JPS62501524A - Video processing device - 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] Video processing device The present invention relates to an image processing device as described in the first part of claim 1. It is something.

For example, measuring the characteristics of the image of an object sensed by a television camera. Great efforts are made to determine the location and orientation of objects. It has been paid and is now being paid. In addition to military, medical, and space engineering applications, Telemetry technology is very attractive for automated work within the mechanical industry.

Interest in the field of workplaces is particularly focused on the fact that many rrL-like tasks are still artificially performed. It arises from the fact that it has to be done. Workers' vision and related lines It is often very difficult to replace dynamic coordination with technical means. relative to the environment At present, it is possible to significantly reduce the labor force by using artificial vision. Until such systems become available at reasonable costs, implementation will be difficult. It seems to be difficult.

The field of mechanical production is a suitable field of image processing, because it is difficult to measure What kind of image of the object is produced by a particular image sensor? This is because in many cases they are well aware of what it looks like.

Also, make sure that the object is visible clearly and with good contrast to the background. In many cases, it is also possible to configure the lighting and background.

Measuring machines can provide fairly good reproducibility for routine measurements, but lighting Due to variations in the image due to variations in and the actual placement of the object relative to the optical axis, Difficulties may arise. The market for simple video processing equipment is limited due to frequent failures. Occupancy rate is small. The development of sophisticated systems in this field of technology is particularly important for industrial robots. It continues with the aim of collaborating with One known robot vision system is It is sold by ASEA of Denmark. Video processing that is part of this system The device uses, among other things, digital filtering and picture noise suppression algorithms to Extracting contour lines from TV camera pictures. This system uses contrast Can handle scenes with very weak From the outline of the object in the picture Location and identity are determined, and certain criteria The orientation of the object relative to the (contrast) direction is determined. During this process, the contour line is It is compared with the contour data previously obtained from the object, processed and stored in the device's memory. be remembered.

Other known devices measure ( A memo containing data detailing the features of the picture of the object (such as spots) In order to perform calculations similar to those described above, the desired processing occurs within the library.

The latter type of system works by using a reference object taken from the production line. - It is too much for the worker who is responsible for “teaching” the equipment how to solve the problem. You will have to make certain demands. An example of such a system was first developed by MIT in the United States. ``Maehfne Vision'' is a ``machine vision'' developed by

Already known systems are complex and expensive. The purpose of the present invention is to to solve a type of work problem for which simplicity and low cost are essential requirements. That's true.

Despite the low cost requirement, this type of system does not require special training. It is desired that the work tool be operated in a manner that is easy to understand and handle.

” The bipedal requirement is fulfilled according to the invention by means of the device characterized in claim 1. It is solved. Other improvements and developments are described in the second and subsequent claims. There is.

Today, graphics processing terminals operate on a so-called rask scan method. In other words, the display file The field is scanned line by line using television technology. Graphic processing terminal market is big. Used to create computer-controlled images Advanced circuits are therefore available at truly reasonable costs.

According to the present invention, an inexpensive yet powerful electronic processing unit is created using graphical display technology. provided. By using the above components, the present invention can achieve A large number of windows (or picture blocks) that can be formed or placed in The fact that it is easy to produce a picture (picture 5ection) Use the fruit. Also, when controlled by a microprocessor, the required window It is also possible to obtain a circuit unit that can generate a pattern at a low cost. be.

Windows commonly occur on television cameras or other similar image-producing sensors Used to control operations on digital video signals. Each window is Processed individually during picture scanning. Once the frame has been scanned, the computer The data processor obtains the set of results of individual window operations and outputs them as output from the device. The combination that is sent produces a tailored final result.

Already known systems for extracting contours perform contour extraction by filtering the image. Trust gradients are encouraged. Due to this filtering operation, the system The sensor is sensitive to noise and such noise does not give information about the position of the contour pixel: a pixel on a video screen that can change color or intensity It becomes necessary to include means for suppressing the minimum area (abbreviation for minimum area). According to the present invention If so, there are several measurement sections (windows) that can be formed individually within the picture frame. (for example, 256 pieces).

The present invention strongly compresses the information content of the video, making it much more random than conventional systems. It works with filters that use cumulative addition, which is much less sensitive to noise.

The invention will now be described in detail with reference to the accompanying drawings.

FIG. 1 shows a block diagram of a first embodiment of the means according to the invention: FIG. 2 shows a first embodiment and a television display on which the object to be measured is placed. – indicates the relative placement of the windows above; Figure 3 shows how numerical interpolation is used to calculate a more accurate value for the azimuth of an object. ; Figure 4 shows numerical interpolation to calculate more accurate angle values when the object rotates. Indicates a mode in which it can be used; FIG. 5 shows an extended variant of the window shown in FIG. Figures 6a and 6b show the type of window shown in Figure 2. formed as disks in the center and off-center of the pattern, respectively. Show the histogram extracted from the object: FIG. 7 shows a block diagram of a second embodiment of the means according to the invention.

The means according to the invention operates in window sets (window 5et), each A window has a distinct form but is part of a pattern formed by all windows It is. According to the invention, the entire window pattern is aligned across the television frame. It is progressively movable and rotationally symmetric. The TV frame also shows its position and orientation. There is an image of the object to be measured. The window pattern has its center inside the object. First, move horizontally (,) and vertically (y) on the TV display until you agree. The function of the means when this scrolling is completed will be explained first. I will clarify. Then some steps to achieve this X-scroll and y-scroll Explain the method.

In the block diagram of FIG. 1, the television camera I is connected to a rotating belt or similar device. The focus is on the object 2 being transported by. camera l directly to object 2 or indirectly through some kind of movable mirror system, said movable mirror system being The motion of the object is corrected in a known manner during the process, and is not important to the invention. , so it will not be discussed here. During the measurement period, the object does not move with respect to the optical axis. , or move so slowly that they do not significantly affect the accuracy of the measurement. camera l The video signal from the preprocessor unit (preprocessor u nit) 4 in a well-known manner and analog-digital (A /D) Converted into a series of digital signals by converter 5. digital window The memory 6 stores the sizes and positions of various windows. In principle, Wi A window is the definition of a certain number of specified but arbitrarily located pixel addresses. be.

This window memory stores a series of digital data (signals) output from the converter 5. Read in sync with power. The contents of the memory allow the operator to draw the specified rectangle. specified for each pixel within a region, and the region covers at least a majority of the picture. ), each pixel within the rectangle is then It will have a unique counterpart in the operator memory 7 connected to the processor 8.

The window located in this area has a function for the equivalent pixel grayscale value. Used to specify dynamic operations. The output of window memory 6 is addition memory Connected to - of the input of 9. Addition memory 9 stores dedicated memo for each window. Has a refield. One of the inputs of the adder circuit IO is the A/D converter 23i5. The second human power is connected to the output of the ^/D converter 5.

A window operator dedicated to each pixel within a window Begin cumulative addition of all grayscale values of included pixels. These values are A value is entered into the accumulation register designated to the window and forming part of the addition memory 9. The adder circuit 10 adds the signals in the order shown in FIG. The operator that characterizes the window is The current pixel value is added to the last recorded sum. Consists of a dress.

A video signal representing the entire television picture (or the portion defined by the entire active rectangle) When the number is processed by the window operator (in real time), the addition memory 9 There is a sum stored for each window in the register. The sum of these is the object The image contains information characterizing its position and orientation. However, the window is formed and positioned in a suitable manner within the picture frame. Piku is displayed on a monitor unit 8I, which monitor unit preferably It is a Levi screen.

One condition that can be attached to the shape of the window is that the image of the object is almost This requires generating a series of window sums that are similar and recognizable. When , the object is translated and rotated within a given range. Depending on the video processing plan It must be possible to separate motion and rotation.

Rotate the window around one artificially selected center point in the picture. can be identified by placing them rotationally symmetrically.

In Figure 2, a rectangular (thin) object M is shown, which is near one of the short sides. It has (dark) holes drilled in it.

M is positioned so that its center coincides with the center of the window pattern. this window In the realization of the pattern, the window is a circular sector (c’trcle 5ect or) A　r−A　lt, and these sectors are circles with center point C! form 2. The radius of a circle is determined by the area of the circle whose characteristic details are divided into sectors. should be chosen to be large enough to be covered in minutes.

The number of sectors is - the width of the sector is approximately the size of the characteristic detail defining the orientation of the object. They are chosen so that they have approximately the same dimensions.

Each sector? In order to give as much characteristic information as possible, the inner circular part Area 13 is cut out from the sector.

In summary, the operator assumes that the partial sums provide characteristic information of the object of current interest. form the window Δ,−, which actually performs the filtering function. Select the number of sectors, inner radius, outer radius and radius of the circle for the cake pattern. child These partial sums can be represented, for example, by a histogram such as 4' shown in Figure 3. . These sums A are preferably stored within each truncated cone sector before entering the histogram. Normalize preferably by dividing by the number of pixels included. ) (the same symbols are used here as in Figure 2), and as desired. For example, the highest value is set to a suitable value, e.g. 32 or 64, and all other values are set to this value. It is normalized in that its size is determined linearly. Most selected The high value depends on the resolution of the A/D converter 5 and other factors. Dependent.

Divided by the periphery which is the radial extension of the periphery of the inner circle A　I-A　+　t However, if other circular rings are generated, the U of this subset The window is used to some extent to compensate for non-uniformities in illumination and/or playback. can.

If the background is a uniformly reflective material, the pixel grayscale of the outer ring The normalized sum of values is the illumination level If in each sector sensed by the camera It is a measure of

Such a coarse correction is obtained by dividing the sum Ai by the associated (normalized) sum Bi.

A reference object carefully placed at the desired location and orientation before the actual measurement occurs. The measurement is then carried out. As shown in Figure 3, using normalization and Mi positive or As mentioned above, the histogram of the generated object is processed by the microprocessor. 8 is stored in a memory 14 connected to 8.

During subsequent measurements, camera I is equipped with a new, similar camera with the same moonlight, background optical conditions, etc. Play the objective.

The object is at the same point as in the reference exposure that generated the stored histogram value. Surrounding circular ring A, -Δ4, B+B, located with center relative to +P It is initially assumed that That is, here the window pattern in the TV picture is Assuming scrolling has already been achieved. However, the object to be measured is , the foreground is rotated in an unknown manner compared to the reference picture. Another histogram Measured from this image of the window pattern where the ram A'1 does not change compared to the reference measurement. and calculated. This second diagram is then similar compared to the first one. It will also take a shape with a displacement of 17. This displacement is expressed as a number of sector cone angles. Two bikuji, which is calculated? irL tangential measure of the relative angle between objects at be. This is due to the correlation, from the position a(k) to the maximum of the hit F(1()) A monkey with calculations.

It is important to note that the object contains characteristic details that are appropriately distributed with respect to the sector divisions. If the situation is favorable, for example a second order approximation More accurate angular resolution can be obtained by numerical interpolation using can get.

Another more direct way to obtain more accurate angular resolution is to use two perpendicular lines a, b As in the case between Select a portion of the histogram curvature shown above.

For example, this part of the diagram can be divided into l sample intervals (-1 angle steps). Move to the right (shown as a dashed line in Figure 4) and calculate the area (shaded) between these two curves. If we calculate rI (with shading), we get the measure of sensitivity for rotation 7 expressed in units of area/angle. degree can be obtained. When measuring a new object, correspond to the sector cone angle Orientation is calculated with angular inaccuracy. Then, the part of the reference histogram and Select the same part of the histogram curve and rescale the new curve. ale) L, the new curve and the reference curve have the same amplitude swing (amplit) udesving) and the two curves in the same manner as in the calibration above. Generate area in between. This area value can be directly converted to an angular difference, including the sign. . This method improves the angular resolution by a factor of 2-10, depending on the slope of the histogram. It is possible to do so.

Two steps in the histogram: 1) the amplitude change of the river is limited; Assuming that the curve becomes continuous if the number of vectors is made sufficiently large, as a modified example, This is possible by using numerical interpolation. In that method, it is calculated by numerical interpolation. The sample values are placed equidistantly along the circular ring, so that the sample values are Corresponds to a set of sectors several times more than patterns. The reference curve and the problem curve In favorable circumstances, one obtains a higher color resolution if is lengthened in a corresponding manner. Is possible. The method of Moe-sho is the same in all cases. A suitable number of (N 30), the entire process of calculating this two almost similar objects is 5 Half the frame rate of the standard TV camera of 011z It can be carried out at 20 m5ek or less. This method uses pixel data is processed in real time, resulting in a highly condensed and reduced (7°) amount of data Only the sums Ai and A'i and possibly Bi and B'i are stored.

Another method of realizing the window pattern is shown in Figure 5 1. :It is shown. Yen 15 is It is divided into 32 sectors. Circle 15 is also divided into 8 rings . Each window in this window pattern is therefore generated by a truncated sector, The number of windows is 8x32.256. reference for a certain object to be measured. Then, in order to obtain the histogram according to FIG. In order to correspond to l1 and B l-B +2, make an appropriate ring of the pattern shown in Figure 5. You can choose. The complete window pattern as shown in Figure 5 is i! forever to window memory 6! I can be absorbed. Add after picture exposure As shown in FIG. It is the microprocessor 8 that generates the histogram.

Only permanently placed interfering objects are reproduced in the same scene as the object during the exposure during measurement. can be done. For example, these objects should be It is "deactivated" by positioning and shaping the circular ring window. forever Interfering objects can be created in other ways, such as by reproducing background and interfering objects only in AO2. and the sums Ai and A, respectively, with all other points being the same. 　'　　　　　　l)s can be subtracted from. Background corrections are already known; Further, although it is not a part of the present invention itself, it will not be described further.

In the above, at the moment when ρI is constant, the object is related to the base or the object in the measurement. It is assumed that it is already rotating.

However, the object is in the picture of the window pattern seen on the screen. It is possible that the person will not be taken care of.

However, in class A applications where this means is used, the object to be measured is or other objects translated from the center by a distance of approximately 0.51 times the maximum dimension of the object. Not possible. It is the external tri-lagging of the TV measurement system. A mechanical hand that selects and guides the appropriate moment for exposure (by triggering) This is accomplished using steps. The entire object being measured is the TV frame at the moment of exposure. must be assumed to be within the valid rectangle of .

The object is not centered and the picture is not centered as assumed when the bearing measurement was described above. If we translate and rotate within the camera, the calculation is done in two steps. child The last of these steps corresponds to that described above.

The first step of indicating whether the object has been translated is based on the following. circle When a shaped (bright) object is viewed against a (dark) background, a histogram Δi occurs; In principle, a histogram can be used if the center Ci of the window pattern is the center of the object. , it consists of N bars all of the same height (see Figure 6a). If this When their centers are mutually displaced, the bar diagram becomes a whole sine cycle (whol (see FIG. 6b).

Before the reference curve Ai is recorded, there are several bits in the direction of the sector where the circular reference object is located. The value of the translation of an object that is subsequently measured when moved by a distance equal to the xel (given in units of pixel size) is the correction coefficient calculated as described above. It can be calculated approximately from the curve. The difference between the maximum and minimum values of the correction curve is (uniform object, uniform background), the displacement of the object with respect to the calibrated distance of the object The six variables that make up the multiplicative value (if the direction of α is toward the sector, the sum Ak is the displacement increases in proportion to the square of the value of . At the same time, the sum of opposite sectors decreases as well.

The same holds true when a tailored object moves within the scene. elongated When a shape is translated and rotated, a curve that hardly matches this approximate relationship still exists. bu may occur.

In the general case, an approximation to the translation on the image plane is that the center is relative to the center of sectoring. The measured signal resulting from a measurement with a reference object when it is displaced by a certain distance It can be calculated from a comparison with A'i or A+i. (The reference object is from the center It is also possible to store the image in the reference object when there are various degrees of deviation. . ) To save the cost of the digital memory of the TV frame, the sum A′i is As before, it is advantageous to calculate in real time. The next frame will be after 20m5ek ( The circular ring window pattern is , using the so-called scrolling technique or the entire window matrix on the TV screen. the eye by not moving along and across the line, i.e. in the X and y directions. Moved to a new calculated center near the center of the object. The same calculation is for the object of the reference diagram representing the initial moss quasi-measurements with matching centers and window patterns. is applied again to the new sum A″f, which is corrected again for .

If a sine-shaped curve is noticeable in the obtained correction diagram, Perform another interpolation to obtain a new, more accurate center value. The center is almost correct. No, you can calculate the rotation angle in the same way as Moe. The image of the object is fixed on the display. , the window pattern is scrolled to the new position.

If you need very precise corner and position values, use - pairs of sector windows (small radii) to locate the details in the picture and connect these to the details in question. It is possible to scroll to the predicted position of the center of the section. On the other hand, the center The position of is determined by the method described above using successive interpolations. In this latter case, the The operator may have acted in accordance with human instructions when preparing the reference data. We must assume that. Measuring like this gives you more precision and reliability Therefore, a delay of several frames must be accepted. The delay is the bias at the start of the measurement. Depending on the size of the center distance or the required accuracy, a maximum of 4 frame times and a minimum of 1 frame time can be used. time (20m5ec).

A circular ring can be generated in the following manner. Inexpensive graphics utilized by this means The electronic components are provided with a window via the microprocessor 8 and its memory. Load the definition of dough. Window pattern data transfer is not that fast. , it is possible to transport many windows within 20 m5ec, and this is desirable. On the other hand, scroll-type translations are commanded urgently! frame exposure This can be achieved by exposing the next frame. Generate sector window An effective way to do this is to create several Build sector windows by using connected concentric rings (see Figure 5). It is to raise it. In this way, a kind of polar coordinate system is formed. It takes a patter can be included in the instrument as a standard matrix via FROM . When an operator adapts a means to a new reference objective, his job is to Scenery, illumination 1 novel, for measurement of translation or rotation, and also in complex cases? After moss R Decide which circular ring to use for precision measurements, which may be the next step. Is Rukoto.

Two different sets of window patterns can be stored in FROM, and The - pair is shaped as in Figures 5 or 2, and the other is shaped like an ordinary chess. (Cartesian coordinate system). The second pattern is useful when examining translations.

That is, in the form of two sets of strips oriented in one X direction and one Y direction, respectively. By using other x/y operators, the centroid is calculated. strip are X and y included in the mathematical definition of the center of gravity of a (two-dimensional) body (bod y) is an approximation of Heavy calculations can be done by using two frames alternately or by Two patterns must not be combined with ffi, so two signal processors working in parallel are required. This is preferably carried out by using In the latter case, one processor Adds the pixel values through the rip window FROM, and the other processor adds the y- Strip window Pl? Add the pixel values via OM. X one direction and y PRON with unidirectional strip all together instead of cake Akita ring pattern When used one by one, the centroid mode becomes active in calculating the translational position of the object.

At this time, the FROM with the cake-like pattern becomes active again and enters the target object. It is located in the mind and can thus direct the rotation angle of the object.

However, using a single window pattern to determine the "center of gravity" in a television picture It would be advantageous to be able to find and measure the azimuth. The reason is practical Dedicates the window pattern used in the next frame before exposure in the example This is because the time-consuming and costly process of transferring to the memory area can be avoided. the above Such memory transfers through the processor, such as It is preferable to complete the process in less time than required. this is a tough job Ru.

Figure 7 shows the data that was previously recorded and stored using the window pattern shown in Figure 5. Indicates the translation and rotation of an object with respect to a reference object in a more direct manner than described above. 1 is a block diagram of one implementation of an electronic circuit for illustrating. FIG. The circuit in Figure 7 includes The unit of FIG. 1 including window memory 6, adder memory 9 and adder circuit 10 is shown. However, this unit is formed in another manner.

The video signal from the television camera is processed to include permanent background elements, if necessary. is sent to the A/D converter 20 in a known manner, except for the signal portion containing the sampling rate. convert the incoming analog video signal into a series of digital signals, e.g. The converted signal is sent to the adder 22 corresponding to the adder IO in FIG. Appears at the parallel output of converter 20 connected to the input.

Since the image of the object to be measured is displayed on the screen 23, the output signal of the A/D converter 20 The signal from unit 20 is sent to a circuit 24, which receives the signal from unit 20 from the picture screen. It is converted from digital to analog before being transferred to zone 23.

A window pattern is displayed on the screen 23 during use. D/A converter 2 4 therefore has a second human power to which a signal is sent, and this signal indicates the proper position within the vehicle. represents the window pattern with which the signal from converter 20 is input before D/A conversion Added digitally. I didn't want to make the window pattern picture stand out. In order to operate in such a manner that it can survive.

In the basic measurement of the reference object, all window patterns are different in all their rings. Appears in a deep gray color and allows the operator to identify actively operating rings in the device. The selection is made via the microprocessor 25. Each subsequent measurement cycle for the object Later, a signal marking the selected ring is sent to the second power of the transducer 24 and For example - by making one sector darker and the next sector brighter, you can A sector is indicated.

The graphics control circuit 26 has a window memory 27 on the first address bus. This window memory receives the address signal and stores this address. The signal gives coordinates on the screen, but the coordinates have their origin in the window of the picture. The pattern has been transformed to be properly centered. Window memory 27 Each window has its own number of numbers, e.g. the window shown in FIG. If a pattern is used, it will have numbers 1-256.

The output signal from the window memory 27 is controlled by the command from the graphics control circuit 26. The signal is sent to the parallel branch to the "adder and D/A converter" 24, and the graphic means produces a picture on the screen, and this picture is the input from the television camera. It can be observed superimposed on the next scene.

For each pixel within the window pattern area, a number is sent out; Represents the window to which the cell belongs.

This signal is transmitted to the window result memory (wi ndow result full memory) 28 address.

In this way, by using the window number as an address to another memory 28, When the TV camera scans the pixels in the corresponding window, it occurs.

In this method, data in the window memory 27 is stored in the window result memory 2. 8, the shape of which is individually and arbitrarily selected. Enables processing of large numbers of measurement windows and real-time, immediate measurement of signals. It allows television cameras to simultaneously generate and process video signals.

This results in shorter processing times and separate video memory (frame grabber). IIle grribbler)) is not required.

The data bus from the window result memory 28 is connected to the adder circuit 22, The arithmetic circuit 22 calculates the product of the memory cells addressed by the window memory 27. Receive the calculated value R, add it to the value V sent to the second input of the adder circuit, and add this new value The new value R is stored in the same memory cell belonging to the memory 28 as an integrated value. this is , controlled by the microprocessor 25 during various sequences. part of the circuit Serial commands (Scrta!) between various units and the processor 25 Separate control path lines for It has not been.

A new measurement object is brought into position for the measurement sequence within the field of view of the television camera. Once entered, the first operation measures the value of the translation, which is compared to the position of the reference object. That's true.

This can be accomplished in one of the ways described above or by the examples described below. It will be done. FIG. 7 is the same no matter which embodiment is selected for the measurement of translation. .

The processor 25 determines whether the object to be investigated is rotated or translated in one direction of X or one direction of Y. If so, which mode of operation is appropriate for the graphical control circuit via lead 29? instruct.

When translation in the y direction is instructed, the graphics control circuit 26 generates an address and During the transformation of the window pattern for the television picture, the motion in the y direction is There is no motion in the X direction, but there is a putter in the X direction for the sensed vixel. The center of the pixel now lies almost directly on a vertical line passing through that pixel. This is Of course, it is not possible to view it on a screen because the television screen is scanned. This is because the translation of the pattern follows the path of the electron beam. on the screen , a wide, bright horizontal band can be seen. This smooth one direction The movement will cause the window memory to move closer to the vertical line, perhaps even completely vertical. A window in the window result memory 28 that is slightly close to the left or right of the straight contour line. Only the doe will have an integrated value recorded after scanning all frames. after that, Window riser corresponding to a nearly vertical radial array of cells in window memory A calculation similar to the "center of gravity" calculation is performed using only the cells of root memory 23, and When the information is accumulated and recorded, the displacement in the y direction with respect to the base position is calculated. be remembered. In the next step, the window pattern is set to the expected position in one direction, y. is scrolled to.

During the next picture frame in the cycle, the window pattern transformation is It is not moved in the 3' direction, but the TV screen is moved in the same manner as above in one direction. is moved against. The accumulated sum is an almost horizontal half in window memory. Stored in the cell of the window result memory 23 corresponding to the radially arranged window be done. The displacement in the X direction is calculated, and the window pattern is scrolled in the X direction. Rolled.

According to the present invention, a part of an object having a known position and orientation is moved to a predetermined position of the object. It is also possible to find out whether a feature in a location is striking or apricot. Target position and orientation can be derived using the method described above, but it is necessary to find an object that has an accurate predetermined position and orientation. It is also possible to read. The purpose of the measurement is, for example, for certain types of quality control. Check whether the origin has the correct predetermined location or This can be used to check whether a manufacturing operation has actually been performed before proceeding to the next manufacturing process. You can. In the last mentioned case, the basic work is carried out by machinery or tools. The aim is to avoid damage to the equipment. For example, the right location and/or A hole having a diameter of 1 mm is inserted in front of the threaded tap. It is not possible to automatically cut threads unless it is confirmed that the do not have. Another application of this embodiment of the invention is the tip of a syringe. and the precise positioning of it for placing a cover or the like on it.

To achieve this possibility, the same type of electrical A specific window configuration can be stored in FROM in the child module, and the combicoater Can be programmed to shift window configuration at operator request.

A window configuration particularly suitable for measuring drill holes in objects is shown in FIG.

In this case, the windows are arranged in a cross shape, and the cross shape is - one central window 30 to 6, the remaining windows are placed along arms 31 to 34, and the h windows The doe only has dimensions corresponding to I to a few pixels along the arm, but this In the direction perpendicular to this, there exists a dimension corresponding to at least several pixels.

The center of the hole, which usually appears dark on the TV screen, is the reverse grayscale of the image. in individual cells of window result memory for individual windows using Using the stored sums, independently measure the position of the center in one direction of X and one direction of Y. It can be calculated as

To locate the center point, take a measurement, scroll the window configuration to the calculated position, and then Take a new measurement, scroll again, and repeat these operations until a stable close position is reached. It can be repeated by doing this.

The easiest way to find out the diameter of the hole is with the inverse gray scale of the measurement in question. The relationship between the sum of all windows with the sum of and the sum of one in the reference measurement and Standardize by considering the sum of the windows representing the darkest and whitest areas. That's true. Said relationship is a measure of the relationship between the planes of the holes, and therefore also a measure of the square between the radii. degree.

Another method is to draw the contour lines between black and white on the X-arm and y-arm, respectively. Therefore, the diameter can be calculated directly. Sum xi and yi as simple pixel If it can be considered as a value, it is possible to use several known filters of various types. can be used to determine the location of the hole boundary.

The above example provides an accurate measurement of position and the entire object at a known position and orientation. or using the window arrangement according to the invention to take measurements of parts of an object. Show how the principles can be used. This method works on selected parts of the statue. provides highly compressed and integrated information.

Still other embodiments of the invention include continuously running strings of material. Aim the camera l (see Figure 1) at the path of g of s + material) and move the material. You can check whether there are any defects on the surface. Suitable for such material control The window configuration is shown in FIG. 9, which only includes the window line 35. So each window has dimensions of only I or a few pixels along the line. However, in the direction perpendicular to this, there are at least several pixels, for example about 5 to 15 pixels. It has a width corresponding to a cell. The camera is configured so that the window configuration is in the direction of string motion. The material strip should overlap the picture of the string at least approximately perpendicular to the Directed to ng.

An FIG.5 FIG.8 FIG.9

Claims (10)

[Claims] 1. the shape and/or position and possibly orientation of the object or part of the object with a text directed towards said object or towards a place where said object can be placed. Determining and/or examining using digital video signals from a video camera In the video processing device, The window memory (6; 27) is provided with a plurality of predetermined memory cells each having a predetermined shape. windows are stored, each window representing one of the images obtained from said television camera. at least one memory cell for each window. a result memory (9; 28), said memory is connected to said window memory; addressable; a summing circuit (10; 22), which receives the signal from said television camera to a first input; in the window result memory (9; 28) regarding the number and window. receives the contents of the memory cell from the television camera to a second input; Including the picture point of the problem in No. The result is stored in the same memory cell in the result memory, and its contents are It is given to the second input before the arithmetic operation, and is equipped with a comparison device (8; 25). Eh, this means that after each video recording of the TV camera, the window result memory is Compare the recorded value with the stored value recorded for the reference object or reference object part. , determine at least one parameter regarding shape, position or orientation. An image processing device characterized by the following. 2. The x-direction and y-direction position with respect to the reference object or reference object part is , determined by said comparison device during at least a first image recording, and said comparison device The window configuration is related to the reference object or reference object portion in the reference record. arranged with respect to the object or object portion in substantially the same manner as it was arranged with respect to the object or object portion. The result of the above determination is used to determine the window on the image obtained from the television camera. (c) The device according to claim 1, characterized in that the configuration location can be changed. . 3. How to rotate the object or object part with respect to the reference object or reference object part from the result of at least one image recording after a change in the position of the center of the window configuration. 3. The device according to claim 2, wherein the device is adapted to be derived. 4. Any of the preceding claims characterized in that the window configuration is rotationally symmetrical. The device described in any of the above. 5. Windows are set up in a circle called the window circle, which is divided into sectors. , each window is a circular ring within one of the sectors that has the same center as the window circle 5. Device according to claim 4, characterized in that it is a part. 6. The comparison device has a selected inner diameter and an outer diameter to orient the object. The series of values obtained sequentially around all the rings in the window configuration Then compare it with a set of values obtained under the same conditions and make sure that both of these values are at least approximate. The device according to claim 5, characterized in that the displacement between the two is determined in order to . 7. at least two of the reference objects in order to determine the translational position of the center of the object. records, one of which is when the reference object was centered with respect to the window configuration. and the rest are recorded with the reference object eccentric with respect to the center position. , the comparator also compares the set of values obtained in the actual recording with the opening of the position and orientation measurements. At the beginning, the reference record is compared with each of the set of reference records after at least one image record. The mode and degree of eccentricity of the object relative to the target object and the displacement of the center of the object are also found. An apparatus according to any one of the preceding claims, characterized in that: 8. To determine the translational position of the center of the object in one of the orthogonal coordinates in the television image In addressing the window memory, one of the coordinates is the golden image record. have the same coordinate transformation within, but the other of said coordinates is a change that follows the true value of the coordinate during image recording. The comparator (27) has a coordinate transformation that is approximately the same as the coordinates. In the cell of the window result memory (28) corresponding to the window placed in the direction Claims characterized in that a calculation similar to a center of gravity calculation is performed using the stored value. 7. The apparatus according to any of paragraphs 1-6. 9. Store at least two different window configurations in special memory, each stored The window configuration that has been created can be transferred to the window memory. Claims: An apparatus according to any one of the preceding claims. 10. The window configuration is a row of windows, each window having one has dimensions of a few pixels along its column, but has a substantially larger width; The television camera has a window configuration that overlaps the recorded image in the direction of movement of the material path. oriented substantially perpendicularly to the path of the moving material; The apparatus according to claim 1.