating a connection destination (connection relationship), such as whether wires are connected to each other or which symbol a pipe is connected to. This connection information includes information regarding line segments 5 such as wiring or piping. Note that a vector-format drawing (vector image) is a drawing composed of the coordinates of the starting point and end point of the line segment 5, a function representing the line segment 5, and the like.

Further, the structured drawing 3 is a drawing (vector image) in which at least symbols 4 other than characters and alphanumeric characters, line segments 5, etc. are drawn in vector format. In the structured drawing 3, attribute information and connection information are given to each drawn symbol 4 and line segment 5. Further, character (including alphanumeric) information is associated with the structural element including the symbol 4 and the line segment 5 as attribute information.

Hierarchical information indicating the hierarchy is given to the symbol 4 drawn in the structured drawing 3. Therefore, data including structured symbols can be classified into layers and recorded on a recording medium.

FIG. 5 shows an expanded connection diagram. For example, a relay 6 included in symbol 4 is depicted in this expanded connection diagram. The attribute information in this embodiment is, for example, that the symbol 4 is a relay 6, that this relay 6 has a device name "49X" (character 9), and that it is a connection point with wirings 7a and 7b. This is information indicating that it has "6A" and "6B". Further, the connection information is, for example, information indicating that the connection point "6A" of the relay 6 is connected to the wiring 7a, and that the wiring 7a is further connected to the relay 6 and its connection point. The expanded connection diagram also includes the panel frame 8 and the like. Furthermore, characters such as 9 including alphanumeric characters are also written.

Note that when the symbol 4 is a device such as a valve or a pump, the attribute information is, for example, information such as the type and name of the device such as the pump. In the case of piping, this information is information regarding the type and name of the piping, and the type of fluid. The connection information is information indicating the connection relationship between equipment such as valves or pumps and piping.

Next, the system configuration of the drawing structuring system 1 will be explained with reference to block diagrams shown in FIGS. 1 to 3.

As shown in FIG. 1, the drawing structuring system 1 includes a pre-processing section 10, a vector conversion function section 11, a character recognition function section 12, a symbol recognition function section 13, an erroneous recognition check function section 14, and a structuring function section 15. Erroneous recognition data storage section 16, optimal recognition learning section 17, structured check function section 18, incorrect structured data storage section 19, optimal structured learning section 20, learning drawing generation section 21, and learned model generation section 22. Be prepared. These are realized by the CPU executing programs stored in the memory or HDD.

Further, the drawing structuring system 1 includes a structuring rule definition database 23, a data structure definition database 24, and a symbol recognition database 25. These are collections of information stored in memory or HDD and organized so that they can be searched or stored.

The structured rule definition database 23 is provided with a structured rule information registration section 26, an attribute information registration section 27, a connection information registration section 28, and an optimal structured learning data registration section 29 (FIG. 3). The data structure definition database 24 is provided with a hierarchical information registration section 30 (FIG. 3). The symbol recognition database 25 is provided with a trained model registration section 31 and a template image registration section 32 (FIG. 2).

The drawing structuring system 1 of this embodiment has hardware resources such as a CPU, ROM, RAM, and HDD, and when the CPU executes various programs, information processing by software is realized using the hardware resources. It consists of computers. Furthermore, the drawing structuring method of this embodiment is realized by causing a computer to execute various programs.

Note that the drawing structuring system 1 may be implemented on a personal computer, may be constructed on a server provided inside a predetermined facility, or may be constructed on a server provided in an external facility. It may also be built on a cloud server on a network.

Each component of the drawing structuring system 1 does not necessarily need to be provided in one computer. For example, one drawing structuring system 1 may be implemented using a plurality of computers connected to each other via a network.

Although not particularly illustrated, the drawing structuring system 1 includes an information input unit into which predetermined information is input in response to an operation by a user using the system. This information input unit includes an input device such as a mouse or a keyboard. In other words, predetermined information is input into the drawing structuring system 1 according to the operations of these input devices.

The drawing structuring system 1 of this embodiment includes a device that displays images, such as a display that outputs processing results. In other words, the control of the output of information performed by the drawing structuring system 1 includes the control of images displayed on the display. Note that the display may be separate from the computer main body, or may be integrated with the computer main body. Furthermore, the drawing structuring system 1 may control images displayed on a display included in another computer connected via a network.

Note that in this embodiment, a display is exemplified as a device that displays an image, but other aspects may be used. For example, information may be displayed using a projector. Furthermore, a printer that prints information on a paper medium may be used in place of the display. In other words, a projector or a printer may be included as an object of information output control.

As shown in FIG. 1, the preprocessing unit 10 performs a process of removing and correcting noise such as spots, dirt, distortions, and chips included in the raster format drawing 2. This process is a preliminary process performed before the raster format drawing 2 input to the drawing structuring system 1 is input to the vector conversion function unit 11, character recognition function unit 12, and symbol recognition function unit 13. Note that noise removal and correction are performed by image processing based on preset setting values.

Note that in the case of the raster format drawing 2 generated by reading paper with a scanner, the read drawing may be tilted or the magnification (scale) may be different. Therefore, the pre-processing unit 10 may perform processing to correct the inclination or magnification of the diagram. Further, even if the symbols 4 have the same shape, the symbols 4 may be written tilted or written with different magnifications (scales). Therefore, the pre-processing unit 10 may perform a process of correcting the inclination or magnification of the symbol 4.

The vector conversion function unit 11 executes vector conversion processing to convert the raster format drawing 2 into a vector format drawing. Here, the vector format drawing of this embodiment is obtained by converting the characters other than the characters 9 (including alphanumeric characters) and symbols 4 included in the raster format drawing 2 into vectors.

Before converting the raster format drawing 2 into a vector format drawing, the vector conversion function unit 11 identifies images included in the raster format drawing 2 that meet preset conditions for the characters 9, and converts these images into vector format drawings. You may also specify it as not to be converted. Then, the vector conversion function unit 11 may perform vector conversion of characters other than the characters 9 included in the raster format drawing 2. Note that the preset condition is, for example, that an image within 10 pixels included in the raster format drawing 2 is regarded as the character 9.

In addition, before converting the raster format drawing 2 into a vector format drawing, the vector conversion function unit 11 identifies images included in the raster format drawing 2 that meet preset conditions for the symbol 4, and converts these images. may be specified as not to be subjected to vector conversion. The vector conversion function unit 11 may perform vector conversion on symbols other than the symbols 4 included in the raster format drawing 2. Note that, as a preset condition, for example, an image having a shape having a predetermined feature amount included in the raster format drawing 2 is regarded as the symbol 4.

Note that the vector conversion function unit 11 may convert the raster format drawing 2, including the characters 9 included in the raster format drawing 2, into a vector format drawing. Furthermore, the raster format drawing 2 may be converted into a vector format drawing, including the symbols 4 included in the raster format drawing 2. The character 9 or symbol 4 converted by the vector conversion function section 11 may be replaced with the character 9 or symbol 4 recognized by the character recognition function section 12 or the symbol recognition function section 13. When replacing these, the one with a higher recognition rate may be selected.

The character recognition function unit 12 recognizes characters 9 (including alphanumeric characters) from the raster format drawing 2. Note that after the raster format drawing 2 is converted into a vector format drawing, the characters 9 may be recognized from this vector format drawing. That is, the character recognition function unit 12 is capable of recognizing the characters 9 included in the raster format drawing 2, and is also capable of recognizing the characters 9 included in the vector format drawing converted by the vector conversion function unit 11. There is.

As shown in FIG. 2, the symbol recognition function unit 13 recognizes the symbol 4 drawn on the raster format drawing 2. The symbol recognition function section 13 includes a symbol recognizer 33, a recognition result selection section 34, and a recognition rule setting section 35. Furthermore, the symbol recognizer 33 includes an image resize processing section 36 , a coordinate recognition section 37 , a deep learning image recognition section 38 , and a template image recognition section 39 .

When the raster format drawing 2 is input, the symbol recognition function unit 13 determines the type of the symbol 4, the coordinates of the symbol 4, and the recognition rate of the symbol 4 as the recognition result 40 of the symbol 4 drawn on the raster format drawing 2. Output.

The type of symbol 4 indicates the type of component such as equipment, valves, pumps, electronic elements, circuit elements, etc. The coordinates of the symbol 4 indicate the position of the symbol 4 at the coordinates when absolute coordinates are defined with a predetermined position in the raster format drawing 2 as the origin. The recognition rate of the symbol 4 indicates the matching rate (recognition accuracy) between the information regarding the symbol 4 registered in advance in the symbol recognition database 25 and the information regarding the symbol 4 extracted from the symbol 4 of the raster format drawing 2.

Note that a predetermined threshold value is set in the recognition rule setting section 35 as a recognition setting value when recognizing symbol 4 as a symbol recognition rule. For example, if the recognition setting value is less than or equal to the threshold value, it is determined that the recognition has failed, and if the recognition setting value exceeds the threshold value, it is determined that the recognition has been successful. Further, the recognition setting values may include setting values other than the threshold values. For example, it may include conversion setting values used during vector conversion.

The symbol recognizer 33 executes processing to recognize symbol 4. Here, the image resizing processing unit 36 executes resizing of the image of the raster format drawing 2. The tilt or magnification (scale) of the image is adjusted so that the symbol 4 can be easily recognized.

The coordinate recognition unit 37 recognizes the coordinate position of the symbol 4 in the raster format drawing 2. By recognizing the coordinate position of the symbol 4, the coordinate recognition unit 37 places the symbol 4 recognized by the symbol recognition function unit 13 at the same coordinate position of the vector format drawing converted by the vector conversion function unit 11 described above. can do. Note that the recognition of the coordinate position of the symbol may be performed by the deep learning image recognition unit 38 or the template image recognition unit 39.

The deep learning image recognition unit 38 recognizes the symbol 4 of the raster format drawing 2 based on the trained model stored in the symbol recognition database 25. For example, an image at a position specified by the coordinate recognition unit 37 is input to the trained model of the deep learning image recognition unit 38. Then, it is possible to recognize the symbol 4 included in the input image. In this way, the symbol 4 of the raster format drawing 2 can be recognized using image recognition technology based on deep learning. Recognizing the symbol 4 may include classifying what type of symbol it is.

Note that the deep learning image recognition unit 38 may recognize the coordinates of the symbol 4 in the raster format drawing 2 and recognize the symbol 4 at the same time.

The template image recognition unit 39 recognizes the symbol 4 of the raster format drawing 2 based on the template image stored in the symbol recognition database 25. For example, symbol 4 can be recognized by capturing the feature points of the template image. Note that the template image is an image that serves as a template for the symbol 4 that is the recognition target. A plurality of template images may be registered for one type of symbol 4.

Note that the template image recognition unit 39 may recognize the coordinates of the symbol 4 in the raster format drawing 2 and recognize the symbol 4 at the same time.

The trained model is registered in advance in the trained model registration section 31 of the symbol recognition database 25. These trained models are generated by learning the feature amount of symbol 4 through deep learning. Further, the template image is registered in advance in the template image registration section 32 of the symbol recognition database 25. That is, the symbol recognition database 25 stores the trained model and also stores the template image of the symbol 4.

The symbol recognizer 33 of this embodiment may include recognition units (recognizers) other than the deep learning image recognition unit 38 and the template image recognition unit 39. Further, the symbol 4 may be recognized by combining a plurality of recognition units such as the deep learning image recognition unit 38 and the template image recognition unit 39.

The recognition result selection unit 34 selects the recognition result based on the trained model and the recognition result based on the template image, whichever has a higher recognition rate. In this way, the symbol 4 can be recognized based on the recognition result with the higher recognition rate, so the accuracy of symbol 4 recognition can be improved.

For example, the deep learning image recognition unit 38 (first type classifier) is a method using deep learning. Methods using deep learning require preparing training data and learning its features. Therefore, if learning is insufficient, there is a disadvantage that detection accuracy is low. On the other hand, it has the advantage of being relatively robust against drawing changes or noise.

The template image recognition unit 39 (second type classifier) uses a method called template matching. Template matching is a method of searching for whether an image of symbol 4 prepared as a template image exists in a drawing to be recognized. Since the template image is composed of a group of symbols having a fixed shape, template matching is effective. Template matching has the advantage of being easy to implement because it does not require feature point extraction processing or preparation of training data. On the other hand, it has the disadvantage that it is easily influenced by noise or distortion, or by slight differences in drawings.

In this embodiment, symbol 4 can be detected with high accuracy by combining these two types of classifiers. In addition, in the structuring process described later, in order to synthesize the vector conversion results and symbol recognition results, the process to obtain the scale and reference point coordinates of the raster format drawing 2 is also performed. Also good.

Note that the information regarding symbol 4 registered in the symbol recognition database 25 is also registered in advance in the structured rule definition database 23 (FIG. 3). The information registered in the symbol recognition database 25 and the information registered in the structured rule definition database 23 are linked with respect to the definition of the symbol 4. For example, when symbol 4 named "SA1" is registered in the symbol recognition database 25, definition information of "SA1" is also registered in the structured rule definition database 23.

As shown in FIG. 1, the erroneous recognition check function unit 14 checks whether there is any erroneous conversion in the vector format drawing that has been vector-converted by the vector conversion function unit 11. Further, the misrecognition check function section 14 checks whether or not there is an error in the character 9 recognized by the character recognition function section 12. Furthermore, the erroneous recognition check function section 14 checks whether or not there is an error in the symbol 4 recognized by the symbol recognition function section 13. Confirmation of erroneous recognition (erroneous conversion) may be performed automatically or may be performed manually. Note that if there is any erroneous recognition (erroneous conversion), it is corrected.

The erroneous recognition check function unit 14 of this embodiment outputs conversion confirmation information that can confirm whether or not there is an erroneous conversion in the vector format drawing to the display included in the drawing structuring system 1. Further, the misrecognition check function unit 14 outputs character confirmation information to the display that allows checking whether or not there is an error in the recognized character 9. Further, the misrecognition check function unit 14 outputs symbol confirmation information to the display that allows confirmation of whether or not there is an error in the recognized symbol 4. The user of the drawing structuring system 1 checks each piece of confirmation information output on the display and determines whether there is any misrecognition (erroneous conversion). Note that the confirmation information is, for example, the recognition rate of the symbol 4 or the character 9. When this recognition rate is below a predetermined threshold (for example, below 50%), the corresponding symbol 4 or character 9 is output as confirmation information.

Examples of misrecognition include, for example, wiring in a vector format drawing becoming unconnected, or symbol 4 in a vector format drawing having a polygonal shape instead of the originally circular shape.

When the user determines that there is a misrecognition, the user performs an operation to correct the misrecognition location (misconversion location). The misrecognition check function unit 14 that receives this operation executes processing to correct the misrecognition location. For example, the misrecognition check function unit 14 can receive an operation to correct the symbol 4 corresponding to an erroneous recognition result (conversion result), and execute processing to correct the symbol 4. In this way, it is possible to correct the symbol 4 placed in the vector format drawing based on the erroneous recognition result.

The misrecognition data storage section 16 accumulates misrecognition data including the vector format drawing corrected by the misrecognition check function section 14 and the corresponding conversion setting values used in the vector conversion function section 11. Further, the misrecognition data storage section 16 accumulates misrecognition data including the character 9 corrected by the misrecognition check function section 14 and the corresponding recognition setting value used in the character recognition function section 12. . Further, the erroneous recognition data storage unit 16 stores erroneous recognition data including the recognition result of the symbol 4 corrected by the erroneous recognition check function unit 14 and the corresponding recognition setting value used in the symbol recognition function unit 13. Accumulate. In this way, past erroneous recognition data can be accumulated and used for subsequent adjustments.

The optimal recognition learning unit 17 stores information regarding the erroneous conversion of the vector conversion function unit 11 accumulated in the erroneous recognition data storage unit 16 (the raster format drawing 2 before vector conversion and the result of erroneous conversion by the vector conversion function unit 11). , the conversion result corrected by the erroneous recognition check function unit 14, and the conversion setting value at the time of the erroneous conversion by the vector conversion function unit 11).

Then, the optimal recognition learning unit 17 determines the optimal recognition method that reduces the incidence of erroneous conversion by the vector conversion function unit 11 based on the information regarding the erroneous conversion of the vector conversion function unit 11 inputted from the erroneous recognition data storage unit 16. Learn by calculating conversion settings. Note that if there is a large amount of erroneously converted data, the optimal recognition learning unit 17 may use deep learning using a multilayer neural network. The optimum recognition learning section 17 further feeds back the calculated optimum conversion setting value to the vector conversion function section 11, and causes the vector conversion function section 11 to update the conversion setting value for vector conversion.

The optimal recognition learning unit 17 also contains information regarding the misrecognition of the character recognition function unit 12 accumulated in the misrecognition data storage unit 16 (the raster format drawing 2 before misrecognition and the information about the misrecognition by the character recognition function unit 12). (the recognition result corrected by the erroneous recognition check function section 14, and the recognition setting value when the character recognition function section 12 made the erroneous recognition) are input.

Then, the optimal recognition learning section 17 determines the optimal recognition method to reduce the incidence of misrecognition by the character recognition function section 12 based on the information regarding the misrecognition of the character recognition function section 12 inputted from the misrecognition data storage section 16. Learn by calculating recognition settings. Note that, if there is a large amount of misrecognition data, the optimal recognition learning unit 17 may use deep learning using a multilayer neural network. The optimal recognition learning section 17 further feeds back the calculated optimal recognition setting value to the character recognition function section 12, and causes the character recognition function section 12 to update the recognition setting value of the character 9.

Furthermore, the optimum recognition learning unit 17 stores information regarding the erroneous recognition of the symbol recognition function unit 13 accumulated in the erroneous recognition data storage unit 16 (the raster format drawing 2 before the erroneous recognition and the information about the erroneous recognition by the symbol recognition function unit 13). (the recognition result corrected by the erroneous recognition check function section 14, and the recognition setting value when the symbol recognition function section 13 made the erroneous recognition) are input.

Then, the optimal recognition learning section 17 determines the optimal recognition method that reduces the incidence of misrecognition by the symbol recognition function section 13 based on the information regarding the misrecognition of the symbol recognition function section 13 inputted from the misrecognition data storage section 16. Learn by calculating recognition settings. In addition, when there is a large amount of misrecognition data, the optimal recognition learning unit 17 may use deep learning using a multilayer neural network. The optimum recognition learning unit 17 further feeds back the calculated optimum recognition setting value to the symbol recognition function unit 13, and causes the symbol recognition function unit 13 to update the recognition setting value of the symbol 4.

That is, the optimal recognition learning section 17 adjusts the recognition setting value used in the symbol recognition function section 13 based on the misrecognition data accumulated in the misrecognition data storage section 16. In this way, the accuracy of recognition of the symbol 4 by the symbol recognition function section 13 can be improved based on correction of past recognition errors.

As shown in FIG. 3, the structured rule definition database 23 stores components drawn in vector format drawings (for example, wiring, diagram frames, panel frames, circuit elements, table structures, operation switch development diagrams, crossover wires, etc.). ) is a database that defines rules for structuring.

The structured rule information registration unit 26 is a part of the structured rule definition database 23. This structured rule information registration section 26 has registered information that defines the characteristics of the constituent elements. That is, the pre-structured symbol 4 and feature information indicating the feature of the shape of the symbol 4 or line segment in the vector format drawing are stored in correspondence with each other. Further, feature information indicating the features (shapes) of the symbols 4 in the vector format drawing is registered in association with identification information that allows the types of the symbols 4 to be individually identified. The identification information of the symbol 4 may be, for example, the name of the symbol 4 or the symbol ID.

For example, in a vector format drawing, a feature of the wiring (line segment 5 in FIG. 5) is a solid line with a predetermined length or more. For example, when the circuit element is a relay 6, the predetermined circuit element (symbol 4) is characterized by a combination of a circle and a solid line of less than a predetermined length, as shown in FIG. Regarding the line segment, for example, the board frame 8 is characterized by a two-dot chain line.

Further, attribute information (property) indicating the attribute of the symbol 4 is registered in the attribute information registration unit 27 in association with the identification information of the symbol 4. Further, the connection information registration unit 28 registers rules for defining the connection relationship between the symbol 4 and the line segment (wiring or piping).

The data structure definition database 24 is a database that defines data structure hierarchies (layers) that serve as standards when classifying and recording structured structural element data. For example, in the hierarchy information registration section 30 of the data structure definition database 24, hierarchy information indicating the hierarchy of the data structure is registered in association with the identification information of the symbol 4.

Note that the structured wiring data and the symbol 4 data can be classified and recorded in a hierarchical structure on a recording medium (not shown) based on the data structure hierarchy registered in the data structure definition database 24.

The structuring function unit 15 executes a process of arranging the symbol 4 specified based on the recognition result of the symbol recognition function unit 13 in a vector format drawing. Further, the structuring function unit 15 refers to the attribute information stored in the structuring rule definition database 23, and connects the symbol 4 with the attribute information according to the rule for defining the connection relationship between the symbol 4 and the line segment. A process of structuring a vector format drawing by adding information is executed.

For example, the structuring function unit 15 refers to the structured rule definition database 23 (including the structured rule information registration unit 26), and the symbol 4 recognized by the symbol recognition function unit 13 is stored in the structured rule definition database 23. It is determined whether or not the obtained feature information matches. Then, if the feature information matches, the symbol 4 stored in advance in the structured rule definition database 23 is placed in the vector format drawing. In this way, highly accurate symbols 4 that are structured in advance can be arranged in the vector format drawing.

Furthermore, the structuring function unit 15 calculates the correspondence between the coordinates of the raster format drawing 2 where the symbol 4 is recognized and the vector format drawing, and calculates the correspondence between the coordinates of the raster format drawing 2 where the symbol 4 is recognized and the vector format drawing corresponding to the coordinates specified in the raster format drawing 2. Processing to place symbol 4 at the coordinate position is executed. In this way, the symbols 4 can be arranged based on the coordinates of the raster format drawing 2 and the vector format drawing, thereby improving the accuracy of structuring.

Attribute information when structured will be explained. For example, symbol 4 of raster format drawing 2 is recognized by symbol recognizer 33. Here, it is assumed that the name of symbol 4 has been specified. Here, the shape information and name of the symbol 4 are compared using the structured rule information registration unit 26 that stores the correspondence.

It is also assumed that some values of the attribute information registered in the attribute information registration section 27 are empty. Here, the structured information adding unit 44 enters character information acquired from the raster format drawing 2 into some of the empty values of the attribute information. For example, assume that the symbol conversion function unit 43 replaces a predetermined symbol 4 in a vector format drawing with a structured symbol. Here, the structured information adding unit 44 registers character information near a predetermined symbol 4 in the raster format drawing 2 as attribute information.

Connection information when structured will be explained. For example, assume that the structured information providing unit 44 defines a vectorized line segment as a wiring, and also defines character information near this line segment as attribute information of the wiring. Here, the structured information providing unit 44 defines connection information between the wiring and the symbol 4. For example, the connection information is information indicating that a predetermined wiring is connected to a first appliance, and that this wiring is branched and connected to a second appliance and a third appliance.

Rules related to definitions such as in what cases symbols 4 and line segments are connected and vice versa are registered in the connection information registration unit 28. For example, a rule is registered in the connection information registration unit 28 such that if the distance from the end point of the symbol 4 to a neighboring line segment is 10 mm or more, the symbols 4 are deemed not to be connected to each other.

Further, the structuring function unit 15 classifies structured data such as symbols 4 and line segments 5 arranged in the vector format drawing by hierarchy based on the hierarchy information registered in the data structure definition database 24. It is recorded on a recording medium (not shown).

The structuring function section 15 of this embodiment includes a wiring extraction function section 41, a drawing frame extraction function section 42, a symbol conversion function section 43, a structuring information adding section 44, a table structure extraction function section 45, and an operation switch development diagram extraction function. The drawing drawing function section 46, a drawing drawing function section 47, and a drawing name giving function section 48 are provided.

The wiring extraction function unit 41 of the structuring function unit 15 extracts the wiring (FIG. 5), which is the line segment 5 drawn in the vector format drawing. For example, when extracting wiring, the wiring extraction function unit 41 first searches for and extracts a solid line of a predetermined length or more in a vector format drawing. Next, the wiring extraction function unit 41 determines whether the extracted solid line matches the wiring characteristics (solid line) registered in the structured rule definition database 23. Here, if they match, the extracted solid line is defined as a wiring registered in the structuring rule definition database 23 and structured. Thereafter, the wiring extraction function unit 41 classifies and records the structured wiring data on a recording medium based on the data structure hierarchy defined in the data structure definition database 24, and assigns an identification number ( line number).

The figure frame extraction function unit 42 extracts a figure frame or board frame 8 (FIG. 5) drawn in a vector format drawing. For example, when extracting the board frame 8, the drawing frame extraction function unit 42 first searches for and extracts a two-dot chain line in the vector format drawing. Next, the structuring function unit 15 determines whether the extracted two-dot chain line matches the characteristics of the board frame (two-dot chain line) registered in the structuring rule definition database 23. Here, if they match, the extracted two-dot chain line is defined as the board frame 8 registered in the structuring rule definition database 23 and structured. Thereafter, the structuring function unit 15 classifies and records the structured data of the board frame 8 on a recording medium based on the data structure hierarchy defined in the data structure definition database 24. Furthermore, if necessary, an identification number (wire number) is attached to the board frame. Note that the structuring function unit 15 also defines a picture frame in the same way as the board frame 8, and records the data. Further, the drawing name assigning function unit 48 assigns information regarding the drawing name, drawing creator, system, etc. to the diagram frame.

The symbol conversion function unit 43 arranges and structures the symbols 4 in a vector format drawing. For example, the symbol conversion function unit 43 first matches the symbol 4 recognized by the symbol recognition function unit 13 with the definition of the symbol 4 registered in the structured rule definition database 23 (structured rule information registration unit 26). Determine whether it is correct. Here, if they match, the symbol 4 registered in the structuring rule definition database 23 is placed in the vector format drawing to structure it.

Note that a match may be defined as, for example, the names (symbol IDs) of the symbols 4 matching. For example, if a symbol named "SA1" is recognized by the symbol recognition function unit 13 as symbol 4, but a symbol named "SA1" is also registered in the structured rule definition database 23, the two match. It can be considered as

Thereafter, the structured information adding unit 44 adds structured information including attribute information and connection information registered in the structured rule definition database 23 to the structured symbol 4. Further, the structured information adding unit 44 adds hierarchical information (structured information) registered in the data structure definition database 24 to the structured symbol 4. Then, based on the data structure hierarchy defined in the data structure definition database 24, the data including the symbols 4 are classified and recorded on the recording medium, and each symbol 4 is assigned an identification number.

Note that the table structure extraction function section 45, operation switch development diagram extraction function section 46, drawing drawing function section 47, and drawing name assignment function section 48 included in the structuring function section 15 are similar to the above-mentioned wiring extraction function section 41, etc. Execute the process. For example, the table structure extraction function unit 45 determines whether the features such as the table structure extracted from the vector format drawing match the features such as the table structure registered in the structuring rule definition database 23. Here, if they match, the table structure of the vector format drawing is defined as a table structure registered in the structuring rule definition database 23 and structured. Then, the structuring function unit 15 classifies and records the structured table structure based on the data structure hierarchy of the data structure definition database 24, and attaches an identification number to the table structure, etc., if necessary. Further, the operation switch development diagram extraction function unit 46 extracts an operation switch development diagram that describes the contact configuration of a cam switch or the like in the development connection diagram. Further, the drawing depiction function section 47 adjusts the depiction of the drawing. For example, it corrects the arrangement of symbols, corrects slanted wiring, corrects the position of piping on drawings, etc. Note that these corrections merely change the visual depiction position and improve the visibility of the drawing without changing the connection relationship. Further, the drawing name assigning function unit 48 assigns information regarding the name, drawing creator, lineage, etc. to the drawing.

The structuring check function unit 18 detects unstructured and inappropriate structuring results based on the structuring rule definition database 23 for the structuring result (i.e., the structured drawing 3) structured by the structuring function unit 15. Check whether it is. Confirmation of the structuring result may be performed automatically or manually. Then, if there is an inappropriate structuring result, it is corrected. Note that drawings for which the structuring check function unit 18 determines that there are no inappropriate structuring results are output as structured drawings 3 from the drawing structuring system 1.

Here, inappropriate structuring results include, for example, the wiring (line segment 5) is unconnected, attribute information is not assigned to a structural element that requires attribute information, wiring (line segment 5) Or, structural elements that require identification numbers, such as symbol 4, are not given identification numbers. Furthermore, the fact that the position of symbol 4 deviates from the appropriate coordinates is also included in the inappropriate structuring results.

The structuring check function unit 18 of this embodiment outputs structuring confirmation information that allows checking of inappropriate structuring results to the display included in the drawing structuring system 1. Further, the user of the drawing structuring system 1 checks the structuring confirmation information output on the display and determines whether the structuring result is inappropriate.

When the user determines that the structuring result is inappropriate, the user performs an operation to correct the inappropriate portion. Upon receiving this operation, the structuring check function unit 18 executes processing to correct inappropriate portions.

The erroneously structured data storage unit 19 stores the vector format drawing before structuring and the correction output by the structuring function unit 15 with respect to the event in which the structuring check function unit 18 corrected the structuring result by the structuring function unit 15. The structuring result that requires structuring, the result corrected by the structuring check function unit 18, and the structuring setting values (structuring rules) of the structuring function unit 15 when outputting the structuring result that requires correction. Accumulate. In this way, past inappropriate structured data can be accumulated and used for subsequent adjustments.

The optimal structuring learning unit 20 stores information accumulated in the incorrectly structured data storage unit 19 (vector format drawings before structuring, structuring results that require correction output by the structuring function unit 15, and The result corrected by the structuring check function unit 18 and the structuring setting value of the structuring function unit 15 at the time of outputting the structuring result that requires correction are input.

Then, the optimal structuring learning unit 20 performs optimal structuring to reduce the number of structuring results that require correction by the structuring function unit 15, based on the information input from the incorrectly structured data storage unit 19. Learn by calculating set values. Note that if there is a large amount of inappropriate structured data, the optimal structured learning unit 20 may use deep learning using a multilayer neural network. Furthermore, the optimal structuring learning unit 20 feeds back the calculated optimal structuring setting values to the structured rule definition database 23 (including the structured rule information registration unit 26), and the structure of the structured rule definition database 23 is Update the structuring settings (structuring rules).

As shown in FIG. 1, the learning drawing generation unit 21 divides one raster-format drawing in which a plurality of symbols 4 are drawn into a plurality of learning drawings by dividing the drawing into parts other than the parts where the symbols 4 are drawn. Generate 49. These learning drawings 49 are used for machine learning.

For example, one drawing is illustrated in FIG. In this figure, it is assumed that the area surrounded by the square broken line (frame line) is the learning drawing 49. When a plurality of learning drawings 49 are generated from one drawing, the drawing is divided so that the edges (frame lines) of each learning drawing 49 cross the margins or line segments 5 of the drawing. In other words, the edge of the learning drawing 49 should not cross the symbol 4. Note that the plurality of learning drawings 49 may have overlapping ranges. Each learning drawing 49 is drawn with the shape of the symbol 4 maintained. Therefore, data regarding the shape of the symbol 4 can be created without exception, and learning can be performed efficiently.

As shown in FIG. 1, the trained model generation unit 22 generates a trained model based on the learning drawing 49. In this way, a plurality of learning drawings 49 can be generated from one drawing, and the number of learning drawings 49 necessary for learning can be secured. Note that the trained model generated by the trained model generation unit 22 is registered in the trained model registration unit 31 (FIG. 2) of the symbol recognition database 25. Note that the work of dividing one drawing in raster format into a plurality of learning drawings 49 may be performed automatically or may be performed manually.

Note that the drawing structuring system 1 may include a computer equipped with artificial intelligence (AI) that performs machine learning. For example, the vector conversion function unit 11, the character recognition function unit 12, the symbol recognition function unit 13, the structuring function unit 15, the optimal recognition learning unit 17, the optimal structuring learning unit 20, and the trained model generation unit 22 include deep learning A deep learning unit may be included that extracts a specific pattern from a plurality of patterns based on.

Analysis technology based on artificial intelligence learning can be used for the computer-based analysis of this embodiment. For example, a learning model generated by machine learning using a neural network, a learning model generated by other machine learning, a deep learning algorithm, a mathematical algorithm such as regression analysis, etc. can be used. Further, forms of machine learning include forms such as clustering and deep learning.

The system of this embodiment includes a computer equipped with artificial intelligence that performs machine learning. For example, the system may be configured with one computer equipped with a neural network, or may be configured with multiple computers equipped with neural networks.

Here, the neural network is a mathematical model that expresses the characteristics of brain function through computer simulation. For example, we will present a model in which artificial neurons (nodes) that form a network through synaptic connections change the strength of synaptic connections through learning and acquire problem-solving ability. Furthermore, neural networks acquire problem-solving ability through deep learning.

For example, CNN (Convolution Neural Network), which has a proven track record in image recognition, is used. In this CNN, the middle layer is composed of a convolution layer and a pooling layer. The convolutional layer obtains a feature map by filtering nearby nodes in the previous layer. The pooling layer further reduces the feature map output from the convolution layer to create a new feature map. At this time, by obtaining the maximum value of pixels included in the region of interest in the feature map, it is possible to absorb some deviation in the position of the feature amount.

The convolution layer extracts local features of an image, and the pooling layer performs a process of combining local features. In these processes, the image is reduced while maintaining the characteristics of the input image. In other words, CNN can significantly compress (abstract) the amount of information an image has. Then, using the abstracted image stored in the neural network, the input image can be recognized and the image can be classified.

For example, by using CNN in the deep learning image recognition unit 38 of the symbol recognizer 33, the types of symbols included in the raster format drawing 2 can be classified. First, a trained CNN model is prepared that has learned the correspondence between features such as symbol shapes and symbol types (for example, types of coils, contacts, pumps, etc.). When a raster format image containing symbols is input to this trained model, the type of symbol included in the input image is output.

Note that deep learning includes various methods such as autoencoder, RNN (Recurrent Neural Network), LSTM (Long Short-Term Memory), and GAN (Generative Adversarial Network). These methods may be applied to the deep learning of this embodiment.

Alternatively, an object detection method that classifies objects in an image and detects the position of the objects may be used. This object detection method corresponds to the roles of the coordinate recognition section 37 and the deep learning image recognition section 38. Typical object detection methods include, for example, R-CNN (Regional CNN), SSD (Single Shot MultiBox Detector), and YOLO (You only look once). Furthermore, instead of the object detection method, it is also possible to use a segmentation method that classifies which class the pixels in the image belong to.

Next, the processing executed by the drawing structuring system 1 will be described using flowcharts shown in FIGS. 6 to 11. Note that the block diagrams shown in FIGS. 1 to 3 are referred to as appropriate. The following steps are at least part of the processing executed by the drawing structuring system 1, and other steps may be included in the processing executed by the drawing structuring system 1.

Note that before executing the following processing, various information is stored in advance in the structured rule definition database 23, data structure definition database 24, and symbol recognition database 25. For example, the structured rule definition database 23 stores in advance attribute information indicating attributes of symbols, wiring, and piping, and rules for defining connection relationships between symbols and line segments (wiring or piping).

As shown in FIG. 6, first, in step S11, a raster format drawing 2 is input to the drawing structuring system 1. Here, the preprocessing unit 10 determines whether or not there is noise such as spots or dirt in the input raster format drawing 2. Here, if there is noise in the raster format drawing 2, the preprocessing unit 10 executes a predetermined correction process such as removing noise. Furthermore, the pre-processing unit 10 may perform processing to correct the inclination or magnification of the diagram. Then, the process proceeds to step S12, step S15, and step S18. Note that these steps are executed in parallel.

In step S12, the vector conversion function unit 11 executes a process of converting the raster format drawing 2 into a vector format drawing.

In the next step S13, the erroneous recognition check function unit 14 outputs conversion confirmation information that allows confirmation of whether or not there is an erroneous conversion in the vector format drawing to the display included in the drawing structuring system 1.

In the next step S14, the erroneous recognition check function unit 14 determines whether or not an operation for correcting the erroneous conversion of the vector format drawing has been received, that is, whether or not there is any erroneous conversion in the vector format drawing. Here, if there is no erroneous conversion in the vector format drawing (NO in step S14), the process advances to step S21. On the other hand, if there is an erroneous conversion in the vector format drawing (YES in step S14), the process advances to step S31.

As shown in FIG. 8, in step S31, the misrecognition check function unit 14 executes a process for correcting misconversion of the vector format drawing. Then, the process advances to step S32 and step S35. Note that these steps are executed in parallel.

In step S32, the misrecognition data storage section 16 stores misrecognition data including the vector format drawing corrected by the misrecognition check function section 14 and the corresponding conversion setting values used in the vector conversion function section 11. Accumulate.

In the next step S33, the optimal recognition learning section 17 receives the information regarding the erroneous conversion of the vector conversion function section 11, which is accumulated in the erroneous recognition data accumulation section 16. Here, the optimal recognition learning unit 17 determines the optimal recognition method for reducing the incidence of erroneous conversion by the vector conversion function unit 11 based on the information regarding the erroneous conversion of the vector conversion function unit 11 inputted from the erroneous recognition data storage unit 16. Learn by calculating conversion setting values.

In the next step S34, the optimal recognition learning unit 17 feeds back the calculated optimal conversion setting value to the vector conversion function unit 11, and updates the conversion setting value for vector conversion in this vector conversion function unit 11. Then, the process advances to step S22.

In step S35, the misrecognition check function unit 14 inputs the corrected vector format drawing, which is the correction result, to the structuring function unit 15. Then, the process advances to step S22.

As shown in FIG. 6, in step S15, the character recognition function unit 12 executes a process of recognizing characters 9 drawn on the raster format drawing 2.

In the next step S16, the misrecognition check function unit 14 outputs character confirmation information that allows checking whether or not there is an error in the recognized character 9 to the display included in the drawing structuring system 1.

In the next step S17, the misrecognition check function unit 14 determines whether or not an operation for correcting the misrecognized character 9 has been received, that is, whether or not the character 9 has been misrecognized. Here, if there is no misrecognition of character 9 (NO in step S17), the process advances to step S21. On the other hand, if character 9 is misrecognized (YES in step S17), the process advances to step S41.

As shown in FIG. 9, in step S41, the misrecognition check function unit 14 executes processing to correct the misrecognized character 9. Then, the process advances to step S42 and step S45. Note that these steps are executed in parallel.

In step S42, the misrecognition data storage section 16 stores misrecognition data including the character 9 corrected by the misrecognition check function section 14 and the corresponding recognition setting value used in the character recognition function section 12. accumulate.

In the next step S43, the optimal recognition learning section 17 receives information regarding the misrecognition of the character recognition function section 12, which is accumulated in the misrecognition data accumulation section 16. Here, the optimal recognition learning section 17 determines the optimal recognition method for reducing the incidence of erroneous recognition by the character recognition function section 12 based on the information regarding the erroneous recognition of the character recognition function section 12 inputted from the erroneous recognition data storage section 16. Calculate and learn recognition settings.

In the next step S44, the optimum recognition learning section 17 feeds back the calculated optimum recognition setting value to the character recognition function section 12, and updates the recognition setting value of the character 9 in this character recognition function section 12. Then, the process advances to step S22.

In step S45, the misrecognition check function section 14 inputs the corrected character 9, which is the correction result, to the structuring function section 15. Then, the process advances to step S22.

As shown in FIG. 6, in step S18, the symbol recognition function unit 13 executes a process of recognizing the symbol 4 drawn on the raster format drawing 2.

In the next step S19, the misrecognition check function unit 14 outputs symbol confirmation information that allows checking whether or not there is an error in the recognized symbol 4 to the display included in the drawing structuring system 1.

In the next step S20, the misrecognition check function unit 14 determines whether an operation for correcting the misrecognized symbol 4 has been received, that is, whether or not the symbol 4 has been misrecognized. Here, if there is no misrecognition of symbol 4 (NO in step S20), the process advances to step S21. On the other hand, if symbol 4 is misrecognized (YES in step S20), the process advances to step S51.

As shown in FIG. 10, in step S51, the erroneous recognition check function unit 14 executes a process of correcting the erroneously recognized symbol 4. Then, the process advances to step S52 and step S55. Note that these steps are executed in parallel.

In step S52, the misrecognition data storage section 16 stores misrecognition data including the symbol 4 corrected by the misrecognition check function section 14 and the corresponding recognition setting value used in the symbol recognition function section 13. accumulate.

In the next step S53, the optimal recognition learning section 17 receives the information regarding the misrecognition of the symbol recognition function section 13, which is accumulated in the misrecognition data accumulation section 16. Here, the optimal recognition learning section 17 determines the optimal recognition method for reducing the incidence of misrecognition by the symbol recognition function section 13 based on the information regarding the misrecognition of the symbol recognition function section 13 inputted from the misrecognition data storage section 16. Calculate and learn recognition settings.

In the next step S54, the optimum recognition learning unit 17 feeds back the calculated optimum recognition setting value to the symbol recognition function unit 13, and updates the recognition setting value of the symbol 4 in the symbol recognition function unit 13. Then, the process advances to step S22.

In step S55, the misrecognition check function unit 14 inputs the corrected symbol 4, which is the correction result, to the structuring function unit 15. Then, the process advances to step S22.

As shown in FIG. 7, in step S21, the vector conversion function unit 11 inputs the converted vector format drawing, which is the conversion result, to the structuring function unit 15. Further, the character recognition function unit 12 inputs the recognized symbol 4, which is the recognition result, to the structuring function unit 15. Further, the symbol recognition function unit 13 inputs the recognized symbol 4, which is the recognition result, to the structuring function unit 15.

In the next step S22, the structuring function unit 15 acquires information regarding the data structure of the vector format drawing to be structured from the data structure definition database 24. For example, the hierarchical information registered in the hierarchical information registration section 30 of the data structure definition database 24 is acquired.

In the next step S23, the structuring function unit 15 acquires information regarding rules for structuring the vector format drawing from the structuring rule information registration unit 26 of the structuring rule definition database 23.

In the next step S24, the structuring function unit 15 acquires information regarding the symbol 4 to be placed in the vector format drawing from the structuring rule definition database 23. For example, the structuring function unit 15 acquires the attribute information of the symbol 4 from the attribute information registration unit 27, and also writes rules for defining the connection relationship between the symbol and line segments (wiring or piping) to the connection information registration unit 27. Get from.

In the next step S25, the structuring function unit 15 uses the conversion result (vector format drawing) by the vector conversion function unit 11, the recognition result of the character 9 by the character recognition function unit 12, and the recognition result of the symbol 4 by the symbol recognition function unit 13. Based on the recognition results, the structural elements of the vector format drawing are structured to generate a structured drawing 3. For example, the structuring function unit 15 arranges the symbol 4 specified based on the recognition result of the symbol recognition function unit 13 in the vector format drawing, and also combines the attribute information and connection information stored in the structuring rule definition database 23. is given to symbol 4 to structure the vector format drawing. Furthermore, the structuring function unit 15 classifies this structured data based on the data structure hierarchy of the data structure definition database 24 and records it on a predetermined recording medium.

In the next step S26, the structuring check function unit 18 outputs structuring confirmation information that allows confirmation of inappropriate structuring results to the display included in the drawing structuring system 1.

In the next step S27, the structuring check function unit 18 determines whether an operation for modifying an inappropriate portion has been received, that is, whether or not the structured drawing 3 needs to be modified. Here, if there is no need to modify the structured drawing 3 (NO in step S27), the process proceeds to step S28. On the other hand, if it is necessary to modify the structured drawing 3 (YES in step S28), the process advances to step S61.

In step S28, the structuring check function unit 18 outputs the structured drawing 3 from the drawing structuring system 1. Then, the process ends.

As shown in FIG. 11, in step S61, the structuring check function unit 18 executes processing to correct inappropriate portions of the structured drawing 3. Then, the process advances to step S62 and step S65. Note that these steps are executed in parallel.

In step S62, the incorrectly structured data storage unit 19 stores the vector format drawing, characters 9, and symbols 4 before structuring, the structuring results that need to be corrected, and the results corrected by the structuring check function unit 18. The structuring setting values (structuring rules) of the structuring function unit 15 when outputting the structuring result that requires correction are stored in the incorrectly structured data storage unit 19.

In the next step S63, various information accumulated in the erroneously structured data accumulation section 19 is input to the optimal structured learning section 20. Here, the optimal structuring learning unit 20 performs optimal structuring to reduce the incidence of structured drawings 3 that require correction, based on various information regarding structuring input from the incorrectly structured data storage unit 19. Learn by calculating set values.

In the next step S64, the optimal structuring learning unit 20 feeds back the calculated optimal structuring setting values to the structuring rule definition database 23 to update the structuring setting values in the structuring rule definition database 23. Then, the process ends.

In step S65, the structuring check function unit 18 outputs the modified structured drawing 3, which is the modification result, from the drawing structuring system 1. Then, the process ends.

In addition, in this embodiment, the determination of any value (recognition setting value, conversion setting value, recognition rate, structuring setting value, etc.) using a reference value (threshold value) is performed by determining whether an arbitrary value is greater than or equal to the reference value. It may be determined whether the value exceeds the reference value, or it may be determined whether or not a given value exceeds the reference value. Alternatively, it may be determined whether or not an arbitrary value is less than or equal to a reference value, or it may be determined whether or not an arbitrary value is less than a reference value. Furthermore, the reference value is not fixed, but may be variable. Therefore, instead of the reference value, a value within a predetermined range may be used to determine whether an arbitrary value falls within the predetermined range. Alternatively, errors occurring in the device may be analyzed in advance, and a predetermined range including the error range centered on the reference value may be used for determination.

Note that although the flowchart of this embodiment shows an example in which each step is executed in series, the sequential relationship of each step is not necessarily fixed, and even if the sequential relationship of some steps is swapped. good. Also, some steps may be executed in parallel with other steps.

The system of this embodiment includes a control device with a highly integrated processor such as a dedicated chip, an FPGA (Field Programmable Gate Array), a GPU (Graphics Processing Unit), or a CPU (Central Processing Unit), and a ROM (Read Only memory) or RAM (Random Access Memory), external storage devices such as HDD (Hard Disk Drive) or SSD (Solid State Drive), display devices such as displays, and input devices such as mouse or keyboard. , and a communication interface. This system can be realized with a hardware configuration using a normal computer.

Note that the program executed by the system of this embodiment is provided by being pre-installed in a ROM or the like. Alternatively, this program may be installed as a file in installable or executable format on a non-transitory computer readable storage medium such as a CD-ROM, CD-R, memory card, DVD, or flexible disk (FD). It may also be stored and provided.

Further, the program executed by this system may be stored on a computer connected to a network such as the Internet, and may be downloaded and provided via the network. Furthermore, this system can also be configured by combining separate modules that independently perform the functions of the constituent elements by interconnecting them via a network or dedicated line.

Note that although this embodiment exemplifies a form in which a raster format drawing 2 related to a plant is converted into a structured drawing 3, other forms may be used. For example, a raster format drawing 2 regarding a vehicle, a ship, an aircraft, etc. may be converted into a structured drawing 3.

Note that the drawing structuring system 1 of this embodiment may not include the erroneous recognition data storage section 16, the optimal recognition learning section 17, the erroneous structuring data storage section 19, and the optimal structuring learning section 20. Further, the drawing structuring system 1 may not include the learning drawing generation section 21 and the trained model generation section 22.

According to the embodiments described above, by providing a symbol recognition function unit that recognizes symbols drawn on raster format drawings, it is possible to improve the accuracy of recognition of symbols drawn on drawings.

Although several embodiments of the invention have been described, these embodiments are presented by way of example and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, substitutions, changes, and combinations can be made without departing from the gist of the invention. These embodiments and their modifications are included within the scope and gist of the invention as well as within the scope of the invention described in the claims and its equivalents.

1...Drawing structuring system, 2...Raster format drawing, 3...Structured drawing, 4...Symbol, 5...Line segment, 6...Relay, 7a, 7b...Wiring, 8...Board frame, 9...Character, 10...Preliminary Processing section, 11... Vector conversion function section, 12... Character recognition function section, 13... Symbol recognition function section, 14... Misrecognition check function section, 15... Structuring function section, 16... Misrecognition data storage section, 17... Optimum Recognition learning section, 18... Structured check function section, 19... Misstructured data storage section, 20... Optimal structured learning section, 21... Learning drawing generation section, 22... Learned model generation section, 23... Structuring rule definition database, 24... data structure definition database, 25... symbol recognition database, 26... structured rule information registration unit, 27... attribute information registration unit, 28... connection information registration unit, 29... optimal structured learning data registration unit, 30 ...Hierarchical information registration unit, 31...Learned model registration unit, 32...Template image registration unit, 33...Symbol recognizer, 34...Recognition result selection unit, 35...Recognition rule setting unit, 36...Image resizing processing unit, 37... Coordinate recognition unit, 38... Deep learning image recognition unit, 39... Template image recognition unit, 40... Recognition result, 41... Wiring extraction function unit, 42... Picture frame extraction function unit, 43... Symbol conversion function unit, 44... Structuring Information provision section, 45...Table structure extraction function section, 46...Operation switch development diagram extraction function section, 47...Drawing depiction function section, 48...Drawing name addition function section, 49...Learning drawing.

Claims (10)

a structured rule definition database that stores attribute information indicating attributes of symbols and rules for defining connection relationships between the symbols and line segments;
a symbol recognition function unit that recognizes the symbol drawn on a raster format drawing;
a vector conversion function unit that converts the raster format drawing into a vector format drawing;
The symbol identified based on the recognition result of the symbol recognition function unit is placed in the vector format drawing, and the symbol is a structuring function unit that adds the attribute information to the vector format drawing to structure it;
Equipped with
Drawing structuring system. In the structured rule definition database, the pre-structured symbol and feature information indicating the feature of the shape of the symbol in the vector format drawing are stored in correspondence with each other,
When the symbol recognized by the symbol recognition function unit matches the feature information stored in the structured rule definition database, the structuring function unit is configured to perform the feature information stored in advance in the structured rule definition database. placing symbols on the vector format drawing;
A drawing structuring system according to claim 1. An erroneous recognition check function unit that outputs symbol confirmation information that allows checking whether or not there is an error in the recognition result of the symbol recognition function unit, and receives an operation to correct the symbol corresponding to the recognition result that is erroneous. Equipped with
A drawing structuring system according to claim 1 or claim 2. an erroneous recognition data storage unit that accumulates erroneous recognition data including the recognition result of the symbol corrected by the erroneous recognition check function unit and the corresponding recognition setting value used in the symbol recognition function unit; prepare,
The drawing structuring system according to claim 3. an optimal recognition learning unit that adjusts the recognition setting value used in the symbol recognition function unit based on the erroneous recognition data accumulated in the erroneous recognition data storage unit;
The drawing structuring system according to claim 4. The structuring function unit specifies the coordinates of the raster format drawing where the symbol is recognized, and places the symbol at a position corresponding to the coordinates of the vector format drawing.
A drawing structuring system according to any one of claims 1 to 5. The symbol recognition function section is
a symbol recognition database storing a trained model that has learned feature quantities of the symbol through deep learning;
a symbol recognizer that recognizes the symbol based on the trained model stored in the symbol recognition database;
Equipped with
A drawing structuring system according to any one of claims 1 to 6. a learning drawing generation unit that generates a plurality of learning drawings by dividing one raster-format drawing in which a plurality of the symbols are drawn by parts other than the parts where the symbols are drawn;
a learned model generation unit that generates the learned model based on the learning drawing;
Equipped with
The drawing structuring system according to claim 7. The symbol recognition database stores a template image of the symbol, and the symbol recognizer recognizes the symbol based on the template image stored in the symbol recognition database,
The symbol recognition function unit includes a recognition result selection unit that selects the recognition result based on the learned model and the recognition result based on the template image, whichever has a higher recognition rate.
A drawing structuring system according to claim 7 or claim 8. storing attribute information indicating attributes of symbols and rules for defining connection relationships between the symbols and line segments in a structured rule definition database;
a symbol recognition function unit recognizing the symbol drawn on the raster format drawing;
a vector conversion function unit converting the raster format drawing into a vector format drawing;
A structuring function unit places the symbol specified based on the recognition result of the symbol recognition function unit in the vector format drawing, and uses the attribute information and the rule stored in the structuring rule definition database. structuring the vector format drawing by assigning the attribute information to the symbol;
including,
Drawing structuring method.

Images