JP2004144615A - Appearance image inspection device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To realize a high-speed and highly-accurate appearance image inspection device by a method for determining existence of a foreign material by dividing an inspection image into small windows and calculating a standard deviation. <P>SOLUTION: This appearance image inspection device has following means. (1) The image is divided into the windows having a window size (for example, 16×16 pixels), and the standard deviation is calculated. (2) Productivity is improved by defective image taking and a place mark. (3) Two kinds of means, namely, a means for accumulating and displaying only defective images and a means for accumulating and displaying the whole images, are provided. (4) A mechanism capable of easily setting an image processing parameter of an inspection area is formed. <P>COPYRIGHT: (C)2004,JPO

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention provides an image inspection apparatus that automatically detects an abnormality in image data captured by a camera,
The present invention relates to an "algorithm for automatically detecting an abnormality" and a "reasonable system configuration".
[0002]
[Prior art]
Conventionally, the visual inspection of a product has been mainly performed by human eyes.
Visual inspection means that the product is free from scratches,
It is checked whether there is no dirt, whether there is no color unevenness, whether there is a different part, and so on.
Inspections are sometimes made by devising the lighting so that it is easy to see, or attaching a magnifying lens.
[0003]
The mechanization of this has also been considered and is, in part, well implemented.
Specifically, cameras, lighting, driving devices, image processing hardware,
Consists of computer software.
[0004]
In addition, a technique called pattern matching is usually used in which a normal image and an inspection image are compared to determine whether there is a problem.
[0005]
[Problems to be solved by the invention]
The present invention is trying to solve the following problems.
[0006]
(1) In many cases, human inspection involves small inspection targets.
There were problems such as burden on workers due to fatigue, oversight, and individual differences in judgment.
[0007]
(2) The reason that mechanization spreads only partially
Performance cannot be improved to a truly effective level with a combination of general-purpose devices.
It has been pointed out that a dedicated system is expensive, and it is difficult to change the specification or reuse it for other products.
[0008]
(3) When performing pattern matching,
The time required for image alignment becomes a problem.
To correct vertical and horizontal deviation and rotational direction deviation,
An operation requiring a long processing time such as affine transformation of an inspection image having a large file size is required.
[0009]
(4) The operability of the conventional device was not good. Specifically, for what reason,
Was it determined to be defective? Also, where is the position determined to be defective,
I want to see the image determined to be defective. There was a strong demand from users.
Also, when changing the inspection target, it is necessary to change the parameters to be set,
If it takes a long time, the overall productivity cannot be increased.
[0010]
[Means for Solving the Problems]
The following means are proposed to solve the above problems.
[0011]
(1) To enable high-speed processing,
The inspection target image is divided into a predetermined window size (for example, 16 * 16 pixels),
Calculate the standard deviation of the image data within the window size.
The magnitude of this standard deviation is
When the value is equal to or larger than a predetermined value, a high-precision inspection can be performed without performing pattern matching due to the presence of foreign matter.
[0012]
(2) The window size is switched so that foreign substances of various sizes can be determined.
[0013]
(3) Prepare a reference image for comparison judgment,
The determination is made by comparing the standard deviation value between the reference image and the inspection target image.
This improves accuracy.
[0014]
(4) As an image positioning means,
Find a solid area in the design,
This mark is used as a reference position detection mark.
By extracting an inspection target area by relative coordinates from a reference position mark, only a necessary area is inspected to reduce a time cost.
[0015]
(5) When a defective product occurs,
By taking the defective image into the PC and marking the defective part and displaying the reason for the defect at the same time,
Defect causes and phenomena can be accurately grasped.
[0016]
(6) means for storing and displaying only the defective image;
By having two means for storing and displaying all images, the former is applied since normal operation is only for defect detection.
In a test run, for example, there are 20 samples to be inspected and all images are
The latter is applied because it is necessary to check visually while switching parameters.
[0017]
(7) Specify the area to be inspected,
Display the image data and calculation result of the specified area,
By setting image processing parameters while looking at these values,
The productivity of equipment operation can be greatly increased.
[0018]
BEST MODE FOR CARRYING OUT THE INVENTION
In the present invention, the inspection image is divided into small windows,
By calculating the standard deviation in the window and determining the presence or absence of foreign matter,
A high-speed and high-accuracy visual image inspection device was realized.
In addition, image positioning methods and bad image recognition methods,
By devising the parameter setting method, a highly productive and easy-to-use system was realized.
[0019]
【Example】
FIG. 1 is a block diagram of an embodiment of the present invention.
First, reference numeral 1 denotes a product to be inspected.
Product 1 rotates with the rotating jig attached to the motor,
The whole product can be imported.
The jig and the motor are not shown.
[0020]
2 is a line sensor camera,
This is a device for capturing the rotating 1 as a computer image.
Since the line sensor camera 2 has the light receiving elements arranged in a line,
The product 1 is rotated in synchronization with the capturing operation of the line sensor camera 2,
They are combined in the frame grabber board 9 to constitute the whole.
[0021]
Reference numeral 3 denotes LED lighting, in which LEDs are arranged in a vertical line,
The arrangement is such that they are arranged in the same direction as the row of the line sensor cameras 2. The LED illumination 3 has a built-in condenser lens, and is focused on a row of the line sensor camera 2.
[0022]
Reference numeral 4 denotes an LED dimming unit, which is a device for adjusting the amount of light of 3.
The adjustment can be specified from the main computer 7,
This is performed through the line sensor relay unit 6.
[0023]
5 is a line sensor signal unit,
2 is a device that converts the image signal of the camera link specification sent from the camera 2 into a 32-bit RGB image signal.
This device also serves as a power supply for the line sensor camera 2.
[0024]
A line sensor relay unit 6 has a function of performing high-speed image processing of an image from the line sensor camera 2 under the control of the main computer 7 and transferring the image to the main computer 7.
[0025]
Reference numeral 7 denotes a main computer, which sends a control signal to the line sensor relay unit 6 for detection processing and an image sent from the line sensor camera 2.
The user performs an inspection by operating the keyboard and mouse attached to these.
[0026]
8 is a graphic card,
This is a device for displaying the image of the line sensor camera 2 and the state of the device on the computer display 11.
The graphic card 8 is built in the main computer 7.
[0027]
Reference numeral 9 denotes a frame grabber board corresponding to the line sensor camera 2.
The main computer 7 receives the image signal of the line sensor camera 2 through this device,
Image processing is being performed.
[0028]
10 is an input / output board,
This is a device for inputting and outputting control signals between the line sensor relay unit 6 and the main computer 7.
[0029]
11 is a computer display,
This is a device for displaying an image signal output from the graphic card 8 built in the main computer 7.
[0030]
Reference numeral 12 denotes a sequencer to which the operation state of the device for attaching a label is sent.
The line sensor relay unit 6 relays and transfers the data to the main computer 7.
The result of the quality judgment of the product is sent to the sequencer 12 through this signal,
After the product inspection process, defective products are discharged.
[0031]
FIG. 2 is an image (schematic) when a product is taken in.
If you process the imported images at once,
Because the design 100 included in the product may hinder detection,
It is necessary to select an arbitrary part in the product and perform detection.
[0032]
However, for products that are rotating without being synchronized with the capture operation,
It is virtually impossible to specify a specific part of the product,
Even if it is realized, there is a possibility that speeding up of the detection processing may be hindered.
[0033]
To solve this,
First, a mark was printed to identify the start position of the product in the captured image.
[0034]
This prints a fixed reference position 101 in the product,
This is because an arbitrary portion of the product can be specified by recognizing the reference position 101 in the image processing.
[0035]
The reference position 101 is a rectangular area printed in a fixed color in a product at a fixed color.
By setting this reference position 101,
No matter when the rotating product is taken in, you can identify a certain place,
Stable detection processing can be realized.
[0036]
FIG. 3 is a conceptual diagram of an algorithm for detecting the presence or absence of a foreign substance 202 from a standard deviation within a detection range.
By preparing a plurality of means for detecting the foreign matter 202,
We have responded so that stable inspection is possible even for products containing complex designs.
The main method is a detection method that uses the value of the standard deviation.
In order to correspond to the design and shape of the product,
This is dealt with by adopting a method such as changing the size of the determination area or comparing with the reference image.
[0037]
The foreign matter 202 that has entered the product tends to be uniformly black.
Therefore, when calculating the standard deviation of the captured pixels,
In the case where no foreign matter is included, the result 203 becomes larger than the result 201.
By utilizing this property, it is possible to determine a foreign substance in a product.
[0038]
To describe the detailed implementation method, if the detection processing is performed collectively within the selected area,
Since it is considered that color unevenness and luminance unevenness have occurred in the area,
In order to increase the detection accuracy, the inspection area is first divided by a small area (window) 200 of 16 × 16 pixels,
The standard deviation within the window 200 was calculated.
[0039]
If the calculated value at this time exceeds the value of the preset detection set value x constant times,
It is determined that there is a foreign substance 204 in the window 200.
However, there is a problem with this detection method,
If there is a foreign substance 205 exceeding the area of the size of the window 200,
Conversely, a case where the standard deviation does not change from the normal value to result 206 may be considered.
To avoid this phenomenon, change the window size to 64x64 208,
After performing the detection process again, the final determination result is output.
[0040]
Further, a method using comparison with a reference image is proposed as a new embodiment. When making a judgment, read the reference image in advance,
A reference image 301 was prepared as a normal sample.
By comparing the standard deviation 303 of the same position of the reference image 301 with the standard deviation 302 of the currently captured image 300,
Stable detection is realized regardless of where the inspection target is in the product.
[0041]
In the automatic appearance inspection, it is most important to set the inspection apparatus as requested by the user and to have a high degree of freedom.
However, if the operation is complicated or the settings are too complicated,
In addition to the user's inability to use it, there is a possibility that the inspection function cannot be fully exhibited.
In this system, it is easy to set the parameters for detection.
The emphasis is on interlocking with the production management system, and the system has been enhanced as an inspection system.
[0042]
First, how to use inspection images,
In this system, when performing a test operation such as a new product detection operation,
For setting detection parameters and checking visually,
Save all images regardless of good or bad,
On the other hand, in the case of normal operation, it is only used as a material for defect countermeasures at a later date,
It is possible to select to save only NG images.
[0043]
However, even if the number of NG images is small,
If all the images are stored in the past, the disk capacity of the main computer 7 becomes insufficient,
Since it is possible that there is a problem with the detection process,
Disk space is saved by overwriting images exceeding a certain number.
[0044]
FIG. 4 shows a screen for setting the detection parameters.
In this screen, specify the reference position 101 of the inspection,
It is possible to specify an arbitrary position in the product based on 101.
[0045]
If you work with these detection parameters using only detection results and numerical values,
Setting parameters is very complicated,
Not only can you not get stable results,
It may not be used effectively as an inspection system.
However, this system displays the maximum / minimum value, standard deviation and average value of the selected area on the screen.
[0046]
Further, the screen can be simultaneously displayed in the inspection target area by specifying the screen options.
Because of this, setting does not become complicated,
Since the value can be easily set and changed, it can be said that the system can be operated without putting a burden on the user.
[0047]
[Brief description of the drawings]
FIG. 1 is an example of the configuration of the present invention. FIG. 2 is a schematic diagram of product take-in. FIG. 3 is an outline of foreign matter and standard deviation. FIG. 4 is an outline of detection using a reference image. FIG. 5 is an example of parameter setting. Description】
DESCRIPTION OF SYMBOLS 1 Product 2 Line sensor camera 3 LED illumination 4 LED dimming unit 5 Line sensor signal unit 6 Line sensor relay unit 7 Main computer 8 Graphic card 9 Frame grabber board 10 Input / output board 11 Computer display 12 Sequencer 100 Design 101 Reference position 200 Window 201 Result including no foreign matter 202 Foreign matter 203 Result including foreign matter 204 Result obtained as a foreign matter 205 Foreign matter exceeding the window size 206 Result of processing a large foreign matter 208 Window 209 after resizing the size Foreign matter caught in the resized window 300 Currently acquired product image 301 Reference image 302 Standard deviation of currently acquired product image 303 Standard deviation of reference image

Claims (8)

The image of the inspection object is stored in the memory,
Or in the appearance image inspection device that captures in a computer and detects foreign matter in the captured image,
The inspection target image is divided into a predetermined window size (for example, 16 * 16 pixels),
Means for calculating a standard deviation of the image data within the window size,
When the magnitude of the standard deviation is equal to or larger than a predetermined value, it is determined that there is a foreign substance. In the above 1,
An appearance image inspection apparatus comprising: means for switching the window size according to the size of the foreign matter in an inspection object. In the above 1,
Prepare a reference image for comparison judgment,
Setting the window size in the same region of the reference image and the inspection target image,
When the standard deviation value of the inspection target image is equal to or more than a predetermined multiple of the standard deviation value of the reference image, it is determined that there is a foreign substance. The image of the inspection object is stored in the memory,
Or in an appearance image inspection apparatus that detects a foreign substance in the captured image by capturing the captured image,
During the design of the target image, has a means for setting a mark for reference position detection,
The reference position mark is characterized by being determined by a fixed color that is continuous within a predetermined area. In the above 4,
The inspection target area is extracted based on the relative coordinates from the reference position mark. The image of the inspection object is stored in the memory,
Or, in the appearance image inspection device that detects a foreign substance in the captured image by capturing the captured image, when a defective product occurs,
It is characterized in that it has means for taking in the defective image into the PC, marking the defective portion, and displaying the reason for the defect. In the above item 6,
Means for storing and displaying only the defective image;
It is characterized by having two means for storing and displaying all images. The image of the inspection object is stored in the memory,
Or in the appearance image inspection device that captures in a computer and detects foreign matter in the captured image,
It has a means to specify the area to be inspected,
It has a means to display image data and calculation results of the specified area,
It is characterized in that a desired image processing parameter can be set with reference to these numerical values.

Images