JPH024032B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Application The present invention relates to a pattern recognition method for recognizing the position of a two-dimensional pattern without contact.

Conventional Structures and Their Problems In recent years, the demand for pattern recognition in industry to detect the position of objects without contact has been increasing, and some patterns are being used in semiconductor assembly processes and the like. Hereinafter, position recognition by conventional pattern matching as described above will be explained with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of a conventional position recognition method using pattern matching. In FIG. 1, 1 is an imaging device, 2 is a synchronization signal generation circuit for driving the imaging device 1, 3 is a coordinate generation circuit for obtaining the position of the scanning beam, and 4 is the imaging device 1
5 is a temporary storage provision circuit consisting of a shift register; 6 is a pre-processing circuit consisting of a circuit for binarizing the video signal from 5; 6 is a parallel cutout of patterns used for matching degree detection from among the patterns stored in the temporary storage circuit 5; 7 is a standard pattern storage circuit that stores pre-registered patterns in a form that allows for the degree of matching to be detected; 8 is a standard pattern storage circuit 7 that stores the patterns cut out by the 2D pattern extraction circuit 6; 9, a coincidence degree storage circuit that stores the degree of coincidence; 10, a comparison circuit that compares the degree of coincidence; 11;
12 is a gate circuit that transmits the degree of coincidence output from the degree of coincidence detection circuit 8 to the degree of coincidence memory circuit 9; 12 is a gate circuit that conveys the coordinates output from the coordinate generation circuit 3 to the coordinate memory circuit; 13 is a gate circuit that transmits the coordinates of the scanning beam; This is a coordinate storage circuit that stores coordinates.

The operation of the position recognition method using pattern matching configured as described above will be explained below. First, an object such as a semiconductor chip is converted into a video signal by the imaging device 1. The scanning beam at this time is synchronized with the synchronization signal generated by the synchronization signal generation circuit 2, and its coordinates are always obtained from the coordinate generation circuit 3. Next, the video signal is converted by a preprocessing circuit 4 into a binary signal that is easy to handle by a logic circuit, and is input to a temporary storage circuit 5 consisting of a shift register. From this, the next two-dimensional pattern cutting circuit 6
Depending on the horizontal and vertical size of the video information (e.g.
(16 x 16 pixels) is sequentially read out as rectangular pattern information. In the coincidence detection circuit 8, 2
The pattern cut out by the dimensional pattern cutout circuit 6 and the standard pattern stored in the coordinate pattern storage circuit 7 are compared to detect the degree of coincidence.
The detected degree of coincidence is compared with the degree of coincidence stored in the past in the degree of coincidence storage circuit 9 in the comparison circuit 10, and when the detected degree of coincidence is greater, the gate circuit 11 is opened and the contents of the degree of coincidence memory circuit 9 are compared. Update. The output of the comparison circuit 10 is further sent to the gate circuit 12, and the coordinate values of the scanning beam at that time are guided to the coordinate storage circuit 13, thereby updating the previously stored coordinate values.

Repeating this operation, at the end of the frame when scanning ends, the coordinate position X, Y in the image where the pattern that most closely matches the pre-stored standard pattern exists is stored and retained along with the matching degree at that time. .

However, with the above configuration, only the coordinate values that best match the pattern are detected, and if a pattern that should be correctly recognized is defective due to scratches or dirt, other unrelated patterns may match better. In this case, a pattern that should be correctly recognized is not detected.

Purpose of the Invention In view of the above-mentioned drawbacks, the present invention provides pattern recognition that recognizes the pattern that should be correctly recognized and can detect the position with high reliability even if there is a pattern that matches the standard pattern better than the pattern that should be correctly recognized. The present invention provides a method.

Structure of the Invention The present invention stores two or more types of standard patterns and the relative positional relationship between the standard patterns in advance, and the degree of matching obtained by pattern matching for each standard pattern is maximized within one frame image. It consists of the steps of detecting a plurality of coordinate values, verifying the relative positional relationship of possible combinations between the standard patterns, and finding the combination point closest to the relative positional relationship between the standard patterns.

DESCRIPTION OF EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings.

FIG. 2 is a block diagram showing the configuration of a pattern recognition method in one embodiment of the present invention. In FIG. 2, 20 is the pattern matching circuit shown in FIG. 1; 21 is a temporary storage circuit that temporarily stores the degree of matching output from the pattern matching 20;
22 is an 8-neighborhood extraction circuit that extracts the degree of coincidence for 8 pixels surrounding a certain point on the screen, 23a to 23h are circuits that compare the degree of coincidence with the 8 neighboring pixels, and 24 is a circuit that determines whether there is a maximum point based on the comparison results. 25 is a local maximum value ranking determination circuit that determines the ranking of local maximum values; 26 is a first local maximum value storage circuit; 27 is a second local maximum value storage circuit; and 28 is from the pattern matching circuit 20. a first local maximum point memory circuit that stores the corresponding local maximum point; 29 is a second local maximum point storage circuit;
30 is a switching circuit; 31a and 31b are both maximum point memory circuits that store the first and second maximum points; 32 is a circuit that determines two recognition points (one set) from a total of four maximum points. This is a recognition point determination circuit.

The operation of the pattern recognition method configured as described above will be explained below.

Assume that when the object is scanned for one frame, the degree of matching output from the pattern matching circuit 20 is expressed as shown in FIG. In Figure 3, a
The point has the highest degree of coincidence, the point b is a point shifted by one pixel from the point a, and the point c is a point that should be correctly recognized. This state will be further explained with reference to FIG. 1 in FIG. 4 is a standard pattern, and the screen is scanned using this pattern to detect the degree of coincidence at each point. 1 in Figure 4 is the standard pattern, and 4 is the third pattern.
In the example of the pattern at point c in the figure, the pattern should be correctly recognized, but because the pattern at part A is missing, the degree of matching is 80/100, which is from point a to c.
The smallest of the points. 2 and 3 in FIG. 4 are examples of the patterns of points a and b in FIG. 3, respectively; 2 has a matching degree of 91/100, and 3 has a matching degree of 88/100.

In this example, we want to detect point c, but if we select two candidate points in descending order of degree of coincidence, only points a and b will be detected, and point c will not be detected. In order to detect point c as the second candidate point, it is necessary to detect the coordinates at which the degree of matching is maximum.

In FIG. 2, a pattern matching circuit 20
The degree of coincidence outputted from the above is sent to a temporary storage circuit 21 consisting of a shift register.

If it is assumed that the pixel B in Figure 4 represents the coordinate values of the standard pattern, then the degree of coincidence of each of the positions indicated by A to I in Figure 5 when B is scanned is the 8 neighborhood shown in Figure 2. It is cut out by the cutting circuit 22. In the comparison circuits 23a to 23h and the local maximum point determination circuit 24 in FIG.
Compare the degree of coincidence of the extracted point E and the degree of coincidence of the surrounding points (8 neighboring points), and when all of the following conditions are satisfied, the degree of coincidence of point E is determined to be the maximum point. .

E _er > A _er , E _er > B _er E _er > C _er , E _er > D _er E _er F _er , E _er G _er E _er H _er , E _er I _er , where A _{er is} the degree of agreement at point A and below B _er
~ _Ier represents the degree of agreement at each point.

When comparing the degree of agreement at points F to I with the degree of agreement at point E, even if the degree of agreement is equal, it is determined to be the maximum point, so if one point consecutive to point E has the same degree of agreement. To effectively determine a maximum point without missing the maximum point or detecting it redundantly.

If it is determined that it is a local maximum point, the local maximum value and the first
The local maximum value stored in the past in the local maximum value storage circuit 26 is compared with the local maximum value ranking determination circuit 25, and if the local maximum value determined to be the local maximum point is larger, it is ranked as the first local maximum value storage circuit 26 and the first place. The contents of the local maximum point storage circuit 28 are updated. Also, the second local maximum value storage circuit 27
The contents stored in the first local maximum value storage circuit 26 are transferred to the second local maximum point storage circuit 29, and the contents stored in the first local maximum point storage circuit 28 are transferred to the second local maximum point storage circuit 29.

When the local maximum value determined to be local maximum is smaller than the content of the first local maximum value storage circuit 26 and larger than the content of the second local maximum value storage circuit 27, the local maximum value is stored in the second local maximum value storage circuit 27 and the second local maximum point memory. Update the circuit contents.

The contents of the first maximum point storage circuit 28 and the second maximum point storage circuit 29 are stored in the maximum point storage circuit 31a via the switching circuit 30. Further, the next time a local maximum point is detected in a different standard pattern, it is stored in the local maximum point storage circuit 31b. The recognition point determination circuit 32 determines two recognition points (one set) from a total of four maximum points. An example is shown in FIG. 103a is the first maximum point for the standard pattern 101, 103
b is the second maximum point for the standard pattern 101,
104a indicates the first maximum point for the standard pattern 102, and 104b indicates the second maximum point for the standard pattern 102. In this case, first, the difference between the distances of A to D→vectors and the distance r of E→vectors is determined. Next, A → ~ D → the angle between the vector and the x-axis and E →
The difference between the angle θ between the vector and the x-axis is determined, and the maximum points 103a and 104b having the A→vector where the difference in distance and the difference in angle are minimum are determined as a set of recognition points. Also, 103a coordinate value and 101
By calculating the difference between the coordinate value of 104b and the coordinate value of 102, the amount of deviation between the two points of the standard pattern is determined.

As described above, according to this embodiment, two coordinates at which the degree of matching is maximum are detected for one type of standard pattern, and the distance and angle between the maximum points are verified. Even if a pattern that matches the pattern exists, the pattern that should be correctly recognized can be effectively recognized and highly reliable position detection can be performed.

Effects of the Invention As described above, the present invention detects a plurality of coordinate values at which the matching degree obtained by pattern matching is maximum for each type of standard pattern, and By determining one set of recognition points in the test, even if the pattern to be recognized has dirt or scratches and the degree of matching is lower than other similar patterns, the position can be effectively and reliably determined using the minimum number of standard patterns. It can be recognized, and its practical effects are great.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of position recognition using conventional pattern matching, FIG. 2 is a block diagram of an embodiment of the present invention, and FIG. 3 is a diagram showing an example of changes in the degree of matching with respect to screen scanning. Figure 4 1
~4 is a diagram showing the state of the pattern to more specifically explain the change in the degree of matching in Figure 3;
The figure shows eight neighboring points, and FIG. 6 is an explanatory diagram of the test between local maximum points. 21...Temporary storage circuit, 22...8 neighborhood extraction circuit, 22a to 23h...Comparison circuit, 24...Peak determination circuit, 25...Peak value ranking determination circuit,
26...First peak value storage circuit, 27...Second
1st peak point storage circuit, 29 2nd peak point storage circuit, 30
...Switching circuit, 31a... Maximum point memory circuit, 31
b... Maximum point memory circuit, 32... Recognition point determination circuit.

Claims (1)

[Claims] 1. In a method of capturing an image of an object with an imaging device and recognizing the position of the object by pattern matching, a plurality of portions of the object on an imaging screen obtained from the imaging device and storing substantially the same pattern as a standard pattern and storing the relative positional relationship between the standard patterns, and comparing the standard pattern with a partial pattern of the object on the imaging screen to detect the degree of coincidence. a step of detecting, as a local maximum point, the coordinate value of a partial pattern whose degree of agreement is higher than the degree of agreement in its vicinity among the many detected degrees of agreement; a step of detecting and storing a predetermined number of points in descending order of maximum value; and the plurality of stored maximum points obtained for one standard pattern and the plurality of stored maximum points obtained for different standard patterns. calculate the relative positional relationship between two points that can be combined with the standard pattern, and determine the combination point of the maximum points having the relative positional relationship between the two points closest to the relative positional relationship between the standard patterns as the recognition point. A pattern recognition method consisting of a process. 2. The relative positional relationship between two points in the means for determining them as recognition points is defined by the distance between the two points, the angle between the straight line connecting the two points and the reference line, or both the above-mentioned distance and angle. A pattern recognition method according to claim 1.