JP2023152778A - Dimension measuring device, dimension measurement method, dimension measurement program, and pass/fail determining device of welded place - Google Patents

Abstract

To provide a dimension measuring device capable of automatically and easily measuring dimensions of a welding bead when a body material and the welding bead are in an optional shape.SOLUTION: A dimension measuring device 20 includes a data acquisition part 1, a bead region extraction part 2, a contour extraction part 3, an end point acquisition part 5, a reference line imparting part 6, and a dimension calculation part 7. The data acquisition part 1 acquires shape data of a body material 100 in which a welding bead 110 is formed. The bead region extraction part 2 and the contour extraction part 3 extract a formation region and contours of the welding bead 110, respectively, and the end point acquisition part 5 acquires an end point of the welding bead 110. The reference line imparting part 6 imparts reference lines to contour data such that a first reference line crosses middle point positions of a plurality of second reference lines within a predetermined angle range with the end point as a starting point. The dimension calculation part 7 calculates a length of the welding bead 110 on the basis of the first reference line, and calculates a width of the welding bead 110 on the basis of the plurality of second reference lines in a plurality of places along the first reference line.SELECTED DRAWING: Figure 1

Description

The present disclosure relates to a weld bead dimension measuring device, a dimension measuring method, a dimension measuring program, and a welding location quality determination device.

It is important to conduct a visual inspection of a welding location (for example, a weld bead and a predetermined range of base metal including the weld bead) in order to ensure weld quality. In addition to the shape defect inspection, which checks for the presence or absence of shape defects such as holes and pits, and the type of shape defects, the appearance inspection also includes a dimensional inspection to check whether the dimensions of the weld bead are within the desired range. be.

Regarding the visual inspection of welded parts, there is a desire to perform the inspection at a certain level without depending on the skill of the worker conducting the inspection. Another problem in recent years is that there is a shortage of skilled workers.

For these reasons, automation of visual inspection has been progressing in recent years. Among these, various techniques have been proposed for automating dimensional inspection, specifically automating dimensional measurement of weld beads.

For example, in Patent Document 1, by fitting an approximate curve or a straight line to the cross-sectional shapes of the base metal and weld bead obtained by a laser displacement meter, and performing numerical analysis such as differentiation, the area where the weld bead exists in the base metal is A method is disclosed for specifying the dimensions of a weld bead.

Patent Document 2 discloses a method of creating a comparative shape using information on the base material before welding and identifying the joint shape by comparing with this.

In the conventional methods disclosed in Patent Documents 1 and 2, a developer of an inspection method determines inspection rules and inspection conditions in advance, and prepares and sets parameters such as threshold values. A method of performing an appearance inspection of a base material on which a weld bead is formed based on these prepared rules and parameters, and identifying the weld bead formation area, dimensions, etc., is sometimes called a rule-based approach.

Japanese Patent Application Publication No. 2019-124560 Japanese Patent Application Publication No. 2010-256326

However, with the rule-based approach described above, if you want to handle data with rules that have not been set in advance, you need to consider and finalize inspection rules and parameters again, which has the disadvantage of very low scalability. be. For example, in the past, only straight weld beads were inspected, but if it became necessary to inspect curved weld beads, inspection rules and parameters would need to be reconsidered and reset. Further, conventionally, when inspecting weld beads, only a flat plate was inspected as a base material, but similar problems occur when inspecting a weld bead formed on a base material having a curved surface. Furthermore, as the number of inspection rules increases, the number of inspection parameters often also increases. In that case, there is also the problem that setting of inspection rules and inspection parameters is complicated and time-consuming.

The present disclosure has been made in view of the above, and an object of the present disclosure is to provide a dimension measuring device and dimension that can automatically and easily measure the dimensions of a weld bead even when the base material and the weld bead have arbitrary shapes. The object of the present invention is to provide a measuring method, a program for measuring dimensions, and a device for determining the quality of welded parts.

A dimension measuring device according to the present disclosure is a dimension measuring device that measures the dimensions of a weld bead formed on a base material, and includes a data acquisition unit that acquires shape data of the base material on which the weld bead is formed; a bead area extraction unit that extracts the formation area of the weld bead as bead area data from the shape data; a contour extraction unit that extracts the outline of the weld bead as outline data based on the bead area data; and the outline data. a smoothing processing unit that performs a smoothing process on the contour data; an end point acquisition unit that obtains an end point of the weld bead based on the smoothed contour data; and a length measurement unit that measures the length of the weld bead using the end point as a starting point. a reference line providing unit that provides the contour data with a first reference line for measuring the width of the weld bead and a plurality of second reference lines for measuring the width of the weld bead; at least a dimension calculation unit that calculates the length of the bead and calculates the width of the weld bead at a plurality of locations along the first reference line based on the plurality of second reference lines, The line providing unit connects the first reference line and the plurality of second reference lines so that the first reference line and the plurality of second reference lines intersect each other within a predetermined angular range. is added to the contour data.

A dimension measuring method according to the present disclosure is a dimension measuring method of a weld bead formed on a base material, and includes a first step of acquiring shape data of the base material on which the weld bead is formed, and a first step of acquiring shape data of the base material on which the weld bead is formed; a second step of extracting the weld bead formation area as bead area data; a third step of extracting the outline of the weld bead as outline data based on the bead area data; and a smoothing process for the outline data. a fifth step of obtaining an end point of the weld bead based on the smoothed contour data; and a first step of measuring the length of the weld bead starting from the end point. a sixth step of adding a reference line and a plurality of second reference lines for measuring the width of the weld bead to the contour data; and calculating the length of the weld bead based on the first reference line. and a seventh step of calculating the width of the weld bead at a plurality of locations along the first reference line based on the plurality of second reference lines, and in the sixth step, the width of the weld bead is calculated at a plurality of locations along the first reference line. The first reference line and the plurality of second reference lines are added to the contour data so that the first reference line and the plurality of second reference lines intersect each other within a predetermined angular range. It is characterized by

A dimension measurement program according to the present disclosure is a dimension measurement program for causing one or more processors to execute the dimension measurement method.

A welding location quality determination device according to the present disclosure includes at least the dimension measurement device and a shape defect determination device that determines whether or not there is a defect in the welding location including the weld bead, and includes a measurement result of the dimension measurement device. , it is characterized in that it is determined whether the appearance of the welded area satisfies a predetermined standard based on the determination result of the shape defect determination device.

According to the present disclosure, even when the base material and the weld bead have arbitrary shapes, the dimensions of the weld bead can be automatically and easily measured.

1 is a functional block diagram of a dimension measuring device according to Embodiment 1. FIG. FIG. 2 is a schematic diagram of the hardware configuration of the dimension measuring device. FIG. 3 is a schematic diagram showing how shape data of a weld bead is acquired by the appearance inspection device. It is a flowchart which shows the dimension measurement procedure of a weld bead. FIG. 6 is a schematic diagram showing the outline of a weld bead during smoothing processing and end point acquisition. 3 is a flowchart showing a procedure for extracting a weld bead formation area. FIG. 3 is a diagram showing an image of shape data and a weld bead formation region and other regions in the shape data. It is a flowchart which shows the annotation addition procedure for learning data generation. This is an example showing how annotations are added to shape data. This is an example showing an image of shape data after being annotated. It is a flowchart which shows the provision procedure of a 1st reference line and a 2nd reference line. FIG. 6 is a schematic diagram showing a state of first loop processing when providing a first reference line and a second reference line. FIG. 11A is a schematic diagram showing a process subsequent to the process shown in FIG. 11A. FIG. 11B is a schematic diagram showing a process subsequent to the process shown in FIG. 11B. FIG. 7 is a schematic diagram showing a second loop process when providing a first reference line and a second reference line. 7 is a flowchart showing a procedure for correcting the first reference line. This is an example of contour data provided with a first reference line when no correction is performed. This is an example of another contour data provided with a first reference line when no correction is performed. This is an example of contour data when the first reference line is transformed into a temporary reference line. This is another example of contour data when the first reference line is transformed into a temporary reference line. This is an example of contour data to which a first reference line has been added after correction has been completed. This is an example of another contour data to which the first reference line has been added after the correction has been completed. FIG. 3 is a functional block diagram of a welding location quality determination device according to a second embodiment.

Embodiments of the present disclosure will be described below based on the drawings. Note that the following description of preferred embodiments is essentially just an example, and is not intended to limit the present disclosure, its applications, or its uses.

(Embodiment 1)
[1: Configuration of weld bead dimension measuring device]
FIG. 1 shows a functional block diagram of a dimension measuring device according to this embodiment, and FIG. 2 shows a schematic diagram of the hardware configuration of the dimension measuring device.

As shown in FIG. 1, the dimension measuring device 20 includes at least a data acquisition section 1, a bead area extraction section 2, a contour extraction section 3, and a smoothing processing section 4. The dimension measurement device 20 also includes an end point acquisition section 5, a reference line provision section 6, and a dimension calculation section 7. The dimension measuring device 20 also includes a storage section 8 and a display section 9.

The data acquisition unit 1 acquires shape data (hereinafter simply referred to as shape data) of the base material 100 (see FIG. 3) on which the weld bead 110 (see FIG. 3) is formed from the appearance inspection device 30 (see FIG. 3). . Although FIG. 1 shows an example in which shape data is directly input from the visual inspection device 30 to the data acquisition unit 1, the present invention is not particularly limited to this. For example, the shape data may be stored in a storage section (not shown) provided outside the dimension measuring device 20, and the data acquisition section 1 may acquire the shape data from the storage section.

The bead region extraction unit 2 extracts the formation region of the weld bead 110 from the shape data as bead region data. In other words, the bead region extraction unit 2 extracts the formation region of the weld bead 110 from the shape data as bead region data in a two-dimensional data format that does not include height information. As a result, bead area data for extracting the weld bead length and weld bead width can be extracted in a simple two-dimensional data format. Machine learning (deep learning) is used to extract bead area data, so welding methods (arc welding, laser welding, etc.), welding joints ( There is no need for various settings (delicate parameter settings for various reference values) depending on the weld bead shape (straight line, curve, etc.), etc. Therefore, the weld bead region can be easily extracted, and even a new user can have a very high affinity for operation. Note that a method for extracting bead area data will be described later.

The contour extraction unit 3 extracts the contour of the weld bead 110 as contour data based on the bead area data. A method for extracting contour data will be described later.

The smoothing processing unit 4 performs smoothing processing on the contour data.

The end point obtaining unit 5 obtains end points (hereinafter sometimes simply referred to as end points) of the weld bead 110 based on the smoothed contour data. Note that, as described later, a plurality of end points may be obtained on the outline of the weld bead 110.

The reference line providing unit 6 creates a first reference line for measuring the length of the weld bead 110 and a plurality of second reference lines for measuring the width of the weld bead 110, starting from the end point, based on contour data. granted to. It is assumed that the first reference line and the midpoint position of each of the plurality of second reference lines intersect within a predetermined angular range. Note that, as described later, a plurality of second reference lines are arranged along the outline of the weld bead 110 at every first distance (equidistant) so as to intersect the first reference line within a predetermined angle range. Granted. Here, the second reference line is located along the contour of at least one side of the weld bead 110 at a first distance, and the midpoint position of the first reference line and the plurality of second reference lines is within a predetermined angular range. The two points on the contour on both sides of the weld bead 110 where the second reference line intersects are sequentially moved so as to intersect within the two points. By doing so, the first reference line and the first A plurality of second reference lines that intersect each reference line can be easily provided. Note that the first distance may be composed of a plurality of types of equally spaced distances. Further, a specific method of providing the first reference line and the second reference line will be described later.

Note that in this specification, the "predetermined angular range" is 90°±δ (°), where δ is a preset fixed value. For example, δ is set to 5 (°).

The dimension calculation unit 7 calculates the length of the weld bead 110 based on the first reference line. Furthermore, the dimension measurement unit calculates the width of the weld bead 110 at multiple locations along the first reference line based on the plurality of second reference lines. Note that the dimension calculation unit 7 also calculates statistics regarding the calculated widths of the weld bead 110 at multiple locations. Examples of statistics include maximum value, minimum value, average value, variance, and the like.

The storage unit 8 stores a dimension measurement program that describes a method for measuring the dimensions of the weld bead 110. Furthermore, the storage unit 8 stores the dimensional measurement results of the weld bead 110. The storage unit 8 may temporarily store the shape data acquired by the data acquisition unit 1. Furthermore, as will be described later, the bead region extraction unit 2 performs machine learning (deep learning) using learning data when extracting bead region data indicating a weld bead formation region from the shape data.

The display unit 9 displays shape data and bead area data as two-dimensional or three-dimensional image data. Furthermore, the display unit 9 displays contour data before and after the first reference line and the second reference line are provided. The display unit 9 may display the dimensional measurement results in numerical values or tabular form. Alternatively, the display unit 9 may display the dimension measurement results in a form that is added to the contour data after the first reference line and the second reference line have been added. Further, when displaying the dimensional measurement results regarding the plurality of weld beads 110, the display unit 9 may display the results in a graph format.

The learning data generation unit 10 generates learning data used to extract bead area data. The storage unit 8 may store learning data in this case. Note that the storage section 8, the display section 9, and the learning data generation section 10 may be provided outside the dimension measuring device 20.

The hardware configuration of the dimension measuring device 20 shown in FIG. 2 is similar to that of a known personal computer (PC). In other words, the dimension measuring device 20 includes, as hardware, an input port 21, a CPU (Central Processing Unit) 22, a GPU (Graphics Processing Unit) 23, a RAM (Random Access Memory)/ROM (Read Only Memory) 24, and an output port 25. and a data bus 26.

A plurality of functional blocks in the dimension measuring device 20 shown in FIG. 1 correspond to various devices shown in FIG. 2. For example, the RAM/ROM 24 corresponds to the storage section 8 shown in FIG. The input port 21 corresponds to the data acquisition section 1 shown in FIG. The dimension measuring device 20 also includes a display 27 as the display section 9 shown in FIG.

Note that the dimension measuring device 20 may include a keyboard or a touch panel (both not shown) as an input device. Further, the display 27 may be a touch panel and may be used as an input device. Further, the dimension measuring device 20 may include an HDD (Hard Disk Drive) or an SSD (Solid State Drive) in addition to the RAM/ROM 24 as the storage unit 8 shown in FIG.

Note that the respective functions of the bead area extraction section 2, contour extraction section 3, end point acquisition section 5, reference line provision section 6, and dimension calculation section 7 shown in FIG. This is achieved by running a dimension measurement program. Note that the function of the bead area extraction unit 2, that is, the function of extracting the formation area of the weld bead 110 from the shape data, is realized by executing a part of the dimension measurement program in the GPU 23. Other functions may be realized by executing a dimension measurement program in the GPU 23 or by executing a dimension measurement program in the CPU 22. Further, the functions of the learning data generation unit 10 are realized by executing a program different from the dimension measurement program in the GPU 23 or the CPU 22.

Note that the GPU 23 and the CPU 22 may be collectively referred to as a processor. The dimension measuring device 20 may have one processor (GPU 23 or CPU 22). However, it is preferable that the dimension measuring device 20 has a plurality of processors (GPU 23 and CPU 22, each of which may have a plurality of processors) from the viewpoint of improving data processing speed.

Although FIG. 2 shows an example in which various devices are connected to one data bus 26, a plurality of data buses 26 may be provided depending on the purpose, as in a normal PC.

[2: Weld bead dimension measurement procedure]
FIG. 3 is a schematic diagram showing how the appearance inspection device acquires shape data of a weld bead, and FIG. 4 is a flowchart showing a procedure for measuring dimensions of a weld bead. FIG. 5 is a schematic diagram showing the outline of a weld bead during smoothing processing and end point acquisition.

As shown in FIG. 3, the appearance inspection device 30 is configured by a three-dimensional shape measurement sensor 31 attached to a robot 32, and measures the shape of a weld bead 110 formed on a base material 100. Note that the three-dimensional shape measurement sensor 31 may be attached to a welding head (not shown) held by the robot 32. Note that the welding head corresponds to an energy supply unit that supplies energy for welding to the base material 100, and refers to, for example, a welding torch or a laser head.

The three-dimensional shape measurement sensor 31 includes a laser light source (not shown) that is configured to scan the surface of the base material 100 and a camera or a light receiving sensor that images the reflection locus of the laser beam projected onto the surface of the base material 100. (none of which are shown). Note that the camera has a CCD or CMOS image sensor as an image sensor. Further, the configuration of the three-dimensional shape measurement sensor 31 is not particularly limited to the above, and other configurations may be adopted. For example, an optical interferometer may be used instead of a camera or a light receiving sensor.

The three-dimensional shape measurement sensor 31 scans the weld bead 110 and the surrounding base material 100 with a laser beam, and the laser beam reflected by the weld bead 110 and the base material 100 is imaged by a camera or a light receiving sensor, thereby performing welding. The three-dimensional shape of the bead 110 and the surrounding base material 100, that is, the shape data is acquired.

Note that the shape data acquired by the three-dimensional shape measurement sensor 31 may include not only position information but also color information. Further, when the image sensor or the light receiving sensor of the three-dimensional shape measurement sensor 31 acquires a color image or a monochrome image, each pixel includes position information and color information or distance information. That is, the data of each pixel constituting the shape data, that is, the pixel data, includes position information and color information, or position information and distance information, or position information, color information, and distance information.

Here, the "position information" is two-dimensional coordinate information having a predetermined position on the surface of the base material 100 as the origin. For example, an arbitrary direction parallel to the surface of the base material 100 is the X direction, a direction parallel to the surface of the base material 100 and perpendicular to the X direction is the Y direction, and a direction perpendicular to the X direction and the Y direction, respectively. When is the Z direction, the position information is expressed as (X, Y)=(a, b) (a, b are arbitrary values). Specifically, the position information is the position of each pixel on the surface of the base material 100 or the position of each pixel on the surface of the weld bead 110 expressed in coordinates (X, Y).

Furthermore, in the shape data, distance information may be further combined with position information. In the present specification, "distance information" refers to the distance between the three-dimensional shape measurement sensor 31 and the surface position of the base material 100 or the surface position of the weld bead 110 expressed by coordinates (X, Y). .

In this case, coordinate information including distance information is expressed as (X, Y, Z) = (a, b, c) (c is an arbitrary value), and the coordinates (X, Y) and the three-dimensional shape measurement sensor 31 is expressed as distance information Z (=c).

Moreover, "color information" is information on color components included in a signal output from an image sensor or a light receiving sensor of a camera. When the image sensor acquires a color image, the color information included in the pixel data may be, for example, luminance information of each of the R component (red component), G component (green component), and B component (blue component). good. When the image sensor acquires a monochrome image, the color information included in the pixel data is brightness information. Further, the color information may include information corresponding to the brightness of a component (A component) representing transparency. For example, the magnitude of brightness may be normalized.

Further, the "color information" may be information regarding the type of color. For example, if a table or the like is prepared in which a number is assigned for each type of color, the color information corresponds to the number corresponding to the type of color.

In either case, shape data is constructed by combining position information with color information at the position. Furthermore, when capturing an image of the base material 100 including the weld bead 110 with an image sensor, the shape data can also be said to be a collection of pixel data for each pixel of the image sensor. However, shape data is not classified into two-dimensional or three-dimensional data like general image data.

The shape data thus obtained by the appearance inspection device 30 is input to the data acquisition unit 1 of the dimension measuring device 20, as shown in FIG. 4 (step S1).

Next, the bead area extraction unit 2 extracts bead area data from the shape data (step S2). Step S2 includes a plurality of steps, the details of which will be explained later.

Next, the contour extraction unit 3 extracts contour data from the bead area data (step S3). Specifically, contour data is extracted from the bead region data using a library having an image processing function, for example, OpenCV (Open Source Computer Vision Library).

Further, the smoothing processing unit 4 performs smoothing processing on the contour data (step S4). Specifically, as shown in FIG. 5, a moving average filter is applied to the contour data to remove jaggies included in the contour data. Note that jaggies are a type of noise that occurs in digital images, etc., and refer to step-like jaggedness that appears in lines and contours.

Note that the smoothing process may be performed by a method other than the filtering process using a moving average filter.

Next, the end point acquisition unit 5 acquires one or more end points (see FIG. 5) for the contour data after step S4 has been executed (step S5). For example, a point where the contour is inflected, or in other words, a point where the direction of the slope of the contour changes in the opposite direction, is selected as the end point. Note that if smoothing processing is not performed, the corners of the jaggies may be recognized as end points. By performing the smoothing process in step S4, it is possible to suppress the occurrence of erroneous recognition of end points.

Further, the reference line providing unit 6 provides the aforementioned first reference line and second reference line, starting from the end points, to the contour data after the end point acquisition in step S5 has been performed (step S6). In other words, the reference line providing unit 6 includes a first reference line for measuring the length of the weld bead 110 and a plurality of second reference lines for measuring the width of the weld bead 110, starting from the end point. is added to the contour data. In addition, when adding the first reference line and the plurality of second reference lines to the contour data, each of the plurality of second reference lines is within a predetermined angular range at the intersection of the first reference line and the second reference line. so that it intersects the first reference line. Note that step S6 includes a plurality of steps. Details of step S5 and step S6 will be explained later.

Next, the dimension calculation unit 7 calculates the length of the weld bead 110 based on the first reference line. Further, the dimension calculation unit 7 calculates the width of the weld bead 110 at a plurality of locations along the first reference line based on the plurality of second reference lines. Further, the dimension calculation unit 7 calculates the above-mentioned statistical amount regarding the widths of the plurality of locations of the weld bead 110 whose dimensions have been calculated (step S7).

Note that when a plurality of end points are acquired in the end point acquisition in step S5, a plurality of sets of the first reference line and the plurality of second reference lines are generated in step S6. In this case, in the dimension calculation in step S7, the length and width of the weld bead 110 are calculated for each group, and the value of the group in which the length of the weld bead 110 is the longest is selected and used as the final calculation result. . Specifically, among the calculated lengths of the plurality of weld beads 110, the maximum value is set as the determined length of the weld bead 110, and only the first reference line corresponding to this is selected. Further, only the plurality of second reference lines corresponding to the selected first reference line are selected. Based on the selected plurality of second reference lines, the width of the weld bead 110 is calculated at a plurality of locations along the selected first reference line. The remaining calculation results, that is, the calculation results based on the first reference line that does not have a fixed length and the corresponding second reference line, are not stored in the storage unit 8 but are deleted.

The dimensions of the weld bead 110 calculated in the dimension calculation in step S7 are output (output of the calculation results in step S8). The calculation result is output to the storage unit 8 or a device or storage unit (none of which is shown) provided outside the dimension measuring device 20. In the latter case, the calculation result is transmitted via the output port 25. Alternatively, the calculation results may be directly output to the display section 9. The calculation results in that case are displayed on the display section 9 in numerical values, table format, or graph format. Further, the calculation result may be displayed on the display unit 9 together with the contour data to which the first reference line and the second reference line are attached.

[2-1: Weld bead formation area extraction procedure]
FIG. 6 is a flowchart showing a procedure for extracting a weld bead formation area, and corresponds to the process of extracting bead area data in step S2 of FIG.

FIG. 7 is a diagram showing an image of shape data and a weld bead formation region and other regions in the shape data. FIG. 8 is a flowchart showing an annotation adding procedure for generating learning data. FIG. 9A is an example showing how annotations are added to shape data, and FIG. 9B is an example showing an image of the shape data after the annotations are added.

When extracting the formation region of the weld bead 110 from the shape data as bead region data, the shape data to be extracted is prepared as shown in FIG. Further, in executing machine learning (deep learning) in step S12, a necessary number and type of learning data are prepared (step S11). As described above, the learning data may be data stored in the storage unit 8 or may be data input from outside the dimension measuring device 20.

Next, the bead region extraction unit 2 uses the learning data to perform machine learning (deep learning) on an algorithm for extracting bead region data (step S12). Using the learning-enhanced algorithm, the shape data is processed into two parts: bead area data and data regarding other areas (step S13). An algorithm called semantic segmentation is used to execute steps S12 and S13. Semantic segmentation is an algorithm that associates labels and categories with all pixel data in image data, and is used for object detection and the like. In this embodiment, an algorithm called DeepLab v3+ is used, but the algorithm is not particularly limited to this. Further, the algorithm is executed by the GPU 23.

When shape data is viewed as image data acquired by an image sensor, classification is performed on data of each pixel in the shape data. In this embodiment, the aforementioned color information and distance information representing the distance to the three-dimensional shape measurement sensor 31 are combined in the data of each pixel in the shape data. In steps S12 and S13, classification is performed focusing on this color information. More specifically, the shape data is expanded, reduced, or divided to a fixed size, and then input to a CNN (Convolutional Neural Network) implemented in the GPU 23, specifically, DeepLab v3+. Furthermore, calculation processing is performed on the shape data in each layer of the CNN. At this time, based on color information or distance information, classification is performed into two classes: a class corresponding to the formation area of weld bead 110 and a class corresponding to other areas. The bead area extraction unit 2 generates shape data that is given the above-mentioned class and has a format in which bead area data and data in other areas can be distinguished. Specifically, as shown in FIG. 7, an image is generated in which the weld bead 110 formation region and other regions are color-coded with respect to the shape data image.

After execution of step S13, the machining result is stored in the storage unit 8 (saving of the machining result in step S14).

In this way, by using an algorithm enhanced by machine learning (deep learning), it is possible to accurately and reliably extract bead region data, which is the formation region of weld bead 110, from shape data. In addition, since the bead area data is specified using deep learning, the bead area data is extracted using delicate parameter settings, compared to, for example, identifying the feature amount of the external shape of the welded part by removing noise using a low-pass filter. Problems such as unstable accuracy do not occur.

Note that the learning data is generated based on shape data obtained by visually inspecting an actual weld bead 110 prepared separately. Here, a case where learning data is generated by the learning data generating section 10 provided inside the dimension measuring device 20 will be described as an example. However, as described above, the learning data may be generated outside the dimension measuring device 20.

First, as shown in FIG. 8, a figure designation mode for annotation is selected (step S21). In this embodiment, the shape of the weld bead 110 is approximated by a polygon and annotation is provided.

Next, the learning data generation unit 10 calls the shape data for generating learning data, opens the data file of the shape data (step S22), and further displays the opened data file on the display unit 9. , visualize. The display section 9 in this case is a touch panel and also has an input function. Note that the input function may be realized by another device such as a touch pen or a mouse.

After displaying the image of the shape data on the display unit 9, the operator traces the outline of the weld bead 110 while viewing the image of the shape data. In this embodiment, the aforementioned polygon specification mode is selected as the figure specification mode. In this case, as shown in FIG. 9A, the vertices of the polygon are determined by sequentially clicking along the outline with a touch pen or the like. Finally, as shown in FIG. 9B, the weld bead 110 is surrounded by a polygon, the formation area of the weld bead 110 is designated, and the annotation ends (step S23).

The annotated shape data is stored in the storage unit 8 as learning data (step S24).

Next, the learning data generation unit 10 determines whether or not there remains a data file of shape data to which annotation should be added (step S25). If the determination result in step S25 is negative, that is, if there remains no data file of shape data to which annotation should be added, the annotation work is ended. If the judgment result in step S25 is positive, that is, if there remains a data file of shape data to which annotation should be added, the series of processes from steps S22 to S24 is repeatedly executed until the judgment result in step S25 is negative. do. Note that the process in step S25 may be performed by an operator.

[2-2: Procedure for assigning first reference line and second reference line]
FIG. 10 is a flowchart showing the procedure for adding the first reference line and the second reference line. Compatible.

FIG. 11A is a schematic diagram showing the state of the first loop processing when providing the first reference line and the second reference line. FIG. 11B is a schematic diagram showing a process subsequent to the process shown in FIG. 11A. FIG. 11C is a schematic diagram showing a state of processing subsequent to the processing shown in FIG. 11B. FIG. 12 is a schematic diagram showing the state of the second loop processing when providing the first reference line and the second reference line.

First, the end point acquired by the end point acquisition section 5 is set as the end point _A0 on the contour data. As shown in FIGS. 10 and 11A, virtual points B _i and C _i are respectively provided on the end point A ₀ (step S31). Although i is an integer greater than or equal to 0, in step S31, i=0. The virtual points B _i and C _i are virtual points (A0 (B0, C0)) set to overlap the end point A ₀ , respectively. The process of step S31 may be called initial setting process. Further, the processes after step S31 are executed by the reference line providing section 6.

Next, as shown in FIG. 11B, the virtual point B _i is moved along the contour of the weld bead 110 by a predetermined distance (hereinafter referred to as the first distance) in a first direction that is a direction away from the end point _A0 . Proceed. The point after movement is assumed to be virtual point B _(i+1) . At the same time, the virtual point C _i is advanced along the contour of the weld bead 110 by a first distance in a second direction that is away from the end point A ₀ and is different from the first direction. The point after the movement is defined as virtual point C _(i+1) (step S32). The midpoint between virtual point B _(i+1) and virtual point C _(i+1) is set as midpoint _{A (i+1)} (step S33). The processing in steps S32 and S33 may be referred to as first loop processing. As shown in FIG. 11B and then FIG. 11C following FIG. 11B, by sequentially proceeding with the first loop process, the end point A ₀ and the virtual points B ₁ , B ₂ , C ₁ , C ₂ are the outline of the weld bead 110. Above, the midpoints A ₁ , A ₂ are respectively located in the area of the weld bead 110 . The path A ₀ -A ₁ -A ₂ shown in FIG. 11C forms part of the first reference line. Further, the line segment B ₁ -C ₁ and the line segment B ₂ -C ₂ each correspond to a second reference line.

Next, determine whether the line segment B _(i+1) - C _(i+1) on the first reference line and the line segment A _i -A _(i+1) on the second reference line intersect within the predetermined angle range mentioned above. It is determined whether or not (step S34). That is, it is determined whether the line segment B _(i+1) - C _(i+1) and the line segment A _i -A _(i+1) intersect within the range of 90 (°) ± δ (°).

If the determination result in step S34 is negative, that is, if the line segment B _(i+1) -C _(i+1) and the line segment A _i -A _(i+1) intersect beyond the predetermined angular range, step S36 Proceed to.

In step S36, either virtual point B _(i+1) or virtual point C _(i+1) is moved along the outline of weld bead 110 by a first distance in a direction away from end point _A0 . The point after the movement is set as virtual point B _(i+2) or virtual point C _(i+2) , and the process advances to step S37.

In step S37, the intersection angle between the line segment B _(i+2) -C _(i+1) and the line segment A _i -A _(i+1) and the intersection angle between the line segment B _(i+1) -C _(i+2) and the line segment A _i -A _{(i+1) are determined. )} , and select virtual point B _(i+2) or virtual point C _(i+2) whose intersection angle is within or close to a predetermined angle range.

If the selected virtual point is virtual point B _(i+2) , the midpoint A _(i+1) is corrected to the midpoint between virtual point B _(i+2) and virtual point C _(i+1) . Furthermore, the original virtual point B _(i+1) is skipped and virtual point B _(i+2) is newly set as virtual point B _(i+1) .

On the other hand, if the selected virtual point is the moving virtual point C _(i+2) , the midpoint A _(i+1) is corrected to the midpoint between the moving point B _(i+1) and the moving point C _(i+2) . Furthermore, the original moving point C _(i+1) is skipped and the virtual point C _(i+2) is newly set as the virtual point C _(i+1) . After step S37 ends, the process returns to step S34, and the series of processes of step S36 and step S37 are repeatedly executed until the determination result of step S34 becomes affirmative. Note that the processing in step S36 and step S37 may be collectively referred to as second loop processing. As shown in FIG. 12, by sequentially proceeding with the second loop process, the end point A ₀ and the virtual points B ₁ , B ₂ , C ₁ , C ₂ are located on the outline of the weld bead 110, and the midpoints A ₁ , A ₂ are respectively arranged within the area of the weld bead 110. The path A ₀ -A ₁ -A ₂ -...-A _i shown in FIG. 12 forms part of the first reference line. Furthermore, the line segments B ₁ -C ₁ , B ₂ -C ₂ , . . . , B _i -C _i each correspond to the second reference line. Note that, in FIG. 12, for example, the mark of the virtual point B _i that is skipped when the determination result in step S34 is negative is shown by a broken circle line.

On the other hand, if the determination result in step S34 is affirmative, that is, if the line segment B _(i+1) - C _(i+1) and the line segment A _i -A _(i+1) intersect within a predetermined angle range, the route It is determined whether the sum of the length of A ₀ -B _i and the length of path A ₀ -C _i exceeds the length of the outline of weld bead 110 (step S35).

If the judgment result in step S35 is negative, that is, if the sum of the length of the path A ₀ -B _i and the length of the path A ₀ -C _i does not exceed the length of the outline of the weld bead 110, the variable i is counted up to (i+1), the process returns to step S32, and a series of processes from steps S32 to S34 are repeatedly executed until the determination result in step S35 becomes affirmative. Here, the length of the path A ₀ - _{B i} is the distance of the path from the end point A ₀ to the virtual point B _i via the virtual point B _k (k is an integer, 1≦k<i). To tell. The length of the path A ₀ -C _i refers to the distance of the path from the end point A ₀ to the virtual point C _i via the virtual point C _k .

On the other hand, if the determination result in step S35 is affirmative, that is, if the sum of the length of the path A ₀ -B _i and the length of the path A ₀ -C _i exceeds the length of the outline of the weld bead 110, The first reference line and the second reference line are provisionally determined (step S38). The first reference line corresponds to a route starting from the end point A ₀ and passing through each of the plurality of midpoints A ₁ , . . . , Ai set at the end of step S35. Note that at the end of step S35, the end point of the first reference line has reached or intersected the contour of the weld bead 110. Furthermore, each of the line segments B ₁ -C ₁ , . . . , B _i -C _i is the second reference line described above. The number of second reference lines is determined according to the length of the weld bead 110 and the above-mentioned first distance (a predetermined distance in the first direction, which is the direction away from the end point _A0 along the contour of the weld bead 110). be done.

However, in the processing up to step S38, the first reference line may not be appropriately set. Therefore, in this embodiment, the first reference line correction process (step S39), which will be described later, is uniformly performed after execution of step S38 to determine the first reference line and the plurality of second reference lines. The process of step S39 will be further explained.

[2-3: First reference line correction procedure]
FIG. 13 is a flowchart showing the procedure for correcting the first reference line, and corresponds to the process of step S39 shown in FIG. 10.

FIG. 14A is an example of contour data with a first reference line added when no correction is performed, and FIG. 14B is an example of another contour data with a first reference line added when no correction is performed. be.

FIG. 15A is an example of contour data when the first reference line is transformed into a temporary reference line, and FIG. 15B is an example of another contour data when the first reference line is transformed into a temporary reference line.

FIG. 16A is an example of contour data to which a first reference line has been added after correction has been completed, and FIG. 16B is an example of another contour data to which a first reference line has been added after correction has been completed. be.

Note that FIGS. 14A, 15A, and 16A each show the same contour data. Similarly, FIGS. 14B, 15B, and 16B each show the same contour data.

Specifically, when bead area data is extracted from the shape data in the procedure shown in FIG. 6 by the bead area data extraction process in step S2 shown in FIG. 4 (step S13), the bead area data after extraction is Noise may be included (see FIG. 14A). In such a case, when the length of the welding bead 110 is calculated in accordance with the subsequent steps of contour data acquisition in step S3 shown in FIG. 4, a value different from the actual length may be output. Alternatively, in acquiring the end points in step S5 shown in FIG. 4, if the end points are not properly acquired, the first reference line itself may not be appropriately generated (see FIG. 14B).

Even when these occur, the first reference line can be appropriately assigned to the contour data by executing the procedure described below. With this, it is possible to appropriately add a plurality of second reference lines to the contour data. Furthermore, the length and width of weld bead 110 can be calculated accurately. The actual correction procedure will be explained in detail below.

First, both ends of the first reference line are removed by a predetermined length to set a temporary reference line (step S41 in FIG. 13; see also FIGS. 15A and 15B). In this case, it is not determined whether a predetermined length has been correctly removed from each end of the first reference line.

Next, straight lines (hereinafter sometimes referred to as extension lines) according to the respective inclinations near both ends of the temporary reference line are extended from each end of the temporary reference line until they intersect the contour (see FIG. 13). Step S42 (see also FIGS. 16A and 16B).

Furthermore, a line that is a composite of the temporary reference line and extension lines extended from both ends of the temporary reference line is newly set as a first reference line. The first reference line is added to the contour data (step S43 in FIG. 13; see also FIGS. 16A and 16B).

In the dimension calculation in step S7 shown in FIG. 4, the length of the first reference line shown in FIGS. 16A and 16B is calculated as the length of the weld bead 110.

Note that if a plurality of endpoints are acquired in the endpoint acquisition in step S5 shown in FIG. 4, a first reference line and a second reference line are added to the contour data using each endpoint as a starting point in the procedure shown in FIG. It goes without saying that the first reference line is also corrected by the procedure shown in FIG. Further, as described above, the length and width of the weld bead 110 are calculated based on the first reference line having a determined length and the plurality of second reference lines corresponding thereto.

[3: Effects, etc.]
As explained above, the dimension measuring device 20 according to the present embodiment includes the data acquisition unit 1, the bead area extraction unit 2, the contour extraction unit 3, the smoothing processing unit 4, the end point acquisition unit 5, the reference line provision unit 6, and the dimension measurement unit 20. The calculation section 7 is provided at least.

The data acquisition unit 1 acquires shape data of the base material 100 on which the weld bead 110 is formed. The bead region extraction unit 2 extracts the formation region of the weld bead 110 from the shape data as bead region data. The contour extraction unit 3 extracts the contour of the weld bead 110 as contour data based on the bead area data. The smoothing processing unit 4 performs smoothing processing on the contour data.

The end point acquisition unit 5 acquires the end points of the weld bead 110 based on the smoothed contour data. The reference line providing unit 6 has a first reference line for measuring the length of the weld bead 110 and a plurality of second reference lines for measuring the width of the weld bead 110, starting from the end point of the weld bead 110. is added to the contour data. The dimension calculation unit 7 calculates the length of the weld bead 110 based on the first reference line, and calculates the width of the weld bead 110 at a plurality of locations along the first reference line based on the plurality of second reference lines. calculate. Furthermore, the reference line providing unit 6 is configured to arrange the first reference line and the plurality of second reference lines so that the first reference line and the plurality of second reference lines intersect each other within a predetermined angular range. is added to the contour data.

By configuring the dimension measuring device 20 in this way, even if the base material 100 and the weld bead 110 have arbitrary shapes, the dimensions of the weld bead 110 can be automatically determined without setting complicated measurement conditions. It can be measured easily and easily. Furthermore, this makes it possible to easily and reliably detect defective products in which the dimensions of the weld bead 110 deviate from a predetermined standard range, thereby reducing inspection costs. In addition to the linear or arcuate weld bead 110, the weld bead 110 having an arbitrary shape mentioned above may include an S-shaped weld bead 110, a zigzag weld bead 110, a flat raised trapezoidal shape, or a flat weld bead 110. Beads 110 are also included.

When the end point obtaining unit 5 obtains a plurality of end points, the reference line providing unit 6 provides a first reference line to the contour data using each of the plurality of end points as a starting point. Furthermore, a plurality of second reference lines are provided to the contour data corresponding to each of the provided plurality of first reference lines.

The dimension calculation unit 7 calculates the length of the weld bead 110 based on each of the plurality of first reference lines, and determines the maximum value of the weld bead 110 among the plurality of calculated lengths of the weld bead 110. length. Further, based on a plurality of second reference lines corresponding to the first reference line having a determined length, the width of the weld bead 110 is calculated at a plurality of locations along the first reference line.

By configuring the reference line providing section 6 and the dimension calculation section 7 in this way, for example, even if the end point obtaining section 5 cannot select an appropriate end point and a plurality of end points are obtained, the weld bead 110 The dimensions of can be measured automatically, easily, and more accurately.

The reference line providing unit 6 sets the end point to the end point A ₀ and executes an initial setting process to add a virtual point B _i and a virtual point C _i (i is an integer of 0 or 1 or more) on the end point A ₀ . Further, the reference line providing unit 6 sets a point that is a first distance from the virtual point B _i in a first direction that is a direction away from the end point A ₀ along the contour of the weld bead 110 to a virtual point B _(i+1). Then, the virtual point C _(i+1) is a point that advances from the virtual point C _i along the contour by a first distance in a second direction that is away from the end point A ₀ and is a direction different from the first direction. , and further executes at least a first loop process of setting the midpoint between the virtual point B _(i+1) and the virtual point C _(i+1) to the midpoint A _(i+1) .

A line segment connecting virtual point B _(i+1) and virtual point C _(i+1) connects midpoint A _i and midpoint A _(i+1) within a predetermined angular range (90 (°) ± δ (°)). When intersecting the connecting line segment, the reference line providing unit 6 creates a path from the end point A ₀ to the virtual point B _i via the virtual point B _k (k is an integer, 1≦k<i). The first loop process is performed once until the sum of the distance and the distance of the path from the end point _A0 to the virtual point _Ci via the virtual point _Ck exceeds the length of the outline of the weld bead 110. Alternatively, the first reference line and a plurality of second reference lines are added to the contour data by executing the process multiple times.

By configuring the reference line providing section 6 in this way, the first reference line for determining the length of the weld bead 110 can be automatically, simply, and more accurately set. Accordingly, each of the plurality of second reference lines can be automatically, simply, and more accurately set. These allow the dimensions of the weld bead 110 to be measured automatically, easily, and more accurately.

If a line segment connecting virtual point B _(i+1) and virtual point C _(i+1) intersects a line segment connecting midpoint A _i and midpoint A _(i+1) beyond a predetermined angular range, the standard The line adding section 6 executes a second loop process including the following processes (i) to (iv).

In the second loop process, the reference line providing unit 6
(i) Move either virtual point B _(i+1) or virtual point C _(i+1) along the outline of weld bead 110 by a first distance in a direction away from end point _A0 , and move the point after the movement. The virtual point B _(i+2) or the virtual point C _(i+2) is set.

(ii) The intersection angle between the line segment connecting virtual point B _(i+2) and virtual point C _(i+1) and the line segment connecting midpoint A _i and midpoint A _(i+1) , and the virtual point B _(i+1). The intersection angle of the line segment connecting the virtual point C _(i+2) and the line segment connecting the midpoint A _i and the midpoint A _(i+1) is compared, and the virtual intersection angle is within or close to a predetermined angle range. Select point B _(i+2) or virtual point C _(i+2) .

(iii) If the selected virtual point is virtual point B _(i+2) , modify the midpoint A _(i+1) to the midpoint between virtual point B _(i+2) and virtual point C _(i+1) , and _(i+2) is newly set as virtual point B _(i+1) .

(iv) If the selected virtual point is virtual point C _(i+2) , modify the midpoint A _(i+1) to the midpoint between virtual point B _(i+1) and virtual point C _(i+2) , and _(i+2) is newly set as the virtual point C _(i+1) .

The reference line providing unit 6 determines whether a line segment connecting virtual point B _(i+1) and virtual point C _{(i+1) is} a line segment connecting midpoint A _i and midpoint A _(i+1) within a predetermined angular range. The second loop process is repeatedly executed until the intersection occurs.

By configuring the reference line providing section 6 in this way, the first reference line that determines the length of the weld bead 110 can be set more accurately. According to this, each of the plurality of second reference lines can be set more accurately.

Depending on the shape and curvature of weld bead 110, the line segment connecting virtual point B _(i+1) and virtual point C _(i+1) may exceed a predetermined angular range and connect midpoint A _i and midpoint A _(i+1). It may intersect with the connecting line segments. In this case, the midpoint A _(i+1) may not be set at a position approximately equidistant from each of the virtual point B _(i+1) and the virtual point C _(i+1) . That is, the distance of the path from the end point A ₀ to the end point of the weld bead 110 via the middle point A _(i+1) may not accurately reflect the length of the weld bead 110 .

According to this embodiment, attention is paid to the intersection angle between the line segment connecting the virtual point B _(i+1) and the virtual point C _(i+1) and the line segment connecting the midpoint A _i and the midpoint A _(i+1). By resetting either virtual point B _(i+1) or virtual point C _(i+1) and also midpoint A _(i+1) , welding can be performed from end point A ₀ via midpoint A _(i+1). The distance of the path to the end point of the bead 110 can accurately reflect the length of the weld bead 110. These allow the dimensions of the weld bead 110 to be measured automatically, easily, and more accurately.

The reference line providing unit 6 removes a predetermined length from both ends of the first reference line to obtain a temporary reference line. Further, straight lines (extension lines) corresponding to the respective inclinations near both ends of the temporary reference line are extended from both ends of the temporary reference line until they intersect the outline of the weld bead 110.

Further, the reference line providing unit 6 adds a new first reference line to the outline data, which is a line that is a composite of the temporary reference line and straight lines (extension lines) each extending from both ends of the temporary reference line.

The dimension calculation unit 7 calculates the length of the newly set first reference line as the length of the weld bead 110, and calculates the length of the line segment connecting virtual point B _(i+1) and virtual point C _(i+1). The width is calculated as the width of the weld bead 110.

As described above, the extracted bead area data may contain noise, and when the length of the weld bead 110 is calculated, a value different from the actual length may be output. Alternatively, if the end points are not properly acquired by the end point acquisition unit 5, the first reference line itself is not properly generated, and the calculated value of the length of the weld bead 110 is output different from the actual length. may be done.

On the other hand, according to the present embodiment, by deleting both ends of the first reference line initially set and adding an extension line with an appropriate slope so as to reach the outline of the weld bead 110, The first reference line can be applied in an appropriate manner. Moreover, thereby, a plurality of second reference lines can also be appropriately added to the contour data. From the above, the length and width of weld bead 110 can be calculated accurately.

The shape data acquired by the data acquisition unit 1 consists of a plurality of pixel data, and each of the plurality of pixel data includes at least color information or distance information, specifically, information about the three-dimensional shape measurement sensor 31 and the base material 100. Preferably, it includes distance information that is a distance from a predetermined position on the surface or a predetermined position on the surface of weld bead 110.

In this case, the bead area extraction unit 2 extracts bead area data from the shape data using a predetermined algorithm that is trained based on the color information or the distance information and is trained using learning data annotated in advance. It is preferable to extract.

By doing so, bead area data can be extracted from the shape data with high accuracy. The shape of the weld bead 110 is diverse and difficult to generalize using a mathematical formula. For this reason, bead area data can be stabilized as the formation area of the weld bead 110 from the shape data of the base material 100 in which the weld bead 110 is formed, using a conventional rule-based mathematical approach such as a method of extracting changing points. It is difficult to extract the

On the other hand, according to the present embodiment, attention is paid to color information included in pixel data in the shape data, and further, so-called AI (Artificial Intelligence) processing is performed on the shape data using a learning-enhanced algorithm. There is. This makes it possible to accurately extract bead area data from the shape data, regardless of the shape of the weld bead 110 or the base material 100.

The method for measuring dimensions of weld bead 110 according to the present embodiment includes at least the following first to seventh steps.

In the first step (obtaining shape data in step S1 in FIG. 4), shape data of the base material 100 on which the weld bead 110 is formed is obtained, and in the second step (bead area data extraction in step S2 in FIG. 4), , the formation area of the weld bead 110 is extracted from the shape data as bead area data.

In the third step (obtaining contour data in step S3 in FIG. 4), the contour of the weld bead 110 is extracted as contour data based on the bead area data, and in the fourth step (smoothing process in step S4 in FIG. 4), , smoothing processing is performed on the contour data, and in a fifth step (endpoint acquisition in step S5 in FIG. 4), endpoints of the weld bead 110 are acquired based on the smoothed contour data.

In the sixth step (step S6 in FIG. 4), a first reference line for measuring the length of the weld bead 110 and a plurality of reference lines for measuring the width of the weld bead 110 are established starting from the acquired end point. 2 reference lines are added to the contour data. Specifically, in the sixth step, the first reference line and the plurality of second reference lines are arranged such that the first reference line and the plurality of second reference lines intersect each other within a predetermined angular range. and the second reference line is added to the contour data. In the seventh step (dimension calculation in step S7 in FIG. 4), the length of the weld bead 110 is calculated based on the first reference line, and the width of the weld bead 110 is calculated based on the plurality of second reference lines. Calculate at multiple locations along one reference line. Here, the second reference line is located along the contour of at least one side of the weld bead 110 at a first distance, and the midpoint position of the first reference line and the plurality of second reference lines is within a predetermined angular range. The two points on the contour on both sides of the weld bead 110 where the second reference line intersects are sequentially moved so as to intersect within the two points. By doing so, the first reference line and the first A plurality of second reference lines that intersect each reference line can be easily provided. Note that the first distance may be composed of a plurality of types of equally spaced distances.

According to this embodiment, even if the base material 100 and the weld bead 110 have arbitrary shapes, the dimensions of the weld bead 110 can be automatically and easily measured without setting complicated measurement conditions. be able to. Furthermore, this makes it possible to easily and reliably detect defective products in which the dimensions of the weld bead 110 deviate from a predetermined standard range, thereby reducing inspection costs.

If a plurality of endpoints are acquired in the endpoint acquisition in the fifth step, in the sixth step, a first reference line is attached to the contour data using each of the plurality of endpoints as a starting point. Furthermore, a plurality of second reference lines are provided to the contour data corresponding to each of the provided plurality of first reference lines.

In the dimension calculation in the seventh step, the length of the weld bead 110 is calculated based on each of the plurality of first reference lines, and the maximum value of the plurality of calculated lengths of the weld bead 110 is set as the weld bead 110. Let the fixed length be . Further, based on a plurality of second reference lines corresponding to the first reference line having a determined length, the width of the weld bead 110 is calculated at a plurality of locations along the first reference line.

By doing this, for example, even if the end point acquisition unit 5 cannot select an appropriate end point and a plurality of end points are acquired, the dimensions of the weld bead 110 can be automatically, easily, and more accurately measured. can do.

The sixth step is to define the end point of the weld bead 110 as the end point _A0 , and to add a virtual point B _i and a virtual point C _i (i is an integer of 0 or 1 or more) to _the end point A 0 (step S31 in FIG. 10). ) and a first loop process (steps S32 and S33 in FIG. 10).

The first loop process is a step of setting a virtual point B _(i+1) as a point that has proceeded from the virtual point B _i by a first distance in a first direction, which is a direction away from the end point A ₀ , along the outline of the weld bead 110. (Step S32 in FIG. 10). The first loop process starts from a virtual point _Ci along the outline of the welding bead 110, and proceeds by a first distance in a second direction that is a direction away from an end point _A0 and that is different from the first direction. The process further includes a step of setting C _(i+1) as a virtual point (step S32 in FIG. 10). The first loop process further includes a step (step S33 in FIG. 10) of setting the midpoint between virtual point B _(i+1) and virtual point C _(i+1) to midpoint A _(i+1) .

In the sixth step, a line segment connecting virtual point B _(i+1) and virtual point C _(i+1) is set between midpoint A _i and midpoint A within a predetermined angular range (90 (°) ± δ (°)). _(i+1) , the distance of the path from the end point A ₀ to the virtual point B _i via the virtual point B _k (k is an integer, 1≦k<i) , the first loop process is performed one or more times until the sum of the distance of the path from the end point A ₀ to the virtual point C _i via the virtual point C _k exceeds the length of the outline of the weld bead 110. The first reference line and the plurality of second reference lines are added to the contour data.

By doing so, the first reference line for determining the length of the weld bead 110 can be automatically, simply, and more accurately set. Accordingly, each of the plurality of second reference lines can be automatically, simply, and more accurately set. These allow the dimensions of the weld bead 110 to be measured automatically, easily, and more accurately.

If the line segment connecting virtual point B _(i+1) and virtual point C _(i+1) intersects the line segment connecting midpoint A _i and midpoint A _(i+1) beyond a predetermined angular range, Step 6 further includes second loop processing (steps S36 and S37 in FIG. 10).

In the second loop process, one of the virtual points B _(i+1) and C _(i+1) is advanced along the outline of the weld bead 110 by a first distance in a direction away from the end point _A0 , and after the movement, The process includes at least the step of setting the point as virtual point B _(i+2) or virtual point C _(i+2) (step S36 in FIG. 10).

In addition, the second loop process calculates the intersection angle between the line segment connecting virtual point B _(i+2) and virtual point C _(i+1) and the line segment connecting midpoint A _i and midpoint A _(i+1). Then, compare the intersection angles of the line segment connecting virtual point B _(i+1) and virtual point C _(i+2) with the line segment connecting midpoint A _i and midpoint A _(i+1) , and find a predetermined angle. This includes a step (step S37 in FIG. 10) of selecting virtual point B _(i+2) or virtual point C _(i+2) whose intersection angle is within or close to the range.

If the selected virtual point is virtual point B _(i+2) , the second loop process corrects the midpoint A _(i+1) to the midpoint between virtual point B _(i+2) and virtual point C _(i+1) , and The process further includes a step of newly setting virtual point B _(i+2) to virtual point B _(i+1) (step S37 in FIG. 10).

When the selected virtual point is virtual point C _(i+2) , the second loop process corrects the midpoint A _(i+1) to the midpoint between virtual point B _(i+1) and virtual point C _(i+2) , and The process further includes a step (step S37 in FIG. 10) of newly setting the virtual point C _(i+2) to the virtual point C _(i+1) .

In the sixth step, a line segment connecting virtual point B _(i+1) and virtual point C _(i+1) intersects a line segment connecting midpoint A _i and midpoint A _(i+1) within a predetermined angular range. The second loop process is repeatedly executed until.

According to this embodiment, the first reference line that determines the length of the weld bead 110 can be set more accurately. Accordingly, it is possible to more accurately set each of the plurality of second reference lines that determine the respective widths of the weld bead at a plurality of locations.

Also, focusing on the intersection angle between the line segment connecting virtual point B _(i+1) and virtual point C _(i+1) and the line segment connecting midpoint A _i and midpoint A _(i+1) , _(i+1) and the virtual point C _(i+1) , and by resetting the midpoint A _(i+1) , the end point of the weld bead 110 is reached from the end point A ₀ via the midpoint A _(i+1). The length of the weld bead 110 can be accurately reflected by the distance of the path leading to the weld bead 110. These allow the dimensions of the weld bead 110 to be measured automatically, easily, and more accurately.

The sixth step further includes a step of deleting a predetermined length from both ends of the first reference line to obtain a temporary reference line (step S41 in FIG. 13). Further, the sixth step is a step of extending straight lines (extension lines) according to the respective inclinations near both ends of the temporary reference line from each end of the temporary reference line until they intersect the outline of the weld bead 110 (see FIG. 13). Step S42) is included.

The sixth step is a step of adding a new first reference line to the contour data by combining the temporary reference line and straight lines (extension lines) each extending from both ends of the temporary reference line (step of FIG. 13). S43).

In the dimension calculation in the seventh step, the length of the newly set first reference line is calculated as the length of the weld bead 110, and the line segment connecting virtual point B _(i+1) and virtual point C _{(i+1) is} The length of is calculated as the width of the weld bead 110.

According to this embodiment, by deleting both ends of the first reference line that is initially set and adding an extension line with an appropriate inclination to reach the outline of the weld bead 110, an appropriate The first reference line can be appropriately provided in the shape of the image. Moreover, thereby, a plurality of second reference lines can also be appropriately added to the contour data. From the above, the length and width of weld bead 110 can be calculated accurately.

It is preferable that the shape data acquired by the data acquisition unit 1 consist of a plurality of pixel data, and each of the plurality of pixel data includes at least color information or distance information.

In the second step, it is preferable to extract bead area data from the shape data based on the color information and using a predetermined algorithm whose learning has been reinforced using learning data annotated in advance.

According to this embodiment, AI processing is performed on the shape data by focusing on color information or distance information included in pixel data in the shape data, and using a learning-enhanced algorithm. This makes it possible to accurately extract bead area data from the shape data, regardless of the shape of the weld bead 110 or the base material 100.

The dimension measurement program according to this embodiment causes one or more processors to execute the method for measuring the dimensions of weld bead 110 described above.

In this way, even if the base metal 100 and the weld bead 110 have arbitrary shapes, the dimensions of the weld bead 110 can be automatically and easily measured without setting complicated measurement conditions. be able to. Furthermore, this makes it possible to easily and reliably detect defective products in which the dimensions of the weld bead 110 deviate from a predetermined standard range, thereby reducing inspection costs.

(Embodiment 2)
FIG. 17 shows functional blocks of the welding location quality determination device according to this embodiment. Note that in FIG. 17, the same parts as in Embodiment 1 are denoted by the same reference numerals, and detailed explanations are omitted. Further, since the configuration of the functional blocks of the dimension measuring device 20 and the functions of each functional block are the same as those shown in FIG. 1, illustration and detailed explanation will be omitted.

As shown in FIG. 17, the welding location quality determining device 50 includes at least a dimension measuring device 20, a shape defect determining device 40, a quality determining section 51, and a display/output section 52.

The shape defect determination device 40 includes at least a shape data processing section 41, a defect determination section 42, and a storage section 43, and its hardware configuration is the same as that shown in FIG. 2. Note that the display 27 may be omitted.

The shape data processing unit 41 has a function of removing noise from the shape data acquired by the visual inspection device 30. Since the reflectance of the laser beam emitted from the three-dimensional shape measurement sensor 31 differs depending on the material of the base material 100, if the reflectance is too high, halation or the like will occur, resulting in noise, which may affect the shape data. For this reason, for example, predetermined software is executed on the CPU to realize noise filtering processing by the shape data processing section 41.

Note that noise can be similarly removed by providing an optical filter (not shown) in the visual inspection device 30. High-quality shape data can be obtained by using an optical filter and filtering processing on software.

The defect determining unit 42 uses the shape data after noise removal by the shape data processing unit 41 to determine whether there is a shape defect in the welding location including the weld bead 110. Furthermore, the defect determination unit 42 estimates the type of shape defect. Further, the defect determination unit 42 determines the position and size of the shape defect at the welding location. The defect determination unit 42 may determine the number of weld defects. For defect determination, for example, the presence or absence and type of defects in the shape data may be inferred by performing machine learning using learning data for defect determination prepared in advance. In this case, algorithm processing for inference is executed on the shape data on the CPU 22 or GPU 23.

The storage unit 43 stores the shape defect determination result of the welding location. Further, the shape data after noise removal may be temporarily stored. Further, data and software used for determination by the defect determination section 42, such as learning data for defect determination and algorithms for inference, may be stored. The storage unit 43 includes a RAM/ROM (see FIG. 2), an HDD, or an SSD.

The quality determining unit 51 determines whether the appearance of the welded area satisfies a predetermined standard based on the measurement results of the dimension measuring device 20 and the determination results of the shape defect determining device 40. The function of the quality determination unit 51 is realized, for example, by executing predetermined software on the CPU.

The display/output unit 52 displays the determination result of the quality determination unit 51 or outputs it to an external device (not shown). The display/output section 52 includes a display (see FIG. 2), an output port (see FIG. 2), or both.

According to this embodiment, the dimensions of the weld bead 110 and the presence or absence of a shape defect at the welding location can be evaluated simultaneously based on the same shape data. Thereby, it is possible to automatically and easily evaluate whether the appearance of the welded area satisfies a predetermined standard with higher precision.

Note that the quality determining section 51 and the display/output section 52 may be incorporated in either the dimension measuring device 20 or the shape defect determining device 40. Further, the dimension measuring device 20 and the shape defect determining device 40 may use common hardware.

However, it is preferable that the processor that executes the dimension measurement program and the processor that executes the program that determines the presence or absence of shape defects are provided separately. Furthermore, it is preferable that the dimension measurement program and the program for determining the presence or absence of shape defects, etc., can be executed independently of each other. By doing so, the dimension measurement of the weld bead 110 and the determination of the presence or absence of a shape defect in the welded part can be performed simultaneously in parallel, and the time required to judge the appearance of the welded part can be shortened.

(Other embodiments)
In this specification, the dimension measurement of the weld bead 110 has been described as an example, but the measurement target is not particularly limited to this. By using the dimension measuring device 20 shown in FIGS. 1 and 2 and executing the dimension measuring procedure shown in Embodiment 1, it is possible to measure the dimensions of a region with a closed outline, an article, or the like.

Moreover, the visual inspection device 30 may be configured with a jig (not shown) that can be handled by an operator and a three-dimensional shape measurement sensor 31 held by the jig. In this case, the operator holds a jig and manually moves the three-dimensional shape measurement sensor 31 along the weld bead 110 to obtain shape data. In this case as well, by using the dimension measuring device 20 shown in FIGS. 1 and 2 and executing the dimension measuring procedure shown in Embodiment 1, the dimensions of the weld bead 110 can be automatically, simply, and more precisely measured. be able to. In this case, the distance between the three-dimensional shape measurement sensor 31 and the base material 100 or weld bead 110 may vary. Therefore, it is better to perform the above-described processing of the shape data into a bisected state (steps S12 and S13 in FIG. 6) based on the color information included in the pixel data.

Note that the position information included in the shape data is not limited to three-dimensional coordinate information, but may be two-dimensional coordinate information. If the position information is three-dimensional coordinate information, it is also possible to obtain the height and leg length of the weld bead 110 by specifying the formation area of the weld bead 110 and taking the difference in height from the surface of the base material 100. It is.

Further, the dimension measuring device 20 may include a shape data processing section 41 shown in FIG. 17 that has a function of removing noise from acquired shape data. Alternatively, the function of the shape data processing section 41 may be incorporated into the data acquisition section 1.

The weld bead size measuring device of the present disclosure is industrially very useful because it can automatically and easily measure the weld bead size even when the base material and the weld bead have arbitrary shapes.

1 Data acquisition unit 2 Bead area extraction unit 3 Contour extraction unit 4 Smoothing processing unit 5 End point acquisition unit 6 Reference line provision unit 7 Dimension calculation unit 8 Storage unit 9 Display unit 10 Learning data generation unit 20 Dimension measurement device 21 Input port 22 CPU
23 GPU
24 RAM/ROM
25 Output port 26 Data bus 27 Display 30 Visual inspection device 31 Three-dimensional shape measurement sensor 32 Robot 40 Shape defect determination device 41 Shape data processing section 42 Defect determination section 43 Storage section 50 Acceptability determination device 51 Acceptability determination section 52 Display/output section 100 Base metal 110 Weld bead

Claims (16)

A dimension measuring device for measuring the dimensions of a weld bead formed on a base material,
a data acquisition unit that acquires shape data of the base material on which the weld bead is formed;
a bead area extraction unit that extracts the weld bead formation area from the shape data as bead area data;
a contour extraction unit that extracts the contour of the weld bead as contour data based on the bead area data;
a smoothing processing unit that performs smoothing processing on the contour data;
an end point obtaining unit that obtains end points of the weld bead based on the smoothed contour data;
A first reference line for measuring the length of the weld bead and a plurality of second reference lines for measuring the width of the weld bead, starting from the end point, are added to the contour data. A granting section;
A dimension in which the length of the weld bead is calculated based on the first reference line, and the width of the weld bead is calculated at a plurality of locations along the first reference line based on a plurality of second reference lines. comprising at least a calculation unit;
The reference line providing unit is configured to arrange the first reference line and the plurality of second reference lines so that the first reference line and the plurality of second reference lines intersect each other within a predetermined angular range. A dimension measuring device characterized in that a reference line is added to the contour data. The dimension measuring device according to claim 1,
In the reference line providing section, the second reference line is located at a midpoint between the first reference line and the plurality of second reference lines along the contour of at least one side of the weld bead and at a first distance. The dimension is given by sequentially moving two points on the contour on both sides of the weld bead where the second reference line intersects so that the positions intersect within a predetermined angular range. measuring device. The dimension measuring device according to claim 1,
When the end point acquisition unit acquires a plurality of end points,
The reference line providing unit applies the first reference line to the contour data using each of the plurality of end points as a starting point, and the plurality of reference lines corresponding to each of the provided plurality of first reference lines. adding a second reference line to the contour data;
The dimension calculation unit calculates the length of the weld bead based on each of the plurality of first reference lines, and calculates the maximum value of the plurality of calculated weld bead lengths. The width of the weld bead is calculated at a plurality of locations along the first reference line based on the plurality of second reference lines corresponding to the first reference line having the fixed length. A dimension measuring device characterized by: The dimension measuring device according to claim 1,
The reference line providing section is
Initial setting processing for setting the end point as an end point A ₀ and adding a virtual point B ₀ and a virtual point C ₀ on the end point A ₀ , respectively;
A virtual point _B (i ₊₁₎ is a point that advances from the virtual point B i (i is an integer of 0 or 1 or more) along the contour by a first distance in a first direction that is a direction away from the end point A ₀ . ,
A _virtual point _{C (i+1 )} and
further performing at least a first loop process of setting the midpoint between the virtual point B _(i+1) and the virtual point C _(i+1) to the midpoint A _(i+1) ;
When a line segment connecting the virtual point B _(i+1) and the virtual point C _(i+1) intersects a line segment connecting the midpoint A _i and the midpoint A _(i+1) within a predetermined angular range. teeth,
The reference line providing unit determines the distance of a route from the end point A ₀ to the virtual point B _i via the virtual point B _k (k is an integer, 1≦k<i), and the end point A _{0 .} The first loop process is executed one or more times until the sum of the distance from the path from the virtual point C _k to the virtual point C _i exceeds the length of the contour, and A dimension measuring device characterized in that a first reference line and a plurality of the second reference lines are added to the contour data. The dimension measuring device according to claim 4,
A line segment connecting the virtual point B _(i+1) and the virtual point C _(i+1) intersects a line segment connecting the midpoint A _i and the midpoint A _(i+1) beyond the predetermined angular range. If you do,
The reference line providing section is
One of the virtual point B _(i+1) and the virtual point C _(i+1) is advanced by the first distance along the contour and in a direction away from the end point _A0 , and the point after the movement is set as a virtual point. Set B _(i+2) or virtual point C _(i+2) ,
The intersection angle of the line segment connecting the virtual point B _(i+2) and the virtual point C _(i+1) , the line segment connecting the midpoint A _i and the midpoint A _(i+1) , and the virtual point B _{(i+1) )} and the virtual point C _(i+2) and the intersection angle of the line segment connecting the midpoint A _i and the midpoint A _(i+1) are compared, and whether or not it is within the predetermined angle range. Select the virtual point B _(i+2) or the virtual point C _(i+2) that has a close intersection angle,
When the selected virtual point is the virtual point B _(i+2) , the midpoint A _(i+1) is corrected to the midpoint between the virtual point B _(i+2) and the virtual point C _(i+1) , and the virtual point Point B _(i+2) is newly set as the virtual point B _(i+1) ,
If the selected virtual point is the virtual point C _(i+2) , the midpoint A _(i+1) is corrected to the midpoint between the virtual point B _(i+1) and the virtual point C _(i+2) , and the virtual point Execute a second loop process to newly set point C _(i+2) to the virtual point C _(i+1) ,
A line segment connecting the virtual point B _(i+1) and the virtual point C _(i+1) intersects a line segment connecting the midpoint A _i and the midpoint A _(i+1) within the predetermined angular range. A dimension measuring device characterized in that the second loop process is repeatedly executed until the end of the process. The dimension measuring device according to claim 4,
The reference line providing section removes a predetermined length from each of both ends of the first reference line to obtain a temporary reference line, and creates a straight line according to the slope of each of the vicinity of both ends of the temporary reference line. extends from each end of the temporary reference line until it intersects the contour,
Furthermore, the reference line providing unit adds a new line to the outline data as the first reference line, which is a combination of the temporary reference line and straight lines each extending from both ends of the temporary reference line,
The dimension calculation unit calculates the length of the newly set first reference line as the length of the weld bead, and calculates a line connecting the virtual point B _(i+1) and the virtual point C _(i+1). A dimension measuring device characterized in that the length of the weld bead is calculated as the width of the weld bead. The dimension measuring device according to any one of claims 1 to 6,
The shape data consists of a plurality of pixel data, and each of the plurality of pixel data includes at least color information or distance information,
The color information is information on color components included in the pixel data,
The distance information is a distance between the surface position of the base material or the weld bead included in the pixel data and a three-dimensional shape measurement sensor that acquires the shape data,
The bead area extraction unit extracts the bead area data from the shape data based on the color information or the distance information and using a predetermined algorithm whose learning is reinforced by learning data annotated in advance. A dimension measuring device characterized by: A method for measuring dimensions of a weld bead formed in a base material, the method comprising:
a first step of acquiring shape data of the base material on which the weld bead is formed;
a second step of extracting the weld bead formation area from the shape data as bead area data;
a third step of extracting the outline of the weld bead as outline data based on the bead area data;
a fourth step of performing smoothing processing on the contour data;
a fifth step of obtaining end points of the weld bead based on the smoothed contour data;
a sixth reference line for measuring the length of the weld bead and a plurality of second reference lines for measuring the width of the weld bead, starting from the end point, to the contour data; step and
A length of the weld bead is calculated based on the first reference line, and a width of the weld bead is calculated at a plurality of locations along the first reference line based on a plurality of second reference lines. comprising at least 7 steps,
In the sixth step, the first reference line and the plurality of second reference lines are arranged such that the first reference line and the plurality of second reference lines intersect each other within a predetermined angular range. A dimension measuring method characterized in that a line is added to the contour data. In the dimension measuring method according to claim 8,
In the sixth step, the second reference line is located at a midpoint between the first reference line and the plurality of second reference lines along the contour of at least one side of the weld bead and at a first distance. The dimension measurement is performed by sequentially moving two points on the contour on both sides of the weld bead where the second reference line intersects so that the two points intersect within a predetermined angular range. Method. In the dimension measuring method according to claim 8,
If a plurality of the end points are obtained in the fifth step,
In the sixth step, the first reference line is added to the contour data using each of the plurality of end points as a starting point, and the plurality of first reference lines are added to the outline data in correspondence with each of the plurality of added first reference lines. 2. Adding two reference lines to the contour data,
In the seventh step, the length of the weld bead is calculated based on each of the plurality of first reference lines, and the maximum value among the plurality of calculated lengths of the weld bead is set as the length of the weld bead. The width of the weld bead is calculated at a plurality of locations along the first reference line based on the plurality of second reference lines corresponding to the first reference line having the fixed length. A dimension measurement method characterized by: In the dimension measuring method according to claim 9,
The sixth step is
setting the end point as an end point A ₀ , and providing a virtual point B ₀ and a virtual point C ₀ on the end point A ₀ , respectively;
At least including a first loop process,
The first loop processing is
A point that advances from the virtual point B _i (i is an integer of 0 or 1 or more) along the contour by a first distance in a first direction that is a direction away from the end point A ₀ is defined as a virtual point B _(i+1) . the step of
A _virtual point _{C (i+1 )} , and
at least the step of setting a midpoint between the virtual point B _(i+1) and the virtual point C _(i+1) to a midpoint A _(i+1) ,
In the sixth step,
When a line segment connecting the virtual point B _(i+1) and the virtual point C _(i+1) intersects a line segment connecting the midpoint A _i and the midpoint A _(i+1) within a predetermined angular range. teeth,
The distance of the path from the end point A ₀ to the virtual point B _i via the virtual point B _k (k is an integer, 1≦k<i), and from the end point A ₀ to the virtual point C _k The first loop process is performed once or multiple times until the sum of the distance of the path to the virtual point C _i exceeds the length of the contour, and the first reference line and the plurality of A dimension measuring method characterized in that the second reference line is added to the contour data. In the dimension measuring method according to claim 11,
A line segment connecting the virtual point B _(i+1) and the virtual point C _(i+1) intersects a line segment connecting the midpoint A _i and the midpoint A _(i+1) beyond the predetermined angular range. If you do,
The sixth step further includes a second loop process,
The second loop processing is
One of the virtual point B _(i+1) and the virtual point C _(i+1) is advanced by the first distance along the contour and in a direction away from the end point _A0 , and the point after the movement is set as a virtual point. a step of setting B _(i+2) or virtual point C _(i+2) ;
The intersection angle of the line segment connecting the virtual point B _(i+2) and the virtual point C _(i+1) , the line segment connecting the midpoint A _i and the midpoint A _(i+1) , and the virtual point B _{(i+1) )} and the virtual point C _(i+2) and the intersection angle of the line segment connecting the midpoint A _i and the midpoint A _(i+1) are compared, and whether or not it is within the predetermined angle range. selecting the virtual point B _(i+2) or the virtual point C _(i+2) that has a close intersection angle;
When the selected virtual point is the virtual point B _(i+2) , the midpoint A _(i+1) is corrected to the midpoint between the virtual point B _(i+2) and the virtual point C _(i+1) , and the virtual point a step of newly setting point B _(i+2) as the virtual point B _(i+1) ;
If the selected virtual point is the virtual point C _(i+2) , the midpoint A _(i+1) is corrected to the midpoint between the virtual point B _(i+1) and the virtual point C _(i+2) , and the virtual point at least the step of newly setting point C _(i+2) as the virtual point C _(i+1) ,
A line segment connecting the virtual point B _(i+1) and the virtual point C _(i+1) intersects a line segment connecting the midpoint A _i and the midpoint A _(i+1) within the predetermined angular range. A method for measuring dimensions, characterized in that the second loop process is repeatedly executed until the end of the process. In the dimension measuring method according to claim 10,
The sixth step further includes:
A predetermined length is removed from each end of the first reference line to obtain a temporary reference line, and a straight line corresponding to the respective slopes near both ends of the temporary reference line is created at both ends of the temporary reference line. extending from each of the contours until intersecting the contour;
the step of adding a new line to the contour data as the first reference line, a line that is a composite of the temporary reference line and straight lines extended from both ends of the temporary reference line,
In the seventh step, the length of the newly set first reference line is calculated as the length of the weld bead, and a line connecting the virtual point B _(i+1) and the virtual point C _(i+1) is calculated. A method for measuring dimensions, characterized in that the length of the weld bead is calculated as the width of the weld bead. In the dimension measuring method according to any one of claims 8 to 13,
The shape data consists of a plurality of pixel data, and each of the plurality of pixel data includes at least color information or distance information,
The color information is information on color components included in the pixel data,
The distance information is a distance between the surface position of the base material or the weld bead included in the pixel data and a three-dimensional shape measurement sensor that acquires the shape data,
In the second step, the bead area data is extracted from the shape data based on the color information or the distance information and using a predetermined algorithm whose learning is reinforced by learning data annotated in advance. A dimension measurement method characterized by: A dimension measurement program for causing one or more processors to execute the dimension measurement method according to claim 8. A dimension measuring device according to claim 1;
At least a shape defect determination device for determining the presence or absence of a defect in a welding location including the weld bead,
Judgment of quality of a welded part, characterized in that it is determined whether the appearance of the welded part satisfies a predetermined standard based on the measurement result of the dimension measuring device and the judgment result of the shape defect determination device. Device.