JP3829280B2 - Appearance inspection device - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an appearance inspection apparatus, and more particularly to an appearance inspection apparatus for mounting components on a printed circuit board on which electronic components such as semiconductor components are mounted.
[0002]
[Prior art]
Conventionally, this type of visual inspection apparatus has a function of determining whether or not an electronic component, which is an object to be inspected, mounted on a printed circuit board is leaked, reversely attached, a constant difference, and a mounting deviation.
As a conventional technique, Japanese Patent Application Laid-Open No. 9-126727 discloses an image of a component on a board from an image captured by a camera by an image processing unit, and a dimension measurement unit performs inspection processing such as pad dimensions by image processing. A method is described in which the data processing unit that measures the arrangement relationship and receives the inspection information from the dimension measuring unit searches the component library unit for inspection information of a component that matches the data.
In Japanese Patent Laid-Open No. 8-254501, a normal inspection mode and a teaching mode are installed in the inspection apparatus. In the teaching mode, a determination criterion is created using discriminant analysis based on sample data as preprocessing, and then, the inspection mode is entered and an inspection is performed based on the determination criterion. Here, a method is described in which a component that causes an erroneous determination is taught, added to sample data, and a determination criterion is created again. However, each technique is a method for calculating the optimum judgment value within the range of the actual substrate. For the final judgment reference value, a certain amount of inspected parts are tested and empirical factors are used. Tuning is necessary.
[0003]
[Problems to be solved by the invention]
In the above prior art, the dictionary data and determination parameters used in the inspection are set by verification and tuning using an actual board or actual parts, and the past board and parts cannot be verified. In order to register a high dictionary, it is necessary to use a certain amount of examination trials and judgment based on empirical factors. Therefore, there is a problem that it requires individual ability, labor and time.
In addition, in order to register a dictionary with a high recognition rate in a misrecognized inspection target, there is a problem that production data is stopped and verified because the dictionary data is tuned using a real recognition target misrecognized. there were.
[0004]
An object of the present invention is to provide an appearance inspection apparatus capable of registering dictionary data and determination parameters with a high recognition rate independent of competence in a short time without stopping production.
[0005]
[Means for Solving the Problems]
In order to solve the above problems, an image processing unit that performs image processing on a captured image obtained by capturing an object to be inspected , an image storage unit that automatically saves the captured image, a standard dictionary unit having character feature data, and A user dictionary unit, a determination parameter unit for setting a pass / fail threshold value for addition, change, deletion and character matching of data in the user dictionary unit, data processed by the image processing unit, data in the standard dictionary unit and user dictionary unit A data processing unit that performs pass / fail determination processing based on the original collation result and the data of the determination parameter unit, and the pass / fail determination result determined by the data processing unit or the input determination result is stored in the image processing unit. and a test result storage unit that stores in association with the captured image, when registering the characteristics of the characters in the new user dictionary section by determining parameter section, the image storage unit by the data processing unit Inspects simulation using the saved captured image, verifies the set of acceptance threshold value of the character collation as missed does not occur.
Here, when a false alarm for determining a non-defective product as a defective product occurs, the user parameter is changed by changing the data in the user dictionary unit and the pass / fail threshold value for character collation by the determination parameter unit, and performing a plurality of inspection simulations. Change the data of the copy and set the pass / fail threshold for character collation .
[0006]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
FIG. 1 shows the configuration of an appearance inspection apparatus according to an embodiment of the present invention. In FIG. 1, a mounting component 2 that is an object to be inspected is mounted on the upper surface of a printed board 1, and a mounting component 3 that is an object to be inspected is mounted on the lower surface. An annular illumination device 4 is disposed on the upper surface between the CCD camera 6 and the printed circuit board 1 downward, and an annular illumination device 8 is disposed on the lower surface in the same manner on the upper surface. In the appearance inspection apparatus, on the upper surface side, the CCD camera 6 and the annular illumination device 4 are moved along the X axis movement axis 5 and the Y axis movement axis 7 by the XY robot control unit 101, and the mounting component 2 to be inspected is annularly illuminated. Irradiation is performed by the apparatus 4 and an image is captured by the CCD camera 6. On the lower surface side, the CCD camera 10 and the annular illumination device 8 are mounted by the XY robot controller 101 while moving the X-axis movement axis 9 and the Y-axis movement axis 11 independently of the upper surface side, and are mounted components to be inspected. 3 is irradiated with an annular illumination device 8 and an image is picked up by a CCD camera 10.
In addition, the appearance inspection apparatus according to the present embodiment has an image processing unit 103 that extracts features of a captured image by image processing, an image storage unit 201 that stores the captured image, and data such as a character string direction of a component. The component library unit 202, the determination parameter unit 203 having a determination threshold, the standard dictionary unit 204 and the user dictionary unit 205 having character feature data, the component placement unit 206 having component mounting position information, and the determination result are stored. And an inspection result storage unit 207 for generating inspection data for inspection and a data processing unit 102 for performing pass / fail judgment processing.
The captured image is subjected to image processing by the image processing unit 103 based on data in the component library unit 202 having information such as a character string direction and a character color by the data processing unit 102, and features are extracted. Next, based on the data in the standard dictionary unit 204 and the user dictionary unit 205 having dictionary data having character features as data, the image extracted from the features is subjected to character collation by an OCR (Optical Character Reader) method. This is performed by the processing unit 103. Next, the data processing unit 102 performs pass / fail determination processing based on the collation result and the data of the determination parameter unit 203.
The determination parameter unit 203 has information regarding the storage conditions of the captured image, and stores the image in the image storage unit 201 based on the information and the quality determination result. When the captured image is stored, the board ID, the mounting surface, the device, and the image file NO, which are information for managing the captured image, are stored in the inspection result storage unit 207 as image data management data 501 (FIG. 4). The board ID, the mounting surface, and the device are information for associating with the pass / fail determination result stored in the inspection result storage unit 207, and the image file NO is information for associating with the file name of the stored captured image.
[0007]
FIG. 2 shows a flow of saving the captured image and the inspection result stored in the image storage unit 201 in the inspection result storage unit 207.
The inspection is started (301), and when the data processing unit 102 determines that the inspection is defective (302), the data processing unit 102 is temporarily stopped. At this time, since it is unknown whether the determination result of the data processing unit 102 is correct or incorrect, the captured image is displayed on the screen and an operator is called by a buzzer (303). The operator compares the imaging screen with the collation character information to make a final determination of a non-defective product and a defective product (304), and inputs a determination result from the input unit (not shown) to the data processing unit 102 (305). The input result is stored in the inspection result storage unit 207, and the captured image is stored in the image storage unit 201 based on the image storage conditions of the determination parameter unit 203.
In the determination parameter 203, three conditions are stored as the captured image storage conditions: all images are stored, an image that the data processing unit 102 determines to be defective is stored, and an image whose final result input by the operator is a defective product is stored. The component ID is used for distinguishing the types of components. The above conditions are set in advance.
The inspection result is stored as inspection result data 502 (FIG. 4) in the inspection result storage unit 207 after completing the inspection of one substrate (306).
[0008]
FIG. 3 shows a flow of inspection simulation using a stored image stored in the image storage unit 201. Here, the purpose is to register dictionary data and determination parameters with a high recognition rate, and this is a method of determining the final registration data by verifying the recognition rate by inspection simulation.
There are two types of dictionary data for character recognition: standard dictionary data in which character characteristics are registered as default values, and user dictionary data that is additionally registered by the user. The determination is made with these two types of dictionary data. Processing is performed by the data processing unit 102 and the image processing unit 103 using the parameters. The standard dictionary data has such information in the standard dictionary unit 204 and cannot be added, changed, or deleted. On the other hand, the user dictionary data has the information in the user dictionary unit 205, and can be added, changed, and deleted. Therefore, the recognition rate is improved by changing the determination parameter of the determination parameter unit 203 that sets the pass / fail threshold value for addition, change, and deletion of user dictionary data and character collation.
As for the flow of the inspection simulation, after the user dictionary is registered or the determination parameter is changed (401), the simulation range is set (402). Here, which part type is the target, the image stored in the period from what time to what time, all the images, or only the image erroneously recognized by the data processing unit 102 Set conditions such as whether or not to target. Next, an inspection simulation process is executed (403). The inspection simulation is processed by the data processing unit 102 and the image processing unit 103 based on the captured image stored in the image storage unit 201 and the inspection algorithm. As the processing method, the processing after image capturing is performed by the same processing as the normal inspection, and the inspection (good / bad determination) result is output (404). In the inspection simulation, processing is executed based on data such as the inspection simulation data 503 (FIG. 4), and the inspection simulation result 504 (FIG. 4) is output and stored in the inspection result storage unit 207. In the inspection simulation result 504, the state of the false alarm is confirmed, and it is confirmed that there is no oversight for determining a defective product as a non-defective product (405). If an oversight has occurred, user dictionary registration or determination parameter change and inspection simulation are performed again. In addition, for false information for determining a non-defective product as a defective product, a user dictionary registration or determination parameter change and a test simulation are performed a plurality of times, and a test simulation result summary 505 output as comparison data from the data processing unit 102 (FIG. 4), final setting data is determined (406), and the contents of change are registered (407).
[0009]
As described above, in the present embodiment, the captured image captured by the inspection is automatically stored in the image storage unit, the inspection simulation using the captured image stored in the image storage unit is performed, and the determination parameter for determining pass / fail is changed. And the dictionary data that registered the feature of the part is changed, and the recognition rate after the change of the judgment registration condition is inspected by using the captured image stored in the image storage unit without using the actual board or the actual part. This makes it possible to register dictionary data and judgment parameters with a high recognition rate, and adjust the judgment registration conditions when misrecognition occurs without stopping production. Can be made.
[0010]
FIG. 5 shows an application example of inspection simulation. This is an example in which a component 601 mounted on a board is inspected, and an inspection example in which the stamped character of the component 601 is “HG1B” and the non-defective product is “HGLB” of the component 602 is shown. The OCR is a method of collating with an index called a certainty factor indicating a collation rate with characters, but here, if “l” is recognized as “L”, an oversight occurs. In this examination, “l” of “HG1B” indicates an example of a result of “l” with a confidence of 86% and “L” with a confidence of 34%, and this character is recognized as “l”. As a result, this part is determined to be defective because “l” is not “L”.
Next, an example in which the component 603 is recognized as a defective product although it is a non-defective product will be described. In this example, “LV42” is “L” with a certainty factor of 46% and “L” with a certainty factor of 35%. As a result, “LV42” is determined to be “LV42”, which is a non-defective product, but is erroneously recognized as a defective product. In order to increase the recognition rate of the component 603, it is necessary to make an adjustment so that the letter “L” can be recognized as “L”. However, here, the feature of the letter “L” is registered as a user dictionary. An example is shown.
When the character “L” is newly registered in the user dictionary unit 205 so as to increase the recognition rate, the registered result is “L” with a certainty factor of 97%, for example, “L”, and “l” with a certainty factor of 17%. May result. In this case, since “LV42” is determined as “LV42”, it is not a false alarm.
Next, as an effect of registering the dictionary, when the part 601 that has been detected as defective can be inspected, an oversight may occur. Since the dictionary data has been changed from the previous check, the result of “l” of “HG1B” may change to “L” with a confidence of 57% and “l” with a confidence of 47%. Yes, “HG1B” is misrecognized as “HGLB”, and as a result, oversight occurs.
Although these verifications need to be verified retroactively to past inspection parts, in this embodiment, inspection simulation processing is performed using captured images stored in the image storage unit 201 without using an actual substrate or actual parts. Since this method is used for verification, it is possible to prevent such an oversight.
[0011]
【The invention's effect】
As described above, according to the present invention, dictionary data having a high recognition rate and determination parameters can be registered in a short period of time and with little effort without individual differences.
Moreover, since adjustment of the determination registration conditions at the time of misrecognition generation can be performed without stopping production, there is an effect that the equipment operation rate can be improved.
[Brief description of the drawings]
FIG. 1 is a configuration diagram of an appearance inspection apparatus according to an embodiment of the present invention. FIG. 2 is a flowchart for explaining storage of captured images and inspection results according to the present invention. FIG. 4 is an explanatory diagram illustrating an example of inspection data according to the present invention. FIG. 5 is an explanatory diagram illustrating an application example of the inspection simulation according to the present invention.
DESCRIPTION OF SYMBOLS 1 ... Printed circuit board, 2 ... Top surface mounting component, 3 ... Bottom surface mounting component, 4 ... Upper surface side annular illumination device, 5 ... Upper surface side Y-axis movement axis, 6 ... Upper surface side camera (CCD), 7 ... Upper surface side X-axis movement Axis, 8 ... lower surface side annular illumination device, 9 ... lower surface side X axis movement axis, 10 ... lower surface side camera (CCD), 11 ... lower surface side Y axis movement axis, 101 ... XY robot control unit, 102 ... data processing unit, DESCRIPTION OF SYMBOLS 103 ... Image processing part, 201 ... Image storage part, 202 ... Parts library part, 203 ... Determination parameter part, 204 ... Standard dictionary part, 205 ... User dictionary part, 206 ... Component arrangement part, 207 ... Inspection result storage part, 501 ... image data management data, 502 ... inspection result data, 503 ... inspection simulation data, 504 ... inspection simulation result, 505 ... inspection simulation result summary, 601 ... mounted on a substrate Defective parts, 602 ... non-defective parts

Claims (2)

In an appearance inspection apparatus for judging whether or not an electronic component that is an inspection target mounted on a printed board is mounted, reversely mounted, a difference in constant, or a mounting deviation is good, a captured image obtained by capturing the inspection target is imaged An image processing unit for processing, an image storage unit for automatically saving the captured image, a standard dictionary unit and a user dictionary unit having character feature data, and addition, change, deletion and character of data in the user dictionary unit Based on the determination parameter unit for setting a verification pass / fail threshold, the result of the image processing by the image processing unit, the result of the verification based on the data of the standard dictionary unit and the user dictionary unit, and the data of the determination parameter unit coercive in association with the data processing unit for the acceptance judgment processing, the captured image is saved the determined acceptance judgment result or the input determination result by said data processing unit to the image processing unit And a test result storage unit that, when registering the characteristics of the characters in the new to the user dictionary section by the determining parameter section, a test simulation using captured images stored in the image storage unit by the data processing unit An appearance inspection apparatus characterized in that the pass / fail threshold setting of the character collation is verified so as not to be overlooked . In claim 1, when a false alarm for determining a non-defective product as a defective product occurs, the determination parameter unit changes the data in the user dictionary and the pass / fail threshold value for character collation, and a plurality of inspection simulations are performed. And changing the data of the user dictionary and setting the pass / fail threshold value for the character collation .

Images