JPH0483105A - Method and apparatus for detecting welding position - Google Patents

Abstract

PURPOSE:To generate entire images respectively and take a logic product of both entire images to remove optical noise by scanning discontinuous frames of an image pickup means using two times of time-series image pickup to obtain discrete images and interpolating discrete portions. CONSTITUTION:A first or a second groove light cross-sectional image data is A/D-converted 13 and output to a sequential smoothing circuit 14 serially. The circuit 14 removes an especially isolated noise component and outputs to a preprocessing circuit 15. The circuit 15 performs binary processing or a treatment of a differential filter, etc. with respect to the image signal output from the circuit 14 to remove rough disturbance light other than the groove light cross-sectional image, and finally compresses the image data by replacing it by a binary signal. The image data of only odd-numbered or even-numbered frames obtained here is stored in image memories 16, 17 respectively. A central processing unit 18 performs calculation such as interpolation, recovery, logic calculation, thinning and detection of a groove position of the image data, control of data transfer, control of a synchronous signal generating circuit 21, control of external input/output data, etc.

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention relates to a welding position detection method and apparatus for optically automatically detecting a welding location on a workpiece.

In particular, the present invention relates to a welding position detection method and apparatus suitable for avoiding erroneous detection due to optical noise.

[Prior Art] As a typical welding position detection device that optically detects the position to be welded, there is a device that combines a slit light and a two-dimensional light receiving means as shown in FIG. 6, for example. In this welding position detection device, a laser beam 11 sharply focused into a slit shape is irradiated from the light projecting means 2 to the groove portions of the workpieces 8 and 9 preceding the arc point, and the reflected light from the groove portions is emitted. IT 12
By detecting with a two-dimensional light receiving means 3 such as a V camera and analyzing the obtained light section line image Q IQ Z Q 3,
Find the position Q2 to be welded. When such a device is applied to a welding robot to control a welding torch, for example, as shown in FIG. Attach it to the tip of robot 5's arm.

In such an optical welding position aspect ratio device, when disturbance light such as spatter generated during welding enters the field of view of the light receiving means, the sputter trajectory becomes a strong linear image in the optically cut image of the groove. The image will be taken. As a result, when processing a captured image, optical noise such as a sputter image may be mistakenly recognized as an actual light-cut image. There is also the problem that the desired groove position cannot be detected with high accuracy.

A suitable method for removing such optical noise is to take a target image twice in chronological order and calculate the AND of each taken image. Since optical noise occurs irregularly, it is almost impossible for optical noise with the same shape to appear at the same location in both captured images that are temporally shifted, and by adopting this method, the negative effects of optical noise can be reduced. Can be removed.

However, in this method, all image data is input from the imaging means twice and stored in the image memory respectively, so it takes time to input the image and requires a large capacity image memory6. There are problems in applying the method to actual equipment, and it has not been adopted in practice.

For this reason, a method can be considered in which a shield is provided between the arc generation point at the tip of the welding wire and the groove surface to prevent the intrusion of ambient light. However, in this method, a disturbance light shield is installed at the tip of the welding position detection part, which increases the size of the detection part, and there is a risk of collision with the welding workpiece when welding with a welding robot, etc., and the applicable range is limited. The problem is that it is restricted.

Therefore, in practice, as shown in Japanese Patent Laid-Open No. 62-267607, the captured image signal is processed to select the groove light cut image from the image containing the disturbance light. This conventional method detects the rising and falling positions of an image signal on a horizontal scanning line of image data, and determines whether the magnitude of the gradient of the rising and falling parts is equal to or greater than a predetermined light cutting line gradient threshold. If it is less than the threshold value, it is considered as noise, and if it is more than the threshold value, then the light cutting line width is determined. That is, when the line width of the image signal is smaller than a preset light-cutting line width, it is regarded as noise and the light-cutting image signal, and when it is larger, only the light-cutting image signal is selected as noise.

[Problems to be Solved by the Invention] However, in the above conventional method, when the sputtering light overlaps the photocutting image, the necessary photocutting line image itself is also missing, and it is difficult to determine the groove position by image processing. This will impede detection. In addition, spatter generated during welding collides with the workpiece, etc., and is scattered in various directions, so the spatter light image recorded on the screen not only appears in the direction orthogonal to the light section line image, but also In some cases, the disturbance light appears in the same direction (parallel) as the light section line image, and for such a sputter light image, even with the above method, the disturbance light is mistakenly recognized as the light section line.

An object of the present invention is to provide a welding position detection method and device that can use a small-capacity image memory and can reliably and quickly eliminate the effects of optical noise such as spatter and arc light generated during welding. It's about doing.

[Means for Solving the Problem] The above object is to irradiate a workpiece with slit light, image the slit light irradiation location on the workpiece twice in chronological order with an imaging means, and When detecting the position to be welded from the slit light image by calculating the logical product of the captured images and removing optical noise from the slit light image, 1.
In each imaging step, the discontinuous frames of the imaging means are scanned to obtain discrete images, the discrete points are interpolated to generate each whole image, and the two whole images are ANDed to eliminate optical noise. This is achieved by removing.

[Function] Spatter generated during welding occurs irregularly over time,
Its scattering speed is generally as fast as several meters per second. Therefore, the frequency, direction, etc. of optical noise appearing on the screen are not constant for each image input. Therefore, optical noise can be effectively removed by performing a logical product of two temporally shifted images. However, as described above, this method requires time to input images twice.

Moreover, a large capacity image memory is required.

However, in the present invention, in one imaging operation, instead of inputting all image data from the imaging means to the image memory, discrete image data is input by scanning only odd or even frames of the imaging means. As a result, image data can be input in a short time, and the capacity of the image memory can be small.

Since the logical product is performed after interpolating discrete locations from the discrete image, only optical noise can be removed.

[Example] An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a functional configuration diagram of a welding position detection device according to an embodiment of the present invention. In FIG. 1, 3 is a two-dimensional light receiving means such as an ITV (hereinafter simply referred to as a light receiving means), 4 is an image processing device, and 6 is a robot control device. Further, 13 is an A/D converter, 14 is a smoothing circuit, 15 is a preprocessing circuit,
16 and 17 are the 1st. The image processing device 4 is composed of a second image memory, 18 a central processing unit, 19 a third image memory, 20 a synchronization signal generation circuit, 21 an external input/output interface, and 22 a display unit. There is.

The light receiving means 3 receives the slit light image of the object to be welded, and creates image data of only odd frames or only even frames of the light receiving means 3 at fixed time intervals according to commands from the synchronizing signal generating circuit 21. It is output to the D converter 13.

FIG. 2 is a diagram illustrating a method of scanning each pixel on the light-receiving surface of the light-receiving means 3. As shown in Figure 2, for the first captured image, one of each line in the vertical direction
,3. ..., 489 and the horizontal scanning is repeated in the order shown by the solid line, and the discrete image data of the first groove light cutting screen is stored in the first image memory 16 as described later. Thereafter, the screen data is cleared by a command from the synchronization signal generation circuit 21. Regarding the second captured image, 2, 4, . ..., 490 and the horizontal scanning is repeated in the order shown by the broken line, and the discrete image data of the second groove light cutting screen is stored in the second image memory 17 as described later.

As shown in FIG. 1, the first or second bevel light cutting screen data input to the A/D converter 13 are each converted into digital quantities and serially output to the smoothing circuit 14. . The smoothing circuit 14 particularly removes isolated noise components and outputs the result to the preprocessing circuit 15. The image signal up to this point is a multi-level signal, but in the preprocessing circuit 15, the image signal output from the smoothing circuit 14 is subjected to binarization processing or processing using a differential filter, etc. The disturbance light is roughly removed, and the image data is finally replaced with a binary signal to compress the image data. The image data of only odd frames or only even frames obtained here is stored in the first image memory 16 or the second image memory 17, respectively. The central processing unit 18 performs image data interpolation, restoration, logical operations, as will be described in detail later.
Control of calculations and data transfer such as thinning, detection of groove position,
It controls the synchronization signal generation circuit 21 and external input/output data.

FIG. 3 is a flowchart showing the processing procedure executed by the central processing unit 18. In addition, Fig. 4 (a), (b
) are respectively stored in the first image memory 16. This is an example of an image stored in the second image memory 17, and is a captured image obtained during welding when the groove shown in FIG. 6 is a fillet joint. As mentioned above, the image stored in the first image memory 16 (see FIG.
Since a)) is composed of odd-numbered frames, the groove optical cutting lines: Q1Q2Q3 are not continuous line segments, and data of even-numbered frames is missing for each line. Furthermore, since the image stored in the second image memory (FIG. 4(b)) is composed of even frames, the groove light cutting line QIQ
In 2Q3, odd frame data is missing for each line. (In addition, this is emphasized and shown with a broken line in FIG. 4, and is not an accurate diagram).

Note that each line segment N1 to N5 in FIGS. 4(a) and 4(b) is a sputter image stored on the image memory.

In this case as well, each image data has data missing for each line.

The central processing unit 18 performs the processing shown in FIG. 3 using the data stored in the image memories 16 and 17. First, in step 1, missing line data of the image data stored in the first image memory 16 is interpolated to restore the entire image. Details of this method will be described later. Figure 4 (c
) shows the restored image. Next, similarly in step 2,
Missing line data is interpolated from the image data stored in the second image memory 17, and the entire image is restored. An example of the restored entire image is shown in FIG. 4(b).

In step 3, step 1. Using the two whole image data obtained in step 2, a logical product operation is performed between corresponding pixels between the two whole images. The two image data obtained in step 1 and step 2 were captured with a slight time difference, but because the time interval is short, the position of the light-cut image formed on the screen hardly changes. do not.

On the other hand, spatter that occurs during welding occurs irregularly over time, and its scattering speed is generally several meters per second, so it appears on the screen every time an image is input.

The direction is also not constant. In other words, FIG. 4(a).

In both figures (b), line segments QIQ2Q3 have the same shape, but line segments N1 to N5 are different. Therefore, Fig. 4(c)
) and (d), as shown in FIG. 4(e), the sputter images N1 to N5 disappear and only the groove light-cut image QIQ2Q3 is extracted. That will happen. This image data (FIG. 4(e)) is stored in the third image memory 19 in step 4. At this time, data compression is performed.

Next, in step 5, after thinning the thick optically sectioned line image, in step 6, the desired groove position (fourth
Q2 position in the figure) is calculated. Further, in step 7, the lateral ratio result of this groove position in the camera coordinate system is determined by the light receiving means 3.
(and the light projecting means 2) is converted into the coordinate system of the robot to which it is attached, and this data is transferred to the robot control device in step 8.

Next, a method for interpolating and restoring image data performed in steps 1 and 2 in FIG. 3 will be described. As mentioned above, the first. The second image memories 16 and 17 have data missing for each line in the vertical direction of the screen. In this embodiment, interpolation of image data on intermediate lines where there is no data is performed based on the presence or absence of data on the horizontal line before and after the horizontal line where image data exists on each intermediate line, and the arrangement of the data as necessary. We are trying to create a new one. Figure 5 shows the -
This is an example. In FIG. 5, a line with symbol j indicates a vertical intermediate line where no data exists, and symbols j-1 and j+1 each indicate horizontal lines before and after it. Figures 5(a) to (d) show the line j- around 1.
When there is data adjacent to 1 and j+2 by 1 pixel,
This shows how to generate and restore data on intermediate line j where no data exists at that time.

In other words, when the pixels existing on the preceding and succeeding lines j-t and j+1 are adjacent to each other within two pixels in the horizontal direction (as shown in the same figure (a)
) to (c)), new data is generated in the areas indicated by symbols B1 to B4. However, the line before and after
When the pixels existing in 1,j+1 are separated by 3 pixels or more in the horizontal direction ((d) in the same figure), the pixels existing in the previous and succeeding lines j-1,j
+1 data are each considered as a separate line element, and the symbol b1
No data is generated in the area shown in ~b4.

Figures 5(e) to (i) show the front and rear lines j-1゜j+1
This figure shows the generation pattern of data on the intermediate line j in a case where data connected by two pixels each exist in close proximity to each other.

In this case, as above, the previous and subsequent lines j −1,
New data is generated (in the case of symbols B1 to B7) or separated (in the case of symbols b1 to b6), depending on how far apart the pixels existing at j+1 are in the horizontal direction.
That's what I do.

As described above, in this embodiment, the rule of separation between the preceding and subsequent pixel arrangement patterns and the generation of intermediate data is predetermined. Then, by sequentially performing interpolation processing on each line in which vertical data is missing according to the above law, it is possible to restore one complete image.

According to this embodiment, by performing a logical operation on two images captured with a slight time difference, it is possible to reduce noise and reliably extract only a light section image. Therefore, it is possible to detect the groove position with high accuracy. In addition, the above two images are created by creating two screens using only odd frames and only even frames by single-line interlaced scanning, and then interpolating and restoring the image of the missing frame, so the conventional imaging time is 2 resolution. The groove position can be detected without reducing the Furthermore, in this embodiment, a total of three screens of storage images, an input image planning screen and one processing image screen, are required, but since compressed binary image data is stored in each screen, the storage capacity is also small. realizable.

[Effects of the Invention] According to the present invention, optical noise can be removed in a short processing time, the welding position can be detected with high accuracy, and the capacity of the image memory used can be small.

[Brief explanation of the drawing]

FIG. 1 is a functional configuration diagram of a welding position detection device according to an embodiment of the present invention, FIG. 2 is an explanatory diagram of scanning of an imaged light receiving screen, and FIG. 3 is a flowchart showing the processing procedure of a welding position detection method. Figures 4 (a) to (e) are conceptual illustrations of the welding position detection method, Figures 5 (a) to (i) are illustrations of the data interpolation method for missing frames, and Figure 6 is a perspective view of the welding robot's hand. 7 are general views of the welding robot. 1... Welding torch, 2... Light emitting means, 3... Light receiving means. 4... Image processing device, 5... Robot, 6... Robot control device, 7... Welding machine, 8, 9... Welding workpiece, 11... Slit-shaped light beam, 12... Bevel light section image.

Claims (1)

[Claims] 1. A workpiece to be welded is irradiated with a slit light, a spot on the workpiece to be welded where the slit light is irradiated is imaged twice in chronological order by an imaging means, and the logical product of both taken images is In a welding position detection method in which a slit light image is obtained from which optical noise has been removed, and a position to be welded is detected from the slit light image, the discontinuity of the imaging means is detected by two times of time-series imaging. A welding position detection method characterized by scanning a frame to obtain discrete images, interpolating the discrete points to generate each whole image, and removing optical noise by performing a logical product of both whole images. . 2. Irradiate the object to be welded with slit light, take images of the irradiated part of the object with the slit light twice in chronological order using an imaging means, and calculate the optical noise by calculating the AND of both images. In the welding position detection method, which obtains a slit light image from which the slit light image is removed and detects the position to be welded from the slit light image, two time-series images are taken, and once only the odd frames of the imaging means are scanned. The entire image is restored by obtaining discrete images and interpolating the even-numbered frames, and the entire image is restored by scanning only the even-numbered frames of the imaging means once to obtain discrete images and interpolating the odd-numbered frames. , a welding position detection method characterized in that optical noise is removed by performing a logical product of both entire images. 3. A light source that irradiates the object to be welded with slit light, an imaging device that images the slit light irradiation location on the object to be welded, and a discrete image obtained by scanning the discrete frames of the imaging device. an image memory for storing data; an interpolation means for restoring the entire image by interpolating discrete parts of the discrete image; The present invention is characterized by comprising a noise removing means for removing optical noise by taking a logical product of two whole images, and a position detecting means for detecting a position to be welded from a slit light image obtained by the noise removing means. Welding position detection device. 4. A light source that irradiates the object to be welded with slit light, an imaging means that images the slit light irradiation location on the object to be welded twice in chronological order, and odd-numbered frames of the imaging means in the first imaging. a first image memory for storing image data of a discrete image obtained by scanning only an image; and a second image memory for storing image data of a discrete image obtained by scanning only an even-numbered frame of the imaging means in a second imaging. 2 image memories;
interpolation means for interpolating even frame locations of the discrete image stored in the image memory to obtain the entire image, and interpolating odd number frame locations of the discrete image stored in the second image memory to obtain the entire image; It is characterized by comprising a noise removing means for removing optical noise by taking a logical product of both whole images, and a position detecting means for detecting a position to be welded from a slit light image obtained by the noise removing means. Welding position detection device. 5. A welding torch, a light source that irradiates the workpiece with slit light, an imaging means that takes an image of the slit light irradiation location on the workpiece, and a discrete frame obtained by scanning the discrete frames of the imaging means. an image memory for storing image data of the target image;
interpolation means for restoring the entire image by interpolating discrete parts of the discrete image; and at least two
a noise removing means for removing optical noise by taking a logical product of two whole images each imaged twice and restored by the interpolating means, and detecting a position to be welded from the slit light image obtained by the noise removing means. 1. A welding robot comprising: a position detecting means for welding; and a control means for moving the welding torch to a position detected by the position detecting means to perform welding. 6. a welding torch, a light source that irradiates the object to be welded with slit light, an imaging means that images the slit light irradiation location on the object to be welded twice in chronological order; a first image memory for storing image data of discrete images obtained by scanning only odd frames of the imaging means; and image data of discrete images obtained by scanning only even frames of the imaging means in a second imaging. a second image memory for storing even frame portions of the discrete images stored in the first image memory, and obtaining a whole image by interpolating the even frame portions of the discrete images stored in the first image memory, and odd frame portions of the discrete images stored in the second image memory; interpolation means for interpolating to obtain a whole image; noise removal means for removing optical noise by performing a logical product of both whole images; and detecting a position to be welded from a slit light image obtained by the noise removal means. 1. A welding robot comprising: a position detecting means for welding; and a control means for moving the welding torch to a position detected by the position detecting means to perform welding.