JPS59107898A - Method of monitoring work of robot, etc. - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a method for monitoring the work of robots, etc., and more specifically, when checking the presence or absence of an object or monitoring an abnormal state of work using a camera. 'A
The present invention relates to a method of achieving the purpose of monitoring by dividing each pixel into m×n pixels in n columns and comparing the actual state at the time of monitoring with a preset normal state for each pixel.

There are cameras, such as industrial television cameras (ITV).
The monitoring method uses solid-state imaging 't, such as monitoring the work status of the robot 4 or the product inspection process. It is used for purposes such as measuring scratches and dirt, checking the presence of books, monitoring abnormal conditions, and reading characters, symbols, etc.

This monitoring method generally uses 211i video from a camera.
i, that is, black and white, or binary coded (1 or O), the resulting image pattern '1cRT is input to a video J control device, etc., and the image displayed on, for example, a CRTK is visually inspected by a worker, etc. is accomplishing its purpose.

However, a drawback of the above-mentioned monitoring method is that the video pattern of the monitored song area is difficult to match the actual monitored area. For this reason, the monitoring field by ITV was narrowed, the scanning line of the monitoring screen was made denser, and its resolution was improved.
Monitoring accuracy has been completely improved. However, this increases the number of cameras required to be installed and requires a precise video analysis device, raising equipment costs.

It is an object of the present invention to provide a method for monitoring robots, etc., without reducing the monitoring accuracy or increasing equipment costs. The purpose of this invention is to monitor the presence or absence of objects and to monitor abnormal conditions, and the monitoring area is divided into m x n-115 systems, and the basic image is calculated based on the state of each pixel for one cycle of normal operation. The basic idea is to find a pattern, constantly compare the basic video pattern with a first-class video pattern obtained by converting the state of each pixel under monitoring into two, and to generate an abnormal signal if the two do not match. Note that some of the pixels that make up the video pattern change their state during one cycle of work, while others do not, but for pixels whose state changes, there is a case where the basic video pattern and the video pattern do not match even when the work is normal. Therefore, it is excluded from the above comparison.

The following is a detailed description of the invention using embodiments shown in the accompanying drawings.

In Fig. 1, the workpiece 2 on the belt comparer 1 moving in the direction of the arrow is processed by the robots 3 and 4. Monitor the defined monitoring area 6 (
It's true. The output of the ITV 5 is input to the pattern processing device 7.

FIG. 2 is a block diagram for explaining the basic operation of the embodiment, and the pattern processing device 7 is the same as the binarization unit 71 that binarizes video broadcasts from ITV by the method described below. RA where the basic video pattern explained below is stored
It consists of a pattern memory 72 made up of M, etc., and a comparison section 76 that controls generation of an abnormal signal when an abnormality occurs. In the same figure, the protective arrow (*) indicates the flow when setting the basic video pattern to the pattern memory 72, and the solid arrow indicates the entire flow during monitoring work.In other words, the video pattern that is the output of the binarization section 71 is When the basic video pattern is seven, it is input to the pattern memory 72, and during monitoring, it is manually input to the comparison section 73. During the monitoring operation, the comparison section 76
Further, the basic video pattern from the pattern memory 72 is input, and the comparison unit 73 outputs an abnormal signal when the two do not match. Then, based on this abnormal signal, processing such as emitting all alarm sounds is performed.

Next, the creation of the basic video pattern "(ll-1)) and its setting in the pattern memory 72 (memory) which is performed prior to the start of the monitoring work will be explained using FIGS. 6 and 4. First, I
In the exposure aperture state of TV5 (hereinafter referred to as the first state), the robot 3,
A video report of the monitoring area 6 during one cycle of work No. 4 is input to the pattern processing device 7. The image tH information of the push and lever is converted into an image pattern by the pattern processing device 7 as follows.

υIJ As shown in Figure 4, the monitoring area 6 is arranged in m columns vertically and n horizontally.
The column is divided into Xn pixels (m and n are set appropriately depending on the conditions). Next, the state of each pixel (white or black) is determined. In the first state, each pixel IL (i, j) (i = 1 to m, j = 1 to n) of the monitoring/blocking area 6 is II (i, j) when the pixel is always white during one work cycle.
Con=0, black is 11 (i.

J] = 1, and when changing to black and white, zl (i, j) =
2 and stored in the pattern memory 72.

Similarly, the exposure of ITV5 is opened (hereinafter referred to as the second state)
The image lPf information for one cycle of work is input to the pattern processing device 7, and the state of each pixel at this time is stored in the pattern memory 72 as a second image pattern l2(1+j);'2(1+j)=0 to 2. do.

Then, within the pattern 72, the first image pattern 11 (
i, j) and the second video pattern I (i.

j), that is, the robot's work is normal, and no abnormalities such as 10] have occurred.

Then, this basic video pattern I (i, j) is stored in the entire pattern memory 72Vc memory.

In addition, in the table above, "state" or "other" refers to locations where pixels change from white to black or vice versa during one cycle of work, such as the 7-person work area (8 )
etc. This area is recognized as gray when viewed as a whole in one work cycle. In addition, the condition is "black"
``White'' indicates fixedly installed locations within the monitoring area 6 (9U Eva, robot main body, belt conveyor), and locations where there are no installed objects. Therefore, the pattern memory 72 (i.e. basic video pattern I(i,
According to j), the state of each pixel in the monitoring area B is converted into a VC and binarized and stored in a 2-focus memory, for example, white → 01, black → 10, others → 11.

As mentioned above, when creating the basic image pattern, by comparing the 5g1 image pattern and the 2nd image pattern with different exposures, it was possible to achieve the things shown in the table above (the values of ``white'' and ``black'' were clearly defined. In other words, for example, when a pattern recognized as "white" in the first video pattern E1 is recognized as "black" or "gray" in the second video pattern As long as the recognition values of video pattern ■1 and second image pattern pattern noe 2 do not match, all of the basic video pattern ■ will be memorized as "gray".This means that in the example above, In the first video pattern E1,
Even if it is recognized as "white," there may be differences in the degree of whiteness depending on the brightness within the monitoring area due to differences in indoor lighting, and the recognition value in that situation cannot be definitive. Therefore, only when the recognition value for the highly exposed first video pattern E1 and the recognition value for the weakly exposed second video pattern E2 match, the matched recognition value is taken as the basic video pattern E. As a result, valley pixels 11 (1+1) (i=1 to n, j-1 to n
) The pixels corresponding to "white" or "black" in K are clearly determined, and even if there is some variation in brightness within the monitoring area, this will not cause erroneous recognition.

In this way, the basic video pattern (11) is completely created.
Pattern memo IJ 72 Start all monitoring after K memory.

The means for determining the basic image pattern from the first and second image patterns is stored in the pattern memory 72 as described above.
(Also, it is also possible to simply store in the pattern memory 72 what has been obtained in advance by an external operation.The state of each pixel constituting the own iS writing operation/initialization is determined by the state of the operation. Since it changes from white to black or vice versa during one cycle, the comparing section 73V during monitoring, which will be described later.
excluded from comparison by c.

Next, the operation of the embodiment during monitoring will be explained using FIGS. .

W, 5, Fig. 1h) shows the belt conveyor 1, the robot (3 or 4) main body 7.7-mm 31 (41), the movable φn radius 8 of the arm, and the unfinished area 9 during normal operation. Conditions will be monitored by ITV5.

In addition, the monitoring started on 14iJK(・The 6th
The basic image pattern r(i, ]) shown in Figure (Al) is 1'
[This basic image + 1 pattern'("+") is a pattern memo') 72Wl is stored in memory.
, fBJ, the shaded area is I(t,j)-2, the diagonal cross is ■(l+])-1, and the white area is I(1,j)=o
The pixels in the corresponding states are shown.

At the time of monitoring, the basic video pattern information from the pattern memory 72 and the boundary video pattern information from the binarization section 71 are input to the comparison section 73, respectively. Therefore, if the robot is operating according to the Nora 1 fixed program and there is no abnormality such as an object entering the monitoring singing area 6, the comparing section 73V
The two video pattern reports inputted in C are the same, and no abnormal signal is output.
In Figure +Al, each pixel in the shaded area is excluded from comparison.

On the other hand, if the operator's hand 11 enters the monitoring chain acid 6 during monitoring as shown in Figure 5 (BIK), or if the arm moves out of control to the position shown in 51 due to a malfunction of the robot, the binarization is performed. The video pattern signal inputted from the section 71 to the comparison section 73 is as shown in FIG. 6 (Bl).
Pixels in the shaded area indicate pixels that are excluded from comparison as described above. That is, in this case, the state of each pixel in the area corresponding to the position 5 of the operator's hand 11 or the runaway arm inputted from the binarization unit 71 to the comparison unit 73 changes from white to black. Therefore, in these pixels, the comparison unit 7
Since the two values input to J-3 are different, and therefore the two image patterns input to the comparison section 76 are different, the comparison section 76 outputs an abnormal signal.

In addition, the above explanation involves the intervention of an external object (hand).
This is a case where the state of the corresponding pixel in the work area changes from white to black due to abnormal movement of the installed object (runaway arm). Even when there is a change in 1-, the video pattern inputted to the comparator 73vc is different, so the comparator outputs an abnormal signal and this abnormal state is detected.

In addition, in the embodiment, an abnormal signal is output for the purpose of detecting an abnormal state in the work area (-), but it goes without saying that all the states of arbitrary pixels in the work area can be compared (to determine whether an object exists or not). Since it is also possible to confirm the pattern, responsiveness is quick and inexpensive pattern processing is possible.

As described above, according to the present invention, the purpose of monitoring is to check the presence or absence of an object and to detect an abnormal state. The presence or absence of abnormal objects is detected by comparing the patterned images.Therefore, there is no need to use expensive video clips as in the past, and the monitoring area can be expanded arbitrarily. Therefore, the number of camera equipment can be significantly reduced.

[Brief explanation of drawings]

FIG. 1 is a schematic perspective view showing an embodiment of the present invention, FIG. 2 is a block diagram showing the relationship of various parts in the embodiment, and FIG. 3 is a flowchart showing an example of a method for creating a basic movie operation pattern. Fig. 4 is an explanatory diagram showing the method of dividing the monitoring area in the embodiment, Fig. 5 (A), fB) is a plan view showing the monitoring area in the embodiment, Fig. 6 (A), (8) are shown in Figure 5 (AJ, (
B) Explanation 1-1 shows the state in which the monitoring area indicated by Vc is converted into a video pattern. 1... Belt conveyor 3, 4... Robot 6.
・Monitoring area 7 ・Pattern processing device 8 ・
...Work area 9...Absent area 71...2
Value conversion unit 72...Pattern memory 73...
Comparison Department Patent Applicant Nippon Gakki Mfg. Co., Ltd. Patent Application Agent Patent Attorney Sugawara - Figure 6 (A) Figure L6 (f3 J Procedural Amendment (Method) 6 April f, 1980 Patent Office Commissioner Kazuo Wakasugi 1 Indication of the case Patent application No. 1987-214529 2 Name of the invention Work monitoring method for robots, etc. Representative of the person making the amendment Gen Kawakami - Amendment in the "Detailed description of the invention" column of the specification to be amended Page 7 of the statement of contents is corrected as shown in the attached sheet.

Claims (1)

[Claims] The monitoring/heading (6) is divided into mXn pixels arranged in m columns and n columns, and the state of the pixels is binarized, and the state change of each pixel for one cycle of normal operation is Classify each pixel into at least two types, those that change and those that do not change, to obtain a basic image pattern I (1°), and during monitoring, constantly binarize the state of each pixel to obtain the image pattern. A method for monitoring a robot, etc., characterized in that the basic video pattern and the video pattern are compared for the pixels that do not change, and if the two do not match, all abnormal signals are generated.