JP7482686B2 - Measurement device and measurement program - Google Patents

Description

The present invention relates to a measurement device and a measurement program.

Patent Document 1 discloses that steel plates 1 and 2 of various steel types and thicknesses are spot welded under various conditions, and the dimensions of each part of the nugget, such as the nugget diameter dN, are measured by image analysis from an image of a cross section of the nugget in the plate thickness direction at the spot welded part 3. This allows the dimensions of the nugget to be measured automatically, eliminating the need for an operator to measure the dimensions of the nugget with a ruler or the like.

JP 2015-033706 A (for example, paragraphs 0029-0035, tables 2 and 3)

However, Patent Document 1 does not disclose a specific method for measuring the dimensions of each part of the nugget from the above image using image analysis. Therefore, there is a risk that the dimensions of each part of the nugget cannot be accurately measured, for example, by measuring the dimensions from an object other than the nugget in the above image, or by measuring the dimensions at a position on the nugget that is shifted from the position at which the dimensions should be measured.

The present invention has been made to solve the above-mentioned problems, and aims to provide a measurement device and a measurement program that can accurately measure the dimensions of each part of a nugget in the cross section of a test piece formed by spot welding.

In order to achieve this object, the measuring device of the present invention is a device equipped with a measuring means for measuring the dimensions of each part in the cross section of a test piece from an image of the cross section of the test piece formed by overlapping flat steel plates and spot welding the plates, and includes an input image acquiring means for acquiring an input image which is an image of the cross section of the test piece, a nugget detection means for distinguishing and detecting nuggets in spot welding from the input image acquired by the input image acquiring means, a plate thickness acquiring means for acquiring the plate thickness of the inputted steel plate, a plate thickness pixel acquiring means for acquiring the number of pixels of the plate thickness acquired by the plate thickness acquiring means, and a plate thickness pixel acquiring means for acquiring the number of pixels of the plate thickness acquired by the input image acquiring means. The test piece is provided with an end acquisition means for acquiring the coordinates of the positions of the upper and lower ends of the test piece from an input image, and a plate boundary detection means for detecting a position obtained by moving the coordinates of the positions of the upper and lower ends of the test piece acquired by the end acquisition means in the input image acquired by the input image acquisition means by the number of pixels of the plate thickness acquired by the plate thickness pixel acquisition means as a plate boundary, which is the boundary between overlapping steel plates in the test piece, and the measurement means measures the dimensions of each part of the nugget in the input image from the nugget detected by the nugget detection means, based on the plate boundary detected by the plate boundary detection means.

The measurement program of the present invention causes a computer to execute the following steps: an input image acquisition step for acquiring an input image, which is an image of a cross section of a test piece formed by overlapping flat steel plates and spot welding them; a nugget detection step for distinguishing and detecting nuggets in spot welding from the input image acquired in the input image acquisition step; a plate thickness acquisition step for acquiring the plate thickness of the input steel plate; a plate thickness pixel acquisition step for acquiring the number of pixels of the plate thickness acquired in the plate thickness input step; an end acquisition step for acquiring coordinates of the positions of the upper and lower ends of the test piece from the input image acquired in the input image acquisition step; a plate boundary detection step for detecting positions obtained by moving the coordinates of the positions of the upper and lower ends of the test piece acquired in the end acquisition step in the input image acquired in the input image acquisition step by the number of pixels of the plate thickness acquired in the plate thickness pixel acquisition step as a plate boundary that is the boundary between the overlapping steel plates in the test piece; and a measurement step for measuring the dimensions of each part of the nugget in the input image from the nugget detected in the nugget detection step, based on the plate boundary detected in the plate boundary detection step.

According to the measuring device of claim 1, an input image is acquired, which is an image of a test piece formed by overlapping plate-shaped steel sheets and spot welding, and a nugget in spot welding is detected from the acquired input image, and further, a plate boundary between the overlapping steel sheets in the test piece is detected. Then, the dimensions of each part of the detected nugget are measured based on the plate boundary. Here, the nugget is formed so as to straddle the boundary between the overlapping steel sheets, i.e., the plate boundary, and therefore the plate boundary is regarded as a position related to the nugget. Therefore, by measuring the dimensions of each part of the nugget based on the plate boundary, it is possible to accurately measure the dimensions of each part of the nugget. In addition, the plate boundary is detected from a position obtained by moving the coordinates of the positions of the upper and lower ends of the test piece acquired from the input image by the number of pixels of the input plate thickness. This also has the effect of accurately detecting the plate boundary.

The measuring device according to claim 2 has the following effect in addition to the effect of the measuring device according to claim 1. The nugget diameter is calculated from the distance between the intersection of the plate boundary and the nugget. This has the effect of allowing an operator to easily and accurately measure the nugget diameter, which is the standard for the size of the nugget, without actually measuring the test piece with a ruler or the like.

According to the measuring device of claim 3, in addition to the effects of the measuring device of claim 1 or 2, the following effect is achieved. A plurality of parallel lines that are straight lines parallel to the plate boundary are calculated, and among the parallel lines, a parallel line is identified in which the distance between the intersection point of the parallel line and the detected nugget is a predetermined percentage of the nugget diameter, and the penetration depth, which is the distance between the identified parallel line and the plate boundary, is measured. This has the effect of allowing an operator to easily and accurately measure the penetration depth, which is the standard for the size of the nugget, without actually measuring the test piece with a ruler or the like.

According to the measuring device of claim 4, in addition to the effect of the measuring device of any one of claims 1 to 3, the following effect is achieved. The input image includes a reference object that serves as a dimensional reference along with the test piece, a conversion coefficient that represents the relationship between a predetermined dimension of the reference object and the number of pixels of the input image that corresponds to that dimension is obtained, and the dimensions of the nugget are measured based on the conversion coefficient. By using a conversion coefficient according to the dimensions of the reference object, the actual dimensions of the nugget can be measured easily and accurately.

According to the measuring device of claim 5, in addition to the effect of the measuring device of any one of claims 1 to 4, the following effect is achieved. The nugget detection means outputs a nugget area image, which is an image in which the nugget in the input image and objects other than the nugget are drawn in different colors, and the dimensions of each part of the nugget are measured from the nugget area image. Since the nugget and objects other than the nugget are drawn separately in the nugget area image, the nugget can be reliably detected from the nugget area image, and there is an effect that the dimensions of each part of the nugget can be accurately measured.

According to the measuring device described in claim 6, in addition to the effect of the measuring device described in any one of claims 1 to 5, the following effect is achieved. Since the nugget in the spot welding is detected separately from the input image by the artificial intelligence, the effect is that the nugget can be suitably detected from the input image that includes the nugget and objects other than the nugget.

The measurement program described in claim 7 provides the same effect as the measurement device described in claim 1. Also, the measurement program described in claim 8 provides the same effect as the measurement device described in claim 6.

FIG. 2 is a schematic diagram showing a measuring device. 1A is a diagram showing an input image, FIG. 1B is a diagram showing a nugget region image, and FIG. 1C is a diagram showing the dimensions of each portion of the nugget. 1A is a block diagram showing the electrical configuration of the measurement device, FIG. 1B is a schematic diagram showing learning data, FIG. 1C is a schematic diagram showing a result table, and FIG. 1D is a schematic diagram showing a RAM. 1A is a flowchart of a main process, and FIG. 1B is a flowchart of a learning process. 13 is a flowchart of a determination process. 13 is a flowchart of a detection process. 1A is a diagram showing the display content of the display device before nugget detection starts, and FIG. 1B is a diagram showing the display content of the display device after nugget detection starts.

The following describes a preferred embodiment of the present invention with reference to the accompanying drawings. FIG. 1 is a schematic diagram showing a measuring device 1. The measuring device 1 is a device that acquires an image of a cross section of a test piece T formed by spot welding, measures the dimensions of each part of the test piece based on the image, and determines whether the result of spot welding conforms to a predetermined standard (e.g., JRIS W 0161) based on the measured dimensions.

The measuring device 1 is provided with an input device 2 for inputting instructions from a user, a display device 3, and a camera C for acquiring an input image P (see FIG. 2(a)), which is an image of a test piece T. The display device 3 is an output device for displaying the input image P and the dimensions of each part of the test piece T.

The test piece T is formed by overlapping and spot welding at least two steel plates Sp, and is a component for determining whether the spot-welded result complies with a specified standard. The spot-welded portion of the test piece T is cut in the plate thickness direction, and the cut surface is positioned so that it faces the camera C.

A ruler J is placed next to the test piece T. The ruler J is a reference object with scale lines engraved at a predetermined interval (e.g., 1 mm). The ruler J is placed so that its longitudinal side is parallel to the longitudinal side of the test piece T.

The imaging range CA of the camera C is set so that the test piece T and the ruler J are included in the input image P, and the horizontal direction of the input image P coincides with the longitudinal direction of the test piece T (i.e., the two steel plates Sp) and the ruler J. Next, the input image P acquired by the camera C will be described with reference to FIG. 2(a).

Figure 2(a) is a diagram showing an input image P. The input image P includes an image of a test piece T and an image of a ruler J. The nugget N in the test piece T is formed by melting the overlapping portions of two or more steel plates Sp (two in this embodiment) by spot welding. Hereinafter, of the multiple steel plates Sp in the test piece T, the uppermost steel plate Sp in the input image P is referred to as the "upper plate" and the lowermost steel plate Sp is referred to as the "lower plate."

The nugget N is molten metal formed by spot welding, and is formed by melting the overlapping portion of the upper and lower plates so that it straddles the plate boundary B, which is the boundary between the upper and lower plates and will be described later. Therefore, the outer shape of the nugget N is not necessarily a simple straight line or circle, but is an indefinite shape. In addition, the color of the nugget N and the surrounding steel plate Sp captured in the input image P varies depending on the color temperature of the lighting where the camera C is placed, the degree of melting due to spot welding, and even the type of steel of the steel plate Sp. For these reasons, it is difficult to accurately detect the nugget N from the input image P using conventional image processing.

Therefore, in this embodiment, a nugget determination model M, which is a learning model (artificial intelligence) that has learned images of the cross section of a test piece T containing a large number of nuggets N and images of nuggets N to be detected from the corresponding images, is used to create a nugget area image Pn that distinguishes and detects the nugget N from the input image P, and the dimensions of the nugget N are measured based on the nugget area image Pn. The detection and measurement of the dimensions of the nugget N using the nugget determination model M will be described with reference to Figures 2(b) and (c).

Figure 2(b) is a diagram showing a nugget region image Pn. The nugget region image Pn in Figure 2(b) is an image that is created by inputting the input image P in Figure 2(a) into the nugget determination model M, and that depicts the nugget N and objects other than the nugget N in a distinguished manner.

The nugget determination model M is a learning model that is trained to combine an image of a cross section of a test piece T that contains a large number of nuggets N with a teacher image that is an image of the nuggets N to be detected from that image. In this embodiment, a learning model called "U-Net" is used as the nugget determination model M, but other learning models may also be used.

The nugget determination model M is trained on an image of a cross section of a test piece T containing a large number of nuggets N, and a teacher image of the nugget N to be detected from that image. Therefore, based on the nugget determination model M, even if the shape of the nugget N in the input image P and the color of the nugget N and the steel plate Sp vary, the nugget N can be detected from the input image P and distinguished from objects other than the nugget N.

In the nugget region image Pn created from the input image P by such a nugget determination model M, the nugget N in the input image P detected by the nugget determination model M is depicted in a different manner from objects other than the nugget N, as shown in FIG. 2(b). Specifically, the nugget N portion and the portion other than the nugget N are depicted in different colors (for example, the nugget N portion is depicted in white, and the portion other than the nugget N is depicted in black) with the contour Ne, which is the outline of the nugget N, as the boundary.

This allows the nugget N to be clearly distinguished from objects other than the nugget N in the nugget area image Pn, so that the position of the nugget N can be accurately and easily detected from the nugget area image Pn. Based on the position of the nugget N from the nugget area image Pn detected in this manner, the dimensions of each part of the nugget N are measured. Here, the dimensions of each part of the nugget N measured by the measuring device 1 will be described with reference to FIG. 2(c).

Figure 2(c) is a diagram showing the dimensions of each part of the nugget N. Figure 2(c) shows an image P' in which the dimensions of each part of the nugget N are shown in the input image P of Figure 2(a). In this embodiment, the dimensions of the nugget N are measured as the nugget diameter Nhd, the upper plate penetration depth Nup, and the lower plate penetration depth Ndd. Specifically, the nugget diameter Nhd is the distance between intersections B1 and B2 in image P', which are the intersections between the outline Ne of the nugget N and the plate boundary B, which is the boundary between two overlapping steel plates Sp. The nugget diameter Nhd is used as a reference for the size of the nugget N in the plate boundary B direction, i.e., in the longitudinal direction of the steel plate Sp.

As described above, in the input image P, the overlapping direction of the steel plate Sp, i.e., the longitudinal direction of the steel plate Sp, coincides with the horizontal direction of the input image P. Therefore, hereinafter, the length direction of the steel plate Sp in the input image P will be referred to as the "horizontal direction" and the thickness direction will be referred to as the "vertical direction."

The upper plate penetration depth Nup is the distance between the plate boundary B and the parallel line that is parallel to the plate boundary B, and the distance between the intersection points of the parallel line with the contour Ne of the nugget N is 80% of the nugget diameter Nhd, and the parallel line above the plate boundary B (i.e., the parallel line by intersection points B3 and B4 in FIG. 2(c)). The lower plate penetration depth Ndd is the distance between the parallel line below the plate boundary B (i.e., the parallel line by intersection points B5 and B6 in FIG. 2(c)) and the plate boundary B, and the parallel line that is parallel to the plate boundary B and the contour Ne of the nugget N is 80% of the nugget diameter Nhd. The upper plate penetration depth Nup and the lower plate penetration depth Ndd are used as the standard for the size of the nugget N in the direction perpendicular to the plate boundary B and the nugget diameter Nhd, i.e., the vertical direction.

In this embodiment, the nugget diameter Nhd, upper plate penetration depth Nup, and lower plate penetration depth Ndd are measured based on a nugget area image Pn (see FIG. 2(b)) created from the input image P by a nugget determination model M. In the nugget area image Pn, the nugget N is depicted separately from objects other than the nugget N, so that the nugget N can be reliably detected and these values can be measured accurately.

Next, the electrical configuration of the measuring device 1 will be described with reference to FIG. 3. FIG. 3(a) is a block diagram showing the electrical configuration of the measuring device 1. As shown in FIG. 3(a), the measuring device 1 has a CPU 10, a hard disk drive (hereinafter referred to as "HDD") 11, and a RAM 12, which are each connected to an input/output port 14 via a bus line 13. The input device 2, display device 3, and camera C described above are further connected to the input/output port 14.

The CPU 10 is a computing device that controls each part connected by the bus line 13. The HDD 11 is a rewritable non-volatile storage device, and stores the measurement program 11a, the nugget judgment model 11b in which the above-mentioned nugget judgment model M is stored, the learning data 11c, the result table 11d, and the judgment criterion data 11e. When the measurement program 11a is executed by the CPU 10, the main process of FIG. 4(a) is executed. The learning data 11c stores images for learning the nugget judgment model M. The learning data 11c will be described with reference to FIG. 3(b).

Figure 3(b) is a schematic diagram of the learning data 11c. In the learning data 11c, an input image, which is an image of a test piece T including a nugget N, and a teacher image, which is an image of the nugget N to be detected from the input image, are stored in association with each other. Specifically, the teacher image is an image created by the user based on the input image, in which the nugget N to be detected from the input image and objects other than the nugget N are drawn in different colors, i.e., in a manner similar to the nugget area image Pn described in Figure 2(b), are stored.

The learning data 11c stores a large number of combinations (e.g., 10,000 pairs) of input images with various shapes of the nugget N and colors of the nugget N and the steel plate Sp, and teacher images corresponding to the input images. In the learning process described below, the nugget determination model M is trained using the input images and teacher images stored in the learning data 11c.

Returning to FIG. 3(a), the result table 11d is a data table in which the measured nugget diameter Nhd, the upper plate penetration depth Nup, and the lower plate penetration depth Ndd are stored. The result table 11d will be explained with reference to FIG. 3(c).

Figure 3 (c) is a schematic diagram of the result table 11d. The result table 11d stores the nugget diameter Nhd measured from the input image P, the upper plate penetration depth Nup, the lower plate penetration depth Ndd, the plate thickness of the upper plate input from the input device 2, the plate thickness of the lower plate input from the input device 2, the input image P used for the measurement, the result image Q (see Figure 7 (b)) depicting the nugget diameter Nhd, the upper plate penetration depth Nup, and the lower plate penetration depth Ndd measured for the input image P, and the judgment result ("OK" or "NG") of whether the spot welding complies with a predetermined standard based on the nugget diameter Nhd, the upper plate penetration depth Nup, and the lower plate penetration depth Ndd, etc., in association with each other.

Return to FIG. 3(a). The criterion data 11e stores reference values for satisfying the predetermined standards for the nugget diameter Nhd, the upper plate penetration depth Nup, and the lower plate penetration depth Ndd. Since the nugget diameter Nhd, the upper plate penetration depth Nup, and the lower plate penetration depth Ndd that satisfy the predetermined standards differ depending on the thickness of the upper or lower plate of the steel plate Sp, the respective reference values according to the thickness of the upper or lower plate of the steel plate Sp are stored in the criterion data 11e. Note that these reference values are not limited to those stored in the criterion data 11e, and may be calculated from the thickness of the upper or lower plate of the steel plate Sp. In this case, for example, the upper plate penetration depth Nup is specified to be 20 to 90% of the thickness of the upper plate of the steel plate Sp, so the reference value of the upper plate penetration depth Nup can be calculated based on this specification.

RAM 12 is a memory for rewritably storing various work data, flags, etc. when CPU 10 executes measurement program 11a. RAM 12 will be described with reference to FIG. 3(d).

3(d) is a schematic diagram of the RAM 12. As shown in FIG. 3(d), the RAM 12 includes an input image memory 12a in which an input image P as shown in FIG. 2(a) is stored, a nugget area image memory 12b in which a nugget area image Pn as shown in FIG. 2(b) is stored, an upper plate thickness memory 12c in which the actual dimension of the thickness of the upper plate is stored, a lower plate thickness memory 12d in which the actual dimension of the thickness of the lower plate is stored, an upper plate thickness pixel memory 12e in which the number of pixels of the thickness of the upper plate is stored, and a lower plate thickness pixel memory 12f in which the number of pixels of the thickness of the lower plate is stored. There are provided a plate thickness pixel memory 12f, a conversion coefficient memory 12g, a nugget position memory 12h in which the position of the nugget N is stored, a nugget diameter memory 12i in which the above-mentioned nugget diameter Nhd is stored, a plate boundary position memory 12j in which the position of the plate boundary B is stored, an upper plate penetration depth memory 12k in which the upper plate penetration depth Nup is stored, a lower plate penetration depth memory 12m in which the lower plate penetration depth Ndd is stored, and a result image memory 12n in which a result image Q (see FIG. 7(b)) is stored.

The conversion coefficient memory 12g is a memory that stores the conversion coefficients calculated based on the scale lines of the ruler J described above. In this embodiment, the conversion coefficient is the actual distance between adjacent scale lines divided by the number of pixels between the adjacent scale lines in the input image P.

Next, the main processing executed by the CPU 10 of the measuring device 1 will be described with reference to Figures 4 to 7. Figure 4(a) is a flowchart of the main processing. The main processing is processing that is executed after the measuring device 1 is powered on.

The main process first checks the operation mode of the measuring device 1 (S1). In this embodiment, the operation modes of the measuring device 1 include a "learning mode" in which the nugget determination model M is trained based on the learning data 11c (see FIG. 3(b)), and a "determination mode" in which the nugget diameter Nhd, upper plate penetration depth Nup, and lower plate penetration depth Ndd are measured from the input image P acquired from the camera C, and it is determined whether they conform to a predetermined standard. In the process of S1, the operation mode specified by the user via the input device 2 is confirmed.

In the process of S1, if the operation mode is the learning mode (S1: learning mode), the learning process (S2) is executed. Here, the learning process will be explained with reference to FIG. 4(b).

Figure 4 (b) is a flowchart of the learning process. In the learning process, first, the nugget determination model M of the nugget determination model 11b is adjusted by machine learning using the learning data 11c (S10). Specifically, the nugget determination model M is provided with parameters (coefficients) for setting the degree of detection of the nugget N when outputting the nugget area image Pn from the input image, and the parameters are configured to be adjustable. In the process of S10, the nugget area image Pn created using the nugget determination model M from all input images in the learning data 11c is compared with the teacher image of the corresponding learning data 11c, and the parameters of the nugget determination model M are automatically adjusted so that the difference between them is reduced.

After the process of S10, the nugget area image Pn created from all input images in the learning data 11c using the nugget determination model M of the nugget determination model 11b is compared again with the teacher image of the corresponding learning data 11c to confirm whether the difference between them is sufficiently small (S11). If the difference is large in the process of S11 (S11: No), the process of S10 is repeated. By repeating the adjustment of the parameters of the nugget determination model M in this way, the detection accuracy of the nugget N by the nugget determination model M is improved, so that the nugget N can be more suitably detected from the input image P. On the other hand, if the difference is sufficiently small in the process of S11 (S11: Yes), the learning process is terminated.

Returning to FIG. 4(a), in the process of S1, if the operation mode is the judgment mode (S1: judgment mode), the judgment process (S3) is executed. The judgment process will be explained with reference to FIGS. 5 to 7.

Figure 5 is a flowchart of the determination process. In the determination process, first, an input image P is acquired from the camera C, and is stored in the input image memory 12a and displayed on the display device 3 (S20). Here, the display contents of the display device 3 will be described with reference to Figure 7 (a).

Figure 7(a) is a diagram showing the display contents of the display device 3 before detection of the nugget N begins. As shown in Figure 7(a), the display device 3 is provided with an input screen display area 3a, a result image display area 3b, and a numerical display area 3c in which measurement results such as the nugget diameter Nhd are displayed. The input screen display area 3a displays the input image P acquired in the process of S20 in Figure 5. The result image display area 3b displays a result image Q based on the input image P.

The numerical display area 3c includes an upper plate area 3d for inputting the thickness of the upper plate, a lower plate area 3e for inputting the thickness of the lower plate, a detection start button 3f for instructing the start of detection of the nugget N, a nugget diameter area 3g for displaying or inputting the nugget diameter Nhd, an upper plate penetration depth area 3h for displaying or inputting the upper plate penetration depth Nup, a lower plate penetration depth area 3i for displaying or inputting the lower plate penetration depth Ndd, a result area 3j for displaying the judgment result, a next load button 3k for instructing the acquisition of the next input image P, and an end button 3m for instructing the end of the judgment process of FIG. 5.

Of these, the nugget diameter area 3g, the upper plate penetration depth area 3h, and the lower plate penetration depth area 3i display the nugget diameter Nhd, the upper plate penetration depth Nup, and the lower plate penetration depth Ndd measured in the detection process described below, respectively, or display values input by the user via the input device 2.

In FIG. 7(a), before detection of the nugget N is performed, the nugget diameter area 3g, the upper plate penetration depth area 3h, and the lower plate penetration depth area 3i are displayed blank, and the result image display area 3b displays a blank image BL (represented by an "x" in FIG. 7(a)) instead of the result image Q. In addition, in the numerical value display area 3c, the detection start button 3f is displayed with a solid line indicating that it is operable, while the next load button 3k and the end button 3m are displayed with a dashed line indicating that they are not operable. Note that the end button 3m may be displayed with a solid line indicating that it is always operable.

Returning to FIG. 5, after the processing of S20, the values input by the user via the input device 2 into the upper plate area 3d and the lower plate area 3e (both see FIG. 7(a)) are stored in the upper plate thickness memory 12c and the lower plate thickness memory 12d, respectively (S21).

After the process of S21, it is confirmed whether the detection start button 3f has been operated (S22). In the process of S22, if the detection start button 3f has not been operated (S22: No), the process waits, and if the detection start button 3f has been operated (S22: Yes), the detection process (S23) is executed. The detection process of S23 will now be described with reference to FIG. 6.

Figure 6 is a flowchart of the detection process. The detection process is a process of detecting a nugget N from the input image P in the input image memory 12a, and measuring the nugget diameter Nhd, the upper plate penetration depth Nup, and the lower plate penetration depth Ndd based on the detected nugget N.

The detection process begins by inputting the input image P in the input image memory 12a into the nugget determination model M to create a nugget area image Pn, which is then stored in the nugget area image memory 12b (S40). After the process of S40, the position of the nugget N is obtained from the nugget area image Pn in the nugget area image memory 12b, and its coordinates are stored in the nugget position memory 12h (S41).

After the process of S41, in order to obtain the positions and dimensions of the test piece T and the ruler J in the input image P, first, a linear image in the horizontal direction is detected in the input image P (S42). The detection of a linear image in the horizontal direction is performed by known image processing (e.g., Hough transform).

After the process of S42, linear images in the vertical direction (the up and down direction in FIG. 2(a)) are detected in the input image P, and the scale lines of the ruler J are detected from the detected linear images in the vertical direction (S43). The detection of linear images in the vertical direction in the input image P is performed by known image processing (e.g., Hough transform) as in the process of S42. In this embodiment, of the detected linear images in the vertical direction, the linear images in the vertical direction near the bottom end of the input image P are the scale lines of the ruler J, and therefore such images are detected.

After the process of S43, a conversion coefficient is obtained from the actual distance and number of pixels between adjacent scale lines of ruler J detected in the process of S43, and is stored in conversion coefficient memory 12g (S44). Specifically, adjacent scale lines of ruler J detected in the process of S43 are obtained, and the number of pixels in the horizontal direction between adjacent scale lines is obtained. Then, the actual distance between adjacent scale lines (for example, 1 mm) is divided by the number of pixels in the horizontal direction between adjacent scale lines to obtain the obtained value, i.e., the conversion coefficient, and stored in conversion coefficient memory 12g.

After the process of S44, the top end of the upper plate and the bottom end of the lower plate are detected from the horizontally linear image of the input image P detected in the process of S42 (S45). To explain the process of S45 specifically using FIG. 2(c), the first linear image from the top end of the input image P is determined to be the top end of the upper plate, and the second horizontally linear image from the bottom end of the input image P is determined to be the bottom end of the lower plate.

Returning to FIG. 6, after the processing of S45, each value of the upper plate thickness memory 12c and the lower plate thickness memory 12d is converted to the number of pixels using the conversion coefficient of the conversion coefficient memory 12g, and stored in the upper plate thickness pixel memory 12e and the lower plate thickness pixel memory 12f (S46). Specifically, each value of the upper plate thickness memory 12c and the lower plate thickness memory 12d is divided by the conversion coefficient of the conversion coefficient memory 12g, and the value is stored in the upper plate thickness pixel memory 12e and the lower plate thickness pixel memory 12f, respectively.

After the process of S46, the coordinates of plate boundary B are calculated from the upper plate thickness pixel memory 12e and the lower plate thickness pixel memory 12f and the values of the upper end of the upper plate and the lower end of the lower plate converted in the process of S46, and are stored in plate boundary position memory 12j (S47). Specifically, the position moved downward from the coordinates of the upper end of the upper plate by the number of pixels in the upper plate thickness pixel memory 12e, and the position moved upward from the coordinates of the lower end of the lower plate by the number of pixels in the lower plate thickness pixel memory 12f are the positions of plate boundary B, and the coordinates of these positions are stored in plate boundary position memory 12j.

After the process of S47, the nugget diameter Nhd is measured from the coordinates of the nugget position memory 12h and the plate boundary position memory 12j and the conversion coefficient of the conversion coefficient memory 12g, and stored in the nugget diameter memory 12i (S48). Specifically, the nugget diameter Nhd is measured by acquiring the coordinates of the intersections B1 and B2 (see FIG. 2(c)) between the outline Ne of the nugget N in the nugget position memory 12h and the plate boundary B in the plate boundary position memory 12j, and the length between the acquired intersections B1 and B2 is measured as the nugget diameter Nhd. Furthermore, the measured nugget diameter Nhd is converted to the actual dimension of the nugget diameter Nhd by multiplying it by the conversion coefficient of the conversion coefficient memory 12g, and this dimension is stored in the nugget diameter memory 12i.

After the process of S48, multiple parallel straight lines that are parallel to the plate boundary B in the plate boundary position memory 12j are calculated (S49). After the process of S49, among the calculated parallel straight lines, a parallel straight line is identified whose distance between the intersection points with the contour Ne of the nugget N in the nugget position memory 12h is 80% of the nugget diameter Nhd in the nugget diameter memory 12i (S50).

After the process of S50, the upper plate penetration depth Nup and the lower plate penetration depth Ndd are measured from the distance between the identified parallel straight line and the plate boundary B in the plate boundary position memory 12j and the conversion coefficient in the conversion coefficient memory 12g, and are stored in the upper plate penetration depth memory 12k and the lower plate penetration depth memory 12m (S51). Specifically, of the parallel straight lines identified in the process of S50, the distance between the parallel straight line above the plate boundary B in the plate boundary position memory 12j (specifically, the parallel straight line passing through the intersections B3 and B4 in FIG. 2(c)) and the plate boundary B in the plate boundary position memory 12j is measured as the upper plate penetration depth Nup. Furthermore, the measured upper plate penetration depth Nup is converted into the actual dimension of the upper plate penetration depth Nup by multiplying it by the conversion coefficient in the conversion coefficient memory 12g, and this dimension is stored in the upper plate penetration depth memory 12k.

On the other hand, among the parallel straight lines identified in the processing of S50, the distance between the parallel straight line below the plate boundary B in the plate boundary position memory 12j (specifically, the parallel straight line passing through the intersection points B5 and B6 in FIG. 2(c)) and the plate boundary B in the plate boundary position memory 12j is measured as the lower plate penetration depth Ndd. Furthermore, the measured lower plate penetration depth Ndd is multiplied by the conversion coefficient in the conversion coefficient memory 12g to convert it into the actual dimension of the lower plate penetration depth Ndd, and this dimension is stored in the lower plate penetration depth memory 12m.

In the above detection process, first, the nugget N in the input image P is detected from the input image P in the input image memory 12a and the nugget determination model M. The plate boundary B is measured from the upper plate thickness memory 12c and the lower plate thickness memory 12d, and the values of the upper end of the upper plate and the lower end of the lower plate, and the nugget diameter Nhd, the upper plate penetration depth Nup, and the lower plate penetration depth Ndd are measured from the contour Ne of the detected nugget N and the plate boundary B. Here, as shown in FIG. 2(c), the nugget N is formed so as to straddle the boundary between the overlapped steel plates Sp and the steel plates Sp, i.e., the plate boundary B, so that the plate boundary B is regarded as a position related to the nugget N. By measuring the nugget diameter Nhd, the upper plate penetration depth Nup, and the lower plate penetration depth Ndd in the nugget N based on the plate boundary B, these dimensions can be measured accurately.

In addition, horizontally linear images such as the top end of the upper plate and the bottom end of the lower plate in the input image P, and vertically linear images such as the scale lines of a ruler J are detected by image processing. By detecting these linear images, which have simple shapes, by image processing, it is not necessary to prepare a learning model such as the nugget determination model M in advance and refer to the learning model when detecting linear images. This makes it possible to easily and quickly detect linear images.

The nugget determination model M for detecting nuggets N is configured as a learning model based on images of nuggets N of various shapes and colors and teacher images to be detected from those images. By inputting the input image P of the input image memory 12a to such a nugget determination model M, the nugget N can be detected from the input image P in a manner that distinguishes it from objects other than the nugget N, thereby suppressing erroneous detection of the nugget N.

In addition, when calculating the nugget diameter Nhd, the upper plate penetration depth Nup, and the lower plate penetration depth Ndd, the conversion coefficients in the conversion coefficient memory 12g based on the scale lines of the ruler J are used. This allows the actual dimensions of the nugget diameter Nhd, the upper plate penetration depth Nup, and the lower plate penetration depth Ndd, which are based on the number of pixels of the input image P, to be accurately and easily measured.

Returning to FIG. 5, after the detection process of S23, the input image P in the input image memory 12a is combined with the nugget N in the nugget position memory 12h, the positions of the upper end of the upper plate and the lower end of the lower plate calculated in the process of S45 of FIG. 6, and the plate boundary B in the plate boundary position memory 12j, and the dimensional information of the nugget diameter Nhd in the nugget diameter memory 12i, the upper plate penetration depth Nup in the upper plate penetration depth memory 12k, and the lower plate penetration depth Ndd in the lower plate penetration depth memory 12m to generate a result image Q (see FIG. 7(b)), which is then stored in the result image memory 12n (S24).

After the process of S24, the nugget diameter Nhd in the nugget diameter memory 12i, the upper plate penetration depth Nup in the upper plate penetration depth memory 12k, the lower plate penetration depth Ndd in the lower plate penetration depth memory 12m, and the result image Q in the result image memory 12n are displayed on the display device 3 (S25). The display contents of the display device 3 as a result of the process of S25 will be described with reference to FIG. 7(b).

Figure 7(b) is a diagram showing the display contents of the display device 3 after detection of the nugget N has started. As shown in Figure 7(b), after detection of the nugget N has started, the result image Q of the result image memory 12n is displayed in the result image display area 3b of the display device 3 by the process of S25 in Figure 5. By comparing the result image Q in the result image display area 3b with the input image P in the input screen display area 3a, the positions of the nugget N, the plate boundary B, etc., and the respective dimensional information of the nugget diameter Nhd, the upper plate penetration depth Nup, and the lower plate penetration depth Ndd can be visually grasped.

In addition, by processing S25 in FIG. 5, the nugget diameter area 3g in the numerical display area 3c displays the value of the nugget diameter Nhd in the nugget diameter memory 12i, the upper plate penetration depth area 3h displays the value of the upper plate penetration depth Nup in the upper plate penetration depth memory 12k, and the lower plate penetration depth area 3i displays the value of the lower plate penetration depth Ndd in the lower plate penetration depth memory 12m.

Therefore, the worker can check the specific values of the nugget diameter Nhd, the upper plate penetration depth Nup, and the lower plate penetration depth Ndd in the nugget diameter area 3g, the upper plate penetration depth area 3h, and the lower plate penetration depth area 3i, while checking the overview of the detection result of the nugget N in the result image Q in the result image display area 3b. This allows the worker to efficiently verify the nugget diameter Nhd, the upper plate penetration depth Nup, and the lower plate penetration depth Ndd.

In addition, in FIG. 7(b), since the detection of the nugget N has already been performed, the detection start button 3f is displayed with a dashed line indicating that it is inoperable to prevent duplication of the detection of the nugget N. On the other hand, since the detection of the nugget N has already been performed and the nugget diameter Nhd, etc. has been measured, the next load button 3k and the end button 3m are displayed with a solid line indicating that they are operable.

Return to FIG. 5. After the process of S25, it is determined whether each value in the upper plate thickness memory 12c, the lower plate thickness memory 12d, the nugget diameter memory 12i, the upper plate penetration depth memory 12k, and the lower plate penetration depth memory 12m conforms to the reference value of the judgment criteria data 11e, and the result is displayed in the result area 3j (see FIG. 7(b)) of the display device 3 (S26).

Specifically, the reference values of the nugget diameter Nhd, the upper plate penetration depth Nup, and the lower plate penetration depth Ndd corresponding to the plate thicknesses in the upper plate thickness memory 12c and the lower plate thickness memory 12d are obtained from the judgment criteria data 11e. If the nugget diameter Nhd in the nugget diameter memory 12i, the upper plate penetration depth Nup in the upper plate penetration depth memory 12k, and the lower plate penetration depth Ndd in the lower plate penetration depth memory 12m all match the obtained reference values, "OK" is displayed in the result area 3j, and if even one does not match, "NG" is displayed in the result area 3j.

That is, the nugget diameter Nhd, the upper plate penetration depth Nup, and the lower plate penetration depth Ndd are measured based on the nugget N detected from the input image P of the input image memory 12a. Based on the measured nugget diameter Nhd, the upper plate penetration depth Nup, and the lower plate penetration depth Ndd, it is determined whether the spot welding of the test piece T in the input image P complies with a specified standard.

This allows the nugget N used for measurement and the nugget N used for judgment to be the same, so it is possible to accurately judge whether the spot welding complies with the specified standards based on the measured nugget diameter Nhd, upper plate penetration depth Nup, and lower plate penetration depth Ndd. In addition, by simply acquiring the input image P, it is possible to measure the nugget diameter Nhd, upper plate penetration depth Nup, and lower plate penetration depth Ndd, and judge the spot welding to the test piece T, making these measurements and judgments easier and reducing the workload of the workers.

After the process of S26, it is confirmed via the input device 2 whether one or more values of the nugget diameter area 3g, the upper plate penetration depth area 3h, or the lower plate penetration depth area 3i are abnormal (S27). In the process of S27, if one or more values of the nugget diameter area 3g, the upper plate penetration depth area 3h, or the lower plate penetration depth area 3i are abnormal (S27: Yes), the corresponding nugget diameter memory 12i, the upper plate penetration depth memory 12k, and the lower plate penetration depth memory 12m are updated with the values input by the operator to the nugget diameter area 3g, the upper plate penetration depth area 3h, or the lower plate penetration depth area 3i via the input device 2 (S28), and the process from S24 onwards is repeated. The values input by the worker in the process of S28 may be the nugget diameter Nhd, the upper plate penetration depth Nup, or the lower plate penetration depth Ndd measured by the worker himself from the input image P or the result image Q, or may be the nugget diameter Nhd, the upper plate penetration depth Nup, or the lower plate penetration depth Ndd measured using separate image processing software.

In addition to checking whether the nugget diameter area 3g etc. has an abnormal value in the process of S27, if the worker determines that the position of the nugget diameter Nhd, the upper plate penetration depth Nup, or the lower plate penetration depth Ndd displayed in the result image Q of the result image display area 3b is abnormal, the process of S28 is executed to update the corresponding nugget diameter memory 12i etc. with the value input by the worker in the nugget diameter area 3g etc. via the input device 2, and in addition, the display position of the dimensional information such as the nugget diameter Nhd displayed in the result image Q of the result image display area 3b may be corrected by the worker via the input device 2. In this case, after these processes, the result image Q corrected via the input device 2 is saved in the result image memory 12n, and the process of S24 can be omitted, and the processes from S25 onwards can be executed.

In the process of S27, if the values of the nugget diameter area 3g, the upper plate penetration depth area 3h, and the lower plate penetration depth area 3i are all normal (S27: No), the values of the upper plate thickness memory 12c, the lower plate thickness memory 12d, the nugget diameter memory 12i, the upper plate penetration depth memory 12k, and the lower plate penetration depth memory 12m, the result image Q of the result image memory 12n, and the judgment result judged in the process of S26 are added to the result table 11d (S29). By checking the values and the result image Q added to the result table 11d, it is possible to verify whether the spot welding complies with the specified standard even after the judgment process is performed, and to verify the dimensions of the nugget diameter Nhd, the upper plate penetration depth Nup, and the lower plate penetration depth Ndd.

The user also creates corresponding teacher images from the input images added to the result table 11d, and adds these to the learning data 11c before performing the learning process in S3. This allows the nugget determination model M of the nugget determination model 11b to be learned based on the images used in the actual determination process, thereby further improving the detection accuracy of the nugget N from the input image P.

After the process of S29, it is confirmed whether the Next Read button 3k or the End button 3m (both see FIG. 7(b)) has been operated (S30). If the Next Read button 3k has been operated in the process of S30 (S30: Next Read button), the process from S20 onwards is repeated to acquire an input image P for the next test piece T. On the other hand, if the End button 3m has been operated in the process of S30 (S30: End button), the judgment process is terminated.

Return to Figure 4(a). After S2 or S3, the process from S1 onwards is repeated.

The present invention has been described above based on the embodiments, but the present invention is in no way limited to the above-mentioned embodiments, and it can be easily imagined that various improvements and modifications are possible within the scope of the invention without departing from its spirit.

In the above embodiment, the nugget area image Pn depicts the nugget N and objects other than the nugget N in the input image P in different colors. However, the form of the nugget area image Pn is not limited to this, and for example, the nugget N and objects other than the nugget N in the input image P may be depicted in different patterns, or only the contour line of the nugget N may be depicted. In such a case, an image consisting only of the pattern of the corresponding nugget N and objects other than the nugget N to be detected from the input image, or the contour Ne of the nugget N (see FIG. 2(b)) may be stored in the teacher image in the learning data 11c (see FIG. 3(b)), and the learning process of S2 (FIG. 5) may be executed.

In the above embodiment, the nugget area image Pn was obtained using the input image P and the nugget determination model M. However, what is obtained from the input image P and the nugget determination model M is not limited to the nugget area image Pn, and may be information other than an image, such as a numerical value representing the position of the nugget N in the input image P. In such a case, the information to be obtained from the input image P can be stored instead of the teacher image in the learning data 11c (see FIG. 3(b)), and the learning process of S2 can be executed.

In the above embodiment, in the processes of S42 and S43 in FIG. 6, a horizontally or vertically linear image is obtained by image processing. However, the acquisition of a linear image is not limited to image processing. For example, a learning model that has been trained to be able to detect linear images may be created, and a linear image in the input image P may be obtained by using the input image P and this learning model.

In the above embodiment, in the process of S47 in FIG. 6, the plate boundary B is calculated from the values of the upper plate thickness memory 12c, the lower plate thickness memory 12d, the upper end of the upper plate, and the lower end of the lower plate, but this is not limited to the above, and the plate boundary B may be obtained by image processing. In such a case, an image corresponding to the plate boundary B can be obtained from the linear image of the horizontal component in the input image P obtained in the process of S42 in FIG. 6. For example, in FIG. 2(c), the second horizontal linear image from the top in image P' corresponds to the plate boundary B, so this linear image can be taken as the plate boundary B.

In the above embodiment, the dimensions of each part of the nugget N were measured from the input image P, namely the nugget diameter Nhd, the upper plate penetration depth Nup, and the lower plate penetration depth Ndd, but this is not limited to this and the dimensions of other parts of the nugget N may be measured. Furthermore, it is not limited to measuring the dimensions of the nugget N from the input image P, and for example, the dimensions of parts of the test piece T other than the nugget N may be detected.

In the above embodiment, the nugget diameter Nhd of the nugget N formed on two overlapping steel plates Sp was measured, but this is not limited to the above, and the nugget diameter Nhd of the nugget N formed on three or more overlapping steel plates Sp may be measured. For example, when three steel plates Sp are overlapped, the nugget diameter Nhd of the nugget N formed to straddle the three steel plates Sp exists in two places: the upper plate nugget diameter at the upper plate boundary, which is the boundary between the top steel plate Sp (i.e., the upper plate) and the middle steel plate Sp, and the lower plate nugget diameter at the lower plate boundary, which is the boundary between the bottom steel plate Sp (i.e., the lower plate) and the middle steel plate Sp.

3(a) and (c), the result table 11d stores the upper plate nugget diameter and the lower plate nugget diameter instead of the nugget diameter Nhd, and the judgment criteria data 11e stores the reference values of the upper plate nugget diameter and the lower plate nugget diameter according to the plate thickness of the steel plate Sp. Also, the RAM 12 is provided with an upper plate boundary position memory in which the position of the upper plate boundary is stored and a lower plate boundary position memory in which the position of the lower plate boundary is stored instead of the plate boundary position memory 12j, and an upper plate nugget diameter memory in which the upper plate nugget diameter is stored and a lower plate nugget diameter memory in which the lower plate nugget diameter is stored instead of the nugget diameter memory 12i.

Then, in the process of S47 in FIG. 6, the upper plate boundary, which is a position moved downward from the coordinates of the upper end of the upper plate by the plate thickness in the upper plate thickness memory 12c, is measured and stored in the upper plate boundary position memory, and the lower plate boundary, which is a position moved upward from the coordinates of the lower end of the lower plate by the plate thickness in the lower plate thickness memory 12d, is measured and stored in the lower plate boundary position memory. In the process of S48, the upper plate nugget diameter, which is the length between the intersection points of nugget N and the upper plate boundary in the nugget position memory 12h, is measured and stored in the upper plate nugget diameter memory, and the lower plate nugget diameter, which is the length between the intersection points of nugget N and the lower plate boundary, is measured and stored in the lower plate nugget diameter memory.

In the process of S24 in FIG. 5, each dimensional information of the upper plate nugget diameter memory and the lower plate nugget diameter memory is synthesized into a result image Q, in the process of S25, each value of the upper plate nugget diameter memory and the lower plate nugget diameter memory and the result image Q are displayed on the display device 3, in the process of S26, a judgment is made using the upper plate nugget diameter memory and the lower plate nugget diameter memory and the judgment criterion data 11e, and the judgment result is displayed on the display device 3. In addition, in the processes of S27 and S28, the corrected upper plate nugget diameter and lower plate nugget diameter are stored in the upper plate nugget diameter memory and the lower plate nugget diameter memory, and in the process of S29, each value of the upper plate nugget diameter memory and the lower plate nugget diameter memory, the result image Q, and the judgment result are added to the result table 11d.

In the above embodiment, a ruler J is exemplified as a reference object included in the input image P together with the test piece T, but this is not limited thereto, and for example, an official postcard, business card, or other object with predefined dimensions may be used as an appropriate reference object. Also, the placement of a reference object such as a ruler J may be omitted. In that case, for example, the longitudinal side of the test piece T may be set to a predetermined length (e.g., 10 cm), and the conversion coefficient may be obtained from this length and the number of pixels on the longitudinal side of the test piece T in the input image P, or the conversion coefficient may be obtained by other methods.

In the above embodiment, the input image P was acquired from the camera C, but this is not limited thereto. For example, the input image P may be acquired from an imaging device other than the camera C, such as a scanner, or an image stored in the HDD 12 may be acquired as the input image P.

In the above embodiment, the input screen display area 3a, the result image display area 3b, and the numerical display area 3c are displayed on the display device 3 in Figures 7(a) and (b), but this is not limited to this. For example, the display of the numerical display area 3c may be omitted and only the input screen display area 3a and the result image display area 3b may be displayed, or the display of the input screen display area 3a may be omitted and only the result image display area 3b and the numerical display area 3c may be displayed, or the input screen display area 3a and the result image display area 3b may be omitted and only the numerical display area 3c may be displayed.

In the above embodiment, a measuring device 1 incorporating the measurement program 11a is exemplified, but this is not limited thereto, and the measurement program 11a may be executed by an information processing device (computer) such as a personal computer, smartphone, tablet terminal, etc.

1 Measuring device 11a Measuring program P Input image J Ruler (reference object)
N: Nugget Sp: Steel plate T: Test piece S20 Input image acquisition means, input image acquisition step
S21 Plate thickness acquisition means, plate thickness acquisition step
S40 Nugget detection means, nugget detection step S44 Conversion coefficient acquisition means
S45 End acquisition means, end acquisition step
S46: Plate thickness pixel acquisition means, plate thickness pixel acquisition step
S47 Plate boundary detection means, plate boundary detection step S48 Part of measurement means, part of measurement step S49 Part of measurement means, parallel straight line calculation means S50 Part of measurement means, straight line specification means S51 Part of measurement means, part of measurement step, penetration depth measurement means

Claims (8)

A measuring device having a measuring means for measuring dimensions of each part in a cross section of a test piece formed by overlapping flat steel plates and spot welding the test piece from an image of the cross section of the test piece,
an input image acquisition means for acquiring an input image which is an image of a cross section of the test piece;
a nugget detection means for distinguishing and detecting a nugget in spot welding from the input image acquired by the input image acquisition means;
A plate thickness acquisition means for acquiring the input plate thickness of the steel plate;
A board thickness pixel acquisition means for acquiring the number of pixels of the board thickness acquired by the board thickness acquisition means;
an end acquisition means for acquiring coordinates of the positions of the upper and lower ends of the test piece from the input image acquired by the input image acquisition means;
a plate boundary detection means for detecting a position obtained by moving the coordinates of the positions of the upper and lower ends of the test piece acquired by the end acquisition means in the input image acquired by the input image acquisition means by the number of pixels of the plate thickness acquired by the plate thickness pixel acquisition means as a plate boundary that is a boundary between the overlapping steel plates in the test piece;
The measuring device is characterized in that the measuring means measures dimensions of each portion of the nugget in the input image from the nugget detected by the nugget detection means, using the plate boundary detected by the plate boundary detection means as a reference. The measuring device according to claim 1, characterized in that the measuring means measures the nugget diameter, which is the distance between the intersection of the plate boundary detected by the plate boundary detection means and the nugget detected by the nugget detection means. The measuring means is
a parallel line calculation means for calculating a plurality of parallel lines which are lines parallel to the plate boundary detected by the plate boundary detection means;
a line specifying means for specifying, from among the plurality of parallel lines calculated by the parallel line calculating means, a parallel line whose distance between an intersection point of the parallel line and the nugget detected by the nugget detecting means is a predetermined ratio of the nugget diameter,
The measuring device according to claim 2, further comprising a penetration depth measuring means for measuring a penetration depth which is a distance between the parallel straight line specified by the straight line specifying means and the plate boundary detected by the plate boundary detecting means. The input image is captured so as to include a reference object serving as a dimensional reference together with the test piece;
a conversion coefficient acquisition means for acquiring a conversion coefficient representing a relationship between a predetermined dimension of the reference object and the number of pixels of the input image corresponding to the predetermined dimension;
4. The measuring device according to claim 1, wherein the measuring means measures dimensions of each portion of the nugget based on the conversion coefficient acquired by the conversion coefficient acquisition means. the nugget detection means outputs a nugget region image which is an image in which a nugget and an object other than the nugget in the input image are drawn in different colors,
5. The measuring device according to claim 1, wherein the measuring means measures dimensions of each portion of the nugget in the input image based on the nugget region image output by the nugget detection means. 6. The measuring device according to claim 1, wherein the nugget detection means uses artificial intelligence to distinguish and detect nuggets in spot welding from the input image acquired by the input image acquisition means. An input image acquisition step of acquiring an input image which is an image of a cross section of a test piece formed by overlapping flat steel plates and spot welding them;
a nugget detection step of distinguishing and detecting a nugget in spot welding from the input image acquired in the input image acquisition step;
A plate thickness acquisition step of acquiring the input plate thickness of the steel plate;
a plate thickness pixel acquisition step for acquiring the number of pixels of the plate thickness acquired in the plate thickness input step;
an end acquisition step of acquiring coordinates of positions of the upper end and the lower end of the test piece from the input image acquired in the input image acquisition step;
A plate boundary detection step in which the coordinates of the positions of the upper and lower ends of the test piece acquired in the end acquisition step in the input image acquired in the input image acquisition step are moved by the number of pixels of the plate thickness acquired in the plate thickness pixel acquisition step, and the position is detected as a plate boundary that is a boundary between the overlapping steel plates in the test piece;
a measurement step of measuring dimensions of each portion of the nugget in the input image from the nugget detected in the nugget detection step, based on the plate boundary detected in the plate boundary detection step. 8. The measurement program according to claim 7, wherein the nugget detection step distinguishes and detects nuggets in spot welding from the input image acquired in the input image acquisition step by using artificial intelligence.

Images