JPS60120477A - Picture processor - Google Patents

Abstract

PURPOSE:To prevent occurrence of errors regardless of the sealing direction with a picture processor which is suitable for correcting the moving position of a playback type robot, etc., by calculating the deflection between a reading out address and prefixed reference address from both addresses. CONSTITUTION:When the 1st or 2nd edge signal b1 or b2 is inputted from the OR circuit 43 of a selection circuit 39, a latch circuit 44 latches a reading out address RE outputted from the 2nd counter 22 of a reading out circuit 21 at the time as the reading out address X indicating the image forming position of an edge corresponding to the sealing line SL2 of a shadow SH at the one-dimensional sensor of a camera 7. A deflection calculating circuit 45 finds the deflection DELTAX between the reading out address X latched at the latch circuit 44 and a reference address OFFH previsouly written in a data register 46 and, at the same time, outputs numerical data KDELTAX which is obtained by multiplying the deflection DELTAX by K (constant) to a controller 47 which drives a pulse motor 5.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention provides (1) an image processing device suitable for a position correction device for a moving device that moves along a predetermined work trajectory in response to a workpiece, such as a playback robot; Regarding.

BACKGROUND OF THE INVENTION Recently, many attempts have been made to have playback type robots serving as moving devices perform sealing work, which is performed, for example, in the assembly process of automobiles.

By the way, when this type of robot performs relatively precise work such as sealing work, work errors due to the repeatability and teaching accuracy of the robot itself, errors due to workpiece positioning accuracy and errors between workpieces, etc. For this reason, attempts have been made to provide a device for correcting the position of the sealing nozzle at the tip of the robot arm.

By the way, as such a position correction device.

The following can be considered.

That is, a drive mechanism equipped with an imaging means using, for example, a -dimensional image sensor is provided at the tip of the arm of the robot, and a sealing nozzle is attached to this drive mechanism.

However, the direction of movement of the imaging means and sealing nozzle, which are moved relative to the arm tip by this drive mechanism, is as follows:
It is assumed that the direction is perpendicular to the ceiling direction, which is the moving direction of the arm, and that the linear imaging field of the imaging means faces in this direction.

Then, for example, a line-shaped shadow (imaging target) of a constant width created by shining light on a ceiling line of a seam of an iron plate is imaged by an imaging means attached to the above-mentioned drive mechanism.

During playback of the robot, the image signal output from the imaging means is processed by the image processing device, and the deviation between the imaging position of the imaged shadow and the predetermined reference pixel position on the one-dimensional image sensor is detected. Find sequentially,
By moving the aforementioned drive mechanism according to the deviation so that the deviation becomes zero, the positional deviation between the sealing nozzle attached to the drive mechanism and the sealing line is corrected.

The image processing device has a function of detecting a positional deviation between the ceiling nozzle and the ceiling line, and after sequentially storing image data for one frame in an image memory based on the image signal output from the imaging means. , while sequentially reading image data from the image memory, detecting the address corresponding to the imaging position of the shadow of the ceiling line based on the read image data and the address where the image data was stored; The above deviation is calculated based on the address where the image data corresponding to the predetermined reference pixel is stored and the address where the image data corresponding to the predetermined reference pixel is stored.

By the way, in the image processing device as described above, it is possible to detect the address indicating the imaging position of the ceiling line based on one edge of a constant-width shading added to the line. Then, the following problem occurred.

For example, if the detection specification is based on the edge on the left side toward the sealink direction in a shadow of a certain width attached to the sealink line, overlapping workpieces (iron plates) as shown in Figure 1 (a) and (b) are used. In the sealing line SLo at the joint of W+ and W2, if the sealing direction is the direction of arrow Y, it corresponds to the line SLo of the shade SHo of width ρX created by shining light on the ceiling line SLo from the direction of arrow X1. The imaging address of the ceiling line SLo based on the edge EDo can be detected, and by correcting the position of the ceiling nozzle (not shown) according to the deviation between the imaging address and the reference address, the ceiling line SL
A sea link along the o can be performed.

=4− However, as shown in Figures 2 (a) and (b), Figure 1 (
b) Contrary to (b), when the sealing direction is in the direction of arrow Y, which is the same as in Fig. 1, at the sealing line SL1 of the joint of overlapping workpieces W+ and W2, the sealing line SLI is from the direction of arrow X2. Width i when exposed to light!
Since the imaging address of the seal line SL1 is detected based on the edge ED, which does not correspond to the line SL of the line of the shadow SH, of x, the deviation that can be claimed has an error of 11 times the width of the shadow SH This causes a problem in that correct position correction cannot be performed.

This invention has been made in view of the above-mentioned background, and provides an image that can prevent the above-mentioned error from occurring regardless of the sealing direction of the sealing line of any stacked workpieces. The purpose is to provide processing equipment.

Structure Therefore, the image processing apparatus according to the present invention detects two edges of a line-shaped object to be imaged based on the image data sequentially read out from the storage means. A second edge detection means is provided to detect the first edge. Either one of the second edge detection means can be selected, and the deviation between the two is calculated based on the read address of the image data corresponding to the edge detected by the selected edge detection means and a predetermined reference address. Configure it to do so.

Embodiments of the present invention will now be described with reference to FIGS. 3 and subsequent figures of the accompanying drawings.

FIG. 3 is a configuration diagram of a mechanical section of a position correction device to which the image processing device according to the present invention is applied.

In the figure, workpieces (plate materials) W3. Arm 1 of playback type robot 1 as a moving device that moves along a work trajectory predetermined by teaching corresponding to ceiling line SL2 at W4.
At the tip of a, a position correction bag @2 excluding the control system constituted by the various parts described below is attached.

The arm 1a is designed to rotate in the direction of arrow A, but since any known robot including the other parts of the robot 1 can be used as is, the details will be omitted.

In the position correction device 2, the bracket B attached to the tip of the arm 1a is equipped with a pulse motor 5, a speed reducer 6, etc. for swinging and rotating the mounting plate 4 around the axis R. When the pulse motor 5 is rotated in the illustrated posture and the direction of the axis R matches the sealing direction B, which is the moving direction of the robot 1, the mounting plate 4 is rotated in a direction perpendicular to the sealing direction (hereinafter referred to as "
The position correction direction is referred to as "position correction direction").

This mounting board 4 is equipped with an N bit (
For example, a 512-bit) -dimensional image sensor camera (
(hereinafter abbreviated as "camera") 7 and a high-pressure spray type sealing nozzle 8 were installed under the following conditions.

That is, the camera 7 is configured such that the arrangement direction of the one-dimensional image sensors therein is parallel to the rocking rotation direction of the mounting board 4, 7-, that is, the plane containing the position correction direction C, and the bracket 3 is placed in the illustrated posture. , the linear imaging field E of the camera 7 shown by the broken line faces in a direction perpendicular to the ceiling direction B, and the optical axis of the camera 7 and the swing rotation axis R of the mounting board 4 are aligned one-dimensionally. The sealing nozzle is installed so that it intersects at right angles at the center position (center pixel position) of the image sensor, and the optical axis is coaxial with the rotation center of the arm 1a when the mounting board 4 is fixed at the reference position for rocking rotation. 8 is installed so that its tip 8a is located in a plane including the optical axis and the axis R, and is located at a position slightly offset rearward from the optical axis with respect to the ceiling direction B. The filler is sprayed onto the sealing line SL2.

The sealing nozzle 8 is attached by a holder S, and the sealing nozzle 8 is attached by this holder 9.
and a flexible hose 10 for pumping the filler.

Next, the spot lamp 11 is installed at the tip of the stay 12 attached to the fixed speed reducer 6-8, and is directed from diagonally above the front left side of the Ziering twistle 8 in the sea link direction B, which is the moving direction of the robot 1. A parallel light beam (spot light) is irradiated onto a predetermined area including the imaging field of view E,
A shadow SH2 of a constant width is provided on the work W4 side along the ceiling line SL2.

The spot lamp 13 is also provided at the tip of the stay 14 attached to the reducer 6, and is located at a position symmetrical to the spot lamp 11 with respect to the optical axis of the camera 7. Used when the alignment is opposite to that shown in Figure 6.

Next, with reference to FIG. 4, an embodiment of the image processing apparatus according to the present invention used in the control system of the position correction device 2 will be specifically described.

First, to explain the signal output from camera 7,
The camera 7 generates an image (video) shown in FIG. 5(a) according to the brightness of the imaged area that has entered the imaging field of view E in FIG.
It outputs a signal Sv, a clock pulse signal Sc for scanning shown in (b) of the same figure, and a video clear signal So shown in (c) of the same figure.

The image signal Sv is an analog signal, and the brighter the imaged part, the higher the signal level, and conversely, the darker the imaged part, the lower the signal level. Further, the clock pulse signal SC is a pulse signal that occurs N times "t" per one frame of the image signal Sv, and the video clear signal So is a signal that occurs 1'' for a time corresponding to one frame.

However, the scanning direction of the one-dimensional image sensor of the camera 7 is always from the edge pixel located on the left side toward the edge pixel on the right side when facing the sea link direction.
.

Next, the lamp lighting direction detector 15 detects whether or not the spot lamps 11 and 13 shown in FIG. outputs a lamp lighting direction signal a which becomes 0'' when the lamp is lit.

The lamp lighting direction detector 15 can be easily constructed, for example, by a circuit that detects whether or not current is flowing through the spot lamps 11 and 13.

However, the lighting control of the spot lamps 11 and 13 is performed by a control section (not shown) of the control system of the position correction device 2, for example, as follows.

That is, when teaching the robot 1 according to the ceiling line, the instructor operates a teaching lamp switch (not shown) according to the stacking of the workpieces on the ceiling line and the ceiling direction, and turns the spot lamp 11°. 13 is turned on, and a lamp lighting direction signal a output from the lamp lighting direction detector 15 at that time is stored in a memory (not shown) corresponding to the ceiling line concerned.

When the robot 1 plays back, data indicating a ceiling line to be played back is received from the robot control section, and a lamp lighting direction signal a corresponding to the data is read from the memory.

Based on the read lamp lighting direction signal a, a=
If it is "1", the spot lamp 11 is turned on, and if a="O"11-, the spot lamp 13 is turned on.

Next, in the image processing device 16 that performs the positional deviation detection function, the binarization circuit 17 inputs the image signal Sv from the camera 7 and detects the shadow SH2 of the sea link line SL2 in the imaging field of view E in FIG. Image signal Sv corresponding to the projected image
(Refer to Figure 5 (a)) is detected, and a width corresponding to the width of the detected notch is zeroed in one frame.
'' is output as time-series image data Dv (see FIG. 5(d)).

Note that in this time-series image data Dv, the direction of the arrow δ corresponds to the scanning direction of the one-dimensional image sensor of the camera 7.

The data selector 18 is for selecting either the write address WR from the first counter 19 or the read address RE from the second counter 22, which will be described later. (See Figure 5 (c)) is 1'', write address WR is ``o''
The read addresses RE are respectively selected at the time of .

12- The first counter (binary counter) 1B is the camera 7
It is cleared every time the video clear signal So from the camera 7 falls to 0'', and the write address WR, which is a count value, is incremented by +1 every time the clock pulse signal Sc is input from the camera 7.

Therefore, the write address WR output from the first counter 19 changes from "0" to rNJ in synchronization with the clock pulse signal Sc output from the camera 7.

Note that when a 512-bit camera 7 is used, if a binary counter of at least 9 bits is prepared as the first counter 1 day, the write address WR will be 00.
It changes from 0■ to IFFH.

The image memory (capacity of at least NX1 bits) 20 as a storage means is stored when the video clear signal SO from the camera 7 becomes 1'' and the data selector 18 selects the write address WR from the first counter 1S. When you are there,
Based on the write address WR, the image data Dv from the binarization circuit 17 is sequentially stored for l frames, and the video clear signal SO from the camera 7 becomes 0'', and the data selector 18 selects the data from the second counter 22. When the read address RE is selected, the stored image data Dv are sequentially output in the order based on the read address RE.

Although shown in a simplified manner in FIG. 4, in reality, a signal obtained by inverting the video clear signal So, for example, is input to the R/W terminal of the image memory 20, so that the video signal So becomes '1'.
When the value is ``, the write mode is set, and when the value is 0, the read mode is set.

The readout circuit 21 includes a second counter 22 configured by, for example, a presettable 9-bit binary counter;
A lower limit address setter 23 that sets a lower limit address MXo for window processing that defines a reading range within one frame of image data Dv, and a clock pulse ck of a constant period.
an oscillator 24 that outputs a
Successive address R which is the count value of the second counter 22
E and the upper limit address M set in the upper limit address setter 25
x 1 and outputs a clear signal CL only when RE≧M x 1.

NO to invert the W video clear signal SO from camera 7
AND @ path 2 which takes the AND of the T circuit 27, the clock pulse ck from the oscillator 24, and the output SO of the NOT circuit 27.
It is composed of B, etc.

The second counter 22 is preset with the lower limit address Mxo set in the lower limit address setter 23 every time the video clear signal So from the camera 7 falls from 1'' to "o", and the video clear signal So becomes 0. '', and the output So of the NOT circuit 27 becomes 1'', which causes the opening<
Every time the clock pulse ck from the oscillator 24 is input through the AND circuit 28, the successive address RE, which is the count value from the lower limit address M x O, is input at each falling edge.
Increment by +1.

Then, when the read address RE reaches the upper limit address M x H set in the upper limit address setter 25 and 15-RE≧M x 1, the comparator 26 outputs a clear signal C.
L is output to the second counter 22 to clear the read address RE to 00 HJ.

Therefore, the read address RE, which is the count value of the second counter 22, is determined when the video clear signal So falls from 1'' to 0'', that is, when the image data Dv for one frame is stored in the image memory 20. From the point in time, the address changes from the lower limit address Mxo to the upper limit address Mx1 at the cycle of the clock pulse ck.

Note that the fact that the window processing described above can be performed means that there is a guarantee that the imaging position of the shadow SH2 of the ceiling line SL2 will not shift significantly to the vicinity of the end pixels of the one-dimensional image sensor. Yes, and if there is no guarantee, it is sufficient to read out all the image data Dv for one frame.

Next, on the second day of the rising detection circuit as the first edge detection means, although not shown, the video clear signal So is 0''.
It will be cleared when it rises from 1″, and Q,
=゛dotQ2 = -1''-16= The data is preset so that A 2-bit shift register 30 that shifts image data Dv from the image memory 20 at each rising edge every time a clock pulse ck is input, and a Q of this shift register 30.
It consists of an EX-OR circuit 31 that takes the exclusive OR of the output of Q1 and the output of Q2, and an AND circuit 11fi32 that takes the logical product of the output of this EX-OR circuit 31 and the output of Q1.

With this configuration, the image data Dv read out from the image memory 20 in the order of the read address RE described above is
The 01 output of the shift register 34 changes (rises) from the data ``0'' corresponding to the shadow SH2 of the ceiling line SL2 to the data ``1'' corresponding to the portion brightly illuminated by the spot lamp 11, so that the 01 output of the shift register 34 becomes ``1''. '', the AND circuit 32 outputs the first edge signal bl which becomes ``1'' only when the Q2 output becomes ``o''.
One edge of 2 can be detected.

Next, a fall detection circuit 34 as a second edge detection means
Although not shown, the video clear signal So
The data is cleared at the time when it rises from 0'' to 1'', and the data is preset so that Q+ = -0'' and Q2 = -〇'', and each time a clock pulse ck is input via the AND circuit 33. image memory 2 at each rising edge.
A 2-bit shift register 35 that shifts the image data Dv from 0, and an EX-OR circuit 3 that takes the exclusive OR of the Q1 output and Q2 output of this shift register 35.
B, an AND circuit 38, etc., which takes the logical product of the output of the EX-OR circuit 36 and the output of the NOT circuit 37, which inverts the Q1 output.

With this configuration, contrary to the rise detection circuit 2 days, when the image data Dv changes from 1'' to 0'',
The Q1 output of the shift register 35 is 0'' and the Q2 output is 1''.
AND the edge signal b2 which becomes 1'' only when
circuit 38 outputs, thereby causing shade 3
The other edge of 1-12 can be detected.

Next, the selection circuit 6 inverts the lamp lighting direction signal a from the lamp lighting direction detector 15 (N).

A T circuit 40, an AND circuit 41 which takes the logical product of the output i of the NOT circuit 40, and the first edge signal b1 which is the output of the AND circuit 32 of the rise detection circuit 29, It is composed of an AND circuit 42 which takes a logical product with the second edge signal b2 which is the output of the AND circuit 38 of the lower detection circuit 34, and an OR circuit 43 which takes the logical sum of the outputs of the AND circuits 41 and 42.

With this configuration, when the spot lamp 13 shown in FIG. When the day is selected, the spot lamp 1 in Fig. 6 is lit, and the lamp lighting direction signal a is "l", only the AND circuit 42 is opened and the falling detection circuit 19-34 is selected. be done.

Then, the edge signal of the selected detection circuit becomes O
It is output via the R circuit 43.

When the first or second edge signal 1) I r b2 is input from the OR circuit 43 of the selection circuit 3S, the latch circuit 44 uses it as a latch signal and outputs it from the second counter 22 of the readout circuit 21 at that time. Continued address RE
, the shadow SH in the one-dimensional image sensor of the camera 7
It is latched as a read address X indicating the imaging position of the edge corresponding to the second seal line SL2.

The deviation calculation circuit 45 calculates the deviation ΔX between the read address X latched in the latch circuit 44 and the reference address 0FFH (255) written in advance in the data register 46.
At the same time, the deviation ΔX@:K (constant) is multiplied by numerical data and ΔX is output to the controller 47 that drives the pulse motor 5.

As a matter of course, the tip of the seal nozzle 8 and the seal line SL2 coincide with each other, so that the imaging position of one edge of the 20-shade SH2 is at the center of the reference pixel in the one-dimensional image sensor of the camera 7. If the pixel is at the pixel position, the deviation ΔX will be "0", and if the pixel is out of position, it will be a positive or negative value depending on the direction and amount of the positional deviation.

Furthermore, when the constant is given numerical data α to the controller 47, the pulse motor 5 driven by the controller 47 rotates, and the camera 7 also rotates, thereby forming an image of one edge of the shadow SH□. If the position moves by β based on the address value, it is determined by β/α.

The controller 47 outputs a drive pulse according to ΔX to the numerical data from the deviation calculation circuit 45 to the pulse motor 5, and adjusts the optical axis of the camera 7 attached to the mounting board 4 and the position of the tip 8a of the sealing nozzle 8. Make corrections.

Therefore, after performing rough teaching work that only satisfies the condition that the ceiling line SL2 intersects the linear imaging field of view E of the camera 7, using the robot 1 equipped with the position correction device configured as described above. Even if the robot 1 is played back based on the teaching data, the image processing device 16 determines the imaging position and center pixel position of one edge of the shadow SH2 of the ceiling line SL2 captured by the camera 7 on the one-dimensional image sensor. The deviation ΔX from the deviation ΔX is promptly detected, and the numerical data corresponding to the deviation ΔX is outputted to the controller 47, and the controller 47 rotates the pulse motor 5 according to the numerical data and ΔX to rotate the sea link nozzle 8. Since the tip of the seal is always positioned on the sealing line SL2, accurate sealing can be performed along the sealing line SL2.

Of course, even if there is an error in overlapping the workpieces W3 and W4 or an error in positioning them to the work area, such errors can be sufficiently absorbed.

Moreover, in the image processing device 16, the work w3,
If the overlapping and sealing direction of W4 is as shown in FIG. 3 and the spot lamp 11 is lit,
The fall detection circuit 34 is selected by the selection circuit 3,
The second edge signal b2 outputted by the fall detection circuit 34
The image data D read out from the image memory 20 by
The latch circuit 44 stores the read address can be latched to,
For example, only the superposition of workpieces W'31 w and i is the third one.
When the spot lamp 13 is on, contrary to the state shown in the figure, the rise detection circuit 29 is selected and the image data Dv rises from 0'' to 1'' by the first edge signal b1. In other words, the read address X when the image data Dv indicates the edge ED3 of the shadow SH3 corresponding to the seal line SL3 shown in FIG. 7 can be latched in the latch circuit 44. Erroneous corrections that occur will not be made.

In addition, work W3. The superposition of W4 is shown in Figure 3 and m16.
If the sea link direction is reversed from the sea link direction B shown in FIG. The scanning direction of the image sensor is
Even if the edge ED2 of the shadow SH2 corresponding to the ceiling line SL2 shown in FIG. 6 corresponds to the fall of the image data Dv from O'' to 1'' by 180° reversal with respect to the work surface, In this case, the spot lamp 13 lights up and the start-up detection circuit selects the 2nd day, so there is no problem.

Next, another embodiment of the image processing device will be described.

FIG. 8 is a block diagram showing an embodiment of an image processing device using a microcomputer.

In the figure, the image processing device 48 is a central processing unit (C
PU) 4B, program memory (ROM) 50, data memory (RAM) 51, input port (I/P) 52.
The CPU 49 executes a program 24 based on the flowchart shown in FIG. By executing this, it not only performs the same function as the image processing device 16 of the previous embodiment, but also performs the following line switching function, which will be described later.

Note that the input port 52 of the microcomputer 54 receives image data Dv from the binarization circuit 17.

A clock pulse signal Sc and a video clear signal So from the camera 7, a lamp lighting direction signal a from the lamp lighting direction detector 15, and a switching signal e described later from a robot control section (not shown), for example, are inputted to the output port 53.
From there, numerical data ΔI (corresponding to ΔX in the previous embodiment) is output to the controller 47.

The flowchart shown in FIG. 9 will be explained step by step below.

5TEP l Write the start address Mx of the image data storage area of the RAM 51 to the counter CN.

5TEP 2 Video clear signal S from camera 7.

(See FIG. 5(c)) waits until it rises from 0'' to 1''.

5TEP 3 Clock pulse signal Sc from camera 7
Check the fall from 1'' to "o" (see Figure 5 (b)), and when it falls, proceed to 5TEP 4.

The image data Dv from the 5TEP4 binarization circuit 17 is taken in.

5TEP 5 Stores the image data Dv captured in 5TEP 4 at the address of the RAM 51 indicated by the counter CN.

5TEP 6 Counter CN address (address) +1
Increment.

5TEP 7 Video clear signal So is from 1" o
'' and check whether the capture of one frame's worth of image data Dv has been completed. If not, return to 5TEP 3 and repeat the processes from 5TEP 3 to 7 to complete the process. If so, proceed to 5TEP 8.

In this way, the image data Dv for the frame is RA
It is stored in the image data storage area (corresponding to image memory) of M51.

However, when acquiring data for each pixel of a one-dimensional image sensor as described above, the period of the clock pulse signal Sc (for example, 1.36μ5ec) cannot be processed by a normal CPU, so for example, if the clock pulse signal SC is At the same time, the serial image data Dv is divided into m-pitch 1-
It is preferable to convert it into parallel data of m bits and take in the m-bit parallel data at once in synchronization with the m-divided signal.

5TEP 8 Lower limit address M x o to counter CN
and the start address MY of the storage area of the change point address data of the image data Dv are written into the pointer G, respectively.

5TEP 9 Take in the lamp lighting direction signal a from the lamp lighting direction detector 15.

5TEPIO5TEP Check whether the lamp lighting direction signal a taken in 9 is "1" or not, and if it is "1", 5
Go to TEP 11, and if "o" go to 5 TEP 12, respectively.

5TEP1.1 Set register F to 0''.

5TEP t2 Set register F to 1''.

5TIEP13 The following switching signal e is taken in from a robot control section (not shown).

27- For example, as shown in FIG. 10, the seal line 5LxO, 5 formed by overlapping the works W5 to W7
When sealing the lines S L xO and S Lx2 in the two directions indicated by the arrows in Lxl and 5Lx2, for example, the robot control unit selects a point Px~1 between the point Px where the line 5LxO branches into the lines 5Lxl and 5Lx2. A switching signal e, which becomes l'' only while the robot 1 is moving to a point Px++, is generated based on the moving position of the robot 1, and this switching signal e is taken in.

5TEP14 Checks whether the switching signal e taken in by 5TEP13 is ``1'', and if it is ``1'', 5TEP1.
If it is 5 ni or "o", proceed to 5TEPi6 respectively.

5TEP15 Invert the value of register F (if O='l
= .

If it is 1″, set it to ‘O’l.

5TEP16 Write the image data of the address (initially Mx□) indicated by the value of counter CN to register T.

5TEP17 Writes image data at address -28=address indicated by the value of counter CN to register D.

5TEP18 It is checked whether the contents of registers D and T match. If they match, it is assumed that the image data has not changed and the process proceeds to 5TEP22. If they do not match, it is assumed that there has been a change and the process proceeds to 5TEP19.

5TEP1.9 Write the contents of register D to register T.

5TEP20 Check whether the contents of registers D and F match or not, and if they match at 0″, change from 1″ to 0″
If there is a falling change from "o" to "l", and if they match at "1", then there is a rising change from "o" to "l", then 5TEP2.
Proceed to 1, and if there is no match, proceed to 5TEP22.

5TEP21 Address data of the falling or rising change point of the image data indicated by the value of counter CN■
is stored in the address of pointer G (initially My), and then the value of pointer G is incremented by +2.

The reason why the value of pointer G is incremented by +2 is because if the camera 7 is 512 bits, the address data (2) will be at least 10-bit specification data. This is because an address area of 2 by 1 is required.

5TEP22 Increment counter CN by +1.

5TEP23 Check whether the address indicated by the value of counter CN matches the upper limit address M x l. If not, return to 5TEP17 and 5TEP1
Repeat steps 7 to 23 and if they match, 5TEP2
Proceed to step 4.

Therefore, the outline of the processing from 5TEP 8 to 5TEP 23 in Section 2 is summarized as follows.

That is, the robot 1 moves the seal line 5LxO in the overlapping works W5 to w7 as shown in FIG.
, 5Lx2 in the two directions indicated by the arrows, the spot lamp 11 shown in FIG. 3 lights up and the lamp ignition direction signal a is "
1'', the switching signal e is 0'' and the register F is also 0'' until the robot 1 reaches point Px-1.
It is.

Therefore, during that time, the sealing line 5L in Figure 10
Edge E DxO of the shadow 5HXO corresponding to xO Address data (2) at the rising change point from 1'' to "o" of the image data Dv indicating the edge E DxO is stored in the address (My) indicating the value of the pointer G.

When the robot 1 reaches the point Px-1, the switching signal e
becomes 1'' and reregister F also becomes 1''.

Therefore, during the second period, until robot 1 or point Pxx in Figure 10 is crossed, the edge ED of the shadow 5HXO
Only the address data ■ at the rising change point from ``0'' to 1'' of the image data Dv indicating ``10'' is at the point l'-P.
Beyond xx, the edge EDy1 of the shadow S Hxl on the workpiece W6 and the edge EDy of the shadow 5t-1x2
Two address data I, I at the rising change point from 0'' to 1'' of the image data Dv indicating 2, respectively, are stored at the addresses (My, My+2) indicated by the value of the pointer G, respectively.

31- Then, when robot 1 or point Px++ is exceeded,
The seal line 5Lx remains unchanged until the switching signal e becomes 0'' and the register F also becomes ``o'', thereby causing the shadow 5Hx1 in FIG.
Image data Dv showing shading 5Hxl corresponding to 1,5Lx2 and edges E Dxl and E Dx2 of 5Hx2, respectively.
When the shadow 5Hx1 disappears from the imaging field of view E, only the address data ■ indicating the edge E D x 2 is shown from the value of pointer G. Address (My, My+23
is memorized.

Incidentally, if the spot lamp 13 in FIG. 3 is lit, the lamp lighting direction signal a becomes 0', and the switching signal e is 0', the image data Dv changes from 0'' to 1''. Address data I at the falling change point is stored, and if a="O" and e="1", address data I at the falling change point from '1' to 0' is stored.

5TEP24 Search for the one closest to the reference address Io (OFFH) from among the address data 32- stored in the RAM 51, and based on the searched address I, reference address ■o, and constant μ, numerical data ΔI= μ
(■Io) is calculated.

5TEP25 Output the numerical data ΔI obtained in 5TEP24 to the controller j to the roller 47 and return to 5TEP1.

By doing this, not only the same effect as in the previous embodiment can be obtained, but also, for example, as shown in FIG.
Even if the line is branched into x+ and 5Lx2, the position correction device @2 will not lose sight of the line to be followed due to the following action.

That is, up to the point Pxx between the points Px-I and Px++ that sandwich the branch point I-Px shown in FIG.
= o of image data Dv indicating O, E D 'I 2
Address data at the change point from ``to 1''■
Based on the reference address ■o, the sealing nozzle 8
Since the position is corrected so that the tip of the edge is along the edge ED y o r E D y 2, even if there are two falling change points due to the edges EDy and I EDY2 after reaching the point Pxx, the one closest to the reference address o is By searching, the address data {circle around (2)} corresponding to the edge EDy2 is selected, and thereby the position is corrected so as to continue along the edge EDy2 up to the point Px↑Nawate.

Then, when the position is corrected along the edge EDy2, after reaching the point Px↑1, the shadows 5Hx1, 5Hx2
Edge EDx1. Image data Dv corresponding to EDx2
Even if a falling change point from I'' to 0'' is detected, the edge E corresponding to the sea link line 5Lx2 can be found by searching for the one closer to the reference address IO from among those address data. The address data (2) of D x 2 can be selected, thereby preventing the sight link line 5Lx2 from being lost.

Note that there is a shadow SH between points Px-+ and Px+1
x.

There will be a work error by the width of 5Hx2, but the point Px
There is no problem if the distance between -+ and Px+1 is made as narrow as possible.

Further, in each of the above embodiments, an example in which the image data Dv formed by the binarization circuit 17 is processed has been specifically described, but the invention is not limited to this, and instead of the binarization circuit 17, the image data Dv is processed. The image data may be formed using a D converter.

However, in this case, instead of detecting the rise or fall, the line is detected by detecting whether the image data indicating the brightness or darkness of the imaged object has changed in the direction of increasing or decreasing by more than a certain value. It is necessary to detect the edges of the object to be imaged.

Effects As described above, in the image processing apparatus according to the present invention, the first and second edges of the line-shaped imaging target are respectively detected based on the image data sequentially read out from the storage means. A second edge detection means is provided to detect the first edge.
Either one of the second edge detection means can be selected, and the deviation between the two is determined by the read address of the image data corresponding to the edge detected by the selected edge detection means and a predetermined reference address. Since we decided to calculate
For example, if the position correction device using this image processing device corrects the position of the sealing nozzle of a sealing robot, the sealing line of any stacked workpieces will be free from errors regardless of the sealing direction. Particularly, even when the light irradiation angle of the spot lamp in the embodiment described above becomes shallow and the width of the shadow becomes wide, accurate sealing work can be performed regardless of the width of the shadow.

[Brief explanation of the drawing]

1 and 2 are diagrams for explaining the background of the present invention, respectively. FIG. 3 is a configuration diagram of a mechanism section of a position correction device to which an image processing device according to the present invention is applied. Embodiment 36 of the image processing device according to the present invention used in the control system of the position correction device shown in FIG. 7 is a diagram for explaining the operation of FIG. 4, FIG. 8 is a block diagram showing another embodiment of the image processing apparatus according to the present invention, and FIG. 9 is an explanation of the operation of the CPU shown in FIG. 8. FIG. 10 is a flowchart used to explain the operation of FIG. 8. 1... Robot 2... Position correction device 5... Pulse motor 7... Camera (imaging means) 8... Ceiling nozzle 11, IES... Spot lamp 16... Image processing device 17.・Binarization circuit 20・
... Image memory 21 ... Readout circuit 2 days ... Rise detection circuit (first edge detection means) 34 ... Fall detection circuit (second edge detection means) 3 days ... Selection circuit 45
... Deviation calculation circuit 47 ... Controller Fig. 6 Fig. 7 JP-A-120477 (14)

Claims (1)

[Claims] 1. 1 frame 4 minutes based on an image signal output from an imaging means using a one-dimensional image sensor that images a linear imaging target with a linear imaging field directed in a direction intersecting the linear imaging target. a storage means for sequentially storing image data of the image data, a reading means for sequentially reading the image data from the storage means, and detecting one edge of the line-shaped imaging target based on the image data sequentially read by the reading means. a first edge detecting means for detecting the other edge of the line-shaped imaging target based on the image data sequentially read by the reading means; a selection means for selecting one of the edge detection means; a readout address of image data corresponding to an edge detected by the first or second edge detection means selected by the selection means; and a predetermined reference address; 1. An image processing device comprising: deviation calculation means for calculating a deviation between the image processing device and the image processing device.