JPH056713B2 - - Google Patents

Description

[Detailed Description of the Invention] [Industrial Application Field] This invention extracts a still image of a moving object, performs image processing to inspect the shape and defects of the object, and determines whether it is good or bad and sorts it. The present invention relates to a binarization method for a pattern inspection/discrimination device.

[Conventional technology]

FIG. 5 is a block diagram showing a conventional example of this type of pattern inspection/discrimination apparatus, and FIG. 6 is a timing chart for explaining its operation. In FIG. 5, 1 is an object, 2 is a conveyor, 3 is an imaging device such as a television camera, 4 is an illumination device, 5 is a position sensor consisting of, for example, a light emitting element and a light receiving element, 6 is an amplifier 7, and a binary value. conversion circuit 8, arithmetic determination section 9
This is an image processing device including a timing control circuit 10 and the like. The calculation/judgment section 9 further includes an image memory 91, a feature extraction circuit 92, a calculation processing circuit 93, a position detection/correction circuit 94, a setting device 95, and the like.

The operation of the conventional device will be described below with reference to FIGS. 5 and 6.

The object 1 is conveyed by a belt conveyor 2,
When this reaches a predetermined position within the field of view of the television camera 3, this is detected by the position sensor 5. The position sensor output is used as a test start signal, as shown in FIG. 6B, for example. Now, suppose that the television camera 3 has a built-in shutter and that this shutter is driven in synchronization with the vertical blanking period (vertical blanking period) V _BLK of the television camera 3. In this case, the shutter is 6th
As shown in Figure C, it is in the "open" state only once. Note that the vertical blanking period is shown as in FIG. 6A.
When the shutter is opened in this manner, the optical image of the object 1 is converted into an image signal (video signal) and extracted by the television camera 3. In other words, the camera output of one field (1V; screen) immediately after the shutter is opened is valid, and this is taken into the image processing device 6 as an image signal. The timing and duration of this image capture is shown as D in FIG. The image signal captured in this way is amplified by an amplifier 7 and binarized by a binarization circuit 8. 2
The digitized image signal is stored in an image memory 91, and a feature extraction circuit 92 extracts various feature quantities of the object from this memory 91. The arithmetic processing circuit 93 compares and discriminates each feature value with a set value preset by the setting device 95, thereby determining whether the object is good or bad and sorting (sorting) the object, and outputs the results to a display device or the like. . The timing and period of this determination/calculation are shown as E in FIG. In addition,
A position detection/correction circuit 94 shown in FIG. 5 is provided to detect the amount of positional deviation of the object from the reference point and correct it.

The above is for a TV camera with a shutter, but when using a TV camera without a shutter, prepare a strobe instead of the shutter, and fire this strobe only once at the same timing as the shutter "opens". Accordingly, inspection and discrimination can be performed in the same manner as above.

In this way, shutters and strobes convert continuously moving images into still images, thereby extracting clear images without image blurring and ensuring highly accurate inspection and discrimination.

[Problem that the invention seeks to solve]

By the way, the threshold value of the binarization circuit in such devices is usually set at a fixed level set in advance, but this is because the circuit configuration and processing are easier and the processing time is shorter. It is. However, the lighting device (reference numeral 4 in Fig. 5)
If the brightness of the object (see ) changes over time, or if the reflectance of the object itself changes depending on the individual, the image signal from the camera will naturally change as well. If you try to binarize this at a fixed level, the object image will be significantly distorted, and as a result, the big problem is that the object will be considered a defective product even though it is normal. have.

Note that when there is one piece of image information, for example, the video signal is converted into a digital amount of a predetermined number of bits, the gray level information is stored in a memory in units of pixels, and this is read out again to perform the necessary information processing. There is also a method of determining the optimal level. However, this method has the problem that not only the circuit configuration becomes enormous, but also the processing time for determining the threshold value becomes long.

Therefore, the present invention provides a stable and highly accurate binarization method that is economical in that it takes a short processing time, and that does not cause erroneous judgments due to variations in brightness of a lighting device or the type of object. The purpose is to

[Means for solving problems]

an image extracting means that includes a two-dimensional imaging means for capturing an image of a moving object and extracting a still image of the object; and an image extraction means that detects the position or rotational deviation amount of the object with respect to a reference position within the imaging field of view and positions based on the deviation amount a position correction means for normalizing the object image; a window generation means for generating at least one predetermined shape window at a predetermined position of the normalized object image; and a detection section for detecting at least the gray level of the image within the window area. and a binarization level determining means for determining the binarization level accordingly.

[Effect]

The image extracting means extracts at least two consecutive still images based on a predetermined time, the position correcting means corrects the position of the object image for the temporally previous image, and the window generating means corrects the position of the object image. A window is generated at the position, the binarization level determining means determines the optimum binarization level from the gray level of the image within the window area, and the temporally later image is created using the determined binarization level. It is a simple and inexpensive configuration, and the optimal 2
Make it possible to determine the valorization level.

〔Example〕

FIG. 1 is a block diagram showing an embodiment of the present invention. In the same figure, 11 is a low pass filter (LPF), 12 is a peak hold circuit (PH circuit),
13 is an analog/digital (A/D) conversion circuit, 14 is an arithmetic circuit, 15 is a comparison level selection circuit, 16 is a digital/analog (D/A) conversion circuit, 17 is a binary output control circuit, and 18 is a PH level. 19 is a standard value output circuit; 21 is a window generation circuit; 22 is a scanning point coordinate generation circuit;
3 is a window setting value storage circuit, 24 is a PH reset circuit, 25 is a reset timing circuit, 26
27 is an effective screen control circuit, 27 is a coordinate correction circuit, and 28 is an effective screen control circuit.
29 is a fixed binarization circuit, 29 is a positional deviation detection circuit, and the other parts are the same as in FIG.

Therefore, as in the case of FIG. 5, the object 1 is conveyed by the belt conveyor 2 so as to come within the field of view of the television camera 3. Assuming that the television camera 3 used here is equipped with a shutter, for example, the shutter is configured to perform an exposure operation every vertical blanking period (V _BLK ) of the television camera 3, and to perform a light shielding operation except for this period. Assuming that the television camera 3 performs raster scanning in the same way as normal, it is provided with appropriate illuminance from the illumination device 4, which consists of, for example, a tungsten lamp lit by direct current, so that if an object is present within the camera field of view, , an object image is taken out from the shutter-equipped television camera 3 every vertical scan (that is, every V). Of course, unless the object is conveyed, the image information on the conveyor is essentially meaningless information. In addition, a strobe is used instead of a television camera with a shutter, and its ignition is as described above.
It goes without saying that by synchronizing with the V _BLK signal, a still image can be obtained in exactly the same manner as above.

The image signal from the television camera 3 obtained in this manner is applied to an analog amplifier 7 of the image processing device, where it is amplified to an appropriate level. The signal from amplifier 7 is input to binarization circuit 8 and low-pass filter 11. The LPF 11 cuts noise and small image signals (high frequency) of one to several pixels from the image signal of the television camera 3.
Only image signals below the intermediate frequency are extracted. This is provided in order to avoid malfunctions based on noise or local false image signals caused by objects (eg, caused by halation, etc.). this
The output of the LPF 11 is led to a peak hold circuit (PH circuit) 12, which holds the peak value of the image signal at a predetermined timing. The operation timing of the PH circuit 12 is controlled by the output from the hold window generation circuit 21. In other words,
The PH circuit 12 operates only within the window area, and its hold value is determined by the PH reset circuit 24 and the PH
The reset timing circuit 25 erases the data at a predetermined timing. Window generation circuit 21
The method generates a two-dimensional area of a predetermined size as a window at a predetermined position in a preset screen, and the window area is formed by generating a window signal for each scanning line. By the way, it is necessary for this window to prevent the relative position of the object image from changing even if the position of the object image shifts within the camera field of view.

FIG. 2 is a reference diagram for explaining a method for correcting positional deviation of an object.

The scanning point coordinate generation circuit 22 in FIG. 1 generates the x and y coordinates of the current scanning point of the television camera 3, and calculates the y coordinate at the current scanning point by counting the number of horizontal scanning lines. On the other hand, the x coordinate on the current horizontal scanning line is known by counting a high frequency (for example, 6 MHz) reference clock, and from this, the y and x coordinate signals at the scanning time are output for each scanning. On the other hand, the positional deviation detection circuit 29 extracts the tip y coordinate y _i and the left edge x coordinate x _i of the object (image) 1 from the image taken out via the television camera 3 in real time in synchronization with the raster scanning of the television. Then, this is compared with the standard y, x coordinates y _S , x _S obtained in advance by the same means as above for the standard object image, and the positional deviation amount Δy,
Find Δx using the following formula.

Δy＝y _S −y _i Δx＝x _S −x _iFor this purpose, first use a standard object to determine the relative positional relationship between the window and the object image, and the position of object 1 on the camera field of view screen F. Check it out in advance. In other words, the reference point coordinates P _WS (x _d , y _d ) of the window and the reference point coordinates P _S (x _s , y _s ) of the object 1 as shown in Fig. 2 A are determined and memorized in advance. I'll keep it.
Next, the reference point coordinates P _i (x _i , y _i ) of the object to be inspected as shown in _{FIG .}
The deviation amount Δx in the x-axis direction and the deviation amount Δy in the y-axis direction are determined from the difference between P _i (x _i , y _s ) and P i (x i , y _i ). Then, by correcting this shift amount with respect to the reference point coordinate _PWS , a window is always generated at a predetermined position of the object no matter how the object position on the screen F changes. Furthermore, the position of the object 1 on the screen F (x _s , x _i , y _s , y _i )
For example, the imaging signal from the television camera 3 is binarized at a fixed binarization threshold level by the fixed binarization circuit 28, and the projection integral value of this binarized image on the x and y axes is You can ask for it.

The above is performed as follows in FIG.

First, the generation coordinates of a window set using a standard object are stored in the setting circuit 23 in advance. The scanning point coordinate generation circuit 22 generates x at the current scanning time,
Since the positional deviation detection circuit 29 outputs the y-coordinate signal and outputs the positional deviation amounts Δx and Δy as described above, the coordinate correction circuit 27 adds each output from the scanning point coordinate generation circuit 22 and the positional deviation detection circuit 29. Or, by subtraction, normalized coordinate data that has been corrected for deviation is output. Window generation circuit 21
generates window information by matching the output data from the coordinate correction circuit 27 and the data from the set value storage circuit 23. Then, only when this window opens, the peak hold circuit (PH circuit) 12 effectively operates and holds the output of the low pass filter (LPF) circuit 11 during that time.
That is, no matter how the object moves within the imaging field of view, it is possible to always obtain grayscale information at a predetermined position of the object.

Here, window generation and peak hold functions will be specifically explained. FIG. 3 is a reference diagram showing the relative positional relationship on the screen due to the movement of objects, and FIG. 4 is a time chart mainly for explaining the peak hold operation in FIG.

Since the shutter performs an exposure operation every vertical blanking period V _BLK and performs a light shielding operation outside this period, the start time of the period V _BLK is t _i-2 , t _i-1 , t _i
Considering this, the positional relationship of objects on the screen at each time is shown, for example, as shown in FIG. 3, A, B, and C. Here, 1 is the object, W is the window, and F
represents the field of view (screen) of the television camera, and V _i-2 , V _i-1 and V _i represent the object images at each time point, respectively.

Now, assuming that the above-mentioned times are shown as shown in FIG. 4A, the position sensor output will be obtained immediately before time t _i , as shown in FIG. 4B. The control circuit 26 generates a binarized output effective screen signal as shown in _FIG . On the other hand, the image signal from the television camera 3 is given to the PH circuit 12 via the amplifier 7 and the LPF 11, but at this time the window generation circuit 21 outputs all the window signals except during the generation period of the effective screen signal. As a result, the peak hold operation is performed every high (H) level period shown in FIG.
Therefore, the PH circuit 12 holds the peak value of the low frequency signal from the television camera 3 until a reset signal is applied. As shown in figure 2, the PH circuit reset signal is output for each V _BLK signal other than the V _BLK signal on which the valid screen signal is formed, and is output to the PH circuit 12.
Perform a reset. Therefore, in the case of FIG. 3 or FIG. 4, the window setting is not sufficient at time t _i-2 , so the output of the PH circuit 12 becomes PH' in the same figure. At time t _i-1 , the object has reached the window position, so
The PH circuit 12 is used as the LPF circuit 1 for grayscale information of the object.
The peak value of the output corresponding to the window area among the outputs of 1 is held as V _H as shown in the same figure. Note that this value V _H is reset by a reset signal output at time t _{i +1} . Furthermore, if there is an object image at time t _i+1 , its peak value is held as shown in PH'' in the figure.

In other words, the peak value of a predetermined area can be known from the image information just before the position sensor output is obtained (t _i-1 in this case), so the appropriate binarization level can be determined based on this. If this is done, it will be possible to binarize the image information immediately after the sensor output is obtained (t _i in this case) at the determined binarization level, and then perform inspection based on the binarized image thus extracted. , the processing time for binarization is shortened and the accuracy of inspection or discrimination is improved.

Next, a method of determining the binarization level from the peak value obtained in this way will be explained.

First, image the standard object at the standard position,
Adjust the binarization threshold to obtain an appropriate binary image, and set the appropriate binarization threshold L _S to correspond to this.
Find the peak hold value _VHS of the PH circuit 12,
These are stored in the arithmetic circuit 14 as a reference. Note that the detection of the relative position between the object and the window as described above is also performed at this point. Next, a similar operation is performed on the object to be inspected to obtain a peak hold value V _HX , which is grayscale information within the window area of the image immediately before determination V _i-1 . In this way, _from the relative relationship of the peak hold values, the binarization threshold candidate L _X of the object to _be inspected can be set as L _X = V _H An appropriate binarization threshold value can be determined by performing the calculation in the above equation. At this time, this also applies to the peak hold value V _HX itself.
Check whether it is within the range of 0.5V _HS ~ 1.5V _HS ,
We are trying to judge its appropriateness. The PH level discrimination circuit 18 makes this judgment.
If V _HX is within the above range (Y), the comparison level selection circuit 15 selects the output _L given as a threshold. On the other hand, if V _HX is outside the above range (N), the selection circuit 15 operates to select the output from the standard value output circuit 19. Note that the standard value output circuit 19 is
Normally, L _S is output as the standard value. In other words, the reason for doing this is that when V _HX is abnormal, it is dangerous to use the binarization threshold calculated by calculation.
This is because it is considered safer to adopt the standard value _VHS .

The binary level signal determined in this way is
It becomes one input of the binarization circuit 8 via the A conversion circuit 16. The image signal at time i is input to the other side of the binarization circuit 8, and this image signal is binarized using the binarization level from the D/A conversion circuit 16. At this time, the control circuit 26 generates the effective screen signal as described above and supplies it to the binary output control circuit 17, so that the circuit 27 only uses the binary image information effective for inspection and discrimination out of the output from the binary conversion circuit. Control is performed to output . The arithmetic control unit 9 executes predetermined image processing using the binary image information obtained in this way, and inspects and discriminates the object.

In the above, we have explained the correction of parallel deviation of the two axes of x and y as position correction, but in the case of rotation (posture) deviation, it can be performed in the same way as above by providing a rotation amount detection circuit and a coordinate conversion circuit by rotation. I can do it. In addition, in the above, the binarization circuit has one channel and one window is set for it, but in order to cope with the shading of the object, the binarization circuit is made to have multiple channels, and correspondingly, multiple channels are set. It is also possible to provide windows and individually control the binarization threshold by taking correspondence with each window. Of course 1
It is also possible to set multiple windows for one binarization circuit. Further, although the peak value of the image signal within the window is used, the average value of the gradation information may be used. Further, the shape of the window is not limited to a rectangle, but can be any shape.

〔Effect of the invention〕

According to this invention, two temporally adjacent image signals are extracted, a binarization level is determined from the temporally previous image signal, and the temporally subsequent image signal is binarized. Therefore, even if the reflectance of the objects differs from one another, the binarization threshold value is selected accordingly, providing the advantage that stable and highly accurate determination is possible at all times. Note that this point also applies to changes in illuminance of the lighting device.
In addition, even if the position or orientation of the object changes, the amount of deviation is detected and normalized, and the binarization level is determined based on this, so a highly accurate binarization level can be achieved. Obtainable. Furthermore, since a moving object is converted into a still image, a clear image without image blurring can be obtained, and as a result, the accuracy of inspection or discrimination can be improved. Furthermore, since a high-speed A/D conversion circuit, a memory circuit, and an enormous processing circuit are not required as hardware, this type of device can be made simple and inexpensive. Furthermore, this invention determines the binarization level by using the image information immediately preceding the image information to be determined, which has conventionally been overlooked as waste. Therefore, the processing time does not substantially increase, and therefore high speed equivalent to the conventional method is guaranteed. Further, in determining the binarization threshold, a certain limit is set on the grayscale value of the image, so that malfunctions due to unexpected objects or the like can be avoided.

[Brief explanation of drawings]

Fig. 1 is a configuration diagram showing an embodiment of this invention;
The figure is a reference diagram for explaining the method of correcting the positional deviation of the target object, Figure 3 is a reference diagram showing the relative position change on the screen due to the movement of the target object, and Figure 4 is mainly the peak hold operation in Figure 1. FIG. 5 is a block diagram showing a conventional example of an object inspection/discrimination device, and FIG. 6 is a time chart for explaining the operation of FIG. Code explanation, 1...Object, 2...Conveyor, 3
...TV camera, 4...Lighting device, 5...Position sensor, 6...Image processing device, 7...Amplifier, 8
... Binarization circuit, 9 ... Calculation determination unit, 10 ... Timing control circuit, 11 ... Low pass filter (LPF), 12 ... Peak hold circuit (PH circuit), 13 ... A/D conversion circuit, 14...Arithmetic circuit, 15...Comparison level selection circuit, 16...D/
A conversion circuit, 17... Binary output control circuit, 18...
...PH level discrimination circuit, 19...Standard value output circuit, 21...Window generation circuit, 22...Scanning point coordinate generation circuit, 23...Window setting value storage circuit, 24...PH reset circuit, 25...Reset Timing circuit, 26... Effective screen control circuit, 27... Coordinate correction circuit, 28... Fixed binarization circuit, 29... Position shift detection circuit, 91... Memory, 93... Arithmetic processing unit, 94... Position detection/
Correction circuit, 95...Setting circuit, F...TV camera field of view, W...Window.

Claims (1)

[Claims] 1. Image extraction means that includes a two-dimensional imaging means that images a moving object and takes out a still image of the object, and detects the position or rotational deviation amount of the object with respect to a reference position within the imaging field of view. position correction means for normalizing the position based on the amount of deviation; window generation means for generating at least one predetermined shape window at a predetermined position of the normalized object image; and at least an image within the window area. binarization level determining means for determining a binarization level according to the gray level, the image retrieval means including a detection unit for detecting a gray level of at least two successive images based on a predetermined time. A still image is extracted, the position correction means corrects the position of the object image for the temporally previous image, the window generation means generates a window at the predetermined position, and the binarization level determination means adjusts the window area. An image binarization method characterized in that a binarization level is determined from the gray level of an inner image, and a temporally later image is binarized using the determined binarization level. 2. In the image binarization method according to claim 1, the image signal for determining the binarization level is extracted by applying a low-pass filter to the output from the image extraction means. Image binarization method. 3. In the image binarization method according to claim 1 or 2, when the gray level of the image detected by the detection unit exceeds a predetermined range, a separately provided standard value is set. An image binarization method characterized by employing a binarization level.