JP3119581B2 - Defect inspection device and defect inspection method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to
The present invention relates to a defect inspection apparatus and a defect inspection method that detect defects such as dirt, scratches, black spots, pinholes, and fish eyes to determine good or defective.

[0002]

2. Description of the Related Art In a manufacturing / inspection process of a film-like object or a sheet-like object, various types of defect inspection apparatuses are known as a defect inspection apparatus for determining whether or not a surface of the object has a defect. . Generally, in order to inspect for defects such as film-like objects and sheet-like objects, a line C is used.
A line image sensor such as a CD camera is used, and the present inventors have already used a run length encoding process using a binary signal of an output signal of the line image sensor and a run length code of a plurality of rows. A defect inspection apparatus capable of performing a high-speed defect measurement by executing the connectivity processing in parallel has been developed and a patent application has been filed (Japanese Patent Laid-Open No.
-67159).

A film-like object, a sheet-like object, or the like as an inspection target is usually run at a predetermined speed, and the running inspection target is scanned by a line image sensor in a direction perpendicular to the running direction. It is common to perform an inspection. The traveling inspection object may meander during traveling. In order to accurately inspect defects even if there is meandering,
It is necessary to take measures such as automatically following the meandering of the inspection object. In a defect inspection apparatus using a conventional line CCD camera, there is a defect inspection apparatus that can automatically set an inspection range by automatically following a meandering of an inspection target.
The output signal of the line CCD camera is differentiated to detect an edge to be inspected, and the inside of the edge coordinates is set to be inspected by a predetermined amount. The present inventors have also disclosed a defect inspection apparatus using a line CCD camera in Japanese Patent Application Laid-Open No. 05-332950 or the like.
After the output signal of the line CCD camera is corrected, binarized, and run-length coded, there is a function of individually recognizing defects by connectivity processing.

[0004]

In the above-mentioned prior art, those which can be inspected by automatically following the meandering of the inspection object are subjected to differential processing. It is impossible to measure the shape. On the other hand, in the above-mentioned conventional technology of the present inventors, it is possible to perform individual recognition of a defect and measure the size and shape of the defect, but it is not possible to automatically follow the meandering of the inspection target and inspect it. is there.

Accordingly, the present invention provides a defect inspection apparatus capable of performing an inspection by automatically following the meandering of an inspection object while having the defect individual recognition function in the prior art developed earlier by the present inventors. An object of the present invention is to provide a defect inspection method.

[0006]

In order to achieve the above object, according to the present invention, an edge of an object to be inspected is detected from an output signal from a line image sensor for obtaining image data at an inspection position of the object to be inspected. A threshold value for binarization is set to a predetermined value so that a defect is not detected regardless of the value of image data outside of the vicinity of the edge coordinates, and the inside of the edge coordinates is specified in advance. The set value or the function value of the image data is set as a threshold value, stored in first and second threshold value storage means, and an output signal from the line image sensor is stored in the first and second threshold value storage means. Binarization is performed using the threshold value stored in the threshold value storage means.

That is, according to the present invention, a line image sensor that scans in a direction substantially perpendicular to the traveling direction of a traveling inspection object to obtain image data at an inspection position of the inspection object, and an output image of the line image sensor Edge detecting means for detecting an edge of the inspection object from data, binarizing means for binarizing output image data of the line image sensor, and no need for the binarizing means to perform binarization The first and second threshold value storing means for storing a threshold value and the vicinity of the coordinates of the edge obtained by the edge detecting means are bordered so that no defect is detected regardless of the value of the image data. The threshold value is set to a predetermined value, and a predetermined value or a function value of the image data is set as the threshold value for the inside of the edge coordinates. Means, a run-length encoding means for obtaining a change point address of the binary image data obtained by the binarization means, and a plurality of lines in the line image sensor output from the run-length encoding means. First and second run-length code storage means for respectively storing the run-length codes, and a connectivity process for the run-length codes stored in the first and second run-length code storage means. A connectivity processing means for detecting a defect of an object, and a run-length code output from the run-length coding means to one of the first and second run-length code storage means after the run-length code has been input. Reading the length code, causing the connectivity processing means to execute connectivity processing, and storing the run length in the other of the first and second run length code storage means. After the input of the run-length code output from the encoding unit is completed, the run-length code is read out, and the connectivity processing unit executes the connectivity process, whereby the first run-length code storage unit stores the run-length code. An input and a connectivity process using a run-length code from the second run-length code storage are substantially parallel-processed, and an input to the second run-length code storage and the first run-length code are processed. A first control unit for controlling the connectivity processing using the run-length code from the storage unit to be performed in a substantially parallel manner; and a threshold value of one of the first and second threshold value storage units. 2 above
When the binarization unit is executing binarization, a threshold value is determined after the connectivity processing by the connectivity processing unit is completed, and the determined threshold value is determined by the first and second threshold values. By writing to the other one of the threshold value storage means, the reading of the threshold value from the first threshold value storage means and the writing of the threshold value to the second threshold value storage means are substantially performed in parallel. And controlling the reading of the threshold value from the second threshold value storage means and the writing of the threshold value to the first threshold value storage means so as to be performed substantially in parallel.
A defect inspection apparatus having control means is provided.

Further, according to the present invention, the object to be inspected is scanned from a line image sensor that obtains image data at an inspection position of the object by scanning in a direction substantially perpendicular to the traveling direction of the object to be inspected. And a step of detecting a defect outside the vicinity of the coordinates of the edge obtained by the step of detecting the edge, regardless of the value of the image data.
A threshold for value conversion is set to a predetermined value, and a predetermined value or a function value of the image data is set as the threshold for the inside of the edge coordinates. Storing in threshold storage means of
Binarizing the output signal from the line image sensor using one of the threshold values stored in the first and second threshold value storage means; Encoding and storing it in one of the first and second run-length code storage means; and storing a plurality of rows of run-length codes of the image data of the line image sensor stored in the one run-length code storage means. Restarting the binarizing step during execution of the connectivity processing by using
Encoding the binarized image data into run-length codes and storing the run-length encoded image data in the other of the first and second run-length code storage means; and using the result of the connectivity processing to set a threshold value for binarization. When the step of determining and performing the binarization using the threshold value of the one threshold value storage means is being executed, the determined threshold value is stored in the first and second threshold value storage means. Storing the defect in the other of the means.

The method of edge detection according to the present invention is as follows. An output signal of a line CCD camera as a line image sensor converted into a digital signal by an A / D converter is defined as multi-valued image data. The edge detecting means for detecting the edge of the inspection object from the multi-valued image data reads out the inspection object from the multi-valued image (Di i = 1 to the number of elements of the line CCD camera) obtained by reading the inspection object. Detect an edge part. Since there is a possibility that the multivalued image may include a portion which is erroneously recognized as an edge portion due to a defect, the leftmost and rightmost edges in the multivalued image data are detected. The left edge detection is performed in the effective range in the multi-valued image data from the left end to the right end. The edge detection at the right end is performed in the effective range in the multi-valued image data from the right end to the left end. The edge position is a coordinate that first satisfies Equation 1 shown in Equation 3 or Equation 2 shown in Equation 4.

[0010]

Equation 1 d ≧ | D _i -D _i + _k | When processing from the left end d: edge determination criterion i: s to (ek) s: start point of effective range e: end point of effective range k: Reference element distance for edge detection

[0011]

Equation 2 d ≧ | D _i −D _i − _k | When processing from the right end i: eｅ (s + k)

The threshold value setting means clears the outside of the coordinates obtained by the edge detecting means as a non-inspection area. Here, to clear the threshold data means to set a threshold that does not detect a defect regardless of the value of the multi-valued image data. For example, when the A / D converter is configured to convert the data into 8-bit data, the value is set to 255 in the case of detecting a bright defect and to 0 in the case of detecting a dark defect. When the threshold memory has this value, any multi-valued image input from the line CCD camera will not be detected as a defect.

In a preferred embodiment, the range for clearing the threshold value is inside the edge coordinates by the correction value. The magnitude of the correction value is set in consideration of the meandering amount of the inspection object within the threshold value setting time interval and the variation of the line CCD camera output at the edge portion. The means for setting threshold data sets the inside of the coordinates obtained by the edge detecting means to an appropriate threshold as the inspection area. The value specified in advance (0 if the A / D converter is 8 bits)
To 255) or a ratio specified in advance (100% or more for a bright defect, 100% or less for a dark defect) with respect to the multi-valued image data.

The run-length encoding means obtains a change point address of data obtained by binarizing the multi-valued image based on the threshold data. The data format of the run-length code will be described later.

The run-length code storage means comprises a memory for storing two series of run-length codes for a plurality of rows of the line CCD camera output from the run-length coding means. In addition, run-length codes for detecting a bright defect and for detecting a dark defect are stored in the memory for one series. The connectivity processing means measures a defect while confirming the connectivity of two scans of the run length code stored in the run length code storage means. There are two series of run-length code storage means for storing run-length codes by the run-length encoding means for one of the series.
This is for performing parallel processing while alternately switching between the processing by the connectivity processing means and the other series of the run-length code storage means. There are also two sets of threshold value storage means. When one of the sets performs binarization, the other sets and sets the threshold data. Parallel processing.

[0016]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described below with reference to the accompanying drawings. FIG. 1 is a block diagram of a defect inspection apparatus according to an embodiment of the present invention. The apparatus shown in FIG. 1 has a line CCD camera 1 operating as a line image sensor and an image processing apparatus 2 for processing an analog signal output from the line CCD camera. The line CCD camera 1 obtains image data by one-dimensionally scanning the surface of the inspection object (not shown). The number of elements (pixels) is 2048 or 5000, the clock frequency is 20 MHz, and the scanning cycle is A maximum of 0.11 ms can be used.

An analog signal obtained by imaging by the line CCD camera 1 is converted into, for example, 8-bit digital data by the A / D converter 3. The digital data is binarized by the comparator 4 based on a threshold value and a detection logic for each element of the line CCD camera 1. As this threshold, those stored in two threshold memories 11A and 11B as first and second threshold storage means are used. Writing and reading will be described later. Note that there are two types of detection logic. In the case of white detection, an area having an output level higher than the threshold is set to "1", and in the case of black detection, an area having an output level lower than the threshold is set to "1". 1 ". Therefore, the area corresponding to the defective part becomes “1” and the area corresponding to the normal part becomes “0” by binarization by the comparator 4. The binarized data by the comparator 4 changes from “0” to “0”.
Run-length encoding is performed by a run-length encoding circuit 5 for obtaining an address of a transition point from 1 "or" 1 "to" 0 ".

The reference value memory 12 stores one line of data from the A / D converter when an object having a uniform transmittance or reflectance is captured by the line CCD camera 1. This data is used to determine a threshold as reference data. That is, the product of the reference data for each pixel and the threshold value for each pixel obtained during the defect inspection is calculated, and this product is stored as a new threshold value. This calculation is performed by a threshold value determining program stored in the ROM 14 based on the reference data of each element. When a new threshold value is determined, the new threshold value is stored in the threshold value memory 11. Therefore, the comparator 4 can binarize the output signal of each element by correcting sensitivity variations among the elements of the line CCD camera 1, illumination unevenness and lens distortion. As described later, since there is a possibility that a defective portion is included in the image data, the influence of the defect is reduced by an averaging process or the like using the image data of a plurality of rows.

The run-length encoding obtained by the run-length encoding circuit 5 is a two-system run-length buffer 9A as first and second run-length code storage means.
9B is temporarily stored. Each run length buffer 9
A and 9B having a capacity of, for example, 64 kB can be used. The run-length codes temporarily stored in the run-length buffers 9A and 9B are alternately subjected to connectivity processing by a CPU (central processing unit) 15. Note that the connectivity processing refers to processing while comparing data in a plurality of continuous scanning lines between the lines.
By performing this connectivity processing, it is possible to identify defects that should be judged to be good by gathering defects that should be judged to be good alone.

The two threshold memories 11A and 11B are used to correct sensitivity variations between the elements of the line CCD camera 1, illumination unevenness in the main scanning direction, and lens distortion.
A threshold for binarization for each element of the line CCD camera 1 is written as needed. That is, the CPU 15 uses the reference data in the reference value memory 12 and the result of the connectivity process to read the ROM
The threshold value is determined by a threshold value determination program stored in 14 and is recorded in threshold value memories 11A and 11B.

The CPU 15 executes an inspection program for determining whether or not a defect is good, in addition to executing the connectivity processing and the threshold value determination program. The ROM 14 stores control commands of the CPU 15 such as the inspection program and the threshold value determination program. The image processing apparatus 2 also has an input / output interface for transferring data such as inspection conditions and inspection results between a RAM 13 for temporarily storing numerical values and the like used in a program stored in a ROM 14 and a host computer 17. It has a face (I / O) 16. In the host computer 17, the inspection conditions of the defect inspection apparatus are set and the inspection result is displayed.
RS-232C or GP-IB can be used as the I / O 16.

Assuming that the two threshold memories 11A and 11B are for the first series and the second series, the data of the first series threshold memory 11A is read and the comparator 4
When the binarization process is being performed, the second-series threshold memory 11B is used alternately so that the newly determined threshold is written. As described above, by performing the parallel processing of the threshold reading and the writing during the defect measurement, even if there is a temporal change in the transmittance or the reflectance of the line CCD camera 1, the illumination, and the inspection object, it can be performed with high accuracy. Defect measurement becomes possible.

The run-length codes are stored in the run-length buffers 9A and 9B. The data format of the run-length code is as follows. The data length is 2 bytes / piece.

[0024]

[Table 1] {Ln, Sn1, En1, Sn2, En2, {Snm, Enm}}, Ln + 1}, Lk Ln: nth line, where n = 1 to k, MSB = 1, k = Specified number of lines Snm: Start point coordinates of the m-th defect part of the n-th line Enm: End point coordinates of the m-th defect part of the n-th line

The number k of lines to be stored in the run-length buffers 9A and 9B is specified in advance, and is usually 10 to
A value of about 100 is set in advance. Assuming that the two run-length buffers 9A and 9B are used for the first stream and the second stream, the run-length code obtained by the run-length encoding circuit 5 is input to the run-length buffer 9A for the first stream. The data of the run length buffer 9B for the second stream is read out and used alternately so that the CPU 15 performs the connectivity processing.

FIG. 2 is a diagram showing the timing of the run-length encoding process, the connectivity process, and the threshold setting. The defect inspection apparatus of the present invention can perform the run-length encoding processing, the connectivity processing, and the threshold setting processing at preferable timings when the defect inspection is performed under the following conditions. Note that the threshold setting processing refers to determination of the threshold and writing to the threshold memory 11A or 11B. FIG. 2 shows the timing when a defect inspection is performed in a state where this condition is satisfied.

[0027]

[Table 2] (scanning cycle of line CCD camera × number of lines stored in run-length buffer) ≧ (connectivity processing time + threshold setting time) Memory capacity of run-length buffer ≧ data amount of run-length code

Line C whose output timing is shown in FIG.
The output signal 31 of the CD camera 1 is an analog signal obtained by reading an image to be inspected for one line (row), and is continuously output to the image processing device 2 according to the scanning cycle of the line CCD camera 1. A predetermined clock is supplied from a clock generator (not shown) to the line CCD camera 1, the A / D converter 3, the CPU 15, and the like. First run-length buffer 9A, second run-length buffer 9
The input data 32 and 33 to B are obtained by binarizing the image signal 31 and then performing run-length encoding, thereby forming a series of lines for a predetermined number of lines. In FIG. 2, each line is illustrated by two lines. The connectivity processing 34 includes the first
Processing the run-length code starting from the time T ₄ the input 32 to the run length buffer 9A is completed, it is to measure the defects.

Run-length buffer sequence switching command 3
6 is output at time T ₆ when the connectivity process 34 ends. This switching command 36, if it is output before the input end T ₈ of the run length buffer 9B for the second series, it can be seen that the real-time processing is being performed. The threshold memory write 37 is for the first series,
Connectivity processing 34 is performed by using the binarized image by using the data series for the threshold memory 11B is performed after the time T ₆ to end. Threshold memory write 3
7 is a threshold memory series switching command 39 is output at time T ₇ ended. As described above, writing to the first threshold value memory 11A is performed when the run-length buffer input 3
Because have ended before the time point T ₈ to the end of 3, every specified number of lines k, it can be seen that review of the threshold setting has been performed.

As described above, by continuously setting the threshold value during the inspection and correcting the threshold value, a long and stable defect inspection can be performed. In each of the above processes, real-time processing is realized by alternately performing processes for two systems. Further, the connectivity processing and the writing to the threshold memory are executed for contradictory sequences. The setting of the threshold, that is, the method of setting the threshold for obtaining a binary value in the inspection area will be described later.

When the number of defects is very large and some abnormality occurs, the above condition may not be satisfied. In this case, since the real-time processing cannot be performed, the following processing is performed. If the above condition is not satisfied, the setting of the threshold is interrupted, and the connectivity processing of the next stream is prioritized. Thereafter, when idle time occurs, the threshold setting process is continued. In this case, a threshold value is set for each of a plurality of series processes. However, since the threshold value is periodically corrected, this poses no practical problem. Only the connectivity processing time, (scanning cycle of line CCD camera x number of lines stored in run-length buffer)
If this is the case, it will not be possible to switch the series,
Inspection is continued by re-inputting the specified number of run-length buffers to the buffer that is currently inputting the run-length buffer. If this state is temporary, the process returns to the timing shown in FIG. 2 at the stage when the number of run-length codes has decreased, and real-time processing can be performed. However, when the real-time property is important, the test is transferred to the host computer 17 as occurrence of an error so that the connectivity processing is not performed. Further, the process is continued from the inspection of the next run length buffer.

If the above condition is not satisfied, an error is transferred to the host computer 17 as in the above case, and the connectivity processing is not performed. Further, the process is continued from the inspection of the next run length buffer. For example,
In the case of 100 designated lines, if there is a defect of 163 or more / row, the run-length buffers 9A and 9B (64 kB)
Overflows, and the above condition is not satisfied. However, even in the case of a film having a considerably large number of defects, the average is 0.05 / line, and when this state occurs, there is no problem in practically treating it as an error.

Next, the operation of the CPU 15 will be described with reference to the flowcharts of FIGS. The flowchart shown in FIG. 3 shows a main processing procedure for inspecting a defect. An edge of the inspection object which is a feature of the present invention is detected, and a threshold value for binarization is determined. The method of clearing the non-inspection area is shown in the flowchart of FIG. The flow of FIG. 4 is executed at any time as an interrupt process or a subroutine, such as being executed during idle time when the CPU 15 is executing the process of FIG.

First, the flowchart of FIG. 3 which is the main routine will be described. When the power is turned on, the memories 11A, 11B, 12 and the buffer 9 are stored in step S1.
Initialization for clearing A, 9B, etc. is performed, and it is determined in step S2 whether a start pulse has been detected. This start pulse indicates the head of data of each line (row) from the line CCD camera 1. When the start pulse is detected, the count is started. The count value is represented by C, and C = 1 when the first start pulse is counted. In the next step S4, it is determined whether or not the input of the run length code to the run length buffer (RLB) 9A or 9B has been completed. This judgment is C =
It is performed by checking whether the number is three or not. Upon completion of the input, a step S5 connectivity process is performed. When the connectivity process is completed, a predetermined flag is set, and by looking at this flag, the next process completion determination step S6 is executed.

When the connectivity processing is completed, in step S7, a command for switching the run-length buffers 9A and 9B (FIG.
36) is output and writing to the n-th memory is performed. n is 1 or 2, and one of the first threshold memory 11A and the second threshold memory 11B is designated. In the next step S8, it is determined whether or not the writing to the memory 11A or 11B has been completed. If it is determined that the writing has been completed, a threshold memory series switching command (39 in FIG. 2) is output. If n = 1 in the next step S10, n is set in step S11.
= 2, but if n = 1, then n in step S12
= 1. After the end of step S11 or S12, the process returns to step S23.

Next, a procedure for detecting the edge of the inspection object and setting an appropriate threshold value based on the detected edge will be described with reference to the flowchart of FIG. A / D in step S21
The multi-value image data is read from the converter 3. In the next step S22, it is determined whether or not an edge has been detected, and YES
If so, the edge coordinates are stored in step S23, and the edge coordinates are corrected by the correction value in the next step S24. The magnitude of the correction value is stored in advance in the ROM in consideration of the meandering amount of the inspection object within the threshold value setting time interval and the variation in the output of the line CCD camera at the edge portion, or the correction value is stored in the host before the start of the inspection. It can also be specified from the computer 17. With this correction, the range of threshold clearing is
It can be inside the edge coordinates by the correction value,
It is more practical than the method of clearing all outside the edge coordinates.

Next, the threshold value of the non-inspection area is cleared in step S25, and the threshold value set in the next step S26 is stored in the first and second threshold value memories 11A and 11B. Although the threshold setting or updating of the inspection area will be described later, the threshold obtained by the calculation is stored in the first and second threshold memories 11A and 11B as needed. The processing of FIG. 4 and the calculation for setting or updating the threshold value of the inspection area can be executed, for example, by performing an interrupt processing during an idle time between processing steps.

FIG. 5 is a diagram schematically showing a technique for edge detection and threshold setting according to the present invention, and FIG. 6 is a signal level showing a data change in one scan of multi-valued image data for explaining such a technique. FIG. Hereinafter, edge detection and threshold value setting according to the present invention will be described with reference to these drawings. FIG. 5 shows the contents of the input image of the line CCD camera. FIG. 5A shows the arrangement of the line CCD camera 1 and the inspection object 32. The inspection field of view of the line CCD camera 1 is set to a range wider than (the width of the inspection object + the amount of meandering) in consideration of the meandering of the inspection object 32.

FIG. 5B shows the contents of the input image of the line CCD camera. The peripheral area indicated by hatching is the non-inspection area, and the central area without hatching therebetween is the inspection area. A light defect 35A and a dark defect 35B are shown at the reading position of the line CCD camera. FIG. 5 (c)
Is a line CCD for one line when the reading position of the line CCD camera 1 in FIG.
3 shows a camera output waveform. Non-inspection area 3 4A, 3 4
Since B has no inspection target, the output of the line CCD camera 1 is low. The inspection area 33 is a line CCD camera output corresponding to the reflectance of the inspection target 32.
Since the dark defect 35B and the non-inspection areas 34A and 34B are low values and cannot be distinguished from the output values of the line CCD camera, the inspection area and the non-inspection area are distinguished by the method shown in FIG.

FIG. 5D shows the line CCD shown in FIG.
The output waveform of the line CCD camera for one line when the camera reading position is read by transmitted illumination is shown. Since the non-inspection area has no inspection object, the output of the line CCD camera is high. The inspection area is a line CCD camera output corresponding to the transmittance of the inspection target. Since the bright defect 35 and the non-inspection areas 34A and 34B are both high values and cannot be distinguished from the output values of the line CCD camera, the inspection area and the non-inspection area are distinguished by the method shown in FIG.

FIG. 6 is a diagram for explaining a threshold value setting method. FIG. 6A shows an input image of the line CCD camera read by the reflection method. Line CCD
The camera output values correspond to the non-inspection areas in the periphery (element Nos. 1 to 6 and element Nos. 2043 to 204 in FIG.
8) is about 20, which is the central inspection area (FIG. 6).
The element No. of FIG. 7 to 2042) is about 100.

[0042]

## EQU5 ## FIG. 6 (b) shows the state of FIG. 6 (a) for edge detection.
Shows an image obtained by processing the reference image shown in FIG. _{Di = D i -D i + k} Di: element No. i line CCD camera input image D _i + _k : element No. (I + k) line CCD camera input image i: element No. 1 to (element N of line CCD camera
ok) k: distance between reference elements for edge detection

FIG. 6C shows the contents of the threshold memory for two series. Bright defect detection is designated as 200% and dark defect detection is designated as 50%. The contents of the threshold memory for detecting a bright defect are as follows.
Indicated by. The contents of the threshold memory in the inspection area are about 200 for detecting a bright defect and about 50 for detecting a dark defect. Further, the contents of the threshold value memory in the peripheral non-inspection area are cleared to 255 when detecting a bright defect and to 0 when detecting a dark defect. When (k = 1, d = 10) is set in Expressions 1 and 2, the edge coordinates i are 6 and 2042. The range of the threshold value clear is set to be inside the edge coordinates by the correction value (c = 5) in consideration of the variation of the line CCD camera output at the edge portion.

FIG. 6D shows an input image of the line CCD camera when a defect is detected.
And the dark defect 35B. FIG. 6 (e)
FIG. 6D shows the input image of the line CCD camera shown in FIG.
2C illustrates a binary image binarized using the threshold value of FIG. The binary image for detecting a bright defect is indicated by a dotted line, and the binary image for detecting a dark defect is indicated by a solid line. In the binary image, only the portions of the light defect 35A and the dark defect 35B are “1”, and the others are “0” including the non-inspection area, and are excluded from defect detection. By performing these processes repeatedly during the inspection, it is possible to perform the defect detection by automatically following the meandering of the inspection target.

FIG. 7 is a diagram for explaining an example of the arrangement of the line CCD camera and the inspection object. 5 and 6 illustrate the case where only one line CCD camera and one inspection object are used. However, in practice, one or both may be plural. FIG. 7A shows a line CC.
The figure shows a case where there is one D camera and two inspection objects 32 and 32 '. For example, this corresponds to the case of inspecting a slit sheet. In this case, it is possible to cope by setting the inspection range of each sheet in advance. The inspection range of the sheet is specified including the meandering range. The number of sheets is not particularly limited, but is currently set to a range of up to 32 sheets.

FIG. 7B shows a line CCD camera 1,
This shows a case where three 1 ′, 1 ″ and one inspection object 32. In order to inspect a wide sheet-like object, a plurality of CCD cameras are usually used. The presence / absence of edge detection is designated for each range.The presence / absence of edge detection has four states, and is designated as unnecessary, implemented only at the right end, implemented only at the left end, and implemented at both ends, as shown in FIG.
In the case of, the line CCD camera 1 specifies that only the left end is performed, the line CCD camera 1 ′ is unnecessary, and the line CCD camera 1 ″ specifies that the right end is only performed.

Next, the threshold value determination program recorded in the ROM 14 will be described. Each element has a number i (i = 1, 2, 2, 3) in the scanning order of the elements of the line CCD camera 1.
.., N) are assigned, and the elements are e (i)
Expressed by The reference data of each element obtained by reading the surface of the object having substantially uniform transmittance or reflectance is D (i).
And From the reference data D (i), threshold calculation data R1 (i) and R2 (i) are created.

In this program, the threshold value of the element e (i) of the line CCD camera 1 is determined as follows. First, when determining the threshold value of the element e (i),
What element e is counted from element e (i) in ascending or descending order
(J) It is determined whether to refer to the threshold value calculation data R1 (j) of (j か i). This is represented by k. And
, R (k) for each of the elements e (1),..., E (k) are determined as shown in Expression 6.

[0049]

R1 (i) = D (i)

Next, the threshold value calculation data R1 of the k-th element e (1) counted from the element e (k + 1) in descending order.
Reference is made to (1). Reference data D of element e (k + 1)
The difference between (k + 1) and the threshold value calculation data R1 (1) is compared with the constraint amount C or −C, and the threshold value calculation data R1 (k + 1) of the element e (k + 1) is determined according to the magnitude relation.
Is obtained from the expressions shown in Expressions 7, 8, and 9. Note that the constraint amount C is determined by the user as a positive constant that limits the range in which the values of the threshold value calculation data R1 (i) and R2 (i) can take.

[0051]

R1 (i) = R1 (ik) -C (R1 (i
−k) −D (i)> C)

[0052]

R1 (i) = R1 (ik) + C (R1 (i
−k) −D (i) <C)

[0053]

R1 (i) = D (i) (−C ≦ R1 (i−
k) -D (i) ≦ C)

For the element e (i) of i = k + 2,..., N, the threshold value calculation data R1 (i) is sequentially determined by Expressions 5, 6, and 7. Next, threshold value calculation data R2 (i) is determined. First, the element e (N−k
+1),..., E (N), the threshold value calculation data R2 (N−k + 1),.

[0055]

R2 (i) = D (i)

Next, the threshold value calculation data R2 of the N-th element e (N) counted from the element e (N−k) in ascending order.
(N) is referred to. Reference data D of element e (N−k)
The difference between (N−k) and the threshold value calculation data R2 (N) is compared with the constraint amount C or −C, and the threshold value calculation data R2 of the element e (N−k) is determined according to the magnitude relationship. (N-k)
Is obtained from the equations shown in Expressions 11, 12, and 13.

[0057]

R2 (i) = R2 (i + k) -C (R2
(When i + k) -D (i)> C

[0058]

R2 (i) = R2 (i + k) + C (R2
(When i + k) -D (i) <-C

[0059]

R2 (i) = D (i) (−C ≦ R2 (i +
k) -D (i) ≦ C)

For the element e (i) of i = N−k−1,..., 1, the threshold value calculation data R2 (i) is successively determined by Expressions 9, 10, and 11. Threshold calculation data R1 (i) and R2 thus obtained
Using (i), the threshold value S (i) of the element e (i) is determined as follows. Among the elements e (i), i = k +
For the elements e (i) of 1,..., N−k, the threshold S (i) is determined by multiplying the average of R1 (i) and R2 (i) by T as in the following equation. .

[0061]

S (i) = {R1 (i) + R2 (i)} · T / 2 T is a constant indicating the transmittance or the reflectance of the surface of the object from which the reference data D (i) has been sampled. For the element e (i) of i = N−k + 1,..., N, the threshold value S (i) of the element e (i) is determined by the following equation.

[0062]

S (i) = R1 (i) · T

For the element e (i) of i = 1,..., K, the threshold value of the element e (i) is obtained by the following equation.

[0064]

S (i) = R2 (i) · T

As described above, the threshold value S of the element e (i)
(I) is determined. In this method, the reference data D (i) of the element e (i) of i = 1,..., K includes a defective portion, and the threshold value calculation data R1 ( i) Even if the whole may be affected by this defective portion, if the reference data D (i) of the element e (i) of i = N,..., N−k + 1 is normal, the element e (i) This effect can be reduced by the threshold value calculation data R2 (i) obtained in descending order, and the threshold value S (i) can be determined. Conversely, reference data D of i = N,..., N−k + 1
(I) includes a defective portion, and the threshold value calculation data R2
(I) Even if the whole may be affected by the defective portion, the influence can be similarly weakened by the threshold value calculation data R1 (i), and the threshold value S (i) can be determined. In this method, the reference data D (i) is the first data D (i).
This corresponds to the next reference data. The threshold value calculation data R1 (i) for i = k + 1,..., N corresponds to the second reference data,
Threshold calculation data R2 (i) for i = N−k,.
Corresponds to the third reference data.

FIGS. 8, 9 and 10 are flow charts of a first example of the threshold value determining program used in the embodiment of the present invention. In steps S1 to S12 of this flowchart, threshold value calculation data R1 (i) of the element e (i) is obtained. Step S
In steps 13 to S23, threshold value calculation data R2 (i) of the element e (i) is obtained. In steps S18 to S27, the threshold value S (i) of the element e (i) is obtained. Among these steps, steps S6 to S12 are performed by the first creating unit, the second creating unit, and the first creating unit.
Step S17 to step S
Reference numeral 23 corresponds to a third creating unit, a fourth creating unit, and a second repeating unit. Steps S28 to S30
Corresponds to threshold value determination means.

This flowchart will be described in the order of steps. In step S1, threshold value calculation data R1 (j) referred to when determining the threshold value of element e (i).
The number of the element e (j) of (j ≠ i) counted from the element e (i) in ascending or descending order is input by the user and substituted into k. The constraint amount C is determined by the user. In step S2, 0 is set to i indicating the number of the element e (i). In step S3, 1 is added to i. In step S4, threshold value calculation data R1
(I) is determined according to Equation 1. In step S5,
It is checked whether i has reached k. If i has not reached k, the process returns to step S3; otherwise, step S3
Proceed to 6.

In step S6, 1 is added to i.
In step S7, the data R1 (i) for calculating the threshold value of the k-th element e (ik) counted from the element e (i) in descending order
It is checked whether or not a value obtained by subtracting the reference data D (i) of the element e (i) from −k) is larger than the constraint amount C. Constraint amount C
If it is larger, the process proceeds to step S8; otherwise, the process proceeds to step S9. In step S8, threshold value calculation data R1 (i) is determined according to equation (2). In step S9, threshold value calculation data R1 (ik)
It is checked whether or not the value obtained by subtracting the reference data D (i) from is smaller than the constraint amount -C. If it is smaller than the constraint amount -C, the process proceeds to step S10, otherwise, step S11.
Proceed to.

In step S10, threshold value calculation data R1 (i) is determined according to equation (6). Step S1
At 1, the threshold value calculation data R1 (i) is calculated according to the equation (7).
Is determined. In step S12, it is checked whether or not i has reached N. If i is N, the process proceeds to step S13; otherwise, the process returns to step S6.

In step S13, the total number N is set to i indicating the number of the element e (i). Step S14
In the equation, the threshold value calculation data R2 (i) is determined according to equation (8). In step S15, 1 is subtracted from i. In step S16, it is checked whether i has reached N−k. If i has not reached N−k, the process returns to step S14; otherwise, the process proceeds to step S17. In step S17, the threshold value calculation data R2 (i + k) of the k-th element e (i + k) counted from the element e (i) in ascending order.
It is checked whether or not a value obtained by subtracting the reference data D (i) of the element e (i) from k) is larger than the constraint amount C. If it is larger than the constraint amount C, the process proceeds to step S18; otherwise, the process proceeds to step S19.

In step S18, threshold value calculation data R2 (i) is determined according to equation (9). Step S1
In step 9, it is checked whether or not a value obtained by subtracting the reference data D (i) from the threshold value calculation data R2 (i + k) is smaller than the constraint amount −C. If it is smaller than the constraint amount -C, the process proceeds to step S20; otherwise, the process proceeds to step S21.
In step S20, threshold value calculation data R2 (i) is determined according to Expression 10. In step S21, threshold value calculation data R2 (i) is determined according to equation (11). In step S22, 1 is subtracted from i. In step S23, it is checked whether i has reached 0.
If i is 0, the process proceeds to a step S24, otherwise,
It returns to step S17.

At step S24, 0 is substituted for i. In step S25, 1 is added to i. In step S26, a threshold value S (i) is obtained according to Equation 11. In step S27, it is checked whether i has reached k. If i is equal to k, the process proceeds to step S28; otherwise, the process returns to step S25.

In step S28, 1 is added to i. In step S29, the threshold value S is calculated according to Equation 9.
(I) is required. In step S30, i is N−k
Is checked to see if it is equal to If i is equal to N−k, the process proceeds to step S31; otherwise, the process proceeds to step S28.
Return to In step S31, 1 is added to i. In step S32, a threshold value S (i) is obtained according to equation (10). In step S33, it is checked whether or not i has reached N. If i has reached N, the program is terminated; otherwise, the process returns to step S31. The threshold determination program is configured as described above.

An example in which a threshold is obtained by the threshold determination program will be described. The graph shown in (a) of FIG. 12 is image data of a normal product used for setting a threshold. The graph shown in (b) of FIG. 12 shows the threshold value obtained by the threshold value determination program using the image data shown in (a) as the reference data D (i). The graph (b) in FIG. 12 is obtained assuming that C = 5 and k = 1. In this way, even if there is a defect in the image data used for the reference data D (i), a stable threshold graph is obtained which is not significantly affected by the defect as shown in FIG. .

Next, a second example of the threshold value determining program will be described. The second example is different from the first example in the threshold value determination program stored in the ROM 14. The threshold determination program used in the second example is
The moving average method and a part of the threshold value determination program of the first example are combined and configured. Specifically, first, the image data used for determining the threshold value is corrected by the moving average method, and from the obtained data, the threshold value calculation data R1 for the elements e (i) of k + 1,... (I)
Is required. Then, the threshold value S (i) is obtained from Expression 8. For the element e (i) for i = 1,..., k, the product of the data obtained by the moving average method and T indicating the transmittance or the reflectance of Equation 8 is obtained as the threshold value S (i). .

The graph (b) in FIG. 13 is obtained by converting the image data shown in the graph (a) in FIG.
(I) shows a threshold value obtained by the threshold value determination program used in the second example. The graph of (b) in FIG. 13 is obtained with C = 5, k = 1. The graph of FIG. 9A is the same as the graph of FIG. As shown in FIG. 13, the position where the level of the image data is minimum in the defective portion is different from the position where the threshold value S (i) of the element e (i) corresponding to this defective portion is minimum. Is a graph of a stable threshold value S (i). in this way,
The threshold value S (i) can be obtained in combination with the moving average method.

Next, a third example of the threshold value determining program will be described. The second example is different from the first and second examples in the threshold value determination program stored in the ROM 14. In the threshold value determination program of the first example, when determining the threshold value of the element e (i) of i = k + 1,..., N−k, the threshold value calculation data R1 (i) and R2 (i) The average was taken and found according to equation 9.
In this case, the reference data D (i) of the element e (i) of i = 1,..., K or the reference data D (i) of the element e (i) of i = N−k + 1,. If one of the reference data D (i) is normal even if a defect portion is included and either the threshold value calculation data R1 (i) or R2 (i) is affected by this, Threshold calculation data R1
By averaging (i) and R2 (i), the influence of the other reference data D (i) including the defective portion is reduced,
The threshold could be determined. The threshold value determination program of the third example is based on the reference data D (i) of the element e (i) of i = 1,..., K or the element e of i = N−k + 1,.
Even if the reference data D (i) of (i) includes a defective portion, the threshold value can be determined by weakening the influence of the defective portion compared to the threshold value determination program of the first example. is there.

In the threshold value determination program of the third example, the threshold value calculation data R1 obtained in the same manner as in the first example.
When there is a gap between the sizes of (i) and R2 (i), i
= 1,..., K, the reference data D (i) and i of the element e (i)
= N−k + 1,..., N, reference data D of element e (i)
It is determined that there is a defective portion in either (i), and the threshold value calculation data R1 (i) or R2
Either (i) is used to determine the threshold value of the element e (i). If i <N / 2, the threshold value calculation data R2 (i) is used to determine the threshold value. If i ≧ N / 2, the threshold value calculation data R1 (i) is the threshold value. Used to determine the value.

As shown in steps S17 to S23 of the threshold value determination program of the first example, as i becomes smaller, threshold value calculation data R2 (i) receives i = N−k + 1,. , N, the influence of the reference data D (i) of the elements e (i) is reduced. Therefore, i =
Even if the reference data of the element e (i) of N−k + 1,..., N includes a defective portion, the threshold value calculation data R2 (i) is used for the determination of the threshold value in the portion where i is small. be able to. Similarly, even if the reference data of the element e (i) of i = 1,..., N includes a defective part, the threshold calculation data R1 (i) is changed to the threshold value at the part where i is large. Can be used for decisions. For the above reasons, i <N /
If 2, the threshold calculation data R2 (i) is used to determine the threshold, and if i ≧ N / 2, the threshold calculation data R1 (i) is used to determine the threshold. Used.

FIG. 11 is a flowchart showing steps S124 to S137 which are a part of the flow of the threshold value determination program of the third example. Among them, steps S128 to S134 will be described.
The first step portion of the threshold value determining program of the third example is the same as steps S1 to S27 of the threshold value determining program of the first example. However, step S1
, The user can determine the value of the variable F in addition to k and C. Steps S135 to S137 are the same as steps S31 to S33 of the threshold value determination program of the first example.

In step S128, 1 is added to i indicating the number of element e (i). In step S129,
The absolute value of the difference between R1 (i) and R2 (i) is determined in step S10.
It is checked whether it is equal to or less than F determined in step 1. If it is less than or equal to F, the process proceeds to step S130; otherwise, the process proceeds to step S131. In step S130, a threshold value is determined according to equation 9.

In step S131, it is checked whether i is smaller than N / 2. If i is smaller than N / 2, the process proceeds to step S132; otherwise, the process proceeds to step S1.
Go to 33. In step S132, a threshold value is determined according to Equation 11. In step S133, a threshold value is determined according to equation (10). After step S132 or step S133, the process proceeds to step S134.

In step S134, it is checked whether i has reached N−k. If N−k has been reached, the process proceeds to step S135, and otherwise returns to step S128.
The above-described flow constitutes the threshold value determination program of the third example.

In the above three examples, k representing the distance between elements used in the threshold value determination program and the constraint amount C
Is determined according to the image to be inspected, an appropriate threshold value can be obtained. Further, a plurality of image data to be used for determining the threshold value are acquired, the levels of the image data for each element are averaged, and this is used as the reference data D (i).
It may be. This is because even if there is a defective portion in one image data, the data of the defective portion may be corrected by another image data. When the moving average method is used in combination as in the second example, a good result can be obtained by determining the number of pixels of the moving average method according to the image to be inspected.

In the above embodiment, the line CCD camera is used as the line image sensor, but another image sensor can be used. Further, it is also possible to realize higher-speed processing by setting the threshold value storage means and the run length code storage means to a number of sequences larger than two.

[0086]

As described above, the defect inspection apparatus and the defect inspection method of the present invention have the following effects because they are configured as described above. In other words, even if the inspection object meanders while traveling, the inspection range can be adjusted by automatically following the meandering of the inspection object, and a high-performance defect inspection apparatus and defect inspection method can be provided. Further, reading of the threshold value from the first threshold value storage means and writing of the threshold value to the second threshold value storage means are substantially performed in parallel,
The reading of the threshold value from the second threshold value storage means and the writing of the threshold value to the first threshold value storage means are controlled so as to be performed substantially in parallel. Even if the line image sensor changes over time or the physical characteristics of the inspection object change, the defect inspection of the inspection object can be performed at high speed and with high accuracy.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of a defect inspection apparatus according to the present invention.

FIG. 2 is a timing chart showing the operation of the defect inspection apparatus of FIG.

FIG. 3 is a flowchart showing a normal defect inspection procedure during the operation of the CPU in FIG. 1;

FIG. 4 is a flowchart showing a procedure for detecting an edge of an inspection object and setting a threshold value during operation of the CPU in FIG. 1;

FIG. 5 is a diagram schematically showing a method of edge detection and threshold setting according to the present invention.

FIG. 6 is a signal level diagram showing a data change in one scan of multi-valued image data for explaining a technique of edge detection and threshold setting according to the present invention.

FIG. 7 is a diagram for explaining an example of arrangement of a line CCD camera and an inspection target according to the present invention.

FIG. 8 is a first part of a flowchart of a program for determining a threshold used in the embodiment of the defect inspection apparatus of the present invention.

FIG. 9 is a second flowchart of a program for determining a threshold used in the embodiment of the defect inspection apparatus according to the present invention;
Part.

FIG. 10 is the last part of the flowchart of the first example program for determining the threshold value used in the embodiment of the defect inspection apparatus of the present invention.

FIG. 11 is the last part of the flowchart of the third example program for determining the threshold value used in the embodiment of the defect inspection apparatus of the present invention.

FIG. 12 shows an embodiment of the defect inspection apparatus according to the present invention;
FIG. 7 is an explanatory diagram showing a relationship between threshold values and reference data for each element obtained in a flowchart showing the first example of FIGS.

FIG. 13 shows a second embodiment of the defect inspection apparatus according to the present invention.
FIG. 4 is an explanatory diagram showing a relationship between threshold values and reference data for each element obtained in an example.

[Explanation of symbols]

Reference Signs List 1 line CCD camera (line image sensor) 2 image processing device 3 A / D converter 4 comparator 5 run-length encoding circuit (run-length encoding means) 9A, 9B run-length buffer (run-length encoding storage means) 11A, 11B Threshold memory (threshold storage means) 12 Reference value memory 13 RAM 14 ROM 15 CPU (edge detection means, threshold setting means, connectivity processing means, first and second control means) 16 I / O 17 host Computer 32 Inspection object 33 Inspection area 34A, 34B Non-inspection area 35A Bright defect 35B Dark defect

──────────────────────────────────────────────────続 き Continuation of the front page (56) References JP-A-9-178664 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G01N 21/84-21/958

Claims (4)

(57) [Claims] 1. A line image sensor that scans in a direction substantially perpendicular to a traveling direction of a traveling inspection target to obtain image data at an inspection position of the inspection target, and the inspection target is obtained from output image data of the line image sensor. Edge detection means for detecting an edge of an object; binarization means for binarizing output image data of the line image sensor; and threshold values required for the binarization means to perform binarization. At the border of the coordinates of the edge obtained by the first and second threshold value storage means and the edge detection means,
For the outside, the threshold is set to a predetermined value so that no defect is detected regardless of the value of the image data, and for the inside of the edge coordinates, a predetermined value or
Threshold value setting means for setting a function value of the image data as the threshold value; run-length encoding means for obtaining a change point address of the binary image data obtained by the binarization means; First and second run-length code storage means for respectively storing a plurality of rows of run-length codes output from the run-length encoding means in the line image sensor; and the first and second run-length code storage means. Connectivity processing means for performing a connectivity process on the run-length code stored in the storage device to detect a defect of the inspection object; and the run-length encoding to one of the first and second run-length code storage means. After the input of the run-length code output from the means is completed, the run-length code is read out and the connectivity processing means executes the connectivity process, and After the run-length code output from the run-length encoding means has been input to the other of the first and second run-length code storage means, the run-length code is read out and the connectivity processing means performs connectivity processing. Is performed, the connectivity to the first run-length code storage unit and the connectivity process using the run-length code from the second run-length code storage unit are substantially parallel-processed, and the second A first control means for controlling the input to the run-length code storage means and the connectivity processing using the run-length code from the first run-length code storage means to perform substantially parallel processing; When the binarization unit performs binarization using one threshold value of the second threshold value storage unit,
The threshold value is determined after the connectivity processing by the connectivity processing means is completed, and the determined threshold value is written to the other of the first and second threshold value storage means, thereby making the first threshold value. The reading of the threshold value from the threshold value storage means and the writing of the threshold value to the second threshold value storage means are performed substantially in parallel, and the threshold value is read from the second threshold value storage means. A defect inspection apparatus comprising: a second control unit configured to control reading of a threshold value and writing of a threshold value into the first threshold value storage unit in substantially parallel processing. 2. An edge of the inspection object is detected from an output signal from a line image sensor that scans in a direction substantially perpendicular to a traveling direction of the traveling inspection object and obtains image data at an inspection position of the inspection object. And a threshold value for binarization outside the vicinity of the coordinates of the edge obtained in the step of detecting the edge, which is determined by a predetermined value so that no defect is detected regardless of the value of the image data. And setting a predetermined value or a function value of the image data as the threshold value for the inside of the edge coordinates, and storing the function value in the first and second threshold value storage means. Binarizing the output signal from the line image sensor using one of the threshold values stored in the first and second threshold value storage means; Run-length encoding of the image data
And storing in one of the second run-length code storage means; and a connectivity process using the run-length codes for a plurality of rows of the image data of the line image sensor stored in the one run-length code storage means. And restarting the binarizing step during the execution of the connectivity process, and run-length encoding the binarized image data, the other of the first and second run-length code storage means And determining a threshold value for binarization using the result of the connectivity processing, and performing the binarization using a threshold value of the one threshold value storage unit. When running,
Storing the determined threshold value in the other of the first and second threshold value storage means. 3. The method according to claim 1, wherein the step of detecting the edge comprises reading an inspection object when the output signal of the line CCD camera as a line image sensor is converted into a digital signal by an A / D converter into multivalued image data. In detecting the edge portion to be inspected from the obtained multi-valued image (Di i = 1 to the number of elements of the line CCD camera), the left edge detection detects the effective range in the multi-valued image data. The edge detection is performed from the left end to the right end, and the edge detection at the right end is performed in the effective range in the multi-valued image data from the right end to the left end. 3. The defect inspection method according to claim 2, wherein coordinates satisfy equation (2) shown in equation (2). Equation 1 d ≧ | D _i −D _i + _k | d: Edge determination criterion i: s to (ek) s: Start point of effective range e: End point of effective range k: To perform edge detection D 2 | ≧ D _i −D _i − _k | i: e〜 (s + k) 4. A time obtained by multiplying a scanning cycle of the line image sensor by the number of rows stored in the run-length code storage means is a time required for the connectivity processing and a time for determining and storing the threshold value. 4. The defect inspection method according to claim 2, further comprising a step of interrupting the operation of determining and storing the threshold value when the sum is smaller than the sum of the threshold values and giving priority to execution of the connectivity processing of the next series.