JPH01136284A - Mechanical recognition method and apparatus for discrimination of numerous individual graphics - Google Patents

Abstract

PURPOSE: To contrive the high speed and high accuracy of the discrimination of a graphic by simultaneously advancing the moving of a base surface and an image processing work and analyzing the center location, the dimension and the shape of a graphic by an image processing unit. CONSTITUTION: On the base surface 1 moved within a plane by a driving device 11, a drawing 20 is fixed. Two photographing devices 3 and 5 are fixed on a table 14 and the schematic image and the precision image of a drawing are photographed by these devices. These photographing devices 3 and 5 are connected with an image processing unit 6. The image processing unit 6 includes an image digitizing device and an image analysis program and performs an image processing and the analysis of drawing data The image processing unit 6 includes the programs controlling the base surface l and the photographing devices 3 and 5. A processing result is displayed on a display 7 for image display and is outputted to a working device 8.

Description

DETAILED DESCRIPTION OF THE INVENTION A kind of machine vision method and apparatus thereof, which is used to interpret objects containing a large number of individual figures, such as hole position maps of printed PC boards. Its structure consists of one driven platform that can move or fix on a plane, two imaging devices installed on the platform, and an image processing unit. The target drawing is placed on a table in it, one of which captures most of the image, and the other provides the imaging device with the detailed image. The electrical signal is output to the image processing unit, which finds the center coordinates of each figure, classifies the figure's size, shape, etc., and generates control data such as dimensions, position, and effective path for movement required for processing.

The present invention relates to a "mechanical recognition method and apparatus," and in particular, uses a kind of photographing device to convert a plan view including a large number of individual figures into an image electrical signal, and then manually inputs each figure to an image processing unit. It refers to finding the center coordinates of shapes and analyzing materials such as the size and shape of shapes and the appropriate connection order.

One example of the above-mentioned drawings is the hole location map of a printed circuit board (a diagram showing the size and location of the large and small holes on the printed circuit board).

The production factory processes the printed circuit board using the hole location map provided by the supplier, and before doing so, it must first analyze the position, size and shape of each hole on the hole location map, and determine the optimal drilling route arrangement. then set the operation order of the drilling machine0
Generally, the processing order for printed circuit boat hole location maps is: (i) Modify the hole location map and produce the mapping on a transparent film.

(d) Classify hole diameters with different colors (artificial processing) (3) Connect the same type of hole positions with the same color (artificial processing) (4) Equation generator (Prograa+m1n3 n+
achine) to find the coordinates of each hole (artificial processing)
(5) Enter the hole position coordinates manually into the computer and find the optimal hole 1.
Arrange the routes.

(6) Manually input data of hole position coordinates and effective drilling route to the drilling machine using numerical 'XjI control program method.

The above-mentioned traditional printed circuit boat hole position determination sequence is artificially intervened, and its speed, accuracy, and reliability are limited; in addition, the equation generator is expensive, increasing the burden on the contractor. Therefore, it is highly desired by the industry to provide a method and apparatus for recognizing printed circuit board hole position drawings that can reduce costs, meet reliable speed requirements, and provide accuracy and reliability.

The present invention solves the above-mentioned problems and demands by developing a machine visual recognition method and g position that has high working speed, high accuracy, and high reliability.

That is, the main object of the present invention is to use a table i [(X - Y table) that can stably move within one plane, place a drawing on it, and supply the photographic device with what is to be imaged. After that, let the image processing unit judge the hole position,
To achieve an economical and practical automatic drawing recognition method and apparatus that completely replaces artificial processing processes.

Another object of the present invention is to take advantage of the structure in which one imaging device captures rough images and one imaging device captures precise images, and adds special image processing and analysis processes. A high working speed automatic drawing recognition method and apparatus are achieved.

Another object of the present invention is to correct the precision image binarization method of the photographing device and the U-imaging geometric aberration, thereby achieving a method and apparatus for automatically interpreting drawings with high precision.

The features of the present invention will be described below in conjunction with a better embodiment and the drawings: The structure of the present invention is as shown in FIG.
is controlled by one driver (11), can be fixed and moved within one plane, a drawing (20) is placed on it, and inside the base (1) there is a built-in light source. A structure that allows you to install a light device and see through transparent drawings; one ti shadow! ! Attachment (3)
is fixed on the table (4) and a schematic image of the drawing (20) is taken from above the drawing (20); another l! Shadow & 1g
1 (5) is also fixed on the table (4) and the drawing (20
); one image processing unit (6) is
includes an image digitization device and an image analysis program;
Image processing and analysis of drawing materials, and one table (1
), imaging ia position (3, 5) and image processing unit (6
In addition to this, the image display display (7) can be circumscribed, and the material can be continuously transferred to a tape puncher or a processing device (8) not shown in the figure. Can output.

The operation process of the device of the present invention includes the following: (i) The photographing device (3) for a larger layer image range
The image processing unit (6) obtains the approximate center position of each figure on the image (2) corresponds to the coordinates of the approximate center of each figure on the image. Convert to tabletop (not shown) coordinates.

(3) Arrange all the figures in an appropriate order and obtain a precise path of each layer image.

(4) The photographing device (5) with a smaller layer image range images each figure one by one according to the path order obtained from step (3), and then determines the exact center position of each figure from the image. Calculate materials such as size, shape, etc.

(5) Convert the precise center position image coordinates and image dimensions of each figure into tabletop coordinates and actual dimensions.

(6) Operation of classifying and arranging figures according to their size and shape, and outputting them to processing equipment f! Decide on the department.

To explain the history in detail for each section: (i) Schematic image processing: The imaging table device 3) performs a schematic layer image on the outside of the image (20),
Divide all the drawings into several blocks, image the layers in an orderly manner, binarize the image, find each figure using the case point jumping test method, and calculate the approximate center position.
゛! , Binarization method: The images to be processed are all black-and-white contrast images of figures and backgrounds, and the gradient decomposition values (thrsho
ld) and binarizes the image, □ converts the multi-layered image into a bright two-layered image with contrast between white and J based on the component decomposition value, that is, the figure range is black and dark, and other than figures. has a light background, or vice versa.

(1) For each image element (Pixel) of image 1, its individual gray level is obtained by the image digitizing device, and the gray level of each pixel is obtained from the image processing program based on the gray level of each image ratio. Statistically, the number of image ratios corresponding to each composition value n+, i=o, 1.2゜3,...m, m=maximum gray composition, already Σn l=
N*N is the total image ratio number, and can be obtained.

(2) Find the average distribution composition value t: t = Σ(Gi・Ni)/N e+8, where Gi = gray degree composition 3) Find each of the following values: N1. N2tt+tt2;
N irises = Σn1 N2 = Σ n.

;; i = 歇し・N Tomi/N2L? =m −(
'm - t) ・N2/N+(4) Find the binarization component demarcation value T: T = 1/2 (t++tt) 2. How to find a figure from a rough image: The ideal way to find a figure is as follows: Contains: , No - (1) Every figure must be processed only once.

(2) Remove processed figures from the image to avoid duplicate processing.

(3) After completing one figure processing, quickly find the next figure and do not omit any one figure.

(4) Hurry up the calculation of figure center coordinates.

In order to satisfy the above four points, the present invention uses the following point-jump inspection method f! 4 made (1) 2nd
As shown in the figure, on every image there are grid points with horizontal spacing ΔX and vertical spacing ΔY, DI2+..."+ 021. D22...
・+D31*... etc. are set, and the brightness of the case points is checked in order to determine the existence of the figure. In order to perform only the limited 11'Wl inspection,
The setting of the case point spacing ΔX and ΔY, which can greatly increase the work speed, is determined by searching for the minimum shape D$ (for example, the diameter of the shape), and on the other hand, the four case points are stored within the minimum shape, Therefore, no matter what shape it is, it always contains two case points, and when checking the case points by kneading, no figure will be overlooked.For convenience of calculation, the present invention takes ΔY=Ds −C1, and ΔX:
Since I/2ΔY, C=2 can be used to apply to the situation where the diameter D of the general figure is s-18.

(2) As shown in Figure 3, when searching for figure P (AI is a dark point) from case point A1, jump downwards by a length of ΔY from the center point of horizontal line step BC within figure Pili that includes point A1. At this time, if the H point is still within the figure P, draw a horizontal line KL that includes the H point within the figure PR, and then draw the dark lines of case points A4 and A5 that can be included on the line. Therefore, subsequent inspections do not overlap, and after inspecting point A1, the next inspection point will be case point A3 after point C, so the difference between AI and C. The possible case point A2 in between may also not be examined.

(3) To find the center of the figure from case point AI, refer to Figure 3, take the center point from line membrane BC, and then point E and F at the lower edge of figure P by pointing upward and downward from D. If the center G is found from the line membrane EF, it becomes the center of the figure P.

(d) Conversion from image coordinates at the center of each figure to trapezoid (figure) coordinates: When converting the image coordinates at the center of each figure to trapezoid coordinates, use a mathematical method to correct the geometric deformation of the image, and then Shoot to deform the image to minimize conversion errors! There are distortions and deformations caused by the lens in IIF, and errors related to the electric device, and these errors are related to the nasal line length between the length in the image and the actual phase leap length not shown. In order to accurately and quickly convert the center coordinates of a figure on an image into trapezoidal coordinates, the present invention uses the one conversion matrix method or the one specific gravity difference method, which will be explained as follows: l. Matrix conversion method: on the trapezoid The image coordinates of these reference points are found on the image using several known reference points, and one transformation matrix (T
), and use this to convert the coordinates of the image and the platform plane,
This matrix (T) is as shown below: When the known coordinates of an arbitrary point on the image are (X, y), the corresponding position [(X.

Y) can be found from the following formula: X=A/l, Y=B/L (2) Regarding how to find the matrix [T], the image coordinates (t
zsvA), j = = 1, 2.3, 4 and the corresponding trapezoidal coordinates (UJ, vj), J = 1.2, 3.4 as an example: Uj = a; / t j, Vt
=b J/ tj (5) To simplify: ``I' book u3 + T12 v j + T Ia-U J
731 books u j - U J books T 32 books vj = [Jj
(6) T21 uj + T22 y J+ T 2
3- V j zero T 31 pieces u 1- V i piece T 32 pieces v
j=V i (7) That is, [T] is determined by the following matrix equation.

In operation, the image is divided into several matrix blocks according to the required accuracy, and one transformation matrix [T] is calculated individually from the four vertices of each matrix block to perform coordinate transformation.

2. Specific gravity deviation method: (1) Locus Ia (uI, v,), i of multiple reference points on Saikou
= 1.2.3. ...n, n are integers, for example 9
, the corresponding pedestal d is (UI@V +),
t = 1s2tL...n.

(2) Calculate Ki, Mi: Kt=U+/ut, i=1 to n; M+=V
+/V i, I = 1 to n; (3) For any point on the iris, the known coordinates are (x*y), and the individual distance di of each reference point (u i, v +) is , calculated from the following formula: d + = J (x - u +) '+ (y - v4)
' From i=1 to n: (4) Calculate K, MilI to 11 K=Σ□/Σ□.

ru・' d+ 6・'d+ II MI 11 1M= Σ □
/ Σ □.

−・' d+ '・' d+(5
) pedestal seat (X, Y) corresponding to the image seat 1m (x, y)
Find: This provides the basis for later imaging each figure one by one. According to the ideal arrangement principle, the figure closest to the first figure is arranged as the second figure, and the figure closest to the second figure is arranged as the third figure, and the rest follows; however, the number of figures is large. When arranging according to the above-mentioned ideal principle, it wastes a lot of time and affects the speed of system work.

Therefore, the present invention presents a unique and fast "segmentation and orientation +9 arrangement method" as follows: 1. Segmentation: In order to increase efficiency, the present invention provides an image processing unit (6)
When calculating precisely a certain graphic material, the table surface (
1) moves to the next drawing, and the image processing unit (6
) has a large amount of time for precisely calculated graphic data, and white surface (
Since the moving speed in 1) is determined by the eye, the principle of 0 division in the present invention is to appropriately divide the figure (20) on the entire surface of one sheet and individually capture and process most of the image. The amount of block movement of the table surface (1) is also far, and the time required for the leap distance of both figures is shorter than or equal to the time required for precise calculation of one figure data of the image processing unit (6). When performing precise layer imaging of each figure one by one, the calculation work of the image processing unit (6) does not need to be interrupted.

2. Arrangement order: As shown in Figure 4, ΔY=D obtained from the jump test mentioned above.
Similarly to the value of s-C, if the image is divided into ΔY square paths and ΔY square paths, there are n figure center loci act (tz, vt), i: at most one path can correspond to the figure center.
For 1, 2, ..., n, n sets of numerical sequences (a is
bi+1) can be established as the basis for the arrangement order, where i is the number of the figure and is an integer from l to n; a
+'= Takes an integer value of (u+ /ΔY) and represents the U-axis orientation of diagram C1; b1= Takes an integer value of (vt /ΔY) and represents the V-axis orientation of diagram CI 0For example: , the number sequence corresponding to figure C1 in Fig. 4 is (2, l,
1), and the number sequence corresponding to figure C2 is (9, 1, 2
), etc. Based on this, the −th of each sequence is
Arrange each sequence of numbers in order from smallest to largest, and obtain the order of figures from the third value i of the sequence; if the sequences of the -th value a, are the same, then the magnitude of the second value, i, is the same. In this case, the existence of a numerical sequence of one previous ai value is determined by a method of changing the magnitude arrangement of b1 values. Taking Fig. 4 as an example, the arrowhead series P1, P2, . . . , P9, etc., which can be explained relatively easily, check whether each route has the execution order of the figures. Since there is no figure in the 21st line with an arrowhead indicating that,
Line 22 continues the inspection in the same direction and detects figure c5, then line 23 changes the inspection direction and detects figure C.
3 is obtained, and the 24th row changes the inspection direction again to obtain figures C3 and C7. Below, in a similar manner, each figure has an arrow mark series Q and a connected Cs, as shown in the figure.
'C++ C31C7C@ * Arranged in C4 and C2 order, and each one is a path of layer image order. In this direction change order arrangement method, the speed to be acquired that approaches the optimal (shortest) path arrangement This is the fastest method, and compared to the method of arranging in the order of CI+C21C:lI...,C7, the movement distance of the platform (1) and the time required can be reduced, and work efficiency can be increased.

(4) Sequential processing of graphical images: The photographing device (5) is small compared to the layered image range.
Using the path obtained in step 3), layer images of each figure (20) one by one. Of course, the layer image is formed when the tabletop (1) (not shown 20) moves to a fixed position. The method of confirmation is
Any known test hand chain surface or l! You can tell whether the work 11rll is stationary or not. After the layer image is completed, the table surface (1) is continuously moved to the next position, and the image processing unit (6) simultaneously proceeds with the image processing. Since the image is relative to the approximate center of the figure, the center point of the image is always within 1!1m of the figure.Therefore, calculate the total points of all sides of the figure from the center point toward the four circumferences, and average the coordinates of the total side points. Find the exact center coordinates of the figure, and then determine the size and shape of the figure.

(5) The method of converting the image coordinates of the exact center position of each figure to the base seat a, that is, converting them into coordinates on the outside of the figure, is the same as in step (d).

(6) Step (1!) Classify the shapes based on the data such as their size and shape found in step 1, and arrange them in an appropriate order. It is possible to provide necessary control materials to the processing equipment.These materials can be output directly to the processing equipment, or recorded with a tape puncher and then used again by the processing equipment.
The method is accomplished using known techniques.

As mentioned above, the present invention utilizes two imaging devices to obtain an outline of the figure and successive layered images of each figure by moving the table, and moves the table and image processing simultaneously, and uses automatic binarization technology. Binarize the image and feed it to the image processing unit's figure center position, dimension and shape analysis calculations, and obtain the corresponding coordinates of the tabletop again using a coordinate transformation method with corrective geometric transformation function, as well as homogeneous division and figure transformation. Through effective formulation and processing using the array method, we have achieved high-speed, high-precision off-the-mark discrimination function.
This method and device created for the first time will surely provide industrial value.

[Brief explanation of the drawing]

FIG. 1 is a configuration diagram of the apparatus of the present invention. FIG. 2 is a jump case diagram of a graphical test in an image roughly captured by the image processing unit in the present invention. FIG. 3 shows the principle of calculating the approximate center of a figure in a rough image, which is included in the present invention. FIG. 4 is an explanatory example of the figure arrangement order in the present invention. Figure 1 Figure 2 Figure 3

Claims (1)

[Scope of Claims] 1. A machine recognition method and apparatus for discriminating a large number of individual figures on a plane drawing, which includes the following steps: (a) Using a photographing device to take a picture of most of the related drawings; A partial map is a block of all or part of a drawing, and includes many individual figures; (b) A partial map is binarized and the figures are contrasted in light and dark against the background. (c) The length is long enough to accommodate one block in the minimum figure image, the width is a progressive value in both vertical and horizontal directions, and a case point jump test is performed on the image, and the location of the figure is determined by the brightness and darkness of the case point. (d) Case point of the detected figure Obtain both edge points of the figure in horizontal orientation, midpoint of both points Obtain another edge point of the figure from the vertical orientation, and set the midpoint to the figure (e) convert the approximate center coordinates of each figure on the image into coordinates on the drawing based on the known coordinates on the image that correspond to the known coordinates of multiple reference points on the drawing; (f) Create an appropriate arrangement order for the entire figure from the approximate center position; (g) From the arrangement order of the figure in step (f) using the photographing device,
Taking the approximate center coordinates of the upper hole position of the figure as the image center in step (e), take precise images of each figure and perform the following procedures: [1] Binarize the image; [2] Image Find all edge points of the figure outward from the center, average the coordinate values, and obtain the image coordinates of the exact center of the drawing; and from the coordinate properties of the edge points, determine the shape of the figure and the length on the image. (h) Convert the accurate center image coordinates and image dimensions of each figure into the figure center coordinates and actual dimensions on the drawing; (i) Classify each figure based on its shape and dimensions, and identify figures of the same type. Arrange them in an appropriate order and use them as data for output operations. 2. In the method described in claim 1, the image binarization processing in steps (b) and (g) therein and the ashness demarcation position T of the image video elements are performed as shown in the lower row. Calculate with formulas; ▲There are mathematical formulas, chemical formulas, tables, etc.▼ ▲There are mathematical formulas, chemical formulas, tables, etc.▼ ▲There are mathematical formulas, chemical formulas, tables, etc.▼ t_1=t・N_1/N_2 t_2=m-(m-t )・N_2/N_1 T=1/2 (t_1+t_2) Wherein, M = maximum grayness value of all video elements (minimum grayness value is 0) G_i = arbitrary grayness value from 0 to m n_i = grayness value is the total number of video elements of G_i N = total number of video elements, that is, ▲ There are mathematical formulas, chemical formulas, tables, etc. ▼ 3. The method as stated in claim 1, steps (e) and ( h) One of the image coordinates (x,
y) to drawing coordinates (X, Y), calculate using the following formula: ▲There are mathematical formulas, chemical formulas, tables, etc.▼ X=A/t Y=B/t Among them, [T]=1 ▲ There are mathematical formulas, chemical formulas, tables, etc., which can be expressed by the following formula. Then, it is obtained from the following order: There are n known reference points (n ≥ 4), and their image coordinates (u_1, v_1 ), (u_2, v_2),...
・, (u_n, v_n) are the object coordinates (U_1, V
_1), (U_2, V_2), ..., (U_n, V_
The matrix equation corresponding to n) may be a mathematical formula, a chemical formula, a table, etc.▼ 4. The method as stated in claim 1, steps (e) and (h) therein , one image coordinate (
x, y) to drawing coordinates (X, Y), calculate using the following formula: Obtain: There are n known reference points, and their image coordinates (u_1, v_1
), (u_2, v_2), ..., (u_n, v_n)
are the object coordinates (U_1, V_1) and (U_2, V_
2) Corresponding to..., (U_n, V_n), then: ▲There are mathematical formulas, chemical formulas, tables, etc.▼ Among them, d_i=√(x-u_i)^2+(y-v_i)^25
, in the method as described in the scope of the patent application, item 1, in steps (f) and (i) therein, the process of figure arrangement order is as below: [1], the smallest figure on the image The spacing is a value slightly smaller than the width, and the center of each figure is placed within the grid of the vertical and horizontal divisions of the interval, that is, the grid forms an arrangement of several rows, in order from one side to the other, the first row, the second row, etc.・Reaching to the last line, the order from one end of each line to the other is first case, second case, and so on until the final case is reached. [2] Check whether there is a figure in each case in order from the first line to the last line, and then arrange the figures according to the detected order of figures.Inspection from each line to each case is as follows. Deciding the order from the first case to the final case, or from the final case to the first case, is ``If there is a figure in a certain line, the next line's inspection direction is the opposite direction of that line; If the two directions are the same, then the direction is the same. 6. A type of machine vision device that reads plan drawings containing many individual figures, such as hole position maps for drilling holes in printed PC boats, and recognizes figures and records information such as positions and dimensions. , including: one machine table; one table, installed on the table, used as a place to place the interpretation drawing; the table, within a certain range in the plane, one drive and control; Due to the action of the device, it can be moved to any position on the table relatively; one first imaging device is fixed on the table, and its imaging axis is perpendicular to the plane of the table surface, providing a relatively large imaging range. when the table moves to a predetermined position.
A second imaging device is fixed on the table, and its imaging axis is perpendicular to the plane of the table, which is relatively easy to use. When the table surface moves to a predetermined position, a small imaging range can be obtained.
It takes an image of the object on the table including a single figure, converts it into an electric signal and outputs it; there is an image processing unit, an image digitization device and a microcomputer device, which contains a process control program; The stereotactic movement of the first and second imaging devices on the table mentioned above is operated to capture the image signals of the first and second imaging devices in a timely manner; the image analysis program includes image binarization, graphic inspection, and graphic center. It includes processes such as calculation, conversion of image coordinates to object coordinates, and figure order arrangement treatment, and in conjunction with operation of the process control program, images most of the drawing from the first imaging device and determines the approximate center of each figure. , provide the second photographing device with the approximate center of taking precise images of each figure one by one in an appropriate order, and obtain accurate data of each figure. 7. The size of the imaging range of the first photographing device in the apparatus as described in claim 6 is defined as "the table surface moves to the two figure positions furthest within its image ( The time required (for the acquisition of the image by the second imaging device) is not greater than the time required by the image processing device to obtain one accurate graphical document.