JP2850007B2 - Reference point recognition method for regular patterns - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method of recognizing a reference point of a regular pattern, and more particularly to a method of recognizing a reference point of a regular pattern that can be used for distortion correction of a picture film.

[Conventional technology]

2. Description of the Related Art In recent years, an extraction simultaneous painting method in which a printing film is put into a mold at the same time as injection molding and only ink is transferred to a molded product has been widely used.

FIG. 1 is a diagram showing a basic configuration of an apparatus for performing such simultaneous injection painting method. A transfer film 2 is wound around the supply roll 1. The transfer film 2 includes a cylinder 3
Is passed through the heater 4 connected to the male die 5 and the female die 6 to be wound by the winding roll 7. On the transfer film 2, a pattern, a character, or the like to be transferred to a molded product is printed in advance. When the pattern, characters, and the like are aligned with the mold, and the male mold 5 is pressed against the female mold 6 or suction is performed from the female mold 6, the transfer film 2 is heated by the heater 4 and easily stretched. , And deforms along the shape of the female mold 6.
Thus, when the resin is injected from the male mold 5, the pattern, characters, and the like of the transfer film are transferred to the molded product during injection molding.

FIG. 2 is a view showing a state where the transfer by the transfer film 2 to the molded article 8 is performed by the above-described method. A printed molded product 9 in which characters are transferred to the upper surface of the three-dimensional molded product is produced, but the printed surface is actually distorted as shown in FIG. This is because the transfer film 2 extends.

In order to correct such distortion, it is necessary to measure the degree of distortion. For this purpose, a regular pattern is printed on the transfer film in advance, and the degree of distortion of the reference pattern when the transfer film is formed along the molded product may be measured. That is, the change of the reference point position of the regular pattern before and after molding may be seen.

[Problems to be solved by the invention]

However, conventionally, the recognition of the reference point of the regular pattern has to be manually performed, which requires a great deal of labor and time. This is because it is generally considered that it is difficult to recognize a reference point of a distorted pattern by uniform judgment. For example, in a square grid pattern, the reference point is the intersection of the vertical and horizontal lines that make up the grid, but if this is distorted, not all the reference points will be the intersection of the vertical and horizontal lines. There may be an intersection between two oblique lines, and the vertical and horizontal positional relationship becomes unclear. For this reason, complicated judgment by humans has been required.

Accordingly, an object of the present invention is to provide a method of recognizing a reference point of a regular pattern, which can automatically recognize the reference point without manual operation.

[Means for Solving the Problems] The present invention is a method of automatically recognizing each reference point using a computer for a regular pattern having a plurality of reference points, comprising: (A) an input step of capturing as a set, a thinning processing step of making the line width of the pattern one pixel for the image captured in the input step, and (a) a thinning processing (B) Recognize the intersection of the two lines intersecting each other, and recognize the position of the pixel at this intersection as the reference point position. (B) Recognize the branch point where one thinned line branches into two. And a reference point recognizing step of recognizing another branch point near the branch point and recognizing a position of a point intermediate between the two branch points as a reference point position.

Also, the present invention is a method of automatically recognizing each reference point using a computer for a regular pattern having a plurality of reference points, comprising: an inputting step of capturing the regular pattern as a set of pixels; A step of recognizing a pattern constituting the line and tracking the regular pattern along the recognized pattern; a step of recognizing an intersection of two patterns constituting the line as a reference point in the course of the tracking; Extracting and recording the position information of the reference point thus obtained and the phase information of the reference point determined based on the tracking process that has reached the reference point.

(Operation)

According to the present invention, first, the regularity pattern is binarized and captured. Therefore, the regular pattern can be taken into a general-purpose computer. The captured regular pattern is thinned so that the line width becomes one pixel, and the intersection of the two lines is recognized as a reference point. Furthermore, after recognizing a branch point where one line branches into two, a pair of the branch points is recognized as one reference point. As described above, in addition to the intersection, the branch point can also be recognized as the reference point. Therefore, it is also possible to recognize a reference point that has been divided into two by processes such as binarization and thinning. Because these recognitions can be processed by computer-based uniform algorithms,
Recognition of all reference points can be automatically performed without manual operation.

In addition, it is also possible to recognize intersections by tracing patterns constituting a line without thinning.

〔Example〕

Hereinafter, the present invention will be described based on the illustrated embodiments.
The embodiment described here is an example in which the reference point recognition method for a regular pattern according to the present invention is applied to a transfer film distortion correction device. In the following examples, §3 is a part particularly relevant to the present invention.

§1 Basic Configuration of Apparatus FIG. 4 (a) is a block diagram showing a configuration of a transfer film distortion correction 3 apparatus according to the present invention, and FIG. 4 (b) is a detailed view of a distortion correction processing unit among them. Distortion pattern image reader 11
Is a device for inputting a distortion pattern as an image, and includes a CCD camera, an ITV, and the like. The image-to-be-corrected input device 12 is a device for inputting a pattern to be transferred to a molded product, and may be a device such as a flat scanner. The pattern is input using the composition or original information. Note that a CAD system can also be used as the corrected image input device 12. In this case, the digital data created by the CAD system is directly input as the picture. The arithmetic processing device 13 is a device that corrects a pattern based on data input from these devices, and has an image processing unit 14 and a distortion correction processing unit 15. The image processing unit 14 has a function of extracting the position coordinates of each reference point based on the pattern input from the distortion pattern image reading device 11, and the operation will be described later in detail. The distortion correction processing unit 15 is based on the data supplied from the image processing unit 14,
It has a function to correct the picture input from the user. A storage device 16 such as a magnetic disk or a magnetic tape is externally connected to the distortion correction processing unit 15, and can store data input once, calculated data, and the like. The storage device 16 may be incorporated in the arithmetic processing device 13. Distortion correction processing unit
The corrected picture corrected in 15 is output by the corrected picture output device 17. As the corrected picture output device 17, devices such as a plotter, a dot impact printer, an ink jet printer, a thermal transfer printer, and a film recorder can be used. In addition, data can be temporarily output to a storage medium such as a floppy disk or a magnetic tape, and the data can be transmitted to another output device offline. In this case, the corrected picture output device 17 is a drive device for each storage medium. The distortion correction processing unit 15 includes a first coordinate system 18 and a second coordinate system, as shown in FIG.
19 and a mapping operation device 20, the operation of which will be described later in detail.

§2 Basic operation of the apparatus Here, first, the operation of the entire apparatus shown in FIG. 4 (a) will be described. As shown in FIG. 2, when the transfer film 2 is formed in conformity with the molded product 8, the pattern is distorted as shown in FIG. Therefore, as shown in FIG. 5 (a), for example, when a square lattice pattern is printed on the transfer film 2 and is actually formed in accordance with the molded product 8, the transfer film 2 is elongated. This pattern is distorted as shown in FIG. The same applies to the case where only the transfer film 2 is molded by sucking into a female mold without actually forming a molded product. The manner of forming the transfer film is clearly shown in the top views of FIGS. 5 (c) and 5 (d). Here, an example using a square lattice pattern is shown.
In short, any pattern can be used as long as it is a reference pattern having a plurality of reference points.
In general, patterns such as a grid pattern, a pattern of oblique coordinates, and a pattern of polar coordinates are preferable. In the case of a square lattice pattern, each intersection of the lattice, that is, the four vertices of each square, is the reference point. In addition, if a mark for positioning with the mold is provided on a part of the transfer film and the mark for positioning is matched with the mold at the time of molding, the positioning in a later step becomes easy.

The operator uses the distortion pattern image reading device 11 to input a distortion pattern as shown in FIG. As will be described later, this distortion pattern image reading device 11 is specifically a device such as a video camera, and the read image data is processed by an image processing unit 14 into an image such as a binarization process and a thinning process. The processing is performed, and the position of each intersection (grid point) is detected. On the other hand, the operator operates the image input device 12 to be corrected.
, The picture is input as image data from the composition or original of the picture to be printed.

The distortion correction processing unit 15 generates a corrected image based on the data processed by the image processing unit 14, the data input by the image input device 12 to be corrected, and a previously stored square lattice pattern without distortion. I do. The corrected picture is output to the outside by the corrected picture output device.

Subsequently, the operation of the image processing unit 14 and the distortion correction processing unit 15 will be described in detail in another chapter.

§3 Operation of image processing unit 3.1 Overall procedure The operation of the image processing unit 14 is shown in the flowchart of FIG. First,
In step S1, a binarization process is performed. The image provided from the distortion pattern image reading device 11 is an image having a gradation. For example, one pixel has any density value between 0 and 255. When this is binarized,
All pixels take either "0" or "1". As the binarization method, generally, a binarization using a fixed threshold and a binarization using a floating threshold are known. The former is a method of dividing each pixel into “1” or “0” at a fixed density value (for example, an average value of density values of all pixels). The latter is a method in which the density value serving as a boundary is changed in each part in the image, and is effective when the brightness distribution by the certification occurs at the time of image reading. FIG. 7A shows an example in which the distortion pattern of the square lattice is binarized.

Subsequently, in step S2, a thinning process is performed.
This is a process of thinning the binarized pattern so that the line width becomes one pixel. Such a thinning process is described in, for example, “Image Processing Subroutine Package SPIDER
USER'S MANUAL ”(Showa 57, published by Kyodo System Development Co., Ltd.), page 491, is a well-known method, and the description is omitted here. FIG. 7B shows an example in which the pattern of FIG. 7A is thinned.

Finally, in step S3, an intersection tracking process is performed. This is a process for determining an intersection V as shown in FIG. 7C from the pattern subjected to the thinning process as shown in FIG. 7B. FIG. 8 is an enlarged view of the thinned pattern shown in FIG. 7B. Here, one pixel is indicated by a circle. The intersection tracking processing determines which pixel is the intersection (that is, the lattice point of the lattice pattern) among the many pixels connected to each other in this manner. Next, details of the intersection tracking processing will be described with reference to the flowcharts of FIGS. 9 and 10.

3.2 Intersection Tracking Processing First, in step S4, the number of connections is calculated for each pixel. Here, the connection number for a certain pixel is
This is a number indicating how many different pixels exist around it. As shown in FIG. 11, the connection number C of the pixel of interest indicated by hatching is 0 to 4 in each of FIGS. Since the thinning processing has been performed, the surrounding pixels are always isolated from each other,
The number of connections C = 4 is the maximum value. The attribute of the point of interest can be determined as follows from the number of links.

C = 0 Isolated point (example: point a in FIG. 8) C = 1 End point (example: point b in FIG. 8) C = 2 Continuous point (example: point c in FIG. 8) C = 3 bifurcation point ( Example: d1 to d4 in FIG. 8) C = 4 intersection (example: point e in FIG. 8) Next, in step S5, an initial tracking start point is designated. The operator can specify a point that can be clearly recognized as an intersection (for example, point e) as the initial tracking start point by viewing the display as shown in FIG. If the operator first designates this point, another intersection is automatically determined by the procedure of step S6. In step S6, a tracking start point and a tracking direction are determined. In this case, the tracking start point is the initial tracking start point specified in step S5. Further, the tracking direction may be set in advance, for example, to the right. In the following procedure, the intersection is sequentially tracked rightward from the tracking start point. FIG. 12 shows a conceptual diagram of such intersection tracking. For example, if the tracking start point is point S1, tracking is performed in the direction of the solid arrow in the figure,
Intersections S2, S3, S4 are determined. When tracking in the right direction becomes impossible, tracking in the reverse direction indicated by the broken arrow in the figure is performed from point S1. When tracking in the horizontal direction like this, tracking is performed only in the right or left direction,
No vertical tracking is performed. Therefore, for example, even if it is recognized that the point S5 is an intersection, tracking to the intersection S5 is not performed at present. Actually, the intersection tracking is performed in the following procedure.

First, an intersection detection process is performed in step S7. Now
In FIG. 13 (a), it is assumed that tracking is performed in the direction of the arrow in the figure, and that each intersection up to intersection a has been tracked. Here, "a certain intersection has been tracked" means that a certain pixel is a lattice point of a lattice pattern,
In this state, the coordinate values are also confirmed, and the topological position in the lattice is also confirmed. The confirmation of the topological position is such that, for example, in FIG. 13 (b), the connection state of the intersection as shown by the solid line is correct, and the connection state as shown by the broken line is incorrect. The intersection detection process of step S7 is a process of detecting three intersections b, c, and d connected to the intersection a. Since the intersection behind the arrow has already been tracked, three intersections b, c, and d remain as untracked intersections adjacent to the intersection a, and these three intersections are obtained in step S7.

Next, in step S8, it is determined whether tracking is successful. Among the three intersections, the intersection (the intersection c in this case) located at the topological center is the intersection to be tracked next, and it is determined whether or not this is appropriate as the tracking intersection. . That is, it is determined whether it is appropriate to trace the intersection c after the intersection a. This may be determined, for example, by determining whether the distance between the points ac is within a predetermined range.

If the intersection c is appropriate, it is determined that tracking has succeeded, and this intersection c is stored as a tracked intersection in step S9. Specifically, a matrix for recording the tracked intersection coordinates is prepared, and the coordinate value of the intersection c is recorded to the right of the coordinate value of the intersection a. By using a matrix in this way, the position of the intersection and the phase relationship can be recorded simultaneously. Subsequently, an untracked intersection is recorded in step S10. here,
The untracked intersections are the intersections b and d in FIG. 13 (a). Since the intersection tracking is performed in the right direction, at this stage the intersections b and d
Are not tracked, but since they were recognized as intersections, we record the fact that these points are intersections but have not been tracked yet.

Thus, the process returns to step S7 again, and the next three intersections e, f, and g
Is detected. Hereinafter, this procedure is repeated, and intersection tracking is performed rightward and rightward. Thus, as shown in FIG. 13 (c), the hatched intersection is recorded as a tracking intersection, and the intersection indicated by a double circle is recorded as an untracked intersection. The intersections indicated by the crosses have not been detected yet.

If the tracking is not successful in step S8, it is determined in step S11 whether the reversal of the tracking direction has been performed so far.If the reversal has not yet been performed, the tracking direction is reversed in step S12, and Tracking continues. That is, tracking is performed in the direction indicated by the dashed arrow in FIG. If the tracking has not been successful and the tracking direction has been reversed, it is determined in step S13 whether or not an untracked intersection remains. If it remains, the process returns to step S6, and tracking is continued with any one of the untracked intersections as a new tracking start point. That is,
In FIG. 13 (c), any one of the intersections indicated by double circles is set as a tracking start point, and first, tracking is performed in the right direction. By this tracking, the intersection that was an untracked intersection (double circle) until now changes to a tracked intersection (hatched), and at the same time an intersection (x mark) that was not detected until now is detected And recorded as a new untracked intersection. Eventually, in the subsequent processing, the intersection sequentially changes from an undetected intersection (x) to an untracked intersection (double circle) and finally to a tracked intersection (hatched). When there are no untracked intersections in this way, all the intersections have been tracked, and the position coordinates of the intersection V as shown in FIG. 7 (c) are obtained.

3.3 Intersection Detection Processing Next, the details of the intersection detection processing in step S7 in FIG.
This will be described with reference to the flowchart in the figure. As described above, this process is a process for detecting intersections b, c, and d from the intersection a in FIG. First, in step S14, a detection direction is detected. In the example of FIG. 13 (a), if the intersection b is detected, the upward direction is the detection direction. Then, in step S15, one pixel is tracked in the detection direction. That is, attention is paid to the pixel g. Then, it is determined whether the attribute of the focused pixel is a branch point (step S16), an intersection point (step S17), an end point (step S18), or a continuous point (step S19). (Because the pixels are sequentially tracked, they are not isolated points.) This attribute can be determined by referring to the number of connected pixels obtained in step S4 as described above. As long as the tracked pixels are continuous points, the process returns to step S15 to continue tracking one pixel. In the example of FIG. 13 (a), the intersection point a to the point b (at this point, it is not recognized that the point b is the intersection point).
The tracking is performed upward one pixel at a time. (1) If a pixel having the attribute "intersection" is found (step S17), it is determined in step S21 that an intersection has been detected. In this case, the coordinate position of the pixel becomes the coordinate position of the intersection as it is. For example, the attribute of point e in FIG. 8 is “intersection point”, and this coordinate position is the coordinate position of the intersection point as it is.

(2) If a pixel having the attribute “branch point” is found (step S16), a pair of branch points is searched for in step S20, and the average coordinates of the two are set as the coordinate position of the intersection. For example, if the point d1 in FIG. 8 is found as a point having the attribute "branch point", a branch point d2 corresponding to this point is searched for. For this, for example, d2 may be searched as another branch point existing within a predetermined radius from the branch point d1. It is considered that such a pair of branch points was originally one intersection, so the line segment d1d2
Is set as the coordinate position of the intersection. The same applies to the branch point pairs d3 and d4.

(3) If a pixel having the attribute "end point" is found (step S18), in step S22, "intersection cannot be detected"
Make a judgment.

In this manner, the detection of the adjacent intersection is performed. As described above, when performing intersection tracking, three adjacent intersections are detected for one intersection. Therefore, until step S23 is completed for all detection directions, step S14
Are repeated, and three intersections are detected.
Note that such intersection detection is processing that is necessary only when no intersection is detected, and is unnecessary processing when an adjacent intersection is already detected as a tracked intersection or an untracked intersection. .

3.4 Another Method of Obtaining Intersection I As described above, the image of the binarized distortion pattern as shown in FIG. 7A is thinned as shown in FIG. The position of the intersection is to be determined as shown in FIG. 3 (c). Another simpler method for obtaining the position of the intersection will be described here. This method does not require any procedures such as thinning and intersection tracking. From the binarized image shown in FIG. 7 (a), the intersection position (to be exact, not the intersection position itself but the position of a point corresponding to the intersection) can be obtained.

As shown in FIG. 7A, the image of the binarized distortion pattern is a set of white or black pixels. The thinning process described above is a process of narrowing the width of a black portion to one pixel, and the intersection tracking process is a process of further determining the position of a black pixel that is an intersection. In any case, it can be said that the processing focuses on the black pixel. Another method described here is a process that focuses on white pixels. An enlarged view of the pattern shown in FIG. 7A is shown in FIG. Here, the white portion is composed of many white pixels, and the black portion is composed of many black pixels. Now, when the geometric center of gravity W is obtained for each independent white portion,
As shown in FIG. 7 (d), the center of gravity W is determined at the center of each white portion. The position coordinates of the center of gravity W can be obtained by a simple arithmetic operation. The feature of this method is that the center of gravity W
Is used instead of the intersection V. As shown in FIG. 7 (d), the center of gravity W also has a lattice arrangement, just like the intersection V has a lattice arrangement. FIG. 7 (e) shows the relationship between the grid consisting of the intersections and the grid consisting of the center of gravity. Assuming that there is a grid consisting of intersections as shown by the solid lines in this figure, the grid consisting of the centers of gravity is a grid shown by the broken lines in the figure. Although each lattice point is displaced, all have substantially the same phase information as a lattice. Therefore, if the original lattice has a distortion, the lattice formed by connecting the centers of gravity of the lattice has the same distortion. After all, instead of finding the intersection V in FIG. 7 (d), finding the center of gravity W and treating this as the intersection does not cause any trouble. This method has the merit that not only the operation is simple, but also that it is less susceptible to noise mixed in during image reading.

3.5 Alternative methods for finding intersections II The above methods are all methods for finding intersections after performing thinning processing. Here, a method for finding fine weather without thinning processing will be described. FIG. 14A shows a part of the pattern before the input pattern is thinned. That is, it corresponds to a partially enlarged view of the pattern shown in FIG. Here, each pixel is represented by a white circle. Although a human can recognize this pattern as a part of a horizontal line having a width W, analysis by a predetermined algorithm must be performed in order for the computer to recognize the pattern.

Therefore, first, a circle having a diameter R is defined. Where R> W
Set so that The area surrounded by the circle is called a spot closed area. This spot closed area is
As shown in FIG. 14 (a), the pixel is superimposed on a part of the pattern, and a pixel on or near the boundary of the spot closed area is extracted as a boundary pixel. FIG. 14 (b) is an enlarged view of FIG. 14 (a). Here, the hatched pixels are the extracted boundary pixels. In this example, the boundary pixels have a distribution divided into two groups, G1 and G2. When the distribution of boundary pixels is divided into two groups in this way, it is determined that the current spot closed area is on a continuous point of the pattern. That is, it is equivalent to FIG. 11 (c) in the above embodiment. After all, the point of this method is that the number of groups in the distribution of boundary pixels should be treated equivalently to the connection number C in the above-described embodiment.
In the case of FIG. 14 (b), the number of groups is two, so it may be treated equivalently to the case where the number of links C = 2.

When the same determination is made for the spot closed area SP3 in FIG. 14 (c), since the boundary pixel is only one group, it becomes equivalent to the case where the number of connections C = 1, and corresponds to FIG. 11 (b). What should I do? That is, it is determined to be an end point.

When the same determination is made for the spot closed area in FIG. 14 (d), the boundary pixels are divided into four groups G1, G2, G3, G4, which is equivalent to the case where the number of connections C = 4. A treatment corresponding to (e) may be performed. That is, the intersection is determined.

When the same determination is made for the spot closed area in FIG. 14 (e), the boundary pixels are divided into three groups G1, G2, G3, which is equivalent to the case where the number of connections C = 3, and FIG. )). That is, it is determined to be a branch point.

As described above, the intersection can be recognized without performing the thinning processing. In FIG. 14 (d), the position of the intersection is a straight line connecting the barycentric position of the group G1 pixel and the barycentric position of the group G2 pixel, the barycentric position of the group G3 pixel and the barycentric position of the group G4 pixel. Intersection PX with a straight line connecting
May be obtained, and these may be used as intersection coordinates.

The tracking from the intersection to the next intersection is shown in Fig. 14 (c).
As shown in (2), when the determination on one spot closed area SP1 is completed, the spot closed area may be moved to SP2 and the same determination processing may be repeated. As shown in FIG. 14 (b), the moving direction of the spot closed area is set to the direction of a straight line 1 connecting the barycentric position g1 of the four pixels of the group G1 and the barycentric position g2 of the pixels of the group G2. The movement pitch Pt may be specified by an operator so that Pt <R. However, when Pt << R, processing takes too much time, which is not preferable. FIG. 14 (e)
FIG. 3 shows an example of a branch point, but since a thinning process is not performed, a branch point does not appear theoretically.

§4 Operation of the distortion correction processing section 4.1 Overall procedure As described above, the distortion correction processing section 15 stores in advance reference pattern data as shown in FIG. 5 (c). In addition,
This reference pattern data may be read from the storage device 16. The distortion correction processing unit 15 includes an image processing unit 14.
The distortion pattern data as shown in FIG. 5 (d) is given, and the data of the image to be corrected is provided from the image input device 12 to be corrected. Two coordinate systems are prepared in the distortion correction processing unit 15 as shown in FIG. 4 (b).

Hereinafter, the operation will be described with reference to the basic configuration diagram of FIG. 4 (b) and the flowchart of FIG. First,
The distortion pattern data is given to the first coordinate system 18 (step
S24), the reference pattern data is given to the second coordinate system 19 (step S25). The corrected pattern data is the first
This is given to the coordinate system 18 (step S26). In this example,
The case where the character “A” is treated as a picture will be described. Therefore, on the first coordinate system 18, a normal character (corrected picture) "A" not distorted in the distorted pattern overlaps. The mapping calculation device 20 converts the mapping of the character “A” into the second based on the correspondence between the reference point of the reference pattern on the first coordinate system 18 and the reference point of the distortion pattern on the second coordinate system 19.
The calculation to be performed on the coordinate system 19 is performed (step S27). This mapping shows the distorted letter "A" as shown in FIG.
(Corrected pattern). The image data of the distorted character “A” is output to the corrected picture output device 17.
In response to this, the corrected picture output device 17 (for example, a plotter) draws the distorted character "A" as a corrected copy (step S28). If the distorted character "A" is printed on the transfer film 2 based on the corrected copy and formed and transferred under the same conditions as the previous time, as shown in FIG. As a result, the letter “A” without distortion is transferred to the upper surface of the printed molded product 9.

4.2 Embodiment of Mapping Operation Next, an embodiment of the mapping operation performed by the mapping operation device 20 will be described. As shown in FIG. 4 (b), the mapping operation device 20
The task of obtaining the mapping of each point constituting the picture on the first coordinate system 18 on the second coordinate system 19 is performed. That is, it is only necessary that the mapping point Q on the second coordinate system 19 can be obtained for any one point P on the first coordinate system 18.

Conventionally, there has been known a method of obtaining a function f for converting a normal pattern on the second coordinate system 19 into a distorted pattern on the first coordinate system 18. However, in order to find the mapping point Q of the point P, it is necessary to find the inverse function g of the function f, which is a mathematically very difficult task. Therefore, a method that does not use such a function is considered. Now, one point P is a reference point (lattice point)
If the point is located at the position, the mapping point Q for this point can be easily obtained. That is, in FIG. 4B, the mapping point of one point P1 is the point Q1. The lattice points corresponding to the topological points of the square lattice are the mapping points. The problem is a method of obtaining a mapping point Q2 for a point such as one point P2 inside the lattice. Here, the second point corresponding to the lattice ABCD to which the point P2 belongs
The grid EFGH on the coordinate system can be readily found as a topologically corresponding grid. In the case of this example, since the grid to which the point P2 belongs is the lower right grid, the corresponding grid on the second coordinate system is also the lower right grid. Subsequently, a point Q2 corresponding to one point P2 in the grid ABCD may be obtained in the grid EFGH.
This point Q is ultimately determined as a point at a position corresponding to the point P in phase. As described above, several methods have been conventionally known for obtaining a topologically corresponding point. However, all of the conventional methods have a problem that a step is formed in a picture. That is, as shown in FIG. 16 (a), when the mapping of each point constituting the pattern extending over two adjacent unit lattices is obtained, the mapping as shown in FIG. 16 (b) is obtained. However, in the conventional method, a step is generated as shown in FIG. The inventor of the present application has devised several specific methods capable of obtaining a mapping in which a pattern does not have a step. Four examples will be described below.

The following four methods apply the same rules to all of them. That is, when calculating a mapping for a point (for example, point P in FIG. 16 (a)) that straddles an adjacent unit cell, two grid points (the 16th point) sandwiching the straddling point are obtained.
The mapping (point Q in FIG. 16 (B)) is determined only by the coordinate values of points (B, C) in FIG. (A). If the mapping is obtained by a method that satisfies such a condition, the problem that a step is formed in the picture can be solved.

<M: n Division Method> First, the first method will be described with reference to FIG. Now, as shown in FIG. 17 (a), one point P in the grid point ABCD
Is mapped to a square lattice EFGH shown in FIG.
Consider the case of seeking within. First, a square ABCD is created by connecting the grid points ABCD. Then, the intersection point X of the straight line AB and DC is connected to the point P by a straight line, and a ratio m: n at which the point P divides a portion of the straight line in the rectangle ABCD is determined. Further, the intersection point Y of the straight lines AD and BC and the point P are connected by a straight line, and the straight line ABCD
A ratio q: r at which the point P divides the portion inside is obtained. On the other hand, in the square lattice EFGH, a straight line connecting two points IJ dividing the sides EF and HG into m: n and a straight line connecting two points KL dividing the sides FG and EH into q: r are drawn, This intersection is designated as point Q. Since each point is given by a two-dimensional coordinate value of (x, y), the above method can be performed by a very easy operation. As shown in FIG. 18, when the opposite side of the square ABCD, for example, the side BC and AD are parallel, the intersection Y is not obtained. In this case, a straight line passing through the point P and parallel to the side BC or AC is drawn. Just think.

<Equal Partition Method> The second method will be described with reference to FIG. First, the 20th
As shown in FIG. 3A, points I and J passing through the point P and dividing the side AB and the side CD at an equal ratio m: n (AI: IB = DJ: JC = m:
n and points K and L passing through the point P and dividing the side BC and the side AD at equal ratios q: r (AK: KD = BL: LC = q: r)
And a straight line l2 passing through. Using the ratios at this time, m: n and q: r, a mapping point Q is obtained as shown in FIG. 20 (b). That is, two points I ′ dividing the sides EF and HG into m: n, respectively.
The intersection point of a straight line connecting J 'and a straight line connecting two points K'L' that divide the sides FG and EH into q: r may be set as a point Q.

An example of a method for calculating m: n by calculation will be described below. Now
The coordinates of the four points ABCD are (xa, ya), (xb, yb),
(Xc, yc), (xd, yd), and the coordinate value of the point P is (xp, yp)
And Here, coordinate values of points I and J are represented by (xi, yi), (xj, y
j), xi = m · (xb-xa) + xa (1) xj = m · (yb-ya) + ya (2) yi = m · (xc-xd) + xd (3) yj = m · (yc −yd) + yd (4) In general, a straight line passing through two points X1 (x1, y1) and X2 (x2, y2) is (y−y1) (x2−x1) = (x−x1) (y2−y1) (5) is represented by Accordingly, the equation of the straight line l1 is as follows: (y-yi) (xj-xi) = (x-xi) (yj-yi) (6) Substituting equations (1) to (4) into this equation, and x, y
By substituting the coordinates (xp, yp) of the point P into the following equation, an expression for m in the form of am ² + bm + c = 0 (7) is obtained. Here, a to c are coefficients obtained from known coordinate values. By solving the equation (7), m satisfying 0 ≦ m ≦ 1 is obtained.

Since n = 1−m (8), the ratio of m: n can be obtained by calculation. q: r
Is similarly obtained.

<Distortion Amount Method> Next, a third method will be described. First, the ratio of m: n and q: r is determined by using the above-described first method or second method. Here, the case where these ratios are obtained by the first method will be described. In FIG. 19, the difference between the x and y coordinate values of the corresponding vertex of the square EFGH (FIG. 4 (b)) is determined for the x and y coordinate values of each point ABCD. For example, when the coordinate value of the point A is (x, y) and the coordinate value of the point E is (x *, y *), the difference is Δ1x = x−x *, Δ
1y = y−y *. For each point of ABCD,
Determine as shown in the figure. Then, the total sums Δx and Δy of the differences are obtained by the following equations.

Δx = Δ1x · n / (m + n) · r / (q + r) + Δ2x · m / (m + n) · r / (q + r) + Δ3x · m / (m + n) · q / (q + r) + Δ4x · n / (m + n) · q / (Q + r) Δy = Δ1y · n / (m + n) · r / (q + r) + Δ2y · m / (m + n) · r / (q + r) + Δ3y · m / (m + n) · q / (q + r) + Δ4y · n / ( (m + n) · q / (q + r) A point Q is obtained at coordinates obtained by moving the point P by the differences Δx and Δy.

<Triangle Vector Ratio Division Method> Finally, the fourth method will be described with reference to FIG. In this method, as shown in FIG. 21 (a), a square to which point P belongs is divided into two triangles, and the triangle to which point P belongs is extracted to obtain a mapping. That is,
Square ABCD and square EF in the previous three methods
Instead of GH, use triangle ABC (Fig. 21 (b)) and right-angled isosceles triangle DEF (Fig. 21 (c)), respectively, and ignore paired triangles indicated by dashed lines in the figure. I just need.

First, the vector ▼ is subtracted from the point a to the point P, and the vector ▼ is represented by ▼ = a ▲ + b ▲, using the vector ▼ and the vector ▼ as unit vectors, to obtain coefficients a and b. Where 0 ≦ a ≦
1, 0 ≦ b ≦ 1. And two unit vectors ▲
A vector DQ represented by ▼ = a ▲ + b ▲ is obtained by ▼ and ▼, and a point Q is obtained as the tip position.

4.3 Supplement on Mapping Operation Finally, a preferred embodiment for performing a specific mapping operation will be supplementarily described.

First, when the pattern data to be corrected is given to the first coordinate system as vector data, it is preferable to obtain a mapping after subdividing this vector data. For example, as shown in FIG. 22 (a), when the pattern to be corrected is given by vectors at five points, when mapping of these five points is obtained and the mapping points are connected by a new vector, fine points between the points are obtained. Information is lost. Therefore, first, as shown in FIG.
After subdividing the vector data and minimizing the length of one vector, as shown in FIG. 3C, if a mapping is obtained in the second coordinate system to obtain a corrected pattern, even fine information between points can be obtained. Will be reproduced.

If the corrected picture data is given as raster data to the first coordinate system, the corrected picture obtained in the second coordinate system may have missing pixels. This is shown in FIG. Here, FIGS. 6A and 6B show a corrected pattern and a distortion pattern given to the first coordinate system, and FIGS. 6C and 6D show a corrected pattern and a corrected pattern given to the second coordinate system. 3 shows a reference pattern. The mapping of the pattern to be corrected shown in FIG. 7A corresponds to the corrected pattern shown in FIG. 7C, but it can be seen that pixel omission has occurred in a portion indicated by a white circle in FIG. This is because a mapping was obtained in the second coordinate system for each pixel in FIG.

One method for dealing with such pixel omission is a method of performing interpolation based on surrounding pixels. For example, a pixel indicated by a black circle in the figure is represented by “1”, and other pixels are represented by “0”.
If the sum of the values of the eight surrounding pixels among the pixels having the value “0” is equal to or more than a predetermined value (for example, 5 or more), the operation of correcting the pixel to “1” is performed. All pixels indicated by white circles in FIG. 9C are corrected to black circles.

Another method for dealing with pixel omission is a method of obtaining an inverse mapping to the first coordinate system and performing interpolation based on a pixel at the position of the inverse mapping. For example, when the inverse mapping to the first coordinate system is obtained for the white circle pixel in FIG.
The mapping should be obtained at the position of any of the black circled pixels. Therefore, if there is a black dot at the position of the inverse mapping, interpolation may be performed so that the original pixel on the second coordinate system is also corrected to a black dot.

It should be noted that such a method of obtaining the inverse mapping can be used not only for interpolation of missing pixels, but also for obtaining the raster data itself of the corrected pattern on the second coordinate system. In this case, for each pixel on the first coordinate system, the operation of obtaining a mapping on the second coordinate system becomes unnecessary. For example, in the example shown in FIG. 23 (c), 10 × 10 pixels are defined on the second coordinate system. It is undecided whether each pixel is "0" or "1". Then, for each pixel, an inverse mapping is obtained on the first coordinate system, and the value of each pixel defined on the second coordinate system is set to “0” based on the value of the pixel at the inverse mapping position. Or "1".

§5 Industrial applicability As described above, an example in which the present invention is applied to the simultaneous injection painting method has been described. However, the present invention can be widely used in general for distortion correction of a transfer film or a laminating film. For example, in the process of manufacturing a can using a molding means or a molded product using a resin (for example, an in-mold molded product or a shrink film), when printing a pattern before molding,
Although the distortion of the picture is caused by the expansion and contraction of the material, even in such a case, if the corrected picture obtained by the present invention is printed, a picture without distortion after molding can be obtained.

In addition, the present invention can be widely used in the technology that needs to recognize the reference point of the regular pattern in short, not only in the field of distortion correction of the picture film,
It can be applied to a wide range of general image processing techniques.

〔The invention's effect〕

As described above, according to the present invention, the reference point of the regular pattern can be automatically recognized by the computer, so that the labor conventionally conventionally performed manually is significantly reduced.

[Brief description of the drawings]

FIG. 1 is a basic configuration diagram of a general apparatus for performing the simultaneous injection painting method, FIG. 2 is an explanatory diagram of the simultaneous injection painting method, and FIG. 3 is a view illustrating a transfer film distorted as a result of performing the simultaneous injection painting method. It is. FIG. 4 (a) is a block diagram showing a basic configuration of a transfer film distortion correcting apparatus according to the present invention, and FIG. 4 (b) is a detailed explanatory diagram of a distortion correction processing unit in the apparatus shown in FIG. 4 (a). ,
FIG. 5 is a diagram showing that the transfer film is deformed by molding, and FIG. 6 is a flowchart showing a processing procedure of the image processing unit in the apparatus shown in FIG. 7 (a) to 7 (c) are diagrams showing processing results along the flowchart of FIG. 6, wherein FIG. 7 (a) is a pattern after binarization processing and FIG. 7 (b) is a thinning processing FIG. 14C shows the pattern after the intersection tracking processing. FIG. 7D is an enlarged view of FIG. 7A, and FIG. 7E is a conceptual view showing that the center of gravity can be used as an intersection. 8 is an enlarged view of the pattern after the thinning processing shown in FIG. 7 (b), FIG. 9 is a flowchart showing the detailed procedure of the intersection tracking processing in FIG. 6, and FIG.
It is a flowchart which shows the detailed procedure of the intersection detection processing in a figure. FIG. 11 is a diagram showing the principle of calculation of the number of links shown in FIG. 9,
FIG. 12 is a conceptual diagram of the intersection tracking processing shown in FIG. 9, and FIG. 13 is an explanatory diagram of the intersection tracking processing shown in FIG. FIG. 14 is an explanatory diagram of a method of obtaining an intersection without performing a thinning process. No.
FIG. 15 is a flowchart showing a processing procedure of a distortion correction processing unit in the apparatus shown in FIG. 4, FIG. 16 is a view for explaining a step generated in a pattern by a mapping operation, and FIGS. 17 and 18 are m: n division methods. , FIG. 19 is an explanatory diagram of a distortion amount space method, FIG. 20 is an explanatory diagram of an equal division method, FIG. 21 is an explanatory diagram of a triangle vector ratio division method, and FIG. 22 is represented by vector data. Illustration of a method for obtaining a mapping after performing vector subdivision on a picture, FIG. 23 shows a picture represented by raster data,
FIG. 4 is an explanatory diagram of a method of performing pixel missing interpolation of a mapping. DESCRIPTION OF SYMBOLS 1 ... Supply roll, 2 ... Transfer film, 3 ... Cylinder, 4 ... Heater, 5 ... Male type, 6 ... Female type, 7 ... Winding roll, 8 ... Molded product, 9 ... Printed products, 11 ……
Distortion pattern image reading device, 12
13 arithmetic processing unit, 14 image processing unit, 15 distortion correction processing unit, 16 storage device, 17 corrected image output device,
18 ... first coordinate system, 19 ... second coordinate system, 20 ... mapping operation device.

Claims (10)

(57) [Claims] 1. A method for automatically recognizing each reference point using a computer for a regular pattern having a plurality of reference points, comprising: an input step of taking an image of the regular pattern as a set of pixels; For the image captured in the input step, a thinning processing step is performed so that the line width of the pattern becomes one pixel. For the pattern subjected to the thinning processing, (a) two thinned lines intersect And the position of the pixel at this intersection is recognized as the reference point position. (B) The branch point where one thinned line branches into two is recognized, and A reference point recognizing step of recognizing another nearby branch point and recognizing a position of a point intermediate between the two branch points as a reference point position. 2. The method according to claim 1, wherein the regular pattern is expressed as a set of a plurality of pixels arranged in a row and a row in the thinning processing step, and four other pixels are arranged in eight directions around a predetermined pixel. In some cases, it is recognized that the predetermined pixel is located at the intersection point, and the other pixels are located in eight directions around the predetermined pixel.
A method for recognizing that a predetermined pixel is located at a branch point position when there is one. 3. The method according to claim 1, further comprising preparing a matrix for recording the coordinate values of the recognized reference points, and using the matrix to store not only the position information of the plurality of reference points but also the phase information between them. A method for recognizing a reference point of a regular pattern, which is recorded simultaneously. 4. The method according to claim 1, wherein the recognized reference points are sequentially tracked along a predetermined direction, and the position information of the tracked reference points is determined based on the tracking direction. A method for recognizing a reference point of a regular pattern, which is recorded together with information. 5. The method according to claim 4, wherein the tracking is continued to a reference point located at the topologically center among a plurality of reference points connected to the tracked reference point. A reference point recognition method for a regular pattern. 6. The method according to claim 5, wherein, for the reference points which are not tracked among the recognized reference points, their position information and phase information are separately recorded as untracked reference points from the tracked reference points. A method for recognizing a reference point of a regular pattern, characterized by recording. 7. A method according to claim 6, wherein when tracking along a predetermined direction becomes impossible, tracking along a reverse direction is performed, and tracking along the reverse direction is performed. When tracking becomes impossible, a reference point recognition of a regular pattern characterized by continuing tracking along a predetermined direction starting from an arbitrary point recorded as an untracked reference point. Method. 8. A method for automatically recognizing each reference point using a computer for a regular pattern having a plurality of reference points, comprising: an inputting step of taking in the regular pattern as a set of pixels; Recognizing a pattern constituting a line from the set and tracking the regular pattern along the recognized pattern; and recognizing an intersection of two patterns constituting the line as a reference point in the tracking process. And extracting and recording the position information of the recognized reference point and the phase information of the reference point determined based on the tracking process that has reached the reference point. Reference point recognition method for regular patterns. 9. The method of claim 8, wherein defining a spot closed area having a dimension larger than a width of a pattern constituting a line, overlapping the spot closed area on a part of a regular pattern,
Extracting a pixel on or near the boundary line of the spot closed area as a boundary pixel; and recognizing an intersection based on a distribution state of the extracted pixel in the spot closed area. A method for recognizing a reference point of a regular pattern, comprising: 10. The regular pattern reference point according to claim 9, wherein the tracking direction of the regular pattern is determined based on the distribution state of the extracted pixels in the spot closed area. Recognition method.