JPS63273047A - Surface irregularity detector - Google Patents

Abstract

PURPOSE:To perform a highly accurate and high speed detection by extracting shadow regions generated when an object to be inspected is irradiated from two oblique directions, detecting the one side boundaries of the shadow regions in different directions and detecting the maximum values of the widths of the shadow regions from the positional relationships of the boundaries. CONSTITUTION:An object 1 to be inspected is irradiated by monochromatic parallel light beam sources 2a and 2b different from each other from oblique directions. An image is detected by a color TV camera 6 arranged perpendicularly to the object 1. Then, the image is separated to images corresponding to illuminating light by a color analyzing circuit 9 and the images are binary-coded by binary-coding circuits 11a and 11b. Then, the + edges (left-hand illumination) or the - edges (right-hand illumination) of the shadow regions are detected by edge detecting circuits 12a or 12b, respectively. The positional relationships between the + edges and the - edges are analyzed in a positional relationship analyzing circuit 13 and the maximum quantities of the lengths of the shadows corresponding to recesses and projections are detected to be outputted. Therefore, a surface irregularity can be detected in real time and at a high speed.

Description

Detailed Description of the Invention [Field of Industrial Application] The present invention relates to a device for identifying and inspecting the uneven state of the surface of an object, for example, through holes drilled in each layer of a multilayer ceramic substrate before sintering. The present invention relates to an uneven state detection device suitable for inspecting the filling state of filled conductors.

[Conventional technology]

Conventionally, as a method for recognizing unevenness, there has been a method that first measures the distance from a detector to each point on the surface of an object, and then processes and recognizes this information (distance information). This type of method is known, for example, as disclosed in Japanese Patent Application No. 116853/1982, but because it takes time to detect distance information, it is difficult to detect the unevenness of through holes, which can reach more than 10,000 points per board. It is extremely difficult to construct a device using this method that can detect defects with sufficient precision and at a speed sufficiently faster than humans.

On the other hand, there is a method that detects the height of an object by illuminating the object from an oblique direction and measuring the length of the shadow produced at that time. For example, JP-A-60-22
No. 611 discloses this type of method, but in this method it is not possible to identify whether an object protrudes from a certain reference plane or is recessed.

There is also a method shown in Japanese Unexamined Patent Publication No. 58-92904, but this method does not mention shadow detection.
Furthermore, there is no recognition that uneven shapes can be identified by analyzing the positional relationship of shadows.

[Problem that the invention seeks to solve]

In the above-mentioned conventional technology, no consideration is given to recognizing the uneven shape of the object at high speed. There was a problem in that it was difficult to automatically inspect the unevenness of more than 10,000 points per board with sufficient precision for defect detection and at a speed sufficiently faster than humans. .

An object of the present invention is to provide an uneven state detection device suitable for realizing high-accuracy and high-speed automatic inspection as described above.

[Means for solving problems]

The above purpose is to extract the shadow area of each illumination that occurs when the object is illuminated from two diagonal directions, detect the boundary line on one side of each shadow area in different directions, and This is achieved by detecting the maximum width of the shadow area in a specific area from the positional relationship.

[Effect]

FIG. 2 is a diagram showing how shadows are formed when illumination light is applied to a specific object from diagonally above.

As shown in FIG. 2, it can be seen that the left edge of the shadow caused by the illumination light from diagonally above the left corresponds to the step position of the object going down from the left to the right.

Conversely, the right edge of the shadow caused by the illumination light diagonally from above and to the right corresponds to the step position of the object going down from the right to the left. Therefore, as shown in Figure 2, the positional relationship between the left edge of the shadow caused by left oblique illumination (hereinafter referred to as the 10th edge) and the right edge of the shadow caused by right oblique illumination (hereinafter referred to as -edge) The unevenness of objects can be identified. That is, from left to right in the figure, if the order is 10 edges, -edge, it is convex, and if the order, -edge, 10 edges, is concave. Furthermore, the length of the shadow from the left edge of the image to the first tenth edge, the length of the shadow from the last edge to the right edge of the image are the convexity, and the length of the shadow between the -edge and the tenth edge is the convexity. Corresponds to rotation. Assuming that the length of each shadow is a, the angle from the reference plane to the optical axis of the illumination light is θ, and shadow detection is performed from the direction perpendicular to the reference plane, the reference plane when the object is convex is The height h from Q is expressed as h = Q tan θ.

On the other hand, in the case of a concave portion, if the width of the concave portion is d, then the following equation is obtained.

Note that h cannot necessarily be found uniquely from 0, but in the detection of defects such as through holes, both d and h are often set as defect detection criteria, so if θ is chosen appropriately, Q can be determined. Since the amount corresponding to the degree of the concave defect can be obtained from the above, there is no problem in detecting the defect.

〔Example〕

The present invention will be explained in detail below.

FIG. 1 is a diagram showing a first embodiment of the present invention.

In FIG. 1, a detection target 1 is, for example, a through-hole filling part of a ceramic substrate before sintering, and monochromatic parallel light sources 2a each having a different wavelength.

2b, the light is illuminated from two diagonal directions. In this embodiment, the monochromatic light source includes a halogen lamp, 3° color glass filter, and 4. It is composed of a lens 5.

The optical axes 7a and 7b of the two light sources are on a plane perpendicular to the reference plane (surface) of the ceramic substrate, and on different sides with respect to the perpendicular to the reference plane of the ceramic substrate. do it like this.

The color TV camera 6 detects an image of the through-hole filling portion that is illuminated from above perpendicularly to the reference plane of the ceramic substrate that is the detection target.

Further, the horizontal scanning direction 15 of the color TV camera 6 is as follows.

The optical axis of the illumination light should be parallel to the plane where it exists.

The color separation circuit 9 separates images into images corresponding to respective illumination lights. In addition, as the color TV camera 6,
If you are using an RGBa tube type (3-plate type) camera and the RGB 3 signals are output independently, the wavelength of the monochromatic illumination light can be set to R.
Alternatively, by selecting a wavelength corresponding to G or B (for example, R and G), the color separation circuit 9 becomes unnecessary.

The shading correction circuits 10a and 10b correct illumination unevenness.

This circuit corrects the uneven sensitivity of the color TV camera 6, and is similar to the circuit disclosed in, for example, Japanese Patent Laid-Open No. 57-35721.

The binarization circuits 11a and llb binarize the shading-corrected image, and set a certain predetermined threshold Th.
Convert the following parts to 1 and the parts exceeding Th to 0. That is, when the input signal value is Vig and the output signal is vo.

By appropriately selecting the threshold Th, the binary image becomes an image corresponding to the shadow area.

The etch detection circuits 12a and 12b detect ten edges or - edges of a shadow area, and convert adjacent input signal values to v ; (x L v t(i + L L
When the output signal is vo(i).

+For sex.

vI, (i) = v; (i)·vI (i+1)-for etch.

vo(i)=v+(i)·vi(x+1).

Here, - represents negation, and . represents logical product.

Incidentally, FIG. 3 shows an embodiment of the binarization circuit 11, and FIG. 4 shows an embodiment of the etch detection circuit 12.

In FIG. 3, 50 is a comparator, and in FIG. 4, φ is a lock common to all circuits, 16 is a D flip-flop, 17 is an inverter, and 18 is an AND gate.

Next, returning to FIG. 1 again, the first positional relationship analysis circuit 13 analyzes the positional relationship between the tenth edge and the first edge, and detects the length of the shadow corresponding to the concave and the length of the shadow corresponding to the convex. Output. That is, as illustrated in FIG. 2 above, for a binary image of a shadow caused by illumination from the right, the maximum length of the shadow from the left edge of the image to the 10th edge is set to 6, and the -edge is set to 6. For a binary image of a shadow caused by illumination from the left, the maximum value of the shadow length between the - edge and the right edge of the image is rotated by 6. , - The maximum value of the length of the shadow between the edge and the ten edge is detected as a rotation.

Next, an embodiment of the positional relationship analysis circuit 13 is shown in FIGS. 5, 6, and 7, and will be described below.

First, FIG. 5 shows a rotation detection circuit. In this circuit, the length of the shadow from the - edge is counted by the counter 19a, and when the ten edge appears next, the count value of the old rotation and the current length of the shadow stored in the flip-flop 20a is counted. and the larger one of them is stored again in the flip-flop 20a. flip flop 20a
The contents of are compared with the contents of flip-flop 20b for each horizontal scan line, and the larger one is stored back in flip-flop 20b. After the processing for one screen is completed, the contents of the flip-flop 20b are stored in the flip-flop 20c in response to the vertical synchronization signal VD.

Through the above processing, the width of the widest part of the shadow area between the - edge and the ten edge is stored in the flip-flop 20c.

It is assumed that the image is scanned from left to right and from top to bottom, and the 5tart signal is a signal that becomes 1 at the left edge of the image.
It is assumed that the End signal represents a signal that becomes 1 at the right end of the image. Also, 21 is a comparator, 22 is a selector, 23
is a D flip-flop, 24 is a flip-flop, 25
is an OR gate, 26 is an AND gate, 27 is a Nant gate, and 28 is an inverter.

Next, FIG. 6 shows a generation circuit for the 5tart signal and the End signal. In FIG. 6, HD is a horizontal synchronizing signal, 2
8 is an inverter, 29 is a comparator, and 30 is a counter.

Next, FIG. 7 shows a six-quantity detection circuit. This circuit is otherwise similar to the daily amount detection circuit of FIG. 5, except that the contents of the counter 19b are directly stored in the flip-flop 20d without comparing them with the contents of the flip-flop 20d. In this case, in a circuit for a binary image of a shadow caused by illumination from the right, a 5tart signal is input to point a in FIG. Then, one edge signal is placed at point a, and En is placed at point b.
d signal is input. Note that in FIG. 7, the same reference numerals as in FIG. 5 indicate components having the same functions.

In this way, the positional relationship analysis circuit 13 calculates the daily amount and 0
The amount is output respectively.

In addition, which rotation and convexity of the left and right illumination should be adopted?
Depends on the purpose of subsequent processing. For example, if it is desired to increase the detection sensitivity of the uneven state and perform highly sensitive defect detection, it is sufficient to take the larger value for each.

In the first embodiment described above, only one object, for example, one through-hole filling part, can be placed on one screen of the color TV camera 6.

By using the color TV camera 6 as the detector, the detection optical system can be easily constructed, and all processing parts are constructed using hardware.

It has the advantage of being able to detect the state of unevenness and the amount of unevenness at high speed in real time. Additionally, the scale of the hardware is small.

Note that as a modification of the first embodiment, another light source such as a laser or a mercury lamp may be used as the light source. Possible options include replacing the color TV camera 6 with a two-dimensional detector using a line sensor and its mechanical scanning, and replacing part or all of the processing of detected images with a computer. It is clear that the object of the present invention can be achieved using either configuration.

In addition, if there are multiple objects on one screen and the areas in which they exist are known, part of the image is stored in memory, the desired area is cut out from there, and the processing circuit described above executes it. That's fine. Such additional changes are easy.

Next, FIG. 8 shows a second embodiment of the present invention.

In FIG. 8, the object to be detected 1 is illuminated by a parallel light source 2a from an oblique direction as shown, and this portion is detected by a line sensor 32a from vertically above through a lens 5d. The plane formed by the optical axis of the illumination system and the optical axis of the detection system is perpendicular to the object surface to be detected. Further, the detection area of the sensor is within a plane defined by the optical axis.

Also, parallel to the illuminated portion of the parallel light source 2a and at a distance i
1t! A portion separated by D is diagonally illuminated from the opposite direction with the parallel light source 2b in the same manner as above. Then, this portion is also detected from vertically above by the line sensor 32b via the lens 5e. In this case as well, the plane formed by the optical axis of the illumination system and the optical axis of the detection system is perpendicular to the object surface to be detected, and the detection area of the sensor lies within this plane. Further, this plane is parallel to the plane formed by the optical axes of the previous pair of detection and illumination systems.

Also.

Each illumination light is transmitted to each line sensor 32a,
The detection portion 32b is shielded from light so as not to overlap.

In the above configuration, a two-dimensional image can be detected by moving the detection target 1 horizontally and in a direction not parallel to the longitudinal direction of the detection portion of the line sensor 32 by the drive mechanism 33.

Then, after passing the output signal of each line sensor 32 through the shading correction circuit 10 and the binarization circuit 11,
If the alignment circuit 31 detects the same location at the same time, then the amount of unevenness can be detected continuously by using the same processing circuit as in the first embodiment.

However, in the case of this embodiment, the color TV shown in FIG.
Since there is no signal corresponding to the vertical synchronization signal VD like the camera 6, a circuit similar to the 5tart and End signal generation circuit shown in FIG. What is necessary is just to configure it so that it outputs an equivalent signal.

Furthermore, since the amount of positional deviation between the two line sensors 32a and 32b is always constant, the alignment circuit 31 can be configured with a delay circuit using a serial input/series output type shift register, such as a video FIFO. This can be easily achieved using memory.

In this embodiment, since the amount of unevenness of a plurality of objects can be detected continuously, there is an advantage that a large amount of inspection can be performed at high speed. Another advantage is that the scale of the processing hardware is small.

It is clear that various modifications can be made to this embodiment as well, similar to the first embodiment.

Furthermore, instead of moving the object in parallel, the same effect can be obtained by simultaneously moving the two sets of detection and illumination optical systems in parallel.

Next, FIG. 9 shows a third embodiment of the present invention.

In FIG. 9, the object to be detected 1 is illuminated from an oblique direction as shown by a parallel light source 2a, and is detected by a line sensor 32a from vertically above.

The plane formed by the optical axes of the illumination system and the detection system is perpendicular to the object surface to be detected. Further, the detection area of the line sensor 32a is perpendicular to this plane.

On the other hand, a portion parallel to and distant from the detection area of the line sensor 32a is illuminated by the parallel light source 2b from an obliquely opposite direction, and this portion is detected by the line sensor 32b. In this case, a total of four optical axes of the two sets of illumination systems and detection systems are arranged on the same plane. Also.

The two light sources 2a and 2b are the two line sensors 32a.

Only one of the detection areas 32b is illuminated.

In the above configuration, the detection target 1 is detected by the line sensor 32.
If the line sensors 32a and 32b are horizontally moved in a direction that is not parallel to the detection longitudinal direction, the respective line sensors 32a.

A two-dimensional image is obtained from 32b. Then, after the output signals of the respective line sensors 32a and 32b are passed through the shading correction circuit 10 and the binarization circuit 11, the alignment circuit 31 corrects the deviation of the line sensor detection position, so that the two line sensors 32a, 32b are simultaneously detecting images illuminated from different directions. Here, similarly to the first embodiment, the edge detection circuits 12a and 12 detect the 10th edge and the 12th edge.
By detecting it in step b and processing it in the positional relationship analysis circuit 13, the amount of unevenness 14 can be detected.

The difference in this embodiment is that since the illumination direction is perpendicular to the detection longitudinal direction of the line sensor, shadows are created in the perpendicular direction compared to the first and second embodiments. For this reason,
The processing direction of the logic circuit changes from the horizontal scanning direction (the self-scanning direction of the line sensor) to the vertical scanning direction (the moving direction of the object).

Next, reference numeral 1011 shows an embodiment of the etch detection circuit 12 in the embodiment of FIG.

In FIG. 10, two pixels are vertically cut out by shift registers 34a and 34b having a length equal to the number of pixels of the line sensor 32, and an etch signal is generated. In addition, 26 is an AND gate, and 28 is an inverter.

Next, FIG. 11 shows an embodiment of the positional relationship analysis circuit 13 that detects rotation in FIG. 9, and FIG. 12 shows an embodiment of the positional relationship analysis circuit 13 that similarly detects six amounts.

11 and 12, the counters, flip-flops, etc. shown in FIGS. 5 and 7 are replaced by the memory 35 and its peripheral circuits (memory 36, adder 37,
8. Comparator 39, flip-flops 40, 41,
Or Gate 42, Nantes Gate 43. Inverter 44)
, so that each pixel is processed in the vertical direction.

Although no control signals regarding addresses are written in the memories 35 in FIGS. 11 and 12, it is assumed that addresses corresponding to pixels of the line sensor 32 are all given by the same address counter. Further, it is assumed that the shift register 37 has a length equal to the number of pixels of the line sensor.

Further, in this embodiment, a detection area generation circuit, which will be described later, generates signals for the upper side (Start signal) and lower side (End signal) of a rectangular area where the detection target exists, as shown in FIG.

Next, FIG. 14 shows an embodiment of the above detection area generation circuit.

In Figure 14, in the memo January 01, 5 tar
The start address (xs, ys) of the t signal, the start address (xat ye) of the End signal, and the width of the detection area are stored. In addition, (Xs+ys) and (xat
ye) is stored in memory in its occurrence class.

This detection area generation circuit generates the 5tart signal and the End signal while sequentially reading out the contents of the memo i month O1. Others: 100 is a counter, 102 is a register, 1
03 is an adder, 104 is a comparator, and 105 is an AND gate.

The positional relationship analysis circuit 13 in FIG.
Regarding the area between the rt signal and the End signal, the same processing as in the embodiment shown in FIG. 5 is performed, and the results are stored in the memory 36.

According to the embodiment shown in FIG. 9 as described above, even if a plurality of objects to be detected exist within a scanning line, the existing regions can be processed separately, so that the embodiment shown in FIG. There is an effect that the irregularities of a larger amount of objects to be detected can be detected simultaneously and at a higher speed than in the example.

It should be noted that in this embodiment as well, various modifications can be made as in the embodiment of FIG. 8, and it is clear that the effects of the present invention can be realized by these modifications.

〔Effect of the invention〕

According to the present invention, since a binary image of a shadow area caused by illumination from two directions is handled, it is possible to identify unevenness and detect its amount with sufficient accuracy for practical use using a small-scale hardware configuration. , and can be carried out at extremely high speed, which has the effect of facilitating defect inspection, shape detection, etc.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing a first embodiment of the invention, FIG. 2 is a diagram showing the principle of the invention, and FIG. 3 is a diagram of an embodiment of a binarization circuit.
Figure 4 is an example of an etch detection circuit, Figure 5 is an example of a positional relationship analysis circuit that detects rotation, and Figure 6 is an example of an etch detection circuit.
Example diagram of tart signal and end signal generation circuit, seventh
The figure is an example diagram of a positional relationship analysis circuit that detects six quantities.
The figures show a second embodiment of the invention, FIG. 9 shows a third embodiment of the invention, and FIG. 10 shows an embodiment of an etch detection circuit. Fig. 11 shows an example of a positional relationship analysis circuit that detects rotation, Fig. 12 shows an example of a positional relationship analysis circuit that detects 6 quantities, and Fig. 13 shows the positional relationship of the 5tart signal and the End signal. FIG. 14 is an embodiment diagram of a 5tart signal and End signal generation circuit. <Explanation of symbols> 1... Detection target 2... Parallel light source 6...
Color TV camera 9 Color separation circuit 10...Shading correction circuit 11...Binarization circuit 12...Edge detection circuit 13...Positional relationship analysis circuit 14...Irregularity amount agent Patent attorney Nakamura Junnosuke 1st figure 1 light 1 thing concave convex 3rd figure n h figure 4 figure 5 figure 6 figure 7 figure 8 figure 9 figure 10 figure 11 figure 12 figure

Claims (1)

[Claims] 1. In a device that detects an optical image of an object and analyzes the optical image to detect the uneven shape of the object's surface, means for illuminating the object from two different diagonal directions;
means for respectively detecting two optical images generated by the illumination and converting them into electrical signals; and means for binarizing the converted two electrical signals and extracting respective shadow areas in the two optical images; means for detecting boundaries on one side of the shadow area in different directions, and detecting the maximum value of the width of the shadow area in a specific area from the positional relationship of the boundaries obtained by the means for detecting the boundaries. What is claimed is: 1. An uneven state detection device comprising means for detecting an amount of unevenness by detecting an amount of unevenness.