JPH03142303A - Method and device for extracting form - Google Patents

Abstract

PURPOSE:To extract the angle of an object surface with a high precision independently of the variance of reflection distribution of the surface by illuminating the object with an illumination system consisting of many various color parts and analyzing the color reflected from the surface. CONSTITUTION:For example, a circle line illumination 2 and a semispherical multicolor food 3 where concentric circles are divided by color are used to illuminate the object with the illumination system consisting of many various color parts. A specular surface object like chip parts 5 having a solder part 6 modulated on a printed board 4 is put in the hood 3 and its image is picked up from above by a color TV camera 1 or the like. Then, the color reflected from the surface of the specular surface object is identified, and angles (for example, theta1 and theta2) of the object surface to a horizontal plane are determined by this color. Since the angle is determined correspondingly to each picture element of reflection, angles of almost all minute faces of the specular surface object are determined with one image picking-up without adopting a special lighting method. A form is restored from points, whose height is known, based on these angles to extract the form of the specular surface object.

Description

[Detailed description of the invention] [Field of application] Is the unexploded man minute? This theory relates to a shape extraction method such as a surface curved shape of a circular object and to remanufacturing. In particular, the present invention relates to a shape extraction method and apparatus suitable for use in measuring the shape of solder on printed circuit boards, measuring the shape of wire bonding for inspection and semiconductors, and measuring the shape of a heat exchanger in an air discharge recording device.

[Conventional technology]

Regarding the method of extracting the shape of such a minute theory tmw body, please refer to IEEE Transactions on Pattern Analysis and Machine Intelligence Volume 10, No. 1 (1988), AK44 Negative to Page 55 (IEEE i”). rarLs, on patter
nAnalysts and Machine Int.
glliytnct, Vol, 10° Shoulder 1 (IS'
8s), pp, a4-55). This method, as shown in Figure 22,
A hemispherical object 50 with a large number of light emitting diodes 45 arranged regularly υ) A surface-shaped object 50 is placed under the hood, and a 1゛V camera 41 is used to detect the object through the hole in the center of the hood. : Capture an image.

At this time, the diode 4 that is emitting light like the 23rd garden
3, it was specularly reflected and captured by the camera 41 - the light on the mound -) From the positional relationship of one point, the three-dimensional position of the reflection point of the euphemistic object can be determined by the principle of triangulation. . Furthermore, the light emitting diode υ distribution that lights up at certain times is
By properly combining them in a divided manner, it is possible to cover a large number of photodiode positions with a small number of lighting units. By this, if there are many υ cases, 10II! It is possible to determine the shape of (2) with a number of auxiliary numbers of about 1000 yen.

In addition, in the 22nd prisoner, 42 is a light emitting diode 430
A controller that controls lighting; 44 is an interface I;
g l1lr, 45 Ha Work Station Nap)
Computer, 46 is input/output #cfII:, 47 is 7"
49 is a color stock memory that stores the image developed by the V camera; 48 is a controller that controls the XYθ table;
is a monitor TV.

In addition, regarding the horizontal placement of soldered electronic components, please refer to IE Transactions on Pattern Analysis and Machine Intelligence 1.
Volume 0, No. 5 (1988), pages 587 to 395 (I
EEE Trans, on patternAn
lysis and Machine Intgl Hence, Vol, 10° 5 (1988)
, pp. 587-595). This method, as shown in No. 24,
II K A two-stage lighting system with a blue ring-shaped light 52 and a red ring-shaped light 55 installed on the lower side is installed, and by lighting the color TV 51, for example, on the printed board 54. Mounted lead components 55 Relead 57υ) It is possible to check the state of the solder by looking at the distribution of red and blue colors that are reflected in the m area of the solder part 56.

If this distribution leads to fishing with 55 -0 children, then the second
In the normal case, as shown in 5 Kōta+, the red, Yoi ring faces one of the leads, and the excavated Sarapuka is the 25th Kō1b.
l, (in abnormal cases as shown in C1, the additional part ρ
・It is said that it is possible to earn money from % signs such as the absence of a circle or its shape not being an inner circle.

[Issues that Hatsuo tries to solve]

However, in the first conventional technology, each light emitting diode cannot be individually controlled, and when trying to obtain a popular continuous shape, the number of light emitting diodes is several 1.
Since there are as many as 00 pieces, the ink face and power supply thread between the light emitting diode and the calculation device become complicated. Also, even if you do not image every light emitting diode individually,
10 degrees of image input from a TV camera is required,
This is a big constraint when it comes to confusion.

Moreover, in the second conventional technique, II! The color distribution on the ll image is detected with an excessive threshold, and its two-dimensional shape is heuristically determined from f (one run 'D), the shape of the mating part, and the lead part. Power) You must not rearrange the determination based on whether the parts are piled up or not. Furthermore, depending on the shape, the placement of the illumination may not be appropriate, and this method cannot overcome the uncertainty in the determination caused by the fact that the shape of the solder itself is not detected. (IJ) However, there are some imitative publications such as excess or deficiency of solder that strongly affect these devices. .

An object of the present invention is to provide an arm circumference shape extraction method and device that can accurately determine the five-dimensional shape of a wide range of points using an illumination system that does not require special light emission control and a single image. It is in.

Another object of the present invention is to provide a shape extraction method and apparatus that can extract angles with an accuracy of [υ](2) regardless of variations in the reflection distribution on the surface of an object.

[Failure to resolve division issues]

In order to achieve the above objectives, the present invention, for example,
As shown in the diagram, Circline Lighting 2 and 17LX
Using a hemispherical υ hood 5 color-coded on a circle, an object is illuminated with an illumination system consisting of a large number of colors that differ from part to part. Then, a mirror object such as a chip component 5 having a solder portion 6 mounted on a print 14 is placed inside the hood 5, and an image is taken from above with a color 1' V camera 21, etc. Identifies the color that will be displayed, and depending on the warp color,
For example, as shown in Figure 3, the angle between the object surface blood C and the horizontal (for example, θ1.

This can be determined to correspond to the valley where the reflection appears, so
In addition, with one imaging without using any special lighting method,
It is possible to determine the angle of most minute surfaces of mirror objects. By restoring the shape from the known points based on this angle, the shape of the object is extracted.

[Effect]

In this way, by viewing concentric color-coded hoods, the shape of the distracting object can be determined with a single m-image taken by the color camera. Therefore, when this method is used for soldering inspection of chip parts, etc., the accurate shape of the solder part can be determined in a short time. .

In addition, it is possible to accurately make various judgments such as the presence or absence of solder, the amount of solder LA, and the quality of the solder profile. In particular, when compared with the conventional method of lighting a large number of diodes, the light from the Shimosaka diode was specularly reflected, and only about 100 good O-coordinate points were gathered. Since the angle of the light source can be determined, the accuracy of the extracted shape is dramatically improved.

Furthermore, in order to improve the shape of the hood, the number of colors arranged on the hood is also important. In particular, in shape inspection applications such as detecting excess or insufficient solder, the detection error can be reduced by a few percent.
I want to hold it. Now, if the angle detection error is Jθ, then
The order of the lengthwise detection 1JI4NJt is as shown in the following equation.

J l=l, * 5LrLAIJ If)
Must be within ±2 to 5 friends. As a result, it is desirable that the number of colors extending from the angle measured up to 45 degrees from the horizontal plane to the hood is 7 to 8 or more. If the colors are divided into about 10 to 20 colors, the detection angle accuracy can be made within ±1 to 2 feet.

〔Example〕

Examples of the present invention will be described below. The first example is an inspection device for a solder portion of an ff[l-mounted component according to the present invention. Circline lighting 2 and multicolor 7-do 3 hood-shaped σ]
The illumination system and the light beam from the color TV camera 1 are installed on an XY table 9υ) on which a printed board 4 is placed. The solder part (6 in FIG. 2) of the chip component 5 on the printed board 4 is imaged by moving the XY table 9 under the control of the table controller 8 and bringing it under the camera 1.
The inspection is performed by analyzing the distribution of reflected light using the image processing device 7. The field of view of the color TV camera) 1 is set to about 10 mm. In addition, the outer diameter of the hood 3 is several hundred mm.
It is. Since the hood 5 is sufficiently large compared to the solder part 6 to be inspected, the position t11t within the screen does not have a large effect on the measurement accuracy in the surface direction. show.

In this device, chip parts, Sup, QFp etc.
Components are inspected, and in the fourth example, a typical chip component 5 is mounted on a printed board 4 and soldered to solder s6. An example of such defective soldering of chip component 5 is %5! N (no solder shown in bl, insufficient solder,
There is an excess of solder shown in the 5th VIC+, a bridge shown in the 51st W(d+), etc. This inspection device is intended to detect such defects with great reliability.

1 of the hood-shaped lighting section? The f configuration is shown in FIG. A color filter 20 is formed on the outside 9111 or the inside 1 of the hemispherical multicolored hood 3, and it emits colored light on the circular axis of the hood S using external lighting (circular lighting 2). . The hood body is made of a white acrylic plate or matte glass 21, and is configured to diffuse colored light in a wide direction.

Here, the color filter 20 is attached in a 1Tlr11 shape along the outer shape of the hood 3, and it is desirable that the color changes from top to bottom in a continuous manner, as shown in Figure 7. Even if a large number of colored bands are arranged side by side, it is not practical to produce or support. In addition, each light source does not necessarily have to be band-shaped, and it does not matter if there are many point light sources. Also, is it okay if the hood 3 is hemispherical? Practically speaking, it can be either a conical shape as shown in Figure 8 (α) or a disc shape as shown in Figure 8 (, 61).Inspecting the root bark?
: Apraxia is possible. For the purpose of simply eating away the solder part in the time-saving direction σ], the 2
It is sufficient to combine several flat plates, such as a million-fold shape or a pyramid shape.

The color of the filter is like a signal: I! ! It is formed into a hood-like shape on the pillow.Continuous changes in stiffness are caused by
Color O 徂 1] is defined on the signal space that performs the foot.

On this signal space? a due to fluctuations in body surface reflectance
In order to prevent the drop in the treetop measurement level, it is necessary to increase the height of the sky.
1lil, Shiinaimori or 0. When using the (R, G, H) signals as fB thread, the overall manliness Y can be expressed as Y = R + G 1 B, but for measurement, d) Tani ji with Y. Divided r=R/Y, 5=G/Y
, b=B/Y is used. Also, on the chromatic space (Y, R
-Y, B-Y), each color is coated, and here too, the processing is done as shown in Figure 10: R-Y/Y, H-Y/Yυ
It's embarrassing to normalize things like this.

By the way, if you use another company's bar signal between color runs, each color will be left at the end of the free color and white, as shown in Figure 10. Here, the angle θ in the figure represents the hue, and the naturalness r + Xe & is the color. On the hood, the color changes with delay due to the filter υ), etc., as shown in the 10th evil (α), and when it changes to the color of the color or the color of the light), as shown in IAI, K (R -Y/Y
, B-Y/Y) 4BQf force@f1M (T and force), (as shown in C1, the hue changes so that θ or θ coincides instantaneously, etc.C7) This is to show K conversion.

This is also 1IllI-like in other signal spaces. An example of coloring σ of the filter 200 is shown in No. 11-.

As shown in No. 12-1-, when the solder cap 6 of the S product 5 is produced with the TV camera 1, the solder portion 6 is close to a sharp circle, so the solder cap is
At this time, the color in the direction of specular reflection enters the TV camera 1. Therefore, by looking at the color of this reflection, the amount of solder blood can be determined. Here, as shown in the figure, there are ψ1 in the case of waiting for a beak that is both strong in the positive and negative forces, and ψ1 in the case of waiting for a beak that is both strong in the positive and negative forces, as shown in the figure, and 0 in the case of taking in the 9 of the comparatively wide moon exhibition. If the coloring is continuously corrected by 8C in the signal space, even if the way of reflection changes slightly, it will be difficult to reduce the measurement angle slightly.

Here, the correspondence between colors and angles is determined by using the calibration tool 11 shown in the 13th IC. The jig 11 is a hemispherical body, which is colored 1' V
When viewed from above at 7I Mela 1, it is the 14th evil (α) Riyo T
V-Blood 12σ, iC polymorphic flesh color υ] 1 Tsuka I out 15 visible. The radius R of the hemisphere of the old tool 11 is rounded on the front side to 41 points, and as shown in k141WL61, the radius R of the hemisphere is I
The short separation r from the beginning is the horizontal motion of the curved line at that point, and the angle θ can be simply extended using the relational expression r = R# 5Lnd. By the way, from that point, the correction of σ such as (RY/Y, H-Y/Y) can be simply watered down, so the first
It is possible to generate a table as shown in 5-. Using this, the angle i can be easily determined when inputting the solder part. In addition, on the table of the 15th evil, the color used for food 3 is dull, so the color of the 15th
The teacher will not be satisfied with the calibration of 4 evils. In this part, enter the angle of the color that is as close as possible to the colors that appear during calibration, or if there is a left-footed angle, it should be detected as an abnormal first shift. .

Ku.

If Solder Miyako, who is actually divorced, is born under the actual lighting, the sled image will be indulging in the mourning of the 16th death. By praising this color, approximately 1
Referring to the table in 5(3), the angle θ from the horizontal position can be fixed as shown in brown 17. In addition, Jjiu! !
Specifically, the plane 0 normal angle θ that can be obtained on the itt side is in the range of 45 degrees to 90 degrees. By the way, in the 15th table
The only degree that can be determined by looking at this is the horizontal angle θ, and in order to determine the spatial llc law, it is necessary to It must be determined whether the other one such as α is σ〕 square thigh. Here, if the shrine has a vertical shape or a flat cio-like curve, then draw the temple θIv5 n on the rest of Figure 17 on the front mountain, and select this song? By lengthening 1-1 for IMK, approximately modulus 1i
The moon α, which forms the X-axis of i11I52, can be called the moon of water.

In this way, two angles (0, α) or ice heart and one side?
:: Uniquely determines ability orientation.

In order to horizontally place the solder fT using this horrendous method, (1
) Thicken the five-dimensional shape of the solder part and use a solder gong to judge the quality. (2) As shown in Figure 19, draw a straight line in the direction of the bolt 6 that was made once, and make a straight line. Determine pass/fail from old Hanta profile. (5) As shown in Fig. 20, place the axis perpendicular to the direction of solder cap 60, and 70) Determine the solder area from the pile of solder parts. (4) In the hood 5 shown in Figs. Since the concave angle from 16 to 6 is colored, the area of the same color is cut out and the warpage distribution is determined to determine the solder size. It is possible to use

In the profile judgment shown in the 19th round here, when the chip crow changes to the younger sister of 1α), θ': 90-tJ
℃, its change (A+) can be marked on the graph.Here, if there is insufficient solder, good scorch, or increased solder, occlusion is the 21st evil (O ' V). αl 1,6
11cI becomes 5, and by detecting this change '1'', it is possible to determine whether it was made or not.

Also, Pile 4 in Figure 211! In the lJ foot, given the pixel size dVc and the normal direction θ, the height fluctuation per pixel Ah can be lengthened by J h =: d a tan θ, which is Starting from the height, i20! The cross-sectional shape of W+, 61 can be determined.

In addition, if there is a part of the Hanta Miyako that has an ocho angle or a part of the Ω~45 exhibition, it is easy to lengthen the cross-sectional shape. This is because this optical system uses specular reflection from the surface of the object to determine the amount of water, and in principle only surfaces between 45 degrees and 90 degrees can be frozen. ρ′−h℃
be. By the way, with ordinary solder fT, the profile of the circle or the imaginary part does not move so much, so by supplementing the center 11m (here, about 23 degrees) as a representative value, it is possible to I can draw shapes.

According to the cross-sectional shape or ice, its area 'l Ct is within the range of one foot as shown in the following formula 〇. ! 3χ can be published. Here, the 20th
If you lengthen CL'5r- with one line like -, this becomes 1' which is almost the same as the volume judgment of (su).
h>ci of 1-total of all lines>7'J where 7'
A and Tl are the upper and lower limits of pass/fail determination.

The above 5tC and a soldering stand? : By selecting shortcut Φ, it is possible to remove the soldering device.

[% male effect]

According to the present invention, the five-dimensional shape of an object waiting for a face-like surface, which has been difficult in the past, can be easily determined by analyzing the color applied to the hood on the surface. Or Xiaomiru. Furthermore, by performing the hood coloring so that it is continuous in the color signal space, it is possible to apply it to the inspection sheet regardless of changes in the shape, reflection, and reflection of the object surface. In signal analysis, by normalizing the 40 components on the color space, we can use the 40 degrees of hue, saturation, taste, and circularity to determine the shape that is affected by changes in the reflectance of the object surface. Extraction is o
J rudder V- becomes.

In this way, it is possible to accurately determine the quality of printed board solder parts by accurately extracting the shape, which has been difficult in the past.
ll: Nimuru.

4. Simple explanation of m-7j The first problem is the soldering device which incorporates the shape extraction method and device according to the present invention. Fake failure of roasted extractor, 51%5
The diagram shows the principle of determining surface direction using color-coded hoods.
The fourth problem is the external appearance of the chip parts. , tCl solder excess, ldl bridge example
The 6th figure is a moving diagram of the illumination thread, and the 7th figure is a band-like color-coded hoodie.
bl shows various alternatives for the hood.
l, ibl is a diagram showing the case reduction χ of flat curved hood,
Figure 10 shows 1! ! , The method of coloring in the left Hirayama is detailed, and the 11th is the coloring of the filter. Fig. 15 is a diagram showing the calibration jig, Fig. 14-(αI, i, 6
1 shows the relationship between the m image and the jig when viewed with a color camera, which is a jig, and the 15th shows an example of the transformation from color to %i storm. The 16th symbol is the soldered chip Toden '4I: I [1111M] seen with a color camera under the multicolored hood. Figure 18 is a continuation of the dimensional method for surface failure.
The garden (cl, lbl, +cq is the 21st test to celebrate the inspection method of tanta bolts by analyzing the cross-sectional shape.
(cl, IbI, tCl, (d+ is the section along the measurement axis and an example of angular change, 22nd
The color is determined by using a large number of light emitting diodes.Figure 23 is a diagram to celebrate the principle of cutting 22-, and Figure 24 is a two-stage lighting string. Figure 25 (cl.
IbI.

(cl, idl is a thorough reading of the foot power method that incorporates the moxibustion tIi of the steel 24 figure.

1...Color IV camera 2...Circline hot space 3...Multicolor hood 4...Print version 5...
・Chip parts 6...Solder capital 7...Image processing ridge device 8...Tiful controller 9...XY Kechiru 1"1...Round shaft pattern 11...Old tool 12...・1°VIIl
llIiI Figure 2 Tick 1 B product 5 (Yoda evil 6 Figure 4 ÷ Tsubu evil product 5 5th circle normal crystal has no 5π゛, missing bridge 6 Tomoe katsu 7 Figure 8 Yo cone shape 10 Figure-Y Figure 12 Child (Shibu Parts 5 Funeral Figure 14 Figure 15 Figure 16 Figure 17 Figure 19 Figure 20 (Hyaku) Possible 1 Tip Shape 21 Figure (α) (b) Normal Crystal Profile 1 Ha/'u da lack: Makini → center, Shii ho Σ Excessive heron Fig. 22 (α) Fig. 25 (b) (C) Excessive reading

Claims (1)

[Claims] 1. The shape of the object is extracted by illuminating the object with an illumination system consisting of a large number of colors that differ for each part, and analyzing the colors reflected from the surface of the object using an image processing device. A shape extraction method characterized by: 2. In the illumination system, a hood that covers a wide range of the upper part of the center, such as a hemisphere, a cone, a pyramid, or a wide plate, is used to enable shape analysis of a wide range of parts of the object. Shape extraction method described. 3. The shape extraction method according to claim 2, wherein a plurality of colors on the hood are arranged concentrically. 4. By arranging many adjacent colors next to each other in a color signal space such as (R, G, B) space or color difference space, highly accurate surface angles can be achieved regardless of variations in the reflection distribution on the object surface. 2. The shape extraction method according to claim 1, using an illumination system that enables extraction of the shape. 5. By arranging a large number of adjacent colors next to each other in a color signal space such as (R, G, B) space or color difference space,
2. The shape extraction method according to claim 1, using an illumination thread that enables highly accurate shape extraction regardless of variations in reflection distribution on the object surface. 6. The colors arranged concentrically according to claim 3 are (R, G
, B) Using an illumination system that enables highly accurate extraction of surface angles regardless of fluctuations in the reflection distribution of the object surface by arranging the color signal to continuously change in a color signal space such as a spatial color difference space. The shape extraction method according to claim 5. 7. The colors arranged concentrically according to claim 3 are (R, G
,B) Claim 3 using an illumination system that enables highly accurate shape extraction regardless of fluctuations in reflection distribution on an object surface by arranging the illumination system so that the color signal changes continuously in a color signal space such as space or color difference space. Shape extraction method described. 8. When detecting the angle and shape of the object surface using color signals, (
Claim 1: In a color signal space such as R, G, B) space or color difference space, the influence of changes in reflectance of an object is eliminated by normalizing brightness and using only hue or saturation or both. Shape extraction method described. 9. A claim in which, when associating a color signal with the direction of an object surface, the correspondence is determined by placing a spherical or cylindrical object with a known diameter at the object position and observing the color of the reflection on the surface. 1. The shape extraction method described in 1. 10. Prepare a conversion table for associating color signals and object surfaces so that conversion can be easily performed, and when creating this conversion table, table elements that cannot be filled in by the correspondence described in claim 9 should be replaced with a table element that is close to it. 10. The shape extraction method according to claim 9, wherein the element is determined by complementing the values of surrounding elements. 11. Claim 10, which makes it possible to detect as an abnormal value when the value of a table element whose value cannot be found is too far from the desired element in the correspondence described in claim 9. Shape extraction method described. 12. Determining the angle of the surface from the horizontal plane based on the color correspondence within the screen by using the multicolor filter configured in concentric circles according to claim 3, and extracting equiangular partial areas based on the angle; A shape extraction method characterized in that the normal direction of the surface is spatially determined by determining an angle other than the angle with the horizontal plane based on the tangent in the screen of the external shape and finding the second angle. . 13. A soldering inspection method for determining the quality of soldering by extracting the shape of a soldered portion of a printed circuit board using the shape extraction method according to claim 1. 14. The soldering inspection method according to claim 13,
13. A soldering inspection method that determines the normal direction of the surface by the inspection method according to claim 12, and determines the volume of the solder portion based on this to determine the quality of the soldering. 15. The soldering inspection method according to claim 13,
In the concentrically colored hood according to claim 3,
A soldering inspection method that uses the fact that the color is determined according to the angle of the surface from the horizontal plane to determine the quality of the soldering from the color pattern. 16. The soldering inspection method according to claim 13,
In the concentrically colored hood according to claim 3, by utilizing the fact that the angle of the surface from the horizontal plane can be determined in correspondence with the color,
Based on this, the angle of the solder part from the horizontal plane is determined, and the quality of the soldering is determined from the pattern of equiangular lines of this angle. 17. The soldering inspection method according to claim 13,
Draw several measurement lines in the direction of the solder, check the angle change of the surface on the measurement line, detect whether it is different from the normal pattern or an abnormal pattern, and determine whether the soldering is good or bad. Soldering inspection method. 18. The soldering inspection method according to claim 13,
Draw some measurement lines in the direction of the solder, check the angle change of the surface on the measurement lines, calculate the external profile of the solder part on the measurement lines based on that, and judge whether the soldering is good or bad based on that pattern. Determine the soldering inspection method. 19. In the soldering inspection method according to claim 13, several measurement lines are drawn in the direction of the solder, angular changes of surfaces on the measurement lines are examined, and based on this, the external profile of the solder portion on the measurement lines is calculated. A soldering inspection method in which the area of the solder part on the measurement line is determined based on this, and the quality of the soldering is determined by checking that it is within a reasonable range. 20. The hood according to claim 2 is an illumination system in which a number of colors are arranged on the surface and a scattering surface or a material with internal scattering is arranged on the inside, or a number of colors are arranged on the inside and a scatterer is arranged on the outside. 3. The shape extraction method according to claim 2, wherein an illumination system or an optical system in which scattering surfaces are arranged both inside and outside the coloring surface is used. 21. An illumination system consisting of a large number of colors that differ from part to part, a color camera that images an object illuminated by the illumination system, and a color reflected from the surface of the object based on the image taken by the color camera. 1. A shape extraction device comprising: an image processing means for extracting the shape of an object by analyzing the shape of the object. 22. The shape extraction device according to claim 21, wherein the illumination system uses a hood that covers a wide area above the center. 23. The shape extraction device according to claim 22, wherein a plurality of colors are arranged concentrically on the hood.