Image processing device, image processing method, and program

Abstract

To provide an information processing apparatus, an information processing method, and a program which can determine an attribute of deformation by using actual size information, while keeping determination accuracy.SOLUTION: An image processing apparatus has: an acquiring unit for acquiring first actual size information of an image including deformation; a selection unit for selecting a leaned model and a parameter used in determination of an attribute of the deformation based on the image and actual size information acquired by the acquiring unit; and a determination unit for determining an attribute of the deformation included in the image by using the first actual size information and a model generated by advanced learning.SELECTED DRAWING: Figure 2

Description

Technical Field

[0001] The present invention relates to an image processing technology for detecting deformations from an image obtained by photographing an inspection target.

Background Art

[0002] For an image obtained by photographing an inspection target such as the wall surface of a concrete structure, there is a method in which a computer device performs machine learning to detect deformations such as cracks and determine the attributes of deformations such as the width of cracks. Further, when determining the attributes of deformations by machine learning, a learned model is created using learning data as an input. However, if the content of the learning data is different, models with different characteristics are created. To improve the determination accuracy, it is desirable to create a plurality of models with different characteristics and select an appropriate model according to the determination target.

[0003] Patent Document 1 describes a method of selecting a learned model used for image analysis, in which a model learned from an image with imaging conditions similar to the imaging conditions consisting of the imaging position and imaging angle of the image is selected.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] As described above, when there are a plurality of models with different learning data contents, it is desirable to select a model learned from learning data similar to the data of the determination target as much as possible. In particular, when determining the deformation width or the like, since the determination result is the actual size (for example, mm), it is desirable to select a model learned from learning data in which the actual size per pixel of the image is similar.

[0006] However, the selection criteria for the model in Patent Document 1 are based solely on the similarity of the image's shooting position and angle, and do not consider the actual size of the image. Furthermore, it is difficult to make accurate judgments when the similarity with the training data is low, and various models with different training data content are needed to judge various images.

[0007] This invention has been made in view of the above problems, and its purpose is to realize a technology that can determine the attributes of deformation using the actual size information of an image while maintaining judgment accuracy. [Means for solving the problem]

[0008] To solve the above problems and achieve the objective, the image processing apparatus of the present invention includes an acquisition means for acquiring first actual size information of an image including deformation, and a database that stores a model generated by prior learning and a second actual size information used to learn the model in association with each other, and then retrieves the first actual size information and the second actual size information from the database. The difference Based on this, for determining the attributes of the deformation contained in the image multiple A selection means for selecting a model, the first actual size information, and the multiple Model and Based on this, in each of the multiple models A determination means for determining the attributes of deformations contained in the aforementioned image, An integration means for integrating the attributes of the deformations determined using the aforementioned multiple models, of have ru. [Effects of the Invention]

[0009] According to the present invention, it becomes possible to determine the attributes of deformation using the actual size information of an image while maintaining judgment accuracy. [Brief explanation of the drawing]

[0010] [Figure 1] Hardware configuration diagram of the image processing device of Embodiment 1. [Figure 2] Functional block diagram of the image processing apparatus of Embodiment 1. [Figure 3] A flowchart showing the process for determining the attributes of deformation in the image processing apparatus of Embodiment 1. [Figure 4]A diagram illustrating the UI screen of the image processing device of Embodiment 1. [Figure 5] A diagram illustrating the database generated by the image processing device of Embodiment 1. [Figure 6] A diagram illustrating the database generated by the image processing device of Embodiment 1. [Figure 7] Functional block diagram of the image processing apparatus of Embodiment 2. [Figure 8] A flowchart illustrating the processing of the image processing apparatus in Embodiment 2. [Figure 9] A diagram illustrating the database generated by the image processing device of Embodiment 2. [Figure 10] A diagram illustrating the database generated by the image processing device of Embodiment 2. [Figure 11] A diagram illustrating the UI screen of the image processing device of Embodiment 2. [Modes for carrying out the invention]

[0011] The embodiments will be described in detail below with reference to the attached drawings. Note that the following embodiments do not limit the invention as defined in the claims. While the embodiments describe multiple features, not all of these features are essential to the invention, and the features may be combined in any way. Furthermore, in the attached drawings, identical or similar configurations are given the same reference numerals, and redundant descriptions are omitted.

[0012] [Embodiment 1] Embodiment 1 describes an example in which a computer device operates as an image processing device and performs machine learning to determine the attributes of an object under inspection based on an image (judgment image) containing the deformation (determination image) and actual size information, as well as a single judgment model.

[0013] Furthermore, "deformation" refers to damage, deterioration, or cracks that occur on the concrete surface of concrete structures such as expressways, bridges, tunnels, and dams due to other factors. "Crack" refers to linear damage with a start point, end point, length, and width that occurs on the walls of structures due to aging, earthquake impact, etc.

[0014] <Hardware Configuration> First, referring to FIG. 1, the hardware configuration of the image processing apparatus according to Embodiment 1 will be described.

[0015] FIG. 1 is a block diagram showing the hardware configuration of the image processing apparatus 100 according to Embodiment 1.

[0016] In Embodiments 1 and 2 described below, the computer device operates as the image processing apparatus 100. Note that the processing of the image processing apparatus of this embodiment may be realized by a single computer device, or each function may be distributed and realized by a plurality of computer devices as necessary. The plurality of computer devices are connected to be communicable with each other.

[0017] The image processing apparatus 100 includes a control unit 101, a non-volatile memory 102, a work memory 103, a storage device 104, an input device 105, an output device 106, a network interface 107, and a system bus 108.

[0018] The control unit 101 includes arithmetic processing processors such as a CPU and an MPU that overall control the entire image processing apparatus 100. The non-volatile memory 102 is a ROM that stores programs and parameters executed by the processor of the control unit 101. Here, the program is a program for executing the processing of Embodiments 1 and 2 described later. The work memory 103 is a RAM that temporarily stores programs and data supplied from external devices and the like. The storage device 104 is an internal device such as a hard disk or a memory card built in the image processing apparatus 100 or an external device such as a hard disk or a memory card detachably connected to the image processing apparatus 100. The storage device 104 includes memory cards, hard disks, etc. composed of semiconductor memories, magnetic disks, etc. Further, the storage device 104 includes a storage medium composed of a disk drive that reads / writes data to / from optical disks such as DVDs and Blue-ray Discs.

[0019] The input device 105 is an operating component such as a mouse, keyboard, or touch panel that accepts user input and outputs operation instructions to the control unit 101. The output device 106 is a display device such as a display or monitor made of an LCD or organic EL, which displays data held by the image processing device 100 or data supplied from external devices. The network interface 107 is connected to a network such as the Internet or a LAN (Local Area Network) for communication. The system bus 108 includes an address bus, a data bus, and a control bus that connect the components 101 to 107 of the image processing device 100 for data exchange.

[0020] The non-volatile memory 102 stores the operating system (OS), which is the basic software executed by the control unit 101, and applications that work in cooperation with the OS to realize advanced functions. In this embodiment, the non-volatile memory 102 also stores applications that enable the image processing device 100 to perform the process of determining the attributes of deformation from the judgment image, which will be described later.

[0021] The processing of the image processing device 100 in this embodiment is achieved by loading software provided by an application. The application is assumed to have software for utilizing the basic functions of the OS installed on the image processing device 100. The OS of the image processing device 100 may also have software for implementing the processing in this embodiment.

[0022] <Functional Configuration> Next, with reference to Figure 2, the functional blocks of the image processing apparatus of Embodiment 1 will be described.

[0023] Figure 2 is a functional block diagram of the image processing apparatus 200 of Embodiment 1.

[0024] The image processing device 200 comprises an acquisition unit 201, a selection unit 202, a determination unit 203, and a correction unit 204. Each function of the image processing device 200 is composed of hardware and software. Note that each functional unit may be composed of one or more computer devices or server devices, and the system may be configured as a network-connected system.

[0025] The acquisition unit 201 acquires a determination image for determining the attributes of the deformation, and actual size information per pixel of the determination image. The determination image and actual size information are read from the storage device 104 or input via the input device 105, but are also received from external devices via the network interface 107.

[0026] The selection unit 202 selects a trained model and parameters to be used to determine the attributes of the deformation based on the judgment image and actual size information acquired by the acquisition unit 201.

[0027] The determination unit 203 performs machine learning based on the determination image and actual size information acquired by the acquisition unit 201 and the determination model selected by the selection unit 202 to determine the attributes of the deformation.

[0028] The correction unit 204 corrects the deformation attributes based on the actual size information per pixel of the judgment image acquired by the acquisition unit 201 and the deformation attributes determined by the judgment unit 203.

[0029] The actual size information per pixel is a converted value representing the actual size (e.g., mm) of one pixel in the judgment image, and is the image size ratio (mm / pixel) that represents the ratio of pixels to the actual size. In addition to the image size ratio, the actual size information is also called the image size conversion, resolution, pixel size, or image size.

[0030] Furthermore, while this embodiment describes an example of determining the width of deformation as an attribute of deformation, it is not limited to this, and the length, depth, thickness, area, volume, etc. of the deformation may also be determined.

[0031] <Process for determining the attributes of the deformation> Next, referring to Figures 3 to 6, the process by which the image processing device 100 of Embodiment 1 determines the attributes of the deformation from the judgment image will be described.

[0032] Figure 3 is a flowchart showing the process by which the image processing device 100 of Embodiment 1 determines the attributes of deformation from the judgment image.

[0033] The process shown in Figure 3 is achieved when the control unit 101 of the image processing device 100 shown in Figure 1 loads the program stored in the non-volatile memory 102 into the work memory 103, executes it to control each component, and performs the function shown in Figure 2. The same applies to Figure 8, which will be described later.

[0034] In S301, the acquisition unit 201 acquires a determination image for determining the attributes of the deformation.

[0035] In S302, the acquisition unit 201 acquires the actual size information of the judgment image acquired in S301 (hereinafter referred to as image actual size information). Multiple image actual size information may be acquired, or image actual size information may be acquired separately for the vertical and horizontal directions.

[0036] Figure 4 illustrates a method for obtaining the actual size information of the judgment image, and shows an example of the display screen of the image processing device 100.

[0037] Image size information can be obtained, for example, by the user operating the input device 105 to input the actual size value into the input field 401 on the display screen shown in Figure 4(a). Alternatively, image size information can be obtained, for example, by the user operating the input device 105 to input a numerical value corresponding to the length of the line segment 403 displayed on the judgment image 402 into the input field 404 on the display screen shown in Figure 4(b). Furthermore, image size information can be obtained by displaying the drawing 405 associated with the judgment image 402 and its actual size value, as shown in the display screen of Figure 4(c), and by the user operating the input device 105 to establish the relationship 406 between the judgment image 402 and the drawing 405. In addition, image size information can be obtained by selecting a judgment image for which the actual size value 407 is known, as shown in the display screen of Figure 4(d). Note that multiple image size information sets may be obtained. For example, in the display screen shown in Figure 4(c), a drawing 405, which is associated with the judgment image 402 and the actual dimensions in the vertical and horizontal directions, may be displayed, and the relationship 406 between the judgment image 402 and the drawing 405 may be made corresponding to obtain the actual dimensions in the vertical and horizontal directions.

[0038] In S303, the selection unit 202 selects a determination model from the determination image to be used to determine the attributes of the deformation.

[0039] The judgment model is created in advance by learning actual size information and images containing deformations close to the actual size information (deformation images). The judgment model stores the actual size information of the deformation images used for learning in association with it. Figure 5 is an example of a part of the database of judgment models selected in S303. The judgment model database 501 shown in Figure 5 is stored in an external storage device that can be accessed via the storage device 104 or the network interface 107. In Figure 5, the judgment model database 501 stores a model ID 502, a model file name 503, and model actual size information 504 for each judgment model. The model ID 502 is the identification information of the judgment model. The model file name 503 is the name of the data file of the judgment model. The model actual size information 504 is the actual size information of the deformation images used for learning, and in this embodiment, it is the image actual size ratio.

[0040] The selection unit 202 selects the judgment model that is closest to the image size information acquired in S302 from the database shown in Figure 5.

[0041] The method for selecting the decision model is as follows:

[0042] First, the distance f(n) is calculated using the following equation 1. Rt is the actual image size information acquired in S302, and Rn is the actual model size information of the judgment model n. (Formula 1) f(n) = |Rt - Rn| Next, the judgment model with the smallest distance f(n) calculated using Equation 1 is selected. If there are multiple judgment models with the smallest distance, the judgment model with the smallest model size information may be selected. Also, if multiple image size information values ​​are obtained in S302, the average of these values ​​may be used as the image size information to select the judgment model.

[0043] In S304, the determination unit 203 performs machine learning based on the determination image and image size information acquired in S301 and S302, and the determination model selected in S303, to determine the attributes of the deformation. The determination unit 203 may also use the deformation location information in addition to the image size information to determine the attributes of the deformation. Here, the deformation location may be detected by the user from the determination image and input via the input device 105, or it may be detected by an external device and input via the network interface 107.

[0044] Figure 6(a) is an example of a part of the database of deformation attributes output as a result of the determination in S304. The deformation attribute database 601 shown in Figure 6(a) is stored in an external storage device accessible via the storage device 104 or the network interface 107. In Figure 6(a), the deformation attribute database 601 stores a deformation ID 602, a start point coordinate 603, an end point coordinate 604, and a deformation attribute 605 for each determined deformation. The deformation ID 602 ​​is the identification information of the determined deformation. The start point coordinate 603 is the start point coordinate of the determined deformation position. The end point coordinate 604 is the end point coordinate of the determined deformation position. The deformation attribute 605 is the attribute of the determined deformation, and in this embodiment, it is the deformation width.

[0045] In S305, the correction unit 204 corrects the attributes of the deformation determined in S304.

[0046] The method for correcting the attributes of the deformation is as follows:

[0047] First, the correction coefficient α is calculated using the following equation 2. Rt is the actual image size information acquired in S302, and Rn is the actual model size information of the judgment model n. (Formula 2) α = Rt / Rn Next, the attribute Cn of the deformation determined in S304 is corrected using the correction coefficient α calculated using Equation 2.

[0048] Then, the attribute C'n of the corrected deformation is calculated using the following equation 3. (Formula 3) C'n = α·Cn The above corrections are determined in S304 and performed for each attribute of the deformation registered in the database 601 shown in Figure 6(a). The attributes of the deformation after correction are registered in the database 606 of the attributes of the corrected deformation shown in Figure 6(b). Figure 6(b) is an example of a part of the database of attributes of the deformation corrected in S305. The database 606 of the attributes of the corrected deformation shown in Figure 6(b) is stored in an external storage device that can be accessed via the storage device 104 or the network interface 107. In Figure 6(b), the database 606 of the attributes of the corrected deformation stores the deformation ID 602, start coordinates 603, end coordinates 604, deformation attribute 605, and attributes of the corrected deformation 607 for each corrected deformation. The deformation ID 602 ​​is the identification information of the corrected deformation. The start coordinates 603 are the start coordinates of the corrected deformation position. The end coordinates 604 are the end coordinates of the corrected deformation position. The deformation attribute 605 is the attribute of the corrected deformation. The attribute 607 of the corrected deformation is the attribute of the corrected deformation, and in this embodiment, it is the deformation width.

[0049] If multiple image size data points are obtained in S302, the average value of these data points may be used to correct the deformation attributes.

[0050] As described above, according to Embodiment 1, the attributes of deformation can be determined using the actual size information of the determination image while maintaining judgment accuracy. This improves the reliability of the process for determining the attributes of deformation and makes it possible to streamline inspection work.

[0051] Note that steps S303 or S305 may be omitted. For example, the selection of the decision model in S303 may be omitted, and a predetermined representative decision model may be used. Also, the correction of the attribute of the deformation in S305 may be omitted.

[0052] Furthermore, in determining the deformation attributes in S304, the determination image may be scaled so that the actual size information of the determination image matches the actual size information of the determination model before determining the deformation attributes.

[0053] Furthermore, if multiple image size information is obtained, or if the image size information includes both vertical and horizontal image size information, the processing from S303 to S305 should be performed for each piece of image size information.

[0054] [Embodiment 2] Next, Embodiment 2 will be described with reference to Figures 7 to 11.

[0055] Embodiment 1 was a process that performed machine learning based on a judgment image, image size information, and a single judgment model to determine the attributes of the deformation. In contrast, Embodiment 2 describes an example in which the attributes of the deformation are determined using multiple judgment models and the attributes of multiple deformations obtained using multiple judgment models are integrated.

[0056] <Device configuration> Figure 7 is a functional block diagram of the image processing apparatus of Embodiment 2. The image processing apparatus 700 of Embodiment 2 has an integration unit 701 added to the image processing apparatus 200 of Figure 2 described in Embodiment 1.

[0057] The integration unit 701 integrates the attributes of multiple deformations determined by the correction unit 204 using multiple determination models.

[0058] The hardware configuration of the image processing device 700 in Embodiment 2 is the same as that shown in Figure 1 of Embodiment 1.

[0059] <Process for determining the attributes of the deformation> Next, referring to Figures 8 to 11, the process by which the image processing apparatus of Embodiment 2 determines the attributes of deformation from the judgment image will be described.

[0060] Figure 8 is a flowchart showing the process by which the image processing device 700 of Embodiment 2 determines the attributes of deformation from the judgment image.

[0061] The processes S801 and S802 in Figure 8 are the same as the processes S301 and S302 in Figure 3 described in Embodiment 1. As described in Embodiment 1, multiple image size information can be acquired, or image size information can be acquired separately for the vertical and horizontal directions.

[0062] In S803, the selection unit 202 selects multiple determination models from the determination image to be used to determine the attributes of the deformation.

[0063] The selection unit 202 calculates the distance f(n) using Equation 1 described in Embodiment 1, and selects N (for example, N=2) judgment models from the database shown in Figure 5 in order of proximity to the calculated distance f(n). If multiple image size information is obtained in S802, the distance f(n) is calculated for each image size information, and the judgment model with the smallest distance f(n) is selected.

[0064] The selected judgment model (hereinafter referred to as the selected model) is stored in association with the image size information of the judgment image acquired in S802. Figure 9 is a diagram illustrating a part of the database of the selected model selected in S803. The database 901 of the selected model shown in Figure 9 is stored in an external storage device that can be accessed via the storage device 104 or the network interface 107. In Figure 9, the database 901 of the selected model stores a model ID 902, model size information 903, and image size information 904 for each selected model. The model ID 902 is the identification information of the selected model. The model size information 903 is the size information of the deformed image used for training, and the image size information 904 is the image size information of the judgment image acquired in S802, which in this embodiment is the image size ratio.

[0065] In S804, the determination unit 203 determines the attributes of the deformation using each selected model selected in S803. The method for determining the attributes of the deformation is the same as in S304 in Figure 3. Figure 10(a) is an example of a part of the database of deformation attributes output as a result of the determination in S804. The database of deformation attributes 1001 shown in Figure 10(a) is stored in an external storage device that can be accessed via the storage device 104 or the network interface 107. In Figure 10(a), the database of deformation attributes 1001 stores the deformation ID 1002, model ID 1003, starting point coordinates 1004, ending point coordinates 1005, and deformation attributes 1006 for each determined deformation. The deformation ID 1002 is the identification information of the determined deformation. The model ID 1003 is the identification information of the selected model used to determine the attributes of the deformation. The starting point coordinates 1004 are the starting point coordinates of the determined deformation location. The endpoint coordinate 1005 is the endpoint coordinate of the determined deformation position. The deformation attribute 1006 is the attribute of the determined deformation, and in this embodiment, it is the deformation width. In Figure 10(a), deformation 1007 and deformation 1008 have the same deformation ID 1002 and different model ID 1003 because the deformation attributes were determined using different determination models for deformations at the same position.

[0066] In S805, the correction unit 204 corrects the attributes of the deformation determined in S804. The correction unit 204 obtains the model size information and image size information of the selected model used to determine the attributes of the deformation by referring to the database 901 of selected models shown in Figure 9 for each model ID 1003 of the deformation ID in the database 1001 shown in Figure 10(a). The correction unit 204 also calculates a correction coefficient α using equation 2 described in Embodiment 1 and corrects the deformation attribute Cn using equation 3 described in Embodiment 1.

[0067] The above corrections are performed for each attribute of the deformation determined in S804 and registered in the database 1001 shown in Figure 10(a). The attributes of the deformation after correction are registered in the database of corrected deformation attributes 1009 shown in Figure 10(b). Figure 10(b) is an example diagram showing a part of the database of corrected deformation attributes in S805. The database of corrected deformation attributes 1009 shown in Figure 10(b) is stored in an external storage device that can be accessed via the storage device 104 or the network interface 107. In Figure 10(b), the database of corrected deformation attributes 1009 stores the deformation ID 1002, model ID 1003, start coordinates 1004, end coordinates 1005, deformation attribute 1006, and corrected deformation attribute 1010 for each corrected deformation. The deformation ID 1002 is the identification information of the corrected deformation. The model ID 1003 is the identification information of the selected model used to determine the deformation attribute. The starting coordinate 1004 is the starting coordinate of the corrected deformation position. The ending coordinate 1005 is the ending coordinate of the corrected deformation position. The deformation attribute 1006 is the attribute of the corrected deformation. The attribute of the corrected deformation 1010 is the attribute of the corrected deformation, and in this embodiment, it is the deformation width.

[0068] In S806, the integration unit 701 integrates the attributes of multiple corrected deformations determined using multiple selection models corrected in S805. For example, the integration unit 701 integrates the attributes of multiple corrected deformations determined using different selection models into the attributes of a corrected deformation specified by the user. Figure 11 is a diagram illustrating a method for integrating the attributes of multiple corrected deformations determined using multiple selection models. The integration screen 1101 shown in Figure 11(a) displays the determination image acquired in S801, each deformation registered in the database 1009 in Figure 10(b), and the attributes of the corrected deformations determined using different selection models for each deformation. Figure 11(a) shows the integration screen 1101 displaying, for example, a determination image 1102 containing the attributes of multiple corrected deformations determined using a selection model with model size information of 0.8 mm / pixel registered in the database 901 in Figure 9, and a determination image 1103 containing the attributes of multiple corrected deformations determined using a selection model with model size information of 1.0 mm / pixel. For example, in the judgment image 1102 of the integrated screen 1101 in Figure 11(a), when the user operates the input device 105 to select deformation 1104 and then operates the confirmation button 1105, the attributes of multiple deformations in the judgment image 1102 are integrated into the attributes of deformation 1104. By repeating the above process, a database is generated in which the attributes of each deformation are integrated for each deformation ID, as shown in Figure 10(b). The attributes before integration may also be appended to the integrated database.

[0069] Furthermore, if the image size information acquired in S802 includes both vertical and horizontal image size information, the database 901 in Figure 9 displays, as shown in the integrated screen 1106 in Figure 11(b), a judgment image 1107 containing the attributes of multiple corrected deformations determined using a selected model with 0.8 mm / pixel model size information selected from the vertical image size information, and a judgment image 1108 containing the attributes of multiple corrected deformations determined using a selected model with 1.0 mm / pixel model size information selected from the horizontal image size information. Then, for example, as shown in the integrated screen 1106 in Figure 11(b), if the user selects a vertical or horizontal deformation in judgment images 1107 and / or 1108 by operating the input device 105 and then operates the confirmation button 1105, the attributes of the multiple deformations in judgment images 1107 and / or 1108 are integrated. In this case, the direction of the deformation can be determined from the starting coordinates 1004 and ending coordinates 1005 registered in the database 1009 in Figure 10(b), and deformations whose direction matches the image size information may be displayed in the pre-selected state 1109. In this way, deformations in the selected state 1109 can be integrated without the user having to perform an operation to select deformations.

[0070] As described above, Embodiment 2 improves the accuracy of determining the attributes of the deformation and the range of the actual image size information of the determined image. This improves the reliability of the process for determining the attributes of the deformation and makes it possible to streamline inspection work.

[0071] Note that steps S803 or S805 may be omitted. For example, the selection of a selection model in S803 may be omitted, and all decision models may be used. Also, the correction of the attribute of the deformation in S805 may be omitted.

[0072] Furthermore, if multiple image size information is obtained, or if the image size information includes both vertical and horizontal image size information, S803 to S 8 You should perform the process described in 06.

[0073] [Other embodiments] The present invention can also be realized by supplying a program that implements one or more functions of each embodiment to a system or device via a network or storage medium, and by having one or more processors in the computer of that system or device read and execute the program. Furthermore, the present invention can also be realized by a circuit (e.g., an ASIC) that implements one or more functions.

[0074] The invention is not limited to the embodiments described above, and various modifications and variations are possible without departing from the spirit and scope of the invention. Accordingly, claims are attached to disclose the scope of the invention. [Explanation of symbols]

[0075] 100, 200, 700...Image processing device, 101...Control unit, 102...Non-volatile memory, 104...Storage device, 201...Acquisition unit, 202...Selection unit, 203...Determination unit, 204...Correction unit, 701...Integration unit

Claims

1. An acquisition means for acquiring first actual size information of an image including deformation, A selection means for selecting multiple models for determining the attributes of deformation contained in the image based on the difference between the first actual size information and the second actual size information, from a database that stores a model generated by pre-training and associates the model with a second actual size information used to train the model. A determination means for determining the attributes of deformations included in the image in each of the plurality of models based on the first actual size information and the plurality of models, An image processing apparatus characterized by having an integration means for integrating the attributes of the deformation determined using the aforementioned plurality of models.

2. The image processing apparatus according to claim 1, further comprising correction means for correcting the attributes of the deformation based on the first actual size information and the second actual size information.

3. The image processing apparatus according to claim 2, characterized in that the correction means corrects the attribute of the deformation based on the ratio of the first actual size information and the second actual size information.

4. The image processing apparatus according to claim 2 or 3, further comprising storage means for storing a database relating the deformation, the location of the deformation, the attributes of the deformation, and the attributes of the corrected deformation.

5. The image processing apparatus according to claim 1, further comprising correction means for correcting the attributes of deformation determined using the plurality of models based on the first actual size information and the second actual size information.

6. The image processing apparatus according to claim 5, further comprising storage means for storing a database relating the deformation, the location of the deformation, the attributes of the deformation, the attributes of the corrected deformation, and the model used to determine the attributes of the deformation.

7. The image processing apparatus according to any one of claims 1 to 6, characterized in that the first actual size information includes actual size information in the vertical and horizontal directions of the image.

8. The image processing apparatus according to claim 7, wherein the selection means selects the model based on the actual size information in the vertical and horizontal directions of the image.

9. The image processing apparatus according to claim 7, wherein the selection means selects the model based on the average of the actual size information in the vertical and horizontal directions of the image.

10. The first actual size information includes the actual size information in the vertical and horizontal directions of the image, The image processing apparatus according to claim 2 or 3, characterized in that the correction means corrects the deformation attribute based on the ratio of the actual size information in the vertical and horizontal directions of the image to the second actual size information.

11. The first actual size information includes the actual size information in the vertical and horizontal directions of the image, The image processing apparatus according to claim 2 or 3, characterized in that the correction means corrects the deformation attribute based on the ratio of the average of the vertical and horizontal actual size information of the image to the second actual size information.

12. The first actual size information includes the actual size information in the vertical and horizontal directions of the image, The image processing apparatus according to claim 5, characterized in that the integration means integrates the attributes of the deformation based on the actual size information in the vertical and horizontal directions of the image and the direction of the deformation.

13. The image processing apparatus according to any one of claims 1 to 12, characterized in that the determination means determines the attribute of the deformation by scaling the image using the first actual size information.

14. The image processing apparatus according to any one of claims 1 to 13, characterized in that the deformation is a crack.

15. The image processing apparatus according to claim 14, characterized in that the attribute of the deformation is the width of the crack.

16. An image processing method, The acquisition means includes the step of acquiring first actual size information of an image including deformation, The selection means includes the step of selecting a plurality of models for determining the attributes of deformation contained in the image based on the difference between the first actual size information and the second actual size information, from a database that stores a model generated by prior training and associates the second actual size information used to train the model. The determination means includes the step of determining the attribute of deformation included in the image in each of the plurality of models based on the first actual size information and the plurality of models, An image processing method characterized by comprising the step of integrating the attributes of the deformation determined using the plurality of models.

17. A program for causing a computer to function as one of the means of an image processing apparatus described in any one of claims 1 to 15.

Images