Symbol extraction device, program, and symbol extraction method

Abstract

To provide a symbol extraction device capable of reducing a burden to find a desired partial image in images.SOLUTION: A symbol extraction device 100 comprises a template setting unit 112 to acquire a template image of a target image extracted from images (refer to design image data 130), an analysis condition setting unit 113 to acquire a threshold used for a successful / failure determination of the extraction, a collation unit 114 to extract a partial image with a similarity to the template image being equal to or above a threshold, from the image, and an extracted result management unit 115 to display the extracted result including at least one item among the extracted partial image, the identification number of the partial image, the position of the image in the partial image, and the similarity between the partial image and the template image.SELECTED DRAWING: Figure 8

Description

[Technical Field]

[0001] The present invention relates to a symbol extraction device, a program, and a symbol extraction method for extracting symbols from image data such as drawings. [Background technology]

[0002] In facility construction such as electrical equipment construction and air conditioning equipment construction, it is necessary to extract the materials and equipment required for the construction from the drawings and estimate the costs. Extracting these materials and equipment manually takes time and places a burden on the person in charge, so there is a need to automate this extraction process (process it by computer).

[0003] To find materials and equipment, it is necessary to search within the drawings for symbols (figures) that represent the materials and equipment. In other words, the system searches within the image data of the drawings for a partial image that matches the image of the symbol. It is rare for the image of a symbol to perfectly match the partial image in the drawing, and it is generally determined that a symbol represents a material or component if the similarity is above a certain threshold. In such a method, if the threshold is large, the probability (accuracy) that the detected partial image is a symbol increases, but the number of symbols missed (detection failures) increases. Conversely, if the threshold is small, fewer symbols will be missed, but the accuracy will decrease.

[0004] The image matching device described in Patent Document 1 is composed of an image input unit that inputs an input image, a template setting unit that sets a desired area in a reference image as a template, a measurement position setting unit that sets an area in the input image that corresponds to the template as a comparison area, a matching threshold setting unit that sets a matching threshold, an image data matching unit that calculates a normalized absolute difference value between the template and the comparison area, a matching judgment unit that compares the matching threshold with the normalized absolute difference value to judge the similarity between the template and the comparison area, a judgment result output unit that outputs the judgment result by the matching judgment means, and a memory that stores each of the data. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Laid-Open No. 2008-204307 Summary of the Invention [Problem to be solved by the invention]

[0006] According to the image matching device described in Patent Document 1, the similarity between two images is expressed as a numerical value between 0 and 1, so there is no need to set a matching threshold for determining the similarity between two images depending on the size of the template area or the brightness of the template or comparison area, and the matching threshold can be easily set. However, setting a threshold is necessary, and the work of confirming that the found (extracted) partial image is a symbol (desired partial image) and searching for overlooked symbols must be done manually, which still places a heavy workload on the person in charge. The present invention has been made in consideration of the above background, and aims to provide a symbol extraction device, program, and symbol extraction method that reduce the burden (human burden, etc.) of searching for a desired partial image within an image. [Means for solving the problem]

[0007] In order to solve the above-mentioned problems, a symbol extraction device according to the present invention includes a template setting unit that sets a template image to be extracted from an image; an analysis condition setting unit that acquires and sets a threshold value used to determine whether extraction is appropriate; a matching unit that extracts from the image a partial image whose similarity to the template image is equal to or greater than the threshold value; and an extraction result management unit that displays an extraction result including at least one item of the extracted partial image, an identification number of the partial image, a position of the partial image within the image, and a similarity between the partial image and the template image. The extraction result management unit has a sorting function based on one selected from the identification number of the displayed partial image, the position of the partial image within the image, and the similarity between the partial image and the template image. a matching unit that extracts partial images from the image whose similarity to the template image is equal to or greater than the threshold; an extraction result management unit that displays extraction results including at least one item of the extracted partial images, their identification numbers, their positions within the image, and their similarity between the partial images and the template image; and a learning unit that generates a learning model that determines whether an image is the template image using learning data including the partial images included in the extraction results, wherein the similarity is calculated based on a template matching technique, and the matching unit extracts partial images whose similarity calculated based on the template matching technique is equal to or greater than the threshold and that are determined to be the template image using the learning model. [Effects of the Invention]

[0008] According to the present invention, it is possible to provide a symbol extraction device, a program, and a symbol extraction method that reduce the burden of searching for a desired partial image within an image. Problems, configurations, and effects other than those described above will become clear from the description of the following embodiments. [Brief explanation of the drawings]

[0009] [Figure 1] 4 is a flowchart showing a procedure for a user who uses the symbol extraction device according to the first embodiment. [Figure 2] 3 shows a screen displaying a drawing read by the symbol extraction device according to the first embodiment. [Figure 3] 1 is a screen of the symbol extraction device displaying a template setting window according to the first embodiment. [Figure 4] 1 is a screen of the symbol extraction device displaying an analysis condition setting window according to the first embodiment. [Figure 5] 10 is a screen showing an extraction result when a threshold value of 0.85 is used according to the first embodiment. [Figure 6] 10 is a screen showing an extraction result when a threshold value of 0.65 is used according to the first embodiment. [Figure 7] 1 is a screen of the symbol extraction device displaying an extraction result window according to the first embodiment. [Figure 8] 1 is a functional block diagram of a symbol extraction device according to a first embodiment. [Figure 9] FIG. 3 is a data configuration diagram of an extraction result database according to the first embodiment. [Figure 10] 4 is a flowchart of a symbol extraction process according to the first embodiment. [Figure 11] 10 is a flowchart of extraction result screen processing according to the first embodiment. [Figure 12] FIG. 10 is a functional block diagram of a symbol extraction device according to a second embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0010] <Outline of the symbol extraction device> A symbol extraction device in a mode (embodiment) for carrying out the present invention will be described below. The symbol extraction device searches an image of a drawing (drawing image) to extract images of symbols contained in the image. The symbol images are set as template images. The symbol extraction device scans the drawing image to find (extract) partial images (symbol images) of the drawing image whose similarity to the template image is equal to or greater than a set threshold (similar to the template image).

[0011] The symbol extractor displays an extraction result window containing a list of the identification numbers of the found partial images and their positions within the drawing image. When an item in this list is specified, the corresponding partial image is displayed. Users of the symbol extraction device set a low similarity threshold to reduce the chance of missing partial images (missed extraction / missed detection). The list in the extraction result window contains incorrect extractions (misrecognitions), but users can select items in the list one after another to display similar partial images one after another, delete the incorrect extractions, and obtain the final extraction results. In this way, users can efficiently extract symbols from drawing images.

[0012] <<How to use the symbol extraction device>> Before describing the configuration of the symbol extraction device 100 (see FIG. 8 described later), we will explain the procedure for using the symbol extraction device 100. Figure 1 is a flowchart showing the procedure for using the symbol extraction device 100 according to the first embodiment. In step S11, the user of the symbol extraction device 100 specifies (the file name of) a design drawing for which symbols are to be searched. The symbol extraction device 100 then reads and displays the design drawing (see FIG. 2, which will be described later). FIG. 2 shows a screen 310 that displays the drawing read by the symbol extraction device 100 according to the first embodiment.

[0013] Returning to FIG. 1, the explanation of the usage procedure will be continued. In step S12, the user sets an image of a symbol as a template. FIG. 3 shows a screen 320 of the symbol extraction device 100 on which a template setting window 321 according to the first embodiment is displayed. The user sets and saves the template image by specifying an area containing a symbol in a drawing (for example, an area containing the symbol indicated by reference numeral 322). The user may set the template image by specifying a file containing the template image, or may set the template image by selecting a template image that has already been set.

[0014] Returning to FIG. 1, the explanation of the usage procedure will be continued. In step S13, the user sets analysis conditions, which are conditions for extracting symbols from a drawing. FIG. 4 shows a screen 330 of the symbol extraction device 100 displaying an analysis condition setting window 331 according to the first embodiment. The image in the upper right corner of the analysis condition setting window 331 is an image (template image) of the symbol to be found (extracted) from the drawing set in step S12. The user sets a threshold value and a rotation angle. The threshold value is a threshold value of the similarity between a partial image of the drawing and the template image, which is referenced when determining whether the partial image of the drawing is a symbol. If the similarity between the partial image of the drawing and the template image is equal to or greater than the set threshold value, the symbol extraction device 100 extracts the partial image as a symbol.

[0015] The rotation angle is a unit of rotation angle when the similarity between a partial image in a drawing and a template image is calculated by rotating the template image. For example, if the rotation angle is specified as 90 degrees, the similarity between the partial image in the drawing and the template image as is, the template image rotated 90 degrees, the template image rotated 180 degrees, and the template image rotated 270 degrees is calculated, and if any similarity is equal to or greater than a threshold, the partial image is extracted as a symbol.

[0016] The user specifies the extraction target area, which is the area in the drawing where symbols are to be found, on the extraction target area specification screen (not shown) that is displayed when the user presses the "Specify extraction target area" button at the bottom of the analysis condition setting window 331. The drawing may contain a symbol legend, and the user specifies the extraction target area so as to exclude the area of ​​the legend. If the drawing contains multiple figures and only some of the figures are the extraction target area, the user specifies those figures as the extraction target area.

[0017] Returning to Figure 1, the explanation of the usage procedure will continue. In step S14, the user presses the "Execute Extraction" button in the analysis condition setting window 331 (see Figure 4) to instruct extraction of a partial image similar to the template image. Note that the user can also instruct extraction by selecting "Execute Extraction" in the "File" menu. In step S15, the user checks the extraction result and if it is OK (step S15→OK), proceeds to step S16, and if it is NG (step S15→NG), return to step S13 and reset the threshold value.

[0018] FIG. 5 is a screen 340 showing the extraction results when a threshold value of 0.85 is used in the first embodiment. The extracted partial images are highlighted by a rectangle, and an identification number is displayed near the rectangle (see reference numeral 343). However, it can be seen on screen 340 that symbols (partial images) that should have been extracted have not been extracted (see reference numerals 342 and 344). The user determines that the current threshold value of 0.85 is too high and many symbols have been missed, so that a reset and re-extraction are necessary. The user returns to step S13, resets the threshold to a value lower than 0.85, for example, 0.65, and instructs re-extraction in step S14. FIG. 6 is a screen 350 showing the extraction results when a threshold value of 0.65 is used in the first embodiment. Since all symbols have been extracted, the user determines that re-extraction is unnecessary and proceeds to step S16.

[0019] Returning to Figure 1, the explanation of the usage procedure will continue. In step S16, the user edits the extraction results. In detail, the user operates an extraction result window 361 (see Figure 7 described later) to remove erroneous extractions contained in the extraction results and add unextracted (omitted) symbols, thereby editing the extraction results. 7 shows a screen 360 of the symbol extraction device 100 displaying an extraction result window 361 according to the first embodiment. The extraction result window 361 displays a list including the following items: identification number (denoted as "#" in FIG. 7), position, angle (denoted as "angle" in FIG. 7), category, and adoption / rejection (denoted as "adoption" in FIG. 7).

[0020] The identification number is a number assigned to the extracted symbol (partial image of the symbol). The position is the position (coordinates) of the symbol in the design drawing. The angle is the rotation angle of the template image when the symbol was extracted (when the similarity is equal to or greater than the threshold) (see the rotation angle in the analysis condition setting window 331 in Figure 4). The classification is either "automatic" or "manual," and is "automatic" for symbols extracted by the symbol extraction device 100. When the user finds an unextracted (undetected) symbol (see reference numeral 368), the user presses the "add" button 363, specifies the area containing the symbol, and adds it to the extraction result. The classification of a symbol added to the extraction result by user specification is "manual."

[0021] The "Accept" or "Reject" indicates whether the extraction result should be adopted as the final extraction result (O / X). When any item (row) in the list is specified, a partial image of the symbol corresponding to that item is displayed in area 362. By displaying the extracted symbol, the user can check whether the extraction result contains any erroneous extractions. If there are any erroneous extractions, the user can press the "Delete" button 364 to remove the erroneous extraction from the extraction results. The extraction result list may include the degree of similarity.

[0022] <Configuration of the symbol extraction device> 8 is a functional block diagram of the symbol extraction device 100 according to the first embodiment. The symbol extraction device 100 is a computer, and includes a control unit 110, a storage unit 120, and an input / output unit 180. User interface devices such as a display, keyboard, and mouse are connected to the input / output unit 180. The input / output unit 180 may include a communication device, enabling transmission and reception of data (e.g., drawings) with other devices. A media drive may also be connected to the input / output unit 180, enabling data exchange using a recording medium.

[0023] <Symbol Extraction Device: Memory Unit> The storage unit 120 is configured to include storage devices such as a ROM (Read Only Memory), a RAM (Random Access Memory), and an SSD (Solid State Drive). The storage unit 120 stores design drawing data 130, an extraction result database 140, template information 121, and a program 128. The template information 121 includes information such as a template image, name, threshold, rotation angle, and extraction target area related to the symbol to be extracted (see the analysis condition setting window 331 in FIG. 4). The program 128 includes a description of the procedure for the symbol extraction process (see FIGS. 10 and 11), which will be described later. The design drawing data 130 is image data of a drawing.

[0024] FIG. 9 is a data configuration diagram of the extraction result database 140 according to the first embodiment. The extraction result database 140 is, for example, data in a table format, with one row (record) containing information related to one extracted partial image. A record in the extraction result database 140 contains columns (attributes) of an identification number (denoted as "#" in FIG. 9), position, angle, category, acceptance / rejection, similarity, and image. The identification number, position, angle, category, and acceptance / rejection correspond to the identification number, position, angle, category, and acceptance / rejection, respectively, in the extraction result window 361 (see FIG. 7). The similarity is the similarity with the template image. The image is data of the extracted partial image.

[0025] <Symbol extraction device: control unit> 8, the control unit 110 will be described. The control unit 110 is configured to include a CPU (Central Processing Unit), and has a drawing display control unit 111, a template setting unit 112, an analysis condition setting unit 113, a matching unit 114, and an extraction result management unit 115.

[0026] <Control unit: Drawing display control unit> The drawing display control unit 111 displays a drawing (see screen 310 in FIG. 2). When symbol extraction is instructed, the drawing display control unit 111 displays a screen including the extraction results (see screens 340 and 350 in FIGS. 5 and 6).

[0027] As described above, the symbol extraction device 100 includes an image display control unit (drawing display control unit 111) that highlights and displays a partial image in an image (an area to be extracted in the drawing) that corresponds to an extracted partial image (partial image of a symbol), and displays an identification number corresponding to the partial image near the partial image (see screens 340 and 350 in FIGS. 5 and 6).

[0028] <Control section: Template setting section, analysis condition setting section> The template setting unit 112 displays a template setting window 321 (see FIG. 3). The template setting unit 112 also acquires a template image of the symbol to be extracted, which is designated by the user, and stores it in the template information 121. The analysis condition setting unit 113 also displays an analysis condition setting window 331 (see FIG. 4). The analysis condition setting unit 113 also acquires the threshold value and rotation angle, which are conditions for extracting symbols specified by the user, and stores them in the template information 121.

[0029] As described above, the symbol extraction device 100 includes the template setting unit 112 that sets a template image (of a symbol image) that is to be extracted from an image (an extraction target region of a drawing). The symbol extraction device 100 also includes an analysis condition setting unit 113 that acquires and sets a threshold value used to determine whether extraction is appropriate. Furthermore, the analysis condition setting unit 113 acquires the rotation angle of the template image.

[0030] <Control unit: Collation unit> The matching unit 114 scans the image of a specified region of the drawing (region to be extracted) and extracts a partial image that matches the template image. More specifically, the matching unit 114 scans the specified region of the drawing and extracts a partial image that is in the specified region of the drawing and has a similarity to the template image that is equal to or greater than a threshold. The matching unit 114 rotates the partial image (see the rotation angle in the analysis condition setting window 331) and calculates the similarity. Methods for calculating the similarity include template matching such as zero-mean normalized cross-correlation, but other methods may also be used.

[0031] If the similarity between a partial image and the template image is equal to or greater than a threshold, the matching unit 114 determines that the partial image matches the template image and adds it to the extraction result database 140 (see FIG. 9). More specifically, the matching unit 114 adds a new record to the extraction result database 140, and stores the identification number, position, angle, category, acceptance / rejection, similarity, image, serial number, coordinates of the partial image in the drawing (for example, the coordinates in the upper left corner), the rotation angle of the partial image when it matches, "automatic", "o", similarity, and partial image.

[0032] As described above, the symbol extraction device 100 includes the matching unit 114 that extracts, from an image (a region to be extracted from a drawing), a partial image whose similarity to a template image is equal to or greater than a threshold value. Furthermore, the matching unit 114 extracts from the image a partial image whose similarity to an image obtained by rotating the template image by one or more rotation angles is equal to or greater than a threshold value. The similarity is calculated based on a template matching method.

[0033] <<Control unit: Extraction result management unit>> The extraction result management unit 115 displays the extraction result window 361 (see FIG. 7) by referring to the extraction result database 140 (see FIG. 9). The displayed contents may be sorted by items such as identification number or position. Furthermore, in accordance with a user's operation (instruction) on the extraction result window 361, the extraction result management unit 115 performs processing such as displaying a partial image corresponding to a selected item, adding a designated omitted extraction symbol (partial image in the drawing, see reference numeral 368 in FIG. 7) to the extraction results, and accepting an instruction on whether to accept or reject the extraction. Details of the processing by the extraction result management unit 115 will be described later with reference to FIG. 11.

[0034] As described above, the symbol extraction device 100 includes an extraction result management unit 115 that displays extraction results (see extraction result window 361) including at least one item among the extracted partial image, the identification number of the partial image, the position of the partial image within the image (the extraction target area of ​​the drawing), and the similarity between the partial image and the template image. The extraction result management unit 115 adds the partial image of the specified image to the extraction result. The extraction result management unit 115 displays the identification number of the partial image. The extraction result management unit 115 receives an instruction as to whether the partial image is to be included in the extraction result. The extraction result management unit 115 has a sorting function based on one selected from the identification number of the displayed partial image, the position of the partial image within the image, and the similarity between the partial image and the template image.

[0035] <Symbol extraction processing> 10 is a flowchart of the symbol extraction process according to the first embodiment. The symbol extraction process executed by the symbol extraction device 100 will be described with reference to FIG. In step S21, the drawing display control unit 111 acquires the drawing designated by the user and displays it on the screen (see FIG. 2).

[0036] In step S22, the drawing display control unit 111 accepts a user operation and branches to a process corresponding to the operation. When the drawing display control unit 111 accepts an operation to instruct template setting (for example, "Template Setting" in the "Settings" menu) (step S22 → Template Setting), the process proceeds to step S23. When the drawing display control unit 111 accepts an operation to instruct analysis condition setting (for example, "Analysis Condition Setting" in the "Settings" menu) (step S22 → Analysis Condition Setting), the process proceeds to step S25. When the drawing display control unit 111 accepts an operation to instruct extraction (for example, the "Extraction Execution" button in the analysis condition setting window 331 shown in FIG. 4 or "Extraction Execution" in the "File" menu) (step S22 → Extraction Execution), the process proceeds to step S27. When the drawing display control unit 111 accepts an operation to instruct display of extraction results (for example, "Edit Extraction Result" in the "Settings" menu) (step S22 → Extraction Result), the process proceeds to step S29. When the drawing display control unit 111 receives an operation to instruct saving of the extraction results (for example, "Save Extraction Results" in the "File" menu) (step S22 → Save), the process proceeds to step S30. When the drawing display control unit 111 receives an operation to instruct termination (for example, "Exit" in the "File" menu) (step S22 → End), the process ends the symbol extraction process.

[0037] In step S23, the template setting unit 112 displays the template setting window 321 (see FIG. 3) (denoted as "setting screen" in FIG. 10). In step S24, the template setting unit 112 stores the template image, which is the content set in the template setting window 321, in the template information 121.

[0038] In step S25, the analysis condition setting unit 113 displays the analysis condition setting window 331 (see FIG. 4) (denoted as "setting screen" in FIG. 10). In step S26, the analysis condition setting unit 113 stores the threshold value, angle setting, and other settings set in the analysis condition setting window 331 in the template information 121.

[0039] In step S27, the matching unit 114 scans the image of the designated area (extraction target area) of the drawing, extracts a partial image that matches the template image, and stores it in the extraction result database 140 (see FIG. 9). In step S28, the drawing display control unit 111 displays a screen (see screens 340 and 350 in FIGS. 5 and 6) showing the extraction results.

[0040] In step S29, the extraction result management unit 115 executes extraction result screen processing (see FIG. 11, which will be described later). In step S30, the extraction result management unit 115 saves the extraction result database 140 (see FIG. 9) in a specified file. For example, the extraction result management unit 115 saves the identification number, position, angle, classification, acceptance / rejection, and similarity in a CSV (Comma-Separated Values) format file, and the image in an image file.

[0041] <<Extraction result screen processing>> 11 is a flowchart of extraction result screen processing according to the first embodiment. Details of step S29 (see FIG. 10) will be described with reference to FIG. In step S41, the extraction result management unit 115 displays the extraction result window 361 (see FIG. 7).

[0042] In step S42, the extraction result management unit 115 accepts an operation on the extraction result window 361 and branches to processing corresponding to the operation. When the extraction result management unit 115 accepts an operation to select an item (a row in the extraction result list on the left side of the extraction result window 361) (step S42 → select item), the process proceeds to step S43. When the extraction result management unit 115 accepts an operation to change the acceptance / rejection (step S42 → change acceptance / rejection), the process proceeds to step S44. When the extraction result management unit 115 accepts pressing of the "add" button 363 (step S42 → add), the process proceeds to step S46. When the extraction result management unit 115 accepts pressing of the "delete" button 364 (step S42 → delete), the process proceeds to step S50. When the extraction result management unit 115 accepts pressing of the "close" button (step S42 → close), the process ends the extraction result screen processing.

[0043] In step S43, the extraction result management unit 115 displays the partial image corresponding to the specified item (row in the extraction result list) in the area 362 (see FIG. 7), and then returns to step S42. In step S44, the extraction result management unit 115 alternately displays whether the specified item is adopted or not. In step S45, the extraction result management unit 115 updates the acceptance / rejection attribute of the record corresponding to the specified item in the extraction result database 140 (see FIG. 9, indicated as "DB" in FIG. 11) so as to be alternated, and then returns to step S42.

[0044] In step S46, the extraction result management unit 115 acquires the area on the drawing designated by the user. In step S47, the extraction result management unit 115 acquires a partial image of the area designated in step S46.

[0045] In step S48, the extraction result management unit 115 displays the partial image acquired in step S47 in area 362. The extraction result management unit 115 also adds information related to the partial image to the extraction result list in the extraction result window 361 and displays it. The position, angle, category, and adoption / rejection in the list are the coordinates of the specified area (for example, the coordinates in the upper right corner), "0," "manual," and "o," respectively. The angle may be the angle at which the template image is rotated to maximize the similarity with the partial image. In step S49, the extraction result management unit 115 adds the displayed content and partial images added to the extraction result list in step S48 to the extraction result database 140 (see FIG. 9), and the process returns to step S42.

[0046] In step S50, the extraction result management unit 115 deletes the currently specified item (row in the extraction result list) and displays the list. In step S51, the extraction result management unit 115 deletes the record from the extraction result database 140 that corresponds to the item deleted in step S50, and the process returns to step S42.

[0047] <Features of the symbol extraction device> The symbol extraction device 100 calculates the similarity between the template image of the specified symbol and a partial image of the drawing, and extracts it if the similarity is equal to or greater than a threshold. The extracted partial image is highlighted on the drawing (see reference numeral 343 in FIG. 5), allowing the user to easily confirm the extraction result. In response to a user's instruction, the symbol extraction device 100 adds partial images of drawings that were not extracted to the extraction results and deletes erroneous detections from the extraction results, thereby editing the extraction results. By adjusting the threshold value to lower the threshold value if there are many extraction omissions and to raise the threshold value if there are many erroneous extractions, the user can efficiently extract symbols by editing the extraction results.

[0048] Second Embodiment The matching unit 114 in the first embodiment extracts a partial image similar to a template image using a template matching method such as zero-mean normalized cross-correlation. The extracted results may be matched using a trained learning model. FIG. 12 is a functional block diagram of a symbol extraction device 100A according to the second embodiment. Compared to the symbol extraction device 100 of the first embodiment, a learning unit 116 is added to the control unit 110, and a learning model 122 is added to the storage unit 120. In addition, the matching unit 114A of the control unit 110 is modified.

[0049] The learning model 122 is a learning model for machine learning technology. The input (explanatory variable) of the learning model 122 is a partial image of the drawing, and the output (objective variable) is whether or not it matches a template image. The learning unit 116 trains the learning model 122 using images in the records of the extraction result database 140 (see FIG. 9) as learning data (makes the learning model 122 learn the learning data).

[0050] The matching unit 114A further determines whether a partial image determined to be similar to the template image using the template matching method matches the template image using the learning model 122. In other words, the matching unit 114A determines that a partial image matches when it is determined to match (be similar to) the template image using both the template matching method and the learning model 122.

[0051] As described above, the symbol extraction device 100A further includes a learning unit 116 that generates a learning model 122 for determining whether an image (a region to be extracted in a drawing) is a template image, using learning data including a partial image included in the extraction result. The learning unit 116 also generates the learning model 122 using the extraction result. The similarity is calculated based on a template matching method. The matching unit 114A included in the symbol extraction device 100A extracts partial images whose similarity calculated based on the template matching method is equal to or greater than a threshold, and which are determined to be template images using the learning model 122.

[0052] <Features of the second embodiment> The symbol extraction device 100A performs matching using machine learning technology, and is therefore able to extract partial images of symbols with higher accuracy than the symbol extraction device 100.

[0053] <<Variations>> Although several embodiments of the present invention have been described above, these embodiments are merely illustrative and do not limit the technical scope of the present invention. For example, the extraction result list on the left side of the extraction result window 361 (see FIG. 7) is displayed in order of identification number, but may be sorted and displayed by similarity. By assuming that there are no erroneous detections in extraction results with high similarity, and checking whether there are any erroneous extractions in extraction results with low similarity, the user can efficiently check the extraction results. Furthermore, the template information 121 and the learning model 122 may be shared by multiple drawings and symbol extraction devices 100 and 100A.

[0054] The present invention can take on various other embodiments, and various modifications such as omissions and substitutions can be made without departing from the spirit of the present invention. These embodiments and modifications are included in the scope and spirit of the invention described in this specification, etc., and are also included in the invention described in the claims and their equivalents. [Explanation of symbols]

[0055] 100,100A Symbol Extractor 111 Drawing display control unit (image display control unit) 112 Template Settings 113 Analysis condition setting section 114 Matching Unit 115 Extraction result management department 116 Learning Department 121 Template Information 122 Learning Model 128 programs 130 Design drawing data (image) 140 Extraction Results Database

Claims

1. a template setting unit that sets a template image to be extracted from the image; an analysis condition setting unit that acquires and sets a threshold value used to determine whether extraction is appropriate; a matching unit that extracts, from the image, a partial image whose similarity to the template image is equal to or greater than the threshold; an extraction result management unit that displays an extraction result including at least one item of an extracted partial image, an identification number of the partial image, a position of the partial image within the image, and a similarity between the partial image and the template image; The extraction result management unit The system has a sorting function based on one selected from the identification number of the displayed partial image, the position of the partial image within the image, and the similarity between the partial image and the template image. Symbol extraction device.

2. a template setting unit that sets a template image to be extracted from the image; an analysis condition setting unit that acquires and sets a threshold value used to determine whether extraction is appropriate; a matching unit that extracts, from the image, a partial image whose similarity to the template image is equal to or greater than the threshold; an extraction result management unit that displays extraction results including at least one item of an extracted partial image, an identification number of the partial image, a position of the partial image within the image, and a similarity between the partial image and the template image; a learning unit that generates a learning model that determines whether an image is the template image using learning data that includes a partial image included in the extraction result, the similarity is calculated based on a template matching technique; The matching unit extracts a partial image whose similarity calculated based on a template matching technique is equal to or greater than the threshold value and which is determined to be the template image using the learning model. Symbol extraction device.

3. The extraction result management unit Add the specified partial image of the image to the extraction result. The symbol extraction device according to claim 1 or 2.

4. The analysis condition setting unit obtaining a rotation angle of the template image; The collation unit A partial image having a similarity to an image obtained by rotating the template image by the rotation angle one or more times is extracted from the image, the similarity being equal to or greater than the threshold value. The symbol extraction device according to claim 1 or 2.

5. further comprising an image display control unit; The extraction result management unit Displaying the identification number of the partial image; The image display control unit A partial image of the image corresponding to the extracted partial image is highlighted and displayed, and an identification number corresponding to the partial image is displayed near the partial image. The symbol extraction device according to claim 1 or 2.

6. The similarity is calculated based on a template matching method. The symbol extraction device according to claim 1 or 2.

7. The extraction result management unit Accept an instruction as to whether the partial image should be included in the extraction result; The learning unit generates the learning model using the extraction result. The symbol extraction device according to claim 2 .

8. Computer, a template setting unit that sets a template image to be extracted from the image; an analysis condition setting unit that acquires and sets a threshold value used to determine whether extraction is appropriate; a matching unit that extracts, from the image, a partial image whose similarity to the template image is equal to or greater than the threshold; an extraction result management unit that displays an extraction result including at least one item of an extracted partial image, an identification number of the partial image, a position of the partial image within the image, and a similarity between the partial image and the template image; The extraction result management unit The system has a sorting function based on one selected from the identification number of the displayed partial image, the position of the partial image within the image, and the similarity between the partial image and the template image. A program that functions as a symbol extraction device.

9. Computer, a template setting unit that sets a template image to be extracted from the image; an analysis condition setting unit that acquires and sets a threshold value used to determine whether extraction is appropriate; a matching unit that extracts, from the image, a partial image whose similarity to the template image is equal to or greater than the threshold; an extraction result management unit that displays extraction results including at least one item of an extracted partial image, an identification number of the partial image, a position of the partial image within the image, and a similarity between the partial image and the template image; a learning unit that generates a learning model that determines whether an image is the template image using learning data that includes a partial image included in the extraction result, the similarity is calculated based on a template matching technique; The matching unit extracts a partial image whose similarity calculated based on a template matching technique is equal to or greater than the threshold value and which is determined to be the template image using the learning model. A program that functions as a symbol extraction device.

10. The symbol extraction device A step of setting a template image to be extracted from the image; A step of acquiring and setting a threshold value used to determine whether extraction is appropriate; extracting a partial image from the image, the partial image having a similarity to the template image equal to or greater than the threshold; displaying an extraction result including at least one item of an extracted partial image, an identification number of the partial image, a position of the partial image within the image, and a similarity between the partial image and the template image; and sorting the extraction results by one selected item from the identification number of the displayed partial image, the position of the partial image within the image, and the similarity between the partial image and the template image. Symbol extraction methods.

11. The symbol extraction device A step of setting a template image to be extracted from the image; A step of acquiring and setting a threshold value used to determine whether extraction is appropriate; extracting a partial image from the image, the partial image having a similarity to the template image equal to or greater than the threshold; displaying an extraction result including at least one item of an extracted partial image, an identification number of the partial image, a position of the partial image within the image, and a similarity between the partial image and the template image; and generating a learning model for determining whether an image is the template image using learning data including the partial image included in the extraction result; the similarity is calculated based on a template matching technique; In the step of extracting the partial image, a partial image whose similarity calculated based on a template matching technique is equal to or greater than the threshold value and which is determined to be the template image using the learning model is extracted. Symbol extraction methods.

Images