JPS62154076A - Picture processing method - Google Patents

Abstract

PURPOSE:To shorten an arithmetic time of a quantity of the feature of a pattern by reducing automatically a preset window according to the size of a pattern in the window or the position to obtain a quadrangle enveloping the pattern. CONSTITUTION:The preset window is made into a quadrangle and set to a little larger in size in considering of position shift of the pattern. Before the quantity of feature is operated, four sides of the set window are moved in parallel until the respective sides contact the pattern, and a rectangular area enveloping the pattern is obtained to make a new window. With respect to the pattern in the newly obtained window, the operation of the quantity of feature is performed. The window is made the rectangular, the four sides of the window are respectively moved in parallel until the respective sides contact the pattern to obtain the rectangular area enveloping the pattern, thereby, among the areas indicated by the preset window, an area which is not required for calculating the quantity of feature of the pattern is deleted and only a required area is left. Since the area for calculating the quantity of feature of the pattern can be deleted, a processing time for the calculation processing of the quantity of feature can be shortened.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to an apparatus for reading characters printed on a material or for inspecting the appearance of an object, and particularly for calculating feature values of patterns within a preset window. The present invention relates to an automatic window reduction method suitable for an apparatus that achieves a purpose by

[Conventional technology]

Conventionally, a simplified image processing gt system has been proposed in Japanese Patent Application Laid-Open No. 1986-9 (
As described in No. 1177, the processing area is not made up of the entire screen, but is made up of a collection of small areas called windows, and processing such as calculation of feature quantities, which requires processing time, is performed only within this window. The aim was to shorten processing time. However, with the above method, when the location of the pattern is clearly known in advance, the window can be set appropriately for the pattern and the processing time can be shortened, but when the positional shift of the pattern is large, It is necessary to increase the size of the window in advance,
As a result, there was a drawback that the processing time could not be reduced significantly.

[Problem that the invention seeks to solve]

The above-mentioned conventional technology does not take into account measures against pattern positional deviation, and as a result, the window must necessarily be made larger, resulting in the problem that the processing time cannot be reduced much.

The purpose of Akira Hongen is to perform high-speed calculations on the screen (even if the pattern is significantly misaligned, the window is corrected according to the pattern position, and the features are not affected by the pattern misalignment). The objective is to provide a method to automatically create windows suitable for patterns.

[Means for solving problems]

The above objective is achieved by performing the following steps.

1) The window to be set in advance should be rectangular, and
Misalignment of pattern: s! Set it to a larger value.

2) Before calculating the feature amount, translate the four sides of the preset window until each side touches the pattern, find a rectangular area that envelops the pattern, and use this as a new window.

3) Perform feature calculations on the pattern within the newly obtained window.

[For production]

Using the window as a footprint, finding a rectangular area that envelops the pattern by moving the four sides of the window in parallel until each side touches the pattern is a method to calculate the feature amount of the pattern out of the area shown in the preset window. It has the effect of deleting unnecessary areas and leaving only the necessary areas. As a result, the feature ffi calculation area of the pattern can be reduced, so that the processing time for the feature value calculation process can be shortened.

〔Example〕

An embodiment of the present invention will be described below with reference to the drawings. Second
In the figure, 1 is an object to be inspected, 2 is a conveyor that conveys the object to be inspected, 3 is a detector that detects the arrival of the object to be inspected, 4 is a TV camera that images the object to be inspected 1, and 5 is a TV camera 4. 6 is an amplifier circuit that amplifies the signal from the amplifier circuit 5 to a predetermined level; 6 is a binarization circuit that binarizes the signal from the amplifier circuit 5 at a predetermined threshold level and quantized or pixelized it according to the predetermined signal; 7 is a binarization circuit that amplifies the signal from the amplifier circuit 5 to a predetermined level; A synchronization signal detection circuit 8 extracts a vertical synchronization signal and a horizontal synchronization signal from the signal from the camera 4, quantizes the signal in the binarization circuit 6, and creates a signal used when creating an address to store the quantized signal; 9 is an image memory that stores the quantized signal from the binarization circuit 6; 10 is a central processing unit (CPU); ), 11 is a memory (ROM) for storing processing procedures, etc.; 11 is a temporary storage memory (RAM);
13 is an input circuit for inputting the signal from the detector 3;
14 is an output circuit for outputting processing results, and 15 is a system bus.

In the operation of this apparatus, first, the conveyor 2 carries out the object 1 to be inspected and a sound is heard. When the inspected object 1 enters the field of view of the TV camera 4, the detector 3 detects the inspected object 1 and outputs a signal to the input circuit 13. Here, the detection accuracy of the detector 3 may be rough (it is sufficient that it can detect that the object to be inspected l is within the field of view of the TV camera 4.C)
When the P U 10 receives a signal from the detector 3 via the input circuit 13 , the TV camera 4 captures an image of the object 1 to be inspected. Here, the object to be inspected 1 and the conveyor 2 which is the background
etc. shall have sufficient contrast. The pattern of the inspected object 1 captured by the TV camera 4 is converted into an electrical signal, amplified to a predetermined level by an amplifier circuit 5, and binarized at a predetermined threshold level by a binarization circuit 6. The signal from the synchronization signal detection circuit 7 is quantized or pixelized.

The signal quantized by the binarization circuit 6 is stored at an address in the image memory 9 generated by the address generation circuit 8 based on the signal from the synchronization signal detection circuit 7. C.P.
After storing the pattern of the inspected object 1 in the image memory 9, U to refers to the window information set in advance in the ROM 11 and reduces the window according to the state of the pattern in the image memory 9. After that, feature values are calculated within the reduced window. CPU 1
0 compares the obtained feature amount with a standard value set in the ROM 11 to determine whether it is good or bad, and outputs the result to the outside via the output device 14.

The window reduction procedure will be explained using FIG. 1st
In the figure, image data is stored in an image memory 9 corresponding to the scanning of the TV camera 4.

FIG. 1 shows a state in which the pattern of the object to be inspected 1 is stored at a black level, as indicated by diagonal lines inside the mouth. 1st
In the figure, 16 indicated by a chain line is a setting window that is set in advance, and 17 indicated by a chain double dot line is a modified window after window reduction processing is performed on the setting window 16.

Here, the origin is the starting point of the image memory 9, that is, the upper left end of the image memory 9 in the figure, and the XItbY axes are determined as shown in the figure. Points P, Q, and R that make up the settings window 16
, S coordinates as (xl, Yl) + (X2+)
'I L (X1+)'2L (X2+72), and the points P', Q', which constitute the correction window 17,
The coordinates of R' and S' are (X3 + Y3 L
(X4.y3)y (X3.y4L(X4+y
, ). Below, the procedure for converting the setting window 16 into the modification window 17 will be shown in steps.

Step 1: Input the pattern of the inspected object 1 into the image memory 9.

Step 2: Make the coordinate values of the correction window 17 coincide with the coordinate values of the setting window 16 as initial values.

That is, x3=xII X4=X21 Y3=Yxr
Let X4=X2.

Step 3: For pixels with Y coordinate value = y3 and X coordinate value in the range of x3 to x4, if all of those pixels are at white level, it is assumed that the upper side P'Q' is not in contact with the pattern of the object to be inspected l. , move on to step 4. Otherwise, i.e. Y coordinate value = y3 and X coordinate value is from x3 to x4
If there is at least one black pixel in the pixel range, it is assumed that the upper side P'Q' is in contact with the pattern of the object 1 to be inspected, and the process moves to step 5.

Step 4: Set Y3=y3+1 and move on to step 3.

Step 5: If all of the pixels in the range of Y coordinate value = y4 and X coordinate value from x3 to x4 are at white level, it is assumed that the lower side R'S' is not in contact with the pattern of the object to be inspected 1. , move on to step 6. Otherwise,
In other words, Y coordinate value = y4 and X coordinate value is from x3 to x4
If there is even one black pixel in the range of pixels, it is assumed that the lower side R'S' is in contact with the pattern of the object to be inspected l, and the process moves to step 7.

Step 6: Set X4=X4 to 1 and proceed to Step 5.

Step 7: If the X coordinate value = x3 and the pixels in the range of X coordinate values from y3 to y4 are all at white level, it is assumed that the left side P/R' is not in contact with the pattern of the object to be inspected 1. , move on to step 8. Otherwise, if the X coordinate value = x3 and there is at least one black pixel among the pixels with the X coordinate value from y3 to y4, then the left side P'R' is in contact with the pattern of the object to be inspected 1. As a result, move on to Step 9.

Step 3: Set x3=x3+1 and move to step 7.

Step 9: If the X coordinate value = X4 and the pixels in the range of X coordinate values from y3 to y4 are all at white level, it is assumed that the right side Q'S' is not in contact with the pattern of the object to be inspected 1. , move to step lO. Otherwise, i.e. the X coordinate value = X 4 and the X coordinate value is from y3 to y4
If there is even one black pixel in the range of pixels, it is assumed that the right side Q'S' is in contact with the pattern of the object to be inspected 1, and the process moves to step 11.

Step 10: Set X4=X4 1 and proceed to step 9.

Step ll: Coordinate value X3°X4 obtained by the above process
+ )'31 The following corrections are made to Y4.

X3:X3 1, x4=x4+1y3=Y3
1 p y4=y<+1 This is a consideration for the process of determining the feature amount by applying neighborhood calculation processing such as 2×2 or 3×3 to the entire window.

That is, as a result of the above correction, the pattern of the object to be inspected 1 exists in the correction window 17 in a state where it is completely surrounded by pixels of the white level. This situation is the first
As shown in the figure.

The subsequent processing, that is, the computation of the feature amount of the pattern, is performed on the corrected window air obtained by the above procedure. Since the area of the modification window 17 can be considerably smaller than that of the setting window 16, feature values can be calculated at high speed.

〔Effect of the invention〕

According to the present invention, the setting window 16 can be reduced to the correction window 17, and thereby feature calculations, which are characterized by long processing times when executed by software, can be performed in a much smaller correction window than the setting window 16. 17, it is possible to automatically set a window of an appropriate size even when there is a positional shift of the inspected object 1, which has the effect of reducing the processing time. The degree of reduction in processing time is due to the fact that window reduction processing takes much less time than feature calculations.
The area of the correction window is l/ compared to the setting window 16.
When it becomes N, it is thought that it becomes about 1/rV+.

[Brief explanation of drawings]

FIG. 1 shows an explanation of the window reduction method, and FIG. 2 shows a block diagram of the appearance inspection device. 1...Object to be inspected, 2...Conveyor, 3
...Detector, 4...TV camera, 5.
...Amplification circuit, 6.. Beating binarization circuit, 7..
... Synchronization signal detection circuit, 8 ... Address generation circuit, 9 ... Image memory, IO ... CPU
, 11...ROM, 12...RAM, 1
3...4 input circuit, 14...output circuit, ts...
Song/system bus, 16...Ol cause

Claims (1)

[Claims] 1. Recognize or reject the object by capturing an image of the object, generating a window for the image information, determining the features within the window, and comparing the features with standard values. In devices that perform judgment, the time required to calculate pattern feature quantities is shortened by automatically reducing a preset window to a quadrilateral that envelops the pattern according to the size or position of the pattern inside the window. An image processing method characterized by: