Drawing information extraction device, drawing information extraction method, and drawing information extraction program

Abstract

To provide a drawing information extraction device capable of identifying the correspondence relation between information defined outside the drawing and information included in the drawing.SOLUTION: An extraction unit 21 extracts a text from a drawing. A feature quantity calculation unit 221 calculates the feature quantity of the text extracted by the extraction unit 21. A classification unit 222 identifies which of multiple items defined outside the drawing corresponds to the text using the text extracted by the extraction unit 21 and the feature quantity calculated by the feature quantity calculation unit 221. An output unit 23 outputs correspondence information 242 associating the text extracted by the extraction unit 21 with the item identified by the classification unit 222.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present disclosure relates to a technique for associating information included in a drawing with information not included in the drawing. [Background technology]

[0002] Drawing recognition technology is known that recognizes characters or symbols present in a drawing and determines the correspondence between the characters or symbols. For example, Patent Document 1 describes a technology that uses deep learning or template image recognition technology and predefined structuring rules to match the shapes of symbols in a drawing and recognize the symbols and the connection relationships between the symbols. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2022-63599 Summary of the Invention [Problem to be solved by the invention]

[0004] However, with the conventional technology described in Patent Document 1, it is not possible to identify the correspondence between information defined outside the drawings and information described in the drawings. For example, with the conventional technology, it is not possible to identify the correspondence between a dimension name defined outside the drawings and a dimension value described in the drawings. The present disclosure aims to make it possible to identify the correspondence between information defined outside a drawing and information included in the drawing. [Means for solving the problem]

[0005] The drawing information extraction device according to the present disclosure comprises: an extractor for extracting text from the drawing; a feature calculation unit that calculates feature amounts of the text extracted by the extraction unit; a classification unit that uses the text and the feature calculated by the feature calculation unit to identify which item of a plurality of items defined outside the drawing the text corresponds to; Equipped with. [Effects of the Invention]

[0006] In this disclosure, text features are used to associate items defined outside the drawing, making it possible to identify the correspondence between information defined outside the drawing and information included in the drawing. [Brief explanation of the drawings]

[0007] [Figure 1] 1 is a hardware configuration diagram of a drawing information extraction device 10 according to a first embodiment. [Figure 2] FIG. 1 is a functional configuration diagram of a drawing information extraction device 10 according to the first embodiment. [Figure 3] 3 is a flowchart showing the flow of processing performed by the drawing information extraction device 10 according to the first embodiment. [Figure 4] FIG. 3 is an explanatory diagram of drawing 30 according to the first embodiment. [Figure 5] FIG. 3 is an explanatory diagram of conversion of coordinate values ​​according to the first embodiment. [Figure 6] FIG. 3 is an explanatory diagram of conversion of notation of text 31 according to the first embodiment. [Figure 7] FIG. 3 is an explanatory diagram of the correspondence between text 31 and items 32 according to the first embodiment. [Figure 8] FIG. 10 is a functional configuration diagram of a drawing information extraction device 10 according to a second embodiment. [Figure 9] 10 is a flowchart showing the flow of processing performed by the drawing information extraction device 10 according to the second embodiment. [Figure 10] 10 is a flowchart of a graphic area identification process according to the second embodiment. [Figure 11] FIG. 10 is an explanatory diagram of an area specifying process according to the second embodiment. [Figure 12] 10 is a flowchart of a group formation process according to the second embodiment. [Figure 13]FIG. 10 is a functional configuration diagram of a drawing information extraction device 10 according to a third embodiment. [Figure 14] 10 is a flowchart showing the flow of processing performed by the drawing information extraction device 10 according to the third embodiment. [Figure 15] 11 is a flowchart of a ranking determination process according to the third embodiment. [Figure 16] FIG. 11 is an explanatory diagram of relationship information 244 according to the third embodiment. [Figure 17] FIG. 11 is an explanatory diagram of a ranking determination process according to the third embodiment. [Figure 18] FIG. 11 is an explanatory diagram of a ranking determination process according to the third embodiment. [Figure 19] FIG. 10 is a functional configuration diagram of a drawing information extraction device 10 according to a fourth embodiment. [Figure 20] 10 is a flowchart showing the flow of processing performed by the drawing information extraction device 10 according to the fourth embodiment. [Figure 21] 13 is a flowchart of an item confirmation process according to the fourth embodiment. [Figure 22] 10 is a flowchart of a re-learning process according to the fourth embodiment. [Figure 23] FIG. 13 is a diagram showing a source drawing and a derived drawing according to the fourth embodiment. [Figure 24] FIG. 13 is a diagram showing an example of training data 245 according to the fourth embodiment. [Figure 25] FIG. 10 is a diagram showing an example of a determination result according to the fourth embodiment. [Figure 26] FIG. 13 is a diagram showing an example of new training data 245 according to the fourth embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0008] Embodiment 1 In the first embodiment, a drawing information extraction device 10 is described that extracts text 31 contained in a drawing and identifies items 32 corresponding to the extracted text 31. The text 31 is, for example, a dimension value and a reference symbol. The items 32 are, for example, names defined outside the drawing data. Drawings of a compressor, an air conditioner, a production robot, etc. can be used as drawings according to the first embodiment.

[0009] ***Configuration Description*** The hardware configuration of a drawing information extraction device 10 according to the first embodiment will be described with reference to FIG. The drawing information extraction device 10 includes a processor 11, a volatile memory 12, a non-volatile memory 13, and an interface device 14.

[0010] The processor 11 is a device that controls the entire drawing information extraction device 10. For example, the processor 11 is a CPU or FPGA. CPU stands for Central Processing Unit. FPGA stands for Field Programmable Gate Array. The processor 11 may be a single processor or multiple processors. The drawing information extraction device 10 may have a processing circuit instead of the processor 11. The processing circuit may be a single circuit or a composite circuit.

[0011] The volatile memory 12 is a main storage device that provides a working area or a storage area for the processor 11 of the drawing information extraction device 10. For example, the volatile memory 12 is a RAM. RAM is an abbreviation for Random Access Memory.

[0012] The nonvolatile memory 13 is an auxiliary storage device that stores data and programs necessary for processing by the drawing information extraction device 10. For example, the nonvolatile memory 13 is an HDD or SSD. HDD is an abbreviation for Hard Disk Drive. SSD is an abbreviation for Solid State Drive.

[0013] The interface device 14 is a communication interface for communicating with an input device for inputting drawing data to the drawing information extraction device 10 and an output device for outputting the drawing data. For example, the interface device 14 is a NIC. NIC stands for Network Interface Card. The input device is a CAD device, an image scanner, an OCR, or the like. OCR stands for Optical Character Reader. The output device is a display, a printer, or the like.

[0014] The functional configuration of the drawing information extraction device 10 according to the first embodiment will be described with reference to FIG. The drawing information extraction device 10 includes, as functional components, an extraction unit 21, an item identification unit 22, an output unit 23, and a storage unit 24. The item identification unit 22 includes a feature calculation unit 221 and a classification unit 222. The storage unit 24 stores a classifier 241 and correspondence information 242.

[0015] The functions of the extraction unit 21, the item identification unit 22, and the output unit 23 are realized by the processor 11. Specifically, a program that realizes the functions of the extraction unit 21, the item identification unit 22, and the output unit 23 is read from the non-volatile memory 13 by the processor 11 and executed. This realizes the functions of the extraction unit 21, the item identification unit 22, and the output unit 23. When the drawing information extraction device 10 includes a processing circuit instead of the processor 11, the functions of the extraction unit 21, the item identification unit 22, and the output unit 23 are realized by the processing circuit. The function of the storage unit 24 is realized by at least one of the volatile memory 12 and the nonvolatile memory 13.

[0016] ***Explanation of Operation*** The operation of the drawing information extraction device 10 according to the first embodiment will be described with reference to FIGS. The operation procedure of the drawing information extraction device 10 according to the embodiment 1 corresponds to the drawing information extraction method according to the embodiment 1. Moreover, the program that realizes the operation of the drawing information extraction device 10 according to the embodiment 1 corresponds to the drawing information extraction program according to the embodiment 1.

[0017] The flow of processing performed by the drawing information extraction device 10 according to the first embodiment will be described with reference to FIG. (Step S101: Extraction process) The extraction unit 21 extracts text 31 and rectangle coordinates 33 from the input drawing 30 . Specifically, the extraction unit 21 extracts a character string of text 31 drawn on the drawing 30, and also extracts rectangular coordinates 33 of the text 31. The rectangular coordinates 33 are the coordinates of the reference point of a rectangle that encloses the text. In this example, the reference point is the upper left corner. The extraction unit 21 outputs the text 31 and the rectangular coordinates 33 to the volatile memory 12 as text information 34.

[0018] Referring to FIG. 4, a diagram 30 according to the first embodiment will be described. As a specific example, the drawing 30 is data created using CAD. CAD stands for Computer Aided Design. The format of the drawing 30 can be DXF format, DWG format, or the like. The drawing 30 may also be data obtained by acquiring a printed drawing as image data using an image scanner or the like and converting the acquired image data into drawing data that can be edited by CAD. DXF stands for Drawing Exchange Format. DWG stands for Drawing. One drawing 30 includes multiple drawings, such as a drawing of the main part of the parts that make up the product and an enlarged view of part of the main part. When a part of the main part in drawing 30 is enlarged, a reference symbol indicating the enlarged area may be included in the drawing.

[0019] For example, AA(B-2)(2:1) in Figure 4 is a reference symbol. The initials "AA" are the reference symbol ID. ID stands for IDentifier. In the "(B-2)(2:1)" that follows, "B-2" is the rough coordinate on the drawing, and 2:1 is the scale ratio between the jump destination and jump source. The range indicated by the reference symbol ID "A" in the part drawing connected by a straight line to the bottom left figure is the jump source, and the figure drawn near AA(B-2)(2:1) containing the initial letter "AA" (the figure on the top left of Figure 4) is the corresponding jump destination. Similarly, the range indicated by the reference symbol ID "C" in the center is the jump source, and the figure drawn near C(D-6)(2:1) containing the initial letter "C" (the figure on the right of Figure 4) is the corresponding jump destination.

[0020] (Steps S102 to S104: Feature Calculation Processing) In step S102, the feature calculation unit 221 refers to the text information 34 output in step S101, and sets each of the one or more texts 31 included in the text information 34 as the target text 31.

[0021] In step S103, the feature amount calculation unit 221 calculates the feature amount 35 of the target text 31. Specifically, the feature calculation unit 221 calculates the feature 35 using at least one of coordinate values ​​indicated by the rectangular coordinates 33 for the target text 31, words included in the target text 31, and attribute information of the target text 31. When calculating the feature 35 using coordinate values, the feature calculation unit 221 can use the coordinate values ​​as the feature 35. When calculating the feature 35 using words included in the target text 31, the feature calculation unit 221 can calculate the feature 35 from the BoW of the word, dimension values, tolerance values, etc. BoW is an abbreviation for Bag of Words. When calculating the feature 35 using attribute information of the text 31, the feature calculation unit 221 can use values ​​such as the angle and layer of the text 31 as the feature 35.

[0022] When calculating the feature amount 35 using coordinate values, the feature amount calculation unit 221 calculates the feature amount 35 after correcting the coordinate values. This will be explained in detail with reference to Fig. 5. The feature calculation unit 221 converts the coordinate values ​​of the target text 31 into relative coordinates with the reference symbol included in the drawing 30 as the origin, and sets the relative coordinates as feature 35. For example, if AA(B-2)(2:1) is the origin, the relative coordinates of "Φ6.2" are (20,15), and the relative coordinates of "Φ5.5" are (50,80). Even drawings of the same series of models and parts may have different coordinates. However, if the positional relationship between the reference symbol and the text is maintained, the relative coordinates will be the same. Therefore, by converting to relative coordinates, it is more likely that the feature quantity 35 calculated from the coordinates will be similar between drawings of the same series of models and parts.

[0023] When calculating the feature amounts 35 using words contained in the target text 31, the feature amount calculation unit 221 calculates the feature amounts 35 after removing variations in spelling. A specific description will be given with reference to Fig. 6. The feature calculation unit 221 refers to synonym information 36, which indicates a set in which the word Φ is registered as synonymous with "Kiri", "Kiritooshi", "DRILL", and "DRILL THRU", and a set in which the word C is registered as synonymous with "Both Ends", "Double Sided Chamfering", and "Chamfering". Then, the feature calculation unit 221 converts synonyms included in the target text 31 into words in the word column. For example, "Kiri", "Kiritooshi", "DRILL", and "DRILL THRU" in the target text 31 are converted into Φ. In this way, the feature calculation unit 221 standardizes dimensional notation and the like between the drawings and eliminates notation variations between the drawings. The feature calculation unit 221 then divides the character string of the text 31 into words, and sets 1 if the word is included, or 0 if it is not included. This generates a vector of the character string of each text 31. In FIG. 6, Φ, kiri, and kiritooshi are treated as synonyms due to the synonym pair, so "Φ" is set to 1 in all text character strings. This vector of the character string of each text 31 can be used as a feature 35.

[0024] In step S104, the feature amount calculation unit 221 completes the feature amount calculation process when the feature amounts 35 for all the texts 31 included in the text information 34 have been calculated.

[0025] (Steps S105 to S107: Classification process) In step S105, the classification unit 222 refers to the text information 34 output in step S101, and sets each of the one or more texts 31 included in the text information 34 as the target text 31.

[0026] In step S106, the classification unit 222 uses the target text 31 and the feature 35 for the target text 31 calculated in step S103 to identify which item 32 of multiple items 32 defined outside the drawing the target text 31 corresponds to. Specifically, the target text 31 and the feature quantities 35 for the target text 31 are input to the classifier 241 stored in the storage unit 24. Then, the classification unit 222 acquires the items 32 corresponding to the target text 31 output from the classifier 241. In this way, the classification unit 222 identifies which items 32 the target text 31 corresponds to. Here, the classifier 241 is an AI learning model that receives text 31 and feature values ​​35 as input and outputs items 32. AI stands for Artificial Intelligence. The classifier 241 is trained using training data including the text 31, feature values ​​35, and items 32 as input.

[0027] For example, as shown in Fig. 7, it is determined which item 32 the target text 31 corresponds to. In Fig. 7, (A) shows a drawing, and dimension values ​​and reference symbols are included as text 31. Also, in Fig. 7, (B) shows the correspondence between text 31 and item 32, and a name is set as item 32. Item 32 is not defined in the drawing, but is defined outside the drawing. The item value "Φ3±0.2" of the item name "Hole diameter" in Figure 7(B) corresponds to the dimension value "3 drilling ±0.2" in Figure 7(A). This shows that Φ and drilling are treated as a synonymous pair, and the variation in the text notation has been resolved.

[0028] The classification unit 222 may identify an item 32 using the similarity between the text 31 and the feature 35 without using the classifier 241. Specifically, the classification unit 222 identifies the text 31 corresponding to each item 32 and the feature 35 of that text 31 by using the correspondence between the text 31 and the items 32 identified from drawings of models and parts of the same series. The classification unit 222 then calculates the similarity between the text 31 and feature 35 corresponding to each item 32 and the target text 31 and the feature 35 of the target text 31. The classification unit 222 then identifies the item 32 with the highest similarity that is equal to or greater than a threshold as the item 32 corresponding to the target text 31.

[0029] In step S107, the classification unit 222 ends the classification process when it has identified which item 32 the target text 31 corresponds to for all texts 31 included in the text information .

[0030] (Step S108: Output process) The output unit 23 outputs the correspondence between the text 31 and the item 32 identified in steps S105 to S107 to the correspondence information 242 in the storage unit 24.

[0031] ***Effects of the First Embodiment*** As described above, the drawing information extraction device 10 according to the first embodiment associates the text 31 with the items 32 defined outside the drawing using the feature quantities 35 of the text 31. This makes it possible to identify the correspondence between information defined outside the drawing and information included in the drawing.

[0032] The drawing information extraction device 10 according to the first embodiment calculates the feature quantity 35 after removing the deviation of coordinate values ​​or the variation in word spelling. This allows the feature quantity 35 to be calculated appropriately. As a result, it becomes possible to appropriately identify the correspondence relationship.

[0033] Embodiment 2 The second embodiment differs from the eleventh embodiment in that each figure constituting a drawing is treated as one component group 37, and the correspondence between the text 31 and the item 32 is specified for each component group 37. In the second embodiment, this difference will be explained, and explanation of the same points will be omitted.

[0034] ***Configuration Description*** The functional configuration of the drawing information extraction device 10 according to the second embodiment will be described with reference to FIG. 2 in that the drawing information extraction device 10 includes a graphic area identification unit 25 and a group formation unit 26 as functional components. The group formation unit 26 includes a reference symbol recognition unit 261, an associated graphic area identification unit 262, and an associated text identification unit 263. The functions of the graphic area identification unit 25 and the group formation unit 26, like the other functional components, are realized by the processor 11 or a processing circuit. 2 in that the drawing information extraction device 10 stores component group information 243 in the storage unit 24. In addition, the drawing information extraction device 10 shown in FIG.

[0035] ***Explanation of Operation*** The operation of the drawing information extraction device 10 according to the second embodiment will be described with reference to FIGS. The operation procedure of the drawing information extraction device 10 according to the second embodiment corresponds to the drawing information extraction method according to the second embodiment. Moreover, the program that realizes the operation of the drawing information extraction device 10 according to the second embodiment corresponds to the drawing information extraction program according to the second embodiment.

[0036] The component group information 243 according to the second embodiment will be described. The part group information 243 is information about each of the multiple parts that make up the product corresponding to the drawing 30. The part group information 243 includes, for each of the multiple parts, a reference symbol that is identification information, and configuration information such as the shape and size of the part. Here, it is assumed that a part group 37 is formed for each part that makes up the product corresponding to the drawing 30.

[0037] The flow of processing performed by the drawing information extraction device 10 according to the second embodiment will be described with reference to FIG. The process of step S201 is the same as the process of step S101 in Fig. 3. The processes of steps S204 to S206 are the same as the processes of steps S102 to S104 in Fig. 3. The process of step S210 is the same as the process of step S108 in Fig. 3.

[0038] (Step S202: Graphic area identification process) The graphic region identification unit 25 forms a plurality of graphic groups 39 by performing expansion processing on a plurality of graphic elements 38 included in the drawing. The graphic elements 38 are elements other than the text 31 in the drawing 30, such as circles, line segments, and leader lines. The graphic region identification unit 25 then identifies, for each of the plurality of graphic groups 39, a group region 40 that includes the graphic elements 38 included in the target graphic group 39.

[0039] (Step S203: Group formation process) The group forming unit 26 groups a plurality of texts 31 and a plurality of graphics 38 to form a plurality of part groups 37 . Specifically, the group formation unit 26 sets each of the plurality of texts 31 as the target text 31. The group formation unit 26 identifies the shape group 39 corresponding to the target text 31 based on the distance between the group region 40 identified for each of the plurality of shape groups 39 in step S202 and the target text 31. The group formation unit 26 identifies the shapes 38 included in each shape group 39 and the text corresponding to that shape group 39 as a component group 37.

[0040] (Steps S207 to S209: Classification process) In step S207, the classification unit 222 sequentially sets each of the component groups 37 identified in step S203 as the target component group 37. At this time, the classification unit 222 sets only the text 31 included in the text information 34 for which the corresponding item 32 has not yet been identified as the target text 31. The processing in steps S208 and S209 is the same as the processing in steps S106 and S107 in Fig. 3. At this time, items 32 for which corresponding text 31 has already been determined may be excluded from the candidates, and an item 32 corresponding to the text 31 may be identified from the candidates. Also, if an item 32 has been determined for each component, only the items 32 for the components corresponding to the target component group 37 may be considered as candidates, and the item 32 corresponding to the text 31 may be identified from the candidates.

[0041] The graphic area specifying process (step S202 in FIG. 9) according to the second embodiment will be described with reference to FIG. (Step S301: Image conversion process) The graphic region specifying unit 25 performs image conversion on the CAD drawing 30 to generate a drawing image 41.

[0042] (Step S302: Grouping process) The graphic region identification unit 25 performs expansion processing on each graphic 38 included in the drawing image 41. The graphic region identification unit 25 groups connected graphics 38 into one graphic group 39. The graphic region identification unit 25 also groups graphics 38 contained within other graphics 38 into the same graphic group 39 as the other graphics 38. If the expansion process is performed excessively, there is a possibility that drawings of different parts will be aggregated into one graphic group 39. Therefore, the weighting of the expansion process may be adjusted depending on the drawing 30 so that the graphic group 39 is formed appropriately.

[0043] (Step S303: Area identification process) The graphic region identification unit 25 sets each graphic group 39 formed in step S302 as a target graphic group 39. As shown in Fig. 11, the graphic region identification unit 25 identifies a group region 40 that includes the target graphic group 39. Specifically, the graphic region identification unit 25 identifies the contour coordinates of the target graphic group 39. In Fig. 11, the region surrounded by a thick line corresponds to the graphic group 39, and the coordinates of the thick line portion are the contour coordinates. Then, the graphic region specifying unit 25 outputs the contour coordinates of each graphic group 39 together with the ID of the graphic group 39 to the volatile memory 12 as graphic group information 42. In FIG.

[0044] The group formation process (step S203 in FIG. 9) according to the second embodiment will be described with reference to FIG. (Steps S401 to S408: Parts group identification process) In step S401, the reference symbol recognition unit 261 refers to the graphic group information 42 and sets each graphic group 39 identified in step S202 as the target graphic group 39.

[0045] In step S402, the reference symbol recognition unit 261 acquires the text 31 contained in the target graphic group 39 from the text information 34 output in step S201.

[0046] In step S403, the reference symbol recognition unit 261 determines whether or not a reference symbol exists in the text 31 acquired in step S402. For example, a format for reference symbols is defined. If the text 31 contains a portion that follows the reference symbol format, the reference symbol recognition unit 261 determines that the text 31 contains a reference symbol. If the reference symbol recognition unit 261 determines that the reference symbol exists in the text 31, the process proceeds to step S404. On the other hand, if the reference symbol recognition unit 261 determines that the reference symbol does not exist in the text 31, the process proceeds to step S406.

[0047] In step S404, the belonging graphic identification unit 262 compares the reference symbol present in the text 31 with the reference symbol for each component included in the component group information 243. In this way, the belonging graphic identification unit 262 determines whether or not the reference symbol present in the text 31 exists in the component group information 243. If the reference symbol present in the text 31 exists in the component group information 243, the belonging feature identification unit 262 advances the process to step S405. On the other hand, if the reference symbol present in the text 31 does not exist in the component group information 243, the belonging feature identification unit 262 advances the process to step S406.

[0048] In step S405, the belonging graphic identification unit 262 identifies the component group 37 of the component corresponding to the reference symbol present in the text 31 as the component group 37 of the target graphic group 39.

[0049] In step S406, the belonging graphic identification unit 262 compares the shape of the part identified from the contour coordinates of the target graphic group 39 with the shapes of each part in the component group information 243 to identify the component group 37 of the part corresponding to the target graphic group 39. At this time, the belonging graphic identification unit 262 may exclude component groups 37 that have already been identified as component groups 37 of other graphic groups 39 from the shape comparison targets.

[0050] In step S407, the belonging graphic identification unit 262 adds the component group 37 identified in step S405 or step S406 to the information about the target graphic group 39 in the graphic group information .

[0051] In step S408, if the processing for all the graphic groups 39 has been completed, the belonging graphic identification unit 262 proceeds to step S409. On the other hand, if there are any graphic groups 39 for which processing has not been completed, the belonging graphic identification unit 262 sets the new graphic group 39 as the target graphic group 39 and executes the processing from step S402.

[0052] (Steps S409 to S413: Belonging text identification process) In step S409, the belonging text specification unit 263 refers to the text information 34 output in step S201, and sets each of the one or more texts 31 included in the text information 34 as the target text 31.

[0053] In step S410, the belonging text specification unit 263 determines whether the rectangular coordinates 33 of the target text 31 are inside the group area 40 of any graphic group 39. If the rectangular coordinates 33 are not inside the group area 40 of any of the graphic groups 39, the belonging text identification unit 263 proceeds to step S411. On the other hand, if the rectangular coordinates 33 are inside the group area 40 of any of the graphic groups 39, the belonging text identification unit 263 proceeds to step S412.

[0054] In step S411, the belonging text identifying unit 263 uses the outline coordinates of each group area 40 to identify the graphic group 39 that is closest to the rectangular coordinates 33 of the target text 31.

[0055] In step S412, the belonging text specification unit 263 specifies the part group 37 to which the target text 31 belongs. A case will be described where the rectangular coordinates 33 are not inside the group area 40 of any of the graphic groups 39. In this case, the belonging text identification unit 263 identifies the component group 37 corresponding to the graphic group 39 identified in step S411 as the component group 37 to which the target text 31 belongs. A case will be described where the rectangular coordinates 33 are inside the group area 40 of one of the graphic groups 39. In this case, the belonging text identification unit 263 identifies the component group 37 corresponding to the graphic group 39 in the group area 40 that contains the rectangular coordinates 33 as the component group 37 to which the target text 31 belongs.

[0056] In step S413, if the processing for all texts 31 included in the text information 34 has been completed, the belonging text identification unit 263 proceeds to step S414. If there is any text 31 for which processing has not been completed, the belonging text identification unit 263 executes the processing from step S410 for the new text 31 as the target text 31.

[0057] (Step S414: Output process) The belonging text specification unit 263 adds the part group 37 specified in step S412 to the information about the target text 31 in the text information .

[0058] ***Effects of the Second Embodiment*** As described above, the drawing information extraction device 10 according to the second embodiment divides a drawing into a plurality of component groups 37, and identifies the correspondence between the text 31 and the items 32 for each component group 37. This makes it easier for the classifier 241 to recognize the correspondence between the text 31 and the items 32. This improves the recognition accuracy of the correspondence between the text 31 and the items 32. Furthermore, since the number of learning targets for the classifier 241 is reduced, the load associated with training the classifier 241 can be reduced.

[0059] Embodiment 3 The third embodiment differs from the second embodiment in that priority is assigned to the component groups 37, and correspondence relationships are identified in descending order of priority. In the third embodiment, this difference will be described, and a description of the same points will be omitted.

[0060] ***Configuration Description*** The functional configuration of the drawing information extraction device 10 according to the third embodiment will be described with reference to FIG. The drawing information extraction device 10 differs from the drawing information extraction device 10 shown in Fig. 8 in that the item identification unit 22 includes a ranking determination unit 223 as a functional component. The drawing information extraction device 10 also differs from the drawing information extraction device 10 shown in Fig. 8 in that relationship information 244 is stored in the storage unit 24.

[0061] ***Explanation of Operation*** The operation of the drawing information extraction device 10 according to the third embodiment will be described with reference to FIGS. The operation procedure of the drawing information extraction device 10 according to the third embodiment corresponds to the drawing information extraction method according to the third embodiment. Moreover, the program that realizes the operation of the drawing information extraction device 10 according to the third embodiment corresponds to the drawing information extraction program according to the third embodiment.

[0062] The flow of processing performed by the drawing information extraction device 10 according to the third embodiment will be described with reference to FIG. The processing from step S501 to step S506 is the same as the processing from step S201 to step S206 in Fig. 9. The processing from step S511 is the same as the processing from step S210 in Fig. 9.

[0063] (Step S507: Ranking determination process) The ranking determination unit 223 identifies the relationships between the component groups 37 from the text 31 included in each component group 37. Then, the ranking determination unit 223 determines the priorities of the multiple component groups 37 based on the relationships between the component groups 37.

[0064] (Steps S508 to S510: Classification process) In step S508, the classification unit 222 sets each of the component groups 37 identified in step S503 as the target component group 37 in descending order of priority determined in step S507. The processes in steps S509 and S510 are the same as those in steps S208 and S209 in FIG.

[0065] The ranking determination process (step S507 in FIG. 14) according to the third embodiment will be described with reference to FIG. (Step S601: Relationship information reference process) The ranking determination unit 223 refers to the relationship information 244 in the storage unit 24 . The relationship information 244 is information that defines the relationship between the component groups 37 in the drawing 30. As a specific example, the relationship information 244 indicates the relationship between the enlargement destination and the enlargement source of the drawing. As shown in FIG. 16, the relationship information 244 includes a relationship, a determination factor, and a priority. The relationship indicates what kind of relationship exists between the component groups 37. The determination factor is information for identifying that there is a relationship. The priority indicates which of the component groups 37 has priority among the related component groups 37. In FIG. 16, the relationship indicates the relationship between the enlargement destination and the enlargement source. The determination factors indicate the relationship between the enlargement destination and the enlargement source by the reference symbols A and C. The priority indicates that the enlargement destination has a higher priority than the enlargement source.

[0066] (Step S602: Definition determination process) The ranking determination unit 223 determines whether or not the relationship information 244 contains a definition of the relationship. If a definition of the relationship exists, the ranking determination unit 223 advances the process to step S603. On the other hand, if a definition of the relationship does not exist, the ranking determination unit 223 advances the process to step S605.

[0067] (Step S603: Parts group determination process) The ranking determination unit 223 determines whether or not a component group 37 corresponding to the relationship defined in the relationship information 244 exists in the drawing 30. Specifically, the ranking determination unit 223 refers to the determination elements of the relationship information 244 and determines whether or not the corresponding component group 37 exists in the drawing 30. If the component group 37 exists in the drawing 30, the ranking determination unit 223 advances the process to step S604. On the other hand, if the component group 37 does not exist in the drawing 30, the ranking determination unit 223 advances the process to step S605.

[0068] (Step S604: First ranking determination process) The ranking determination unit 223 determines the order of priority between the component groups 37 corresponding to the relationships defined in the relationship information 244 in accordance with the priority of the relationship information 244 .

[0069] (Step S605: Second ranking determination process) For component groups 37 for which the priority order cannot be determined based on the relationships between the component groups 37 in step S604, the priority order determination unit 223 determines the priority order so that the less text 31 included, the higher the priority order. Note that here, the priority order is determined taking into consideration only the number of texts 31, but the priority order may also be determined taking into consideration the number of elements other than text 31, so that the fewer the number of elements, the higher the priority order.

[0070] (Step S606: Output process) The ranking determination unit 223 outputs the priorities determined in steps S604 and S605 to the classification unit 222.

[0071] In Figure 17, component group 37(1) and component group 37(2) have a relationship of source of enlargement and destination of enlargement. Similarly, component group 37(3) and component group 37(4) have a relationship of source of enlargement and destination of enlargement. The relationship between component group 37(1) and component group 37(2) is identified by the presence of reference symbol A shown as a determination element in Figure 16. Similarly, the relationship between component group 37(3) and component group 37(4) is identified by the presence of reference symbol C shown as a determination element in Figure 16. Then, in step S604, it is determined that the expansion destination component groups 37(2) and 37(4) have a higher priority than the expansion source component groups 37(1) and 37(3). However, in this case, the priority is not determined between component groups 37(2) and 37(4) and between component groups 37(1) and 37(3). Next, in step S605, the number of text 31 in component group 37(2) and the number of text 31 in component group 37(4) are counted. Here, the number of text 31 in component group 37(2) is four, and the number of text 31 in component group 37(4) is three. Therefore, component group 37(4), which has fewer text 31, is determined to have a higher priority than component group 37(2), which has more text 31. Similarly, the number of text 31 in component group 37(1) and the number of text 31 in component group 37(3) are counted. Here, the number of text 31 in both component group 37(1) and component group 37(3) is two. Therefore, either component group 37(1) or component group 37(3) may have a higher priority. Here, it is assumed that component group 37(1) is determined to have a higher priority than component group 37(3). As a result, as shown in FIG. 18, the order of priority is component group 37(4)>component group 37(2)>component group 37(1)>component group 37(3).

[0072] ***Effects of the Third Embodiment*** As described above, the drawing information extraction device 10 according to the third embodiment identifies correspondences between the text 31 and the items 32 in stages according to the priority order determined based on the relationships between the component groups 37. As a result, correspondences that can be identified in the component groups 37 where correspondences are easier to identify are identified first. This improves the accuracy of recognizing correspondences between the text 31 and the items 32.

[0073] Furthermore, the drawing information extraction device 10 according to the third embodiment assigns a higher priority to a part group 37 having a smaller number of texts 31. The fewer the number of texts 31, the easier it is to identify the correspondence. This improves the accuracy of recognizing the correspondence between the texts 31 and the items 32.

[0074] Embodiment 4 The fourth embodiment differs from the first to third embodiments in that the classifier 241 is retrained using retraining data obtained by excluding from the training data pairs containing text 31 with the same content as the text 31 extracted from the drawing 30. In the fourth embodiment, this difference will be explained, and explanation of the same points will be omitted. In the fourth embodiment, a case where a modification is made to the second embodiment will be described. However, modifications can also be made to the first and third embodiments.

[0075] ***Configuration Description*** The functional configuration of the drawing information extraction device 10 according to the fourth embodiment will be described with reference to FIG. The drawing information extraction device 10 differs from the drawing information extraction device 10 shown in Fig. 8 in that the item identification unit 22 includes, as functional components, an item determination unit 224 and a relearning unit 225. The drawing information extraction device 10 also differs from the drawing information extraction device 10 shown in Fig. 8 in that learning data 245 is stored in the memory unit 24.

[0076] ***Explanation of Operation*** The operation of the drawing information extraction device 10 according to the fourth embodiment will be described with reference to FIGS. The operation procedure of the drawing information extraction device 10 according to the fourth embodiment corresponds to the drawing information extraction method according to the fourth embodiment. Moreover, the program that realizes the operation of the drawing information extraction device 10 according to the fourth embodiment corresponds to the drawing information extraction program according to the fourth embodiment.

[0077] The flow of processing performed by the drawing information extraction device 10 according to the fourth embodiment will be described with reference to FIG. The processing from step S701 to step S706 is the same as the processing from step S201 to step S206 in Fig. 9. The processing from step S712 is the same as the processing from step S210 in Fig. 9.

[0078] (Step S707: Item confirmation process) The item determination unit 224 sets each text 31 extracted from the drawing 30 in step S701 as the target text 31. If the training data 245 contains text 31 with the same content as the target text 31, the item determination unit 224 determines the target text 31 as item 32 in the training data 245. On the other hand, if the training data 245 does not contain text 31 with the same content as the target text 31, the item determination unit 224 determines the target text 31 as undetermined. Here, learning is performed using as input the text 31 and learning data 245 consisting of pairs of feature quantities 35 and items 32.

[0079] (Step S708: Re-learning process) The retraining unit 225 removes the pairs for the text 31 determined in step S708 from the training data 245 and uses the remaining pairs to retrain the classifier 241.

[0080] (Steps S709 to S711: Classification process) In step S709, the classification unit 222 sequentially sets each component group 37 identified in step S703 as the target component group 37. At this time, the classification unit 222 sets only the text 31 included in the text information 34 for which the corresponding item 32 has not yet been identified as the target text 31. Note that the text 31 for which the item 32 has been determined in step S708 is also excluded from the target. The processes in steps S710 and S711 are the same as those in steps S509 and S510 in FIG.

[0081] The item confirmation process (step S708 in FIG. 20) according to the fourth embodiment will be described with reference to FIG. (Step S801: Target setting process) The item determination unit 224 refers to the text information 34 extracted in step S701, and sets each of the one or more pieces of text 31 included in the text information 34 as the target text 31.

[0082] (Step S802: Existence determination process) The item determination unit 224 determines whether or not the training data 245 contains a text 31 having the same content as the target text 31 . If the training data 245 contains text 31 with the same content as the target text 31, the item determination unit 224 proceeds to step S803. On the other hand, if the training data 245 does not contain text 31 with the same content as the target text 31, the item determination unit 224 ends the process.

[0083] (Step S803: Item confirmation process) The item determination unit 224 determines the item 32 corresponding to the target text 31 as the item 32 in the set of texts 31 having the same content as the text 31 in the training data 245 . For example, suppose a set of text 31, item 32, and feature 35 obtained from a source drawing is registered as learning data 245, and text 31 and item 32 in a destination drawing created from the source drawing are to be associated with each other. In this case, for text 31 whose content has not been changed from the source drawing, text 31 with the same content also exists in learning data 245. Therefore, item 32 of that text 31 is determined.

[0084] (Step S804: End determination process) When the item determination unit 224 has completed processing for all text 31 included in the text information 34, the process proceeds to step S805. If there is any text 31 for which processing has not been completed, the item determination unit 224 performs processing from step S802 on the new text 31 as the target text 31.

[0085] (Step S805: Output process) The item determination unit 224 outputs the determined and undetermined results of each text 31 to the re-learning unit 225.

[0086] The re-learning process (step S709 in FIG. 20) according to the fourth embodiment will be described with reference to FIG. (Step S901: Result acquisition process) The re-learning unit 225 acquires the determined and undetermined results for each text 31 output in step S805.

[0087] (Step S902: Final determination process) The re-learning unit 225 uses the results acquired in step S901 to determine whether or not there is text 31 whose item 32 has been confirmed. If there is text 31 with a confirmed item 32, the re-learning unit 225 proceeds to step S903. On the other hand, if there is no text 31 with a confirmed item 32, the re-learning unit 225 ends the process.

[0088] (Step S903: Exclusion process) The re-learning unit 225 removes the sets of texts 31 whose items 32 have been determined from the training data 245, and generates new training data 245 made up of the remaining sets.

[0089] (Step S904: Learning process) The re-learning unit 225 re-learns the classifier 241 using the new training data 245 generated in step S903.

[0090] Assume that the drawing shown in (A) of FIG. 23 is the drawing that is the source of training data 245. Assume that a derived drawing shown in (B) of FIG. 23 is generated using the drawing shown in (A) of FIG. 23 as the source drawing. Assume that the correspondence between text 31 and items 32 of the derived drawing is identified. In this case, training data 245 shown in FIG. 24 is data created from the source drawing. In training data 245 shown in FIG. 24, items 32 corresponding to each text 31 of the source drawing shown in (A) of FIG. 23 are shown. Note that the feature quantities 35 in training data 245 are omitted here. In step S802 of FIG. 21, it is determined whether each text 31 of the derived drawing shown in FIG. 23B exists in the learning data 245 shown in FIG. 24. Then, it is determined that "Φ6.2" and "3-drill ±0.2" exist in the learning data 245. Then, the item 32 of "Φ6.2" is determined to be "outer diameter 2," and the item 32 of "3-drill ±0.2" is determined to be "hole diameter." On the other hand, since "Φ4.5" and "Φ5.2 ±0.1" do not exist in the learning data 245, the items 32 are undetermined. Therefore, as shown in FIG. 25, the determination result for each text 31 of the derived drawing is determined. In this case, in step S903 of Fig. 22, the pair of "Φ6.2" and "3 cut ±0.2" for which item 32 has been determined is excluded from the training data 245. As a result, training data 245 is generated in which the pair of "Φ4.5" and "Φ5.2 ±0.1" remains, as shown in Fig. 26. Then, the classifier 241 is retrained using this training data 245.

[0091] ***Effects of the Fourth Embodiment*** As described above, the drawing information extraction device 10 according to the fourth embodiment retrains the classifier 241 using retrained data obtained by excluding from the training data pairs containing text 31 with the same content as the text 31 extracted from the drawing 30. This allows appropriate training to be performed for the text 31 with undetermined items 32. As a result, by using the classifier 241 after retraining, it is possible to appropriately associate the items 32 with the text 31 with undetermined items 32.

[0092] Various aspects of the present disclosure are summarized below as appendices. (Appendix 1) an extractor for extracting text from the drawing; a feature calculation unit that calculates feature amounts of the text extracted by the extraction unit; a classification unit that uses the text and the feature calculated by the feature calculation unit to identify which item of a plurality of items defined outside the drawing the text corresponds to; A drawing information extraction device comprising: (Appendix 2) The feature calculation unit calculates the feature using at least one of coordinate values ​​of the text, words included in the text, and attribute information of the text. 2. A drawing information extraction device according to claim 1. (Appendix 3) The feature amount calculation unit converts the coordinate values ​​into relative coordinates based on a specific reference symbol in the drawing, and then calculates the feature amounts from the coordinate values. 3. A drawing information extraction device according to claim 2. (Appendix 4) The feature calculation unit converts words included in the text by referring to synonym information in which synonyms are defined, and then calculates the feature from the words. 4. The drawing information extraction device according to claim 2 or 3. (Appendix 5) The classification unit inputs the text and the feature amount into a classifier, which is a learning model, and acquires an item corresponding to the text output by the classifier, thereby identifying which item the text corresponds to. 5. A drawing information extraction device according to any one of appendices 1 to 4. (Appendix 6) the extraction unit extracts a plurality of texts and a plurality of figures from the drawing; The drawing information extraction device further comprises: a group forming unit that groups the plurality of texts and the plurality of graphics to form a plurality of component groups; Equipped with The classification unit identifies, for each part group formed by the group formation unit, to which item the text included in the part group corresponds. 6. A drawing information extraction device according to any one of appendices 1 to 5. (Appendix 7) The drawing information extraction device further comprises: a graphic area specifying unit that performs expansion processing on the plurality of graphics to form a plurality of graphic groups, and specifies, for each of the plurality of graphic groups, a group area that includes a graphic included in the target graphic group; Equipped with The group forming unit, for each of the plurality of texts, identifies a graphic group corresponding to the target text based on the distance between the group area identified for each of the plurality of graphic groups by the graphic area identifying unit and the target text, and identifies, for each of the plurality of graphic groups, a graphic included in the graphic group and the text corresponding to the graphic group as one component group. 7. A drawing information extraction device according to claim 6. (Appendix 8) The drawing information extraction device further comprises: a ranking determination unit that identifies relationships between the component groups from text included in the component groups and determines priorities for the plurality of component groups based on the relationships between the component groups; Equipped with The classification unit identifies, in descending order of the priority of the component groups determined by the ranking determination unit, to which item the text included in each component group corresponds. 8. The drawing information extraction device according to claim 6 or 7. (Appendix 9) The priority determination unit determines the priority of a component group whose priority cannot be determined based on the relationship between the component groups such that the less text included, the higher the priority. 9. The drawing information extraction device according to claim 8. (Appendix 10) the classifier is trained using training data comprising a plurality of text-item pairs; The drawing information extraction device further comprises: a re-training unit that re-trains the classifier using re-training data obtained by excluding from the training data sets any sets containing text with the same content as the text extracted by the extraction unit; 6. The drawing information extraction device according to claim 5, comprising: (Appendix 11) The computer extracts the text from the drawing, The computer calculates features of the text, A drawing information extraction method in which a computer uses the text and the feature amount to identify which item of a plurality of items defined outside the drawing the text corresponds to. (Appendix 12) an extraction process for extracting text from the drawing; a feature calculation process for calculating feature amounts of the text extracted by the extraction process; a classification process for identifying which of a plurality of items defined outside the drawing the text corresponds to, using the text and the feature calculated by the feature calculation process; and A drawing information extraction program that causes a computer to function as a drawing information extraction device that performs the above.

[0093] The embodiments and modifications of the present disclosure have been described above. Some of these embodiments and modifications may be combined and implemented. Also, one or more of them may be implemented partially. Note that the present disclosure is not limited to the above embodiments and modifications, and various modifications are possible as needed. [Explanation of symbols]

[0094] 10 Drawing information extraction device, 11 Processor, 12 Volatile memory, 13 Non-volatile memory, 14 Interface device, 21 Extraction unit, 22 Item identification unit, 221 Feature calculation unit, 222 Classification unit, 223 Rank determination unit, 224 Item confirmation unit, 225 Re-learning unit, 23 Output unit, 24 Memory unit, 241 Classifier, 242 Correspondence information, 243 Component group information, 244 Relationship information, 245 Learning data, 25 Graphic area identification unit, 26 Group formation unit, 261 Reference symbol recognition unit, 262 Belonging graphic identification unit, 263 Belonging text identification unit, 30 Drawing, 31 Text, 32 Item, 33 Rectangular coordinates, 34 Text information, 35 Feature, 36 Synonym information, 37 Component group, 38 Graphic, 39 Graphic group, 40 Group area, 41 Drawing image, 42 Shape group information.

Claims

1. an extractor for extracting text from the drawing; a feature calculation unit that calculates feature amounts of the text extracted by the extraction unit; a classification unit that uses the text and the feature calculated by the feature calculation unit to identify which item of a plurality of items defined outside the drawing the text corresponds to; A drawing information extraction device comprising:

2. The feature calculation unit calculates the feature using at least one of coordinate values ​​of the text, words included in the text, and attribute information of the text. The drawing information extraction device according to claim 1 .

3. The feature amount calculation unit converts the coordinate values ​​into relative coordinates based on a specific reference symbol in the drawing, and then calculates the feature amounts from the coordinate values. The drawing information extraction device according to claim 2.

4. The feature calculation unit converts words included in the text by referring to synonym information in which synonyms are defined, and then calculates the feature from the words. The drawing information extraction device according to claim 2.

5. The classification unit inputs the text and the feature amount into a classifier, which is a learning model, and acquires an item corresponding to the text output by the classifier, thereby identifying which item the text corresponds to. The drawing information extraction device according to claim 1 .

6. the extraction unit extracts a plurality of texts and a plurality of figures from the drawing; The drawing information extraction device further comprises: a group forming unit that groups the plurality of texts and the plurality of graphics to form a plurality of component groups; Equipped with The classification unit identifies, for each part group formed by the group formation unit, to which item the text included in the part group corresponds. The drawing information extraction device according to claim 1 .

7. The drawing information extraction device further comprises: a graphic area specifying unit that performs expansion processing on the plurality of graphics to form a plurality of graphic groups, and specifies, for each of the plurality of graphic groups, a group area that includes a graphic included in the target graphic group; Equipped with The group forming unit, for each of the plurality of texts, identifies a graphic group corresponding to the target text based on the distance between the group area identified for each of the plurality of graphic groups by the graphic area identifying unit and the target text, and identifies, for each of the plurality of graphic groups, a graphic included in the graphic group and the text corresponding to the graphic group as one component group. The drawing information extraction device according to claim 6.

8. The drawing information extraction device further comprises: a ranking determination unit that identifies relationships between the component groups from text included in the component groups and determines priorities of the plurality of component groups based on the relationships between the component groups; Equipped with The classification unit identifies, in descending order of the priority of the component groups determined by the ranking determination unit, to which item the text included in each component group corresponds. The drawing information extraction device according to claim 6.

9. The priority determination unit determines the priority of a component group whose priority cannot be determined based on the relationship between the component groups such that the less text included, the higher the priority. The drawing information extraction device according to claim 8.

10. the classifier is trained using training data comprising a plurality of text-item pairs; The drawing information extraction device further comprises: a re-training unit that re-trains the classifier using re-training data obtained by excluding from the training data sets any sets containing text with the same content as the text extracted by the extraction unit; The drawing information extraction device according to claim 5 , comprising:

11. The computer extracts the text from the drawing, The computer calculates features of the text, A drawing information extraction method in which a computer uses the text and the feature amount to identify which item of a plurality of items defined outside the drawing the text corresponds to.

12. an extraction process for extracting text from the drawing; a feature calculation process for calculating feature amounts of the text extracted by the extraction process; a classification process for identifying which of a plurality of items defined outside the drawing the text corresponds to, using the text and the feature calculated by the feature calculation process; and A drawing information extraction program that causes a computer to function as a drawing information extraction device that performs the above.

Images