KR101643713B1 - Method for inspecting of product using learning type smart camera - Google Patents

Method for inspecting of product using learning type smart camera Download PDF

Info

Publication number
KR101643713B1
KR101643713B1 KR1020150111121A KR20150111121A KR101643713B1 KR 101643713 B1 KR101643713 B1 KR 101643713B1 KR 1020150111121 A KR1020150111121 A KR 1020150111121A KR 20150111121 A KR20150111121 A KR 20150111121A KR 101643713 B1 KR101643713 B1 KR 101643713B1
Authority
KR
South Korea
Prior art keywords
image
boundary area
block
smart camera
matching rate
Prior art date
Application number
KR1020150111121A
Other languages
Korean (ko)
Inventor
유진선
Original Assignee
주식회사 이오비스
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 주식회사 이오비스 filed Critical 주식회사 이오비스
Priority to KR1020150111121A priority Critical patent/KR101643713B1/en
Application granted granted Critical
Publication of KR101643713B1 publication Critical patent/KR101643713B1/en

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/01—Subjecting similar articles in turn to test, e.g. "go/no-go" tests in mass production; Testing objects at points as they pass through a testing station
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84—Systems specially adapted for particular applications
    • G01N21/88—Investigating the presence of flaws or contamination
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84—Systems specially adapted for particular applications
    • G01N21/88—Investigating the presence of flaws or contamination
    • G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges

Abstract

The present invention learns to apply a state determination result determined by the management terminal to an image belonging to a boundary area that is a range between good products and defective areas so as to apply the determination result to an image belonging to a boundary area, The present invention relates to a method of inspecting an object to be inspected using a learning smart camera.
A method for inspecting an object to be inspected using a learning-type smart camera according to the present invention includes a first step of photographing an inspection object located on a stage in a smart camera, a first step of comparing the currently captured image with a pre- A third step of determining whether a matching rate currently calculated in the smart camera belongs to a boundary area between a good product and a defective area, a third step of determining whether the current photographing image belongs to a boundary area belonging to the boundary area, A fourth step of registering the boundary area information including the boundary area image and the matching ratio calculated in the second step in the data memory in the case of the image, Information is read out and displayed, and the result of the determination input by the administrator is displayed on a smart car Storing the mismatched block image as a result of the determination of the boundary area image stored in the data memory of the mosaic module; storing the mismatched block in the mismatched block, A step of calculating a matching rate and further storing and registering the mismatching block and the mismatching block matching rate of the corresponding image in the boundary area image; And a seventh step of performing a defect or good product determination process for the inspection object corresponding to the boundary area image based on the matching block matching rate and the boundary area determination criterion information including the judgment result information, Wherein the step (c) comprises the steps of: Block with the same matching rate cumulatively stored as the boundary area determination criterion information is compared with each other, and if the matching rate for the mismatching block is equal to or greater than the preset matching rate, the corresponding boundary area image When the matching rate between the mismatching block of the current boundary area image and the mismatching block of the boundary area image having the cumulative stored identical matching ratio is less than the preset matching rate A fifth step of providing a determination result input by an administrator through the management terminal as a result of checking the current boundary area image, and the sixth step of calculating and storing the mismatching block matching ratio in the sixth step .

Description

TECHNICAL FIELD [0001] The present invention relates to a method for inspecting an article to be inspected using a learning type smart camera,

The present invention relates to a method and apparatus for performing a determination of whether a product to be inspected is defective or good based on a matching rate with a reference image, The present invention relates to a method of inspecting an article to be inspected using a learning type smart camera, which makes it possible to improve the inspection accuracy with respect to the article to be inspected.

In a mass-produced product, defects such as a crack in the molded product, breakage of the edge portion, attachment of impurities, etc. occur due to changes in molding conditions and working environment during molding.

In order to select defective products with defective products, the molded products are placed on a belt conveyor in the production line. When the defective product is moved according to the operation of the belt conveyor, whether or not defective products such as the trunk or the edge portion has occurred A quality inspection process is performed.

In the quality inspection process, there is a method of inspecting defects through a visual inspection of the inspector, or inspection using a separate inspection system, and defective products are extracted from the belt conveyor according to the inspection results.

However, when the inspector inspects the defects of the product with the naked eye, since the moving speed of the product, that is, the moving speed of the belt conveyor, for visual inspection can not be improved beyond the set speed, productivity is reduced, If a defect is found that can not be detected, the defect rate of the product is increased.

Therefore, inspection systems are installed in production lines of a certain scale or more to perform inspection of products.

An inspection system for inspecting defects of a product generally includes a camera disposed on an upper portion of a belt conveyor, and an inspection device compares a product image photographed by the camera with a previously stored reference image. If the matching rate is less than a reference value, And judges that the matching rate is normal when the matching rate is equal to or greater than the reference value.

However, in the case of inspecting the defective product using the inspection system, since the matching rate of the photographed image and the reference image is compared with the reference value to determine the defective state of the inspection target product, If the rate is in the boundary region of the reference value, there is a problem that the accuracy of the inspection is lowered because the defective product is judged to be a normal product or an error that the normal product judges to be a defective product.

Accordingly, there is a need for an inspection system capable of improving the accuracy of state inspection of a product by performing more accurate state determination of a product having a matching rate between good and defective products.

1. Korean Patent Publication No. 2004-0054893 (entitled " Inspection Method Using 3D Inspection Device " 2. Korean Patent Laid-Open Publication No. 2015-0022249 (Name of the invention: defect inspection apparatus and method)

SUMMARY OF THE INVENTION Accordingly, the present invention has been made in view of the above circumstances, and it is an object of the present invention to provide an image processing apparatus and a method for processing an image, The object of the present invention is to provide a method of inspecting an article to be inspected using a learning type smart camera that can improve inspection accuracy of an inspection subject article by performing inspection processing.

According to an aspect of the present invention, there is provided a method of controlling a smart camera, the method comprising: a first step of photographing an inspection object located on a stage in a smart camera; A third step of determining whether the matching rate currently calculated in the smart camera belongs to a boundary area between good products and bad images, and a third step of determining whether the current captured image is a boundary area image belonging to the boundary area A fourth step of registering the boundary area information including the boundary area image and the matching ratio calculated in the second step in a data memory, and reading the corresponding boundary area image information registered and stored in the data memory of the smart camera at the management terminal And outputs the determination result input by the administrator to the data memory of the smart camera Storing the mapped block image as a determination result of the stored boundary area image, and calculating a mismatch block matching rate for the mismatched block by extracting the mismatch block between the corresponding boundary area image and the reference image in which the determination result is registered in the smart camera A step for further storing and registering the mismatching block and the mismatching block matching rate of the corresponding image with respect to the corresponding boundary area image, And a seventh step of performing a defect or good product determination process on the inspection object corresponding to the boundary area image based on the matching block matching rate and the boundary area judgment reference information including the judgment result information, In step 7, the mismatch block for the current boundary region image and the boundary region If the matching rate for the mismatch block is equal to or greater than a preset matching rate, the determination result of the corresponding boundary area image registered in the boundary area determination reference is compared with the current mismatching block, When the matching rate between the mismatching block of the current boundary area image and the mismatching block of the boundary area image having the cumulative stored identical matching ratio is less than the preset matching rate, And a sixth step of calculating the incoherent block matching rate and storing and registering the inconsistent block matching rate. The learning smart camera according to claim 1, A method of inspecting the article to be inspected is provided.

In the fourth step, the smart camera stores image information belonging to a boundary area in a shared memory area, and in the fifth step, the management terminal accesses the shared memory area and reads image information belonging to the boundary area And a result of the determination of the image information input by the administrator is additionally stored so as to correspond to the corresponding image of the shared memory area, and an inspection method of the inspection object article using the learning smart camera is provided .

delete

According to another aspect of the present invention, there is provided a method of detecting an object to be inspected positioned on a stage in a smart camera, Determining whether a matching rate currently calculated in the smart camera belongs to a boundary area between a good product and a defective area; calculating a matching rate by comparing the current captured image with a border area A step 14 of registering the boundary area information including the boundary area image and the matching area ratio calculated in the twelfth step in the case of the area image and extracting a mismatch block between the boundary area image and the reference image in the smart camera, The method comprising the steps of: (a) The method comprising the steps of: reading out and outputting each mismatch block of the image and registering the determination result input by the manager in correspondence with the mismatch block; And a seventeen step of performing a defective or good quality determination process for the inspection object corresponding to the boundary area image using the determination result of the cumulative registered mismatch block. In the seventeenth step, The inspection result for the boundary region image is provided based on the determination result of the mismatch block extracted from the current boundary region image and the cumulative registered previous mismatch block satisfying the predetermined similarity degree range, A cumulative registered mismatch that satisfies a predetermined similarity degree range in the region image And if it is determined that there is no luck, the determination result input by the administrator through the management terminal is provided as the inspection result for the current boundary area image. do.

In step 17, if the result of the mismatch block determination of the predetermined similarity degree ranges for the plurality of mismatching blocks for the corresponding image is different from each other, And providing a result of the determination of the state as a result of the inspection.

In step 17, if there is a mismatch block having a determination result of a mismatch block that is equal to or higher than the reference similarity degree with a mismatch block extracted from the current image among the cumulatively registered mismatch blocks, A method of inspecting an object to be inspected using a learning smart camera is provided.

According to the present invention, the state determination is performed according to the learning of the determination result determined by the management terminal for an image whose matching rate with the reference image is poor and which belongs to the boundary region between good products, The accuracy can be improved.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a schematic view showing an apparatus for inspecting an inspection object using a learning-type smart camera to which the present invention is applied; FIG.
FIG. 2 is a block diagram showing the internal structure of the smart camera 100 shown in FIG. 1 functionally separated. FIG.
FIG. 3 illustrates a record configuration of the boundary image region 162 and the boundary image determination reference region 164 shown in FIG. 2. FIG.
4 is a flowchart for explaining a method of inspecting an inspection object using a learning smart camera according to the first embodiment of the present invention.
FIG. 5 is a flowchart illustrating details of a status determination result processing step (ST 100) according to the boundary image determination reference shown in FIG.
6 is a view illustrating an image captured by a reference image and a boundary region.
FIG. 7 is a diagram illustrating a result of an examination of a state inspection accuracy using an inspection method of an inspection object article using a learning-type smart camera according to the present invention. FIG.
8 is a flowchart for explaining a method of inspecting an inspection object using a learning smart camera according to a second embodiment of the present invention.
FIG. 9 is a flowchart illustrating details of a state determination result processing step (ST 560) according to the boundary image determination reference shown in FIG.

Hereinafter, a preferred embodiment of the present invention will be described in detail with reference to the accompanying drawings.

FIG. 1 is a view showing a schematic configuration of an inspection object inspection apparatus using a learning-type smart camera to which the present invention is applied.

As shown in FIG. 1, a method of inspecting an object to be inspected using a learning-type smart camera applied to the present invention basically includes a smart camera 100 and a management terminal 200.

The smart camera 100 is positioned on the stage 2 where the article to be inspected 1 is located, for example, on a belt conveyor and photographs the inspected article 1, 1) and outputs the determination result information of the good or defective status to the inspection target article 1 based on the matching rate.

In addition, the smart camera 100 compares the reference image with the photographed image of the currently inspected object (1), and for the image belonging to the boundary region whose matching rate is between the bad region and the good region, (Not shown), and applies the determination result information determined by the management terminal 200 to the inspection area of an image belonging to the boundary area.

The management terminal 200 reads and displays an image belonging to a boundary area stored in a shared memory area (not shown) of the smart camera 100 according to a request from an administrator, (Not shown) of the smart camera 100. The smart camera 100 is configured to record the result of the determination on the image belonging to the region in the shared memory region (not shown).

FIG. 2 is a block diagram showing the internal structure of the smart camera 100 shown in FIG. 1 functionally separated.

2, the smart camera 100 includes an information input unit 110, a sensing sensing unit 120, an image sensing unit 130, an information output unit 140, a communication unit 150, a data memory 160, And an inspection control unit 170.

The information input unit 110 is for inputting inspection-related information from outside or setting inspection-related functions.

The sensing unit 120 is for recognizing a sensing signal indicating that the inspection object 1 is positioned on the stage 2. This receives the position sensing signal provided from the sensor 3 shown in Fig.

The image pickup unit 130 photographs the inspection target article 1 positioned on the stage 2. [

The information output unit 140 outputs inspection result information on the inspection object article 1. [ At this time, the information output unit 140 may include means for outputting or displaying the inspection result information. Further, the inspection result information may be provided in the form of characters, colors, and voices corresponding to "defective" or "good ".

The communication unit 150 is for communicating with an external terminal, and in particular, for communicating with the management terminal 200.

The data memory 160 includes a reference information storage block 161 in which a reference image and status judgment reference information are stored, a boundary area information storage block 162 for storing image related information belonging to the boundary area, And a storage block 163. The determination criterion information is a reference value for determining whether the defect is good or bad, and is set to a matching rate of the reference image and the current image. At this time, the determination criterion information may be a good reference matching rate for determining a good product and a poor reference matching rate for determining a defect. For example, the good reference matching rate is set at 80% or more, and the poor reference matching rate is set at less than 70%. Accordingly, the matching rate belonging to the border area can be automatically set to 70% or more and less than 80%.

In addition, in the data memory 160, the boundary area information storage block 162 includes a matching rate record, a video record belonging to the boundary area, and a determination result record. Here, the matching rate record is a matching rate for an image belonging to the boundary area calculated by the inspection control unit 170, and a video record belonging to the boundary area is an image belonging to the boundary area photographed by the image sensing unit 130 , And the result of the determination includes a defect or good product status information determined by the management terminal 200.

In addition, in the data memory 160, the boundary area determination reference storage block 163 stores a matching rate record for an image belonging to the boundary area, a mismatch block record in which an mismatch image of the image corresponding to the matching rate is stored, And a result record. Here, in the mismatch block record, positions and block images of all mismatch blocks extracted from the corresponding image may be stored.

The inspection control unit 170 controls the image sensing unit 130 to capture the inspection object 1 positioned on the stage 2 when the sensing signal is received from the sensing sensing unit 120, Defective, and boundary area states of the currently inspected article 1 by comparing the calculated matching rate and the reference matching rate with each other. At this time, the inspection control unit 170 outputs the result information corresponding to the good product and the defective state through the information output unit 140.

The inspection control unit 170 stores the image belonging to the boundary region and the matching rate in the data memory 160 and then outputs the determination result information determined by the management terminal 200 to the corresponding And outputs the inspection result to the information output unit 140 as the inspection result for the inspection object article 1. [

In addition, the inspection control unit 170 extracts a block that does not match the reference image with respect to the image belonging to the boundary region in which the determination result information is registered, and outputs the matching rate for the corresponding image and the determination result information determined from the management terminal 200 And is separately stored in the boundary area determination reference storage block 163 of the data memory 160. That is, the inspection control unit 170 accumulates the determination result of the image belonging to the boundary region in which the determination result is determined from the management terminal 200, in the boundary region determination reference storage block 163.

In addition, the inspection control unit 170 performs a defect or good quality judgment on the image belonging to the border area based on the boundary area judgment reference information accumulated in the data memory 160, Or the image belonging to the boundary area in which no judgment is made on the good product is stored in the boundary area information storage block 162 of the data memory 160 and the determination is made from the management terminal 200.

Next, the operation of the inspection object inspection apparatus using the learning smart camera having the above-described configuration will be described with reference to the flowchart shown in FIG.

First, in the smart camera 100, the reference image for the inspection object article 1 is registered, and reference values for status determination are stored. In addition, the smart camera 100 is disposed on the upper part of the belt conveyor so as to have an appropriate photographing angle for photographing the inspection object 1 positioned on the upper surface of the belt conveyor. 6 (A) illustrates a reference image.

When the inspected object 1 located on the upper side of the belt conveyor is moved in accordance with the conveyance of the belt conveyor in the above-described state and the sensor 3 senses that the inspected object 1 is positioned at the corresponding position, And provides a sensing signal to the smart camera 100.

The smart camera 100 photographs the inspection object 1 based on reception of a sensing signal provided from the sensor 3 (ST10).

The smart camera 100 compares the current photographed image with the previously registered reference photographed image and calculates the matching rate (ST20).

The smart camera 100 determines the state of the current captured image based on the matching rate calculated in ST20 (ST30). That is, when the calculated matching rate is equal to or higher than the good reference matching rate, the smart camera 100 determines that the smart camera 100 is in the good condition. If the calculated ratio is less than the bad reference matching rate, If it does not belong to the state, the image belonging to the boundary area is determined.

In step ST40, the smart camera 100 determines whether the determination result of the current captured image belongs to the boundary area in step ST40. If the smart camera 100 determines that the current captured image is not an image belonging to the boundary area, that is, if the smart camera 100 determines that the image is defective or in good condition, the smart camera 100 outputs or outputs the corresponding status information as inspection result information ).

If the determination result of the current captured image is the image belonging to the boundary region in step ST40, the smart camera 100 displays the current captured image and the matching rate thereof in the data memory 160 in the management terminal 200, In the boundary area information storage block 162 (ST60). At this time, the smart camera 100 may provide a message to the management terminal 200 indicating that the boundary image has been generated.

In addition, the management terminal 200 accesses the boundary area information storage block 162 of the smart camera 100 to read an image belonging to the boundary area, and displays the read image through display means (not shown). The result of the determination made by the administrator through the input means (not shown) of the management terminal 200 is stored in the boundary area information storage block 162, which is a shared area of the smart camera 100 . When the determination result is recorded and registered in the boundary image area 162, the corresponding sensing signal is provided to the smart camera 100, or the smart camera 100 sends a determination recording completion message to the management terminal 200 .

That is, when the determination result information about the current image belonging to the border area is registered in the border area information storage block 162 of the data memory 160 by the management terminal 200 (ST70) That is, good or bad status information, through the information output unit 130 (ST80). In FIG. 6, (B) is an image determined as good by the management terminal 200 among the boundary images, and (C) illustrates an image determined as bad by the management terminal 200 among the boundary images.

In step ST80, the smart camera 100 determines the image belonging to the border area through the management terminal 200 and outputs the result information. After that, the smart camera 100 displays the corresponding image in the border area information storage block 162 And registers the mismatching block information, the matching rate for the corresponding image, and the determination result information in the boundary area determination reference storage block 163 of the data memory 160 (ST90). For example, "P" in the image of FIG. 6 (C) represents a mismatch block, and a total of four mismatch block information is stored for the corresponding border region image.

As described above, in a state where judgment reference information for an image belonging to a border area is registered in the border area judgment reference storage block 163, the smart camera 100 judges whether the image of the inspection subject article 1 (ST100), a state determination process for the corresponding image is performed based on the boundary area determination reference information stored in the boundary area determination reference storage block 163 of the data memory 160. [0040] FIG.

FIG. 5 is a detailed flowchart illustrating a state determination processing step 100 for an image belonging to a boundary region according to the boundary region determination criterion shown in FIG.

4, when the smart camera 100 determines that the matching rate between the current captured image and the reference image is an image belonging to the boundary region (ST110) The boundary area determination reference storage block 163 of the data memory 160 searches the boundary area determination reference information for the matching rate that is the same as the matching rate of the current image (ST120). For example, when the matching rate of the image belonging to the current boundary region is 75%, the smart camera 100 may set the boundary region for the image having the matching rate of 75% among the boundary region images registered in the boundary region determination reference storage block 163, All the judgment reference information is searched.

If there is no boundary area determination criterion information having the same matching rate as the current image matching ratio in step ST120, information on the corresponding image is stored in the boundary area information storage block 162 of the data memory 160 and managed And performs a procedure after step ST60 of FIG. 4 for determining a determination result of the corresponding image belonging to the boundary area through the terminal 200. [

If the boundary region determination criterion information having the same matching rate as the current image matching ratio exists in step ST120, the smart camera 100 extracts a mismatch block that does not match the reference image in the current image (ST130) .

In step ST140, the smart camera 100 compares the mismatching block between the mismatch block of the image belonging to the current border area extracted in step ST130 and the border area decision criterion information found in step ST120. At this time, the smart camera 100 compares the mismatch block for each image and the mismatch block extracted from the current image in the searched boundary region determination reference information, and calculates a matching rate for each searched boundary region determination reference image. For example, the first matching rate between the four mismatching blocks for the current image and the first mismatching block for the first bordering area determination reference image is calculated, and the first matching rate between the four mismatching blocks for the current image and the first mismatching block is calculated. 2 The matching rate is calculated.

In step ST140, the smart camera 100 determines whether the matching rate for the mismatch block is greater than or equal to a predetermined mismatch block reference matching rate (ST150).

If the mismatch block matching rate is equal to or greater than the predetermined mismatch block reference matching rate in step ST150, the determination result of the boundary image determination information is determined as a status determination result for the corresponding image (ST160). That is, the smart camera 100 determines the corresponding boundary image determination result of the boundary region determination reference storage block 164, which is compared with the image belonging to the current boundary region, as the state determination result of the current image belonging to the boundary region, 130). For example, when the second matching rate of the four mismatch blocks with respect to the second border area determination reference image is 90% or more, the smart camera 100 determines the "good" determination result registered for the second boundary area determination reference image It is determined as a result of state determination of an image belonging to the current boundary region.

 If it is determined in step ST150 that the mismatch block matching rate is less than the predetermined mismatch block reference matching rate, the information about the corresponding image is stored in the border area information storage block 162 of the data memory 160, And then proceeds to step ST60 of FIG. 4 to determine the determination result of the image.

FIG. 7 is a graph showing a result of an experiment of state inspection accuracy of a specific inspection object using a state inspection apparatus using a learning smart camera according to the present invention.

7, since the inspection result information for the image belonging to the boundary area is accumulated, the automatic processing for the image belonging to the boundary area is enlarged, so that the judgment task of the manager is gradually reduced and the image belonging to the boundary area is stored As the test result information increases, the accuracy of the automatic test result gradually increases, and the average accuracy is 89.3%.

Meanwhile, in the above embodiment, the management terminal 200 registers the determination results of all of the images belonging to the boundary region. However, in the present invention, the management terminal 200 determines the mismatch region And to register the result of the determination of the good product or the failure.

FIG. 8 is a flowchart illustrating an operation of the state inspection apparatus using the learning-type smart camera according to the second embodiment of the present invention. FIG. 8 shows a procedure after ST40 for confirming that the current captured image is a video belonging to the boundary region in the smart camera in FIG.

First, if it is determined that the photographed image currently photographed by the smart camera 100 is an image belonging to the boundary region (ST 510), a mismatch block between the current image and the reference image is extracted, and the boundary including the mismatching block, And stores the area information in the shared area of the data memory 160 (ST 520). At this time, the number of the mismatch blocks may be plural.

The management terminal 200 reads the boundary area information in the shared area of the data memory 160 of the smart camera 100 and stores the determination result information inputted by the manager into each unmatching block so as to correspond to the corresponding unmatching block (ST530). At this time, the administrator can additionally store the determination result information on the image in the data memory 160 through the management terminal 200. [ For example, "good" determination information for a first mismatch block located in a background region for an image belonging to one border region and "bad" determination information for a second mismatch block located in the article region are stored in the data memory 160 ). ≪ / RTI >

Then, the smart camera 100 outputs inspection result of good or bad for the inspection object 1 corresponding to the image, based on the determination result information provided by the management terminal 200 (ST540).

In addition, the smart camera 100 stores the boundary area determination reference information including the determination result information registered by the administrator for each mismatch block in another storage area of the data memory 160, that is, the mismatch block determination reference storage area (ST550). At this time, an identification code for the mismatch block is stored in the mismatch block determination reference storage area and stored. The identification code for the mismatch block may be automatically assigned using the position coordinate value corresponding to the mismatch block. For example, when the center position coordinate of the mismatch block is (30,50), it can be given in the form of "30-50".

In a state in which the determination reference information determined by the management terminal 200 is registered for the mismatch block in the boundary area determination reference storage area as described above, the smart camera 100 determines that the captured image of the inspection object article 1 If it is determined that the image belongs to the boundary region, the state determination process for the corresponding image is performed based on the determination criterion information for each mismatch block (ST560).

FIG. 9 is a detailed flowchart illustrating a state determination processing step 560 for an image belonging to a boundary area according to the block matching block determination criterion shown in FIG.

If the smart camera 100 determines that the current captured image is an image belonging to the boundary region (ST561) in the state that the determination reference information for the mismatch block is registered in ST550 of FIG. 8, the smart camera 100 determines that the mismatch block between the current image and the reference image is And stores the boundary region information including the mismatching block and the corresponding image in the shared area of the data memory 160 (ST 562).

The smart camera 100 compares the mismatch block extracted in step ST562 with the mismatch block registered in the mismatch block determination reference storage area of the data memory 160 to search for the mismatch block having a high degree of similarity (ST563). At this time, the similarity condition for the mismatch block may be set to a position or a size. For example, the smart camera 100 may identify an ID corresponding to a center position coordinate value that satisfies a preset similarity degree range based on a center position coordinate value of a mismatch block for a current image among mismatch blocks registered in the mismatch block determination reference storage area, Search for mismatched blocks with code. For example, an mismatch block of an identification code having an X coordinate value in the range of "25 to 35" and a Y coordinate value in the range of "45 to 55" is searched for the mismatch block which is the center position coordinate value (30, 50).

The smart camera 100 confirms the determination result of the searched incoherent block, and determines and outputs the inspection result for the boundary area image according to the determination result (ST564).

In step ST564, if the smart camera 100 determines that the result of the determination is different for all the mismatch blocks for the corresponding image, the smart camera 100 determines whether the ratio is greater than or equal to 80% Can be output. For example, when eight "good" determination results and two "bad" determination results of ten mismatch blocks extracted for an image belonging to one boundary region are confirmed, "Print the test results.

In step ST564, the smart camera 100 may output a determination result of "bad" as an inspection result when any one of the "bad" determination results is checked for all the mismatch blocks for the corresponding image. This takes into account the state in which the discordance block is present on the item 1 to be inspected.

In another embodiment of the present invention, a first reference matching rate for determining whether the image is good or not and a second reference matching rate for determining an image belonging to the boundary region are stored in the data memory 160, It is also possible to judge whether the image is a good image on the basis of the first reference matching rate with respect to the current captured image and determine whether the image belongs to the boundary area based on the second reference matching rate Do.

According to another embodiment of the present invention, the inspection result statistics on images belonging to the boundary area based on the boundary area judgment reference information registered in the boundary area judgment reference storage block 163 of the data memory 160 And the good / bad judgment process for the image belonging to the border area is performed based on the result of the inspection. In this case, when the number of registered images is equal to or greater than the reference sample number as a result of the determination of the matching rate of the images belonging to the border area, the smart camera 100 determines that the matching rate of images It can be additionally set as a reference matching rate or a bad reference matching rate. For example, when the number of images belonging to the boundary region for the matching rate "79 " is 10,000 or more, which is the number of reference samples, the matching rate" 79 & Can be additionally set with a good reference matching rate.

In addition, it is needless to say that the present invention can be applied to an inspection system in which one management terminal 200 is coupled to a plurality of smart cameras 100 (not shown) as another embodiment.

Although the present invention has been described in connection with the above-mentioned preferred embodiments, it is possible to make various modifications and variations without departing from the spirit and scope of the invention. Accordingly, the scope of the appended claims should include all such modifications and changes as fall within the scope of the present invention.

100: smart camera, 200: management terminal,
110: Information input unit, 120: Sensing sensing unit,
130: imaging section, 140: information output section,
150: communication unit, 160: data memory,
170:
1: Item to be inspected, 2: Stage,
3: Sensor.

Claims (6)

A first step of photographing an inspection object located on a stage in a smart camera,
A second step of comparing a current photographed image with a previously registered reference image in a smart camera to calculate a matching rate,
A third step of determining whether the matching rate currently calculated in the smart camera belongs to a boundary area between good products and bad,
A fourth step of registering the boundary area image including the boundary area image and the boundary area information including the matching ratio calculated in the second step in the data memory when the current captured image is a boundary area image belonging to the boundary area,
The management terminal reads out and displays the corresponding boundary area image information registered and stored in the data memory of the smart camera, and outputs the determination result input by the manager to the determination result of the corresponding boundary area image stored in the data memory of the smart camera A fifth step of storing and registering,
A non-uniform block matching rate for the mismatched block is calculated by extracting a mismatch block between the corresponding boundary region image and the reference image in which the determination result is registered in the smart camera, and the mismatching block and the mismatching block matching rate of the corresponding image, A sixth step of further storing and registering the region image,
The smart camera may further include a feature extraction unit for extracting feature information of the inspection target object corresponding to the boundary area image based on the cumulative registered mismatch block and the block matching ratio and the boundary area determination reference information including the determination result information, And a seventh step of performing a defective or good product determination process,
In step 7, the smart camera compares the mismatch blocks of the boundary area image and the mismatch blocks of the boundary area images having the same matching rate accumulated and accumulated as the boundary area determination reference information, and determines a matching rate for the mismatch block If the matching rate is equal to or greater than the matching rate, the determination result of the corresponding boundary area image registered as the boundary area determination reference is provided as the inspection result for the current boundary area image,
If the matching rate between the mismatching block of the current boundary area image and the mismatching block of the boundary area image having the cumulative stored same accumulated matching ratio is less than the preset matching rate, And a sixth step of calculating and storing the inconsistent block matching rate and the sixth step of performing the sixth step of storing and registering the inconsistent block matching rate. The method according to claim 1,
In the fourth step, the smart camera stores image information belonging to a boundary region in a shared memory region,
In the fifth step, the management terminal accesses the shared memory area to read out and output image information belonging to the boundary area, and outputs the determination result of the corresponding image information input by the manager to the corresponding image of the shared memory area And the additional information is stored so as to correspond to the information of the inspection object. delete An eleventh step of photographing the inspection object located on the stage in the smart camera,
A twelfth step of comparing a current photographed image and a previously registered reference image in a smart camera to calculate a matching rate,
A thirteenth step of determining whether the matching rate currently calculated in the smart camera belongs to a boundary area between good products and bad,
A step 14 of registering boundary region information including the matching boundary region image and the matching ratio calculated in the twelfth step when the current captured image is a boundary region image belonging to the boundary region,
A step 15 of extracting a mismatch block between the boundary region image and the reference image in the smart camera and storing the extracted mismatch block in a data memory,
A step 16 of reading out and outputting each mismatch block of the corresponding border area image stored in the data memory of the smart camera in the management terminal and registering the decision result inputted by the manager in correspondence with the mismatch block,
A step 17 of performing a defect or good article determination process on the inspection object corresponding to the boundary area image using the determination result of the mismatched block cumulatively registered from the management terminal with respect to the boundary area image photographed in the smart camera ≪ / RTI >
In the seventeenth step, the mismatch block extracted from the current boundary region image among the mismatch blocks accumulated in the data memory in the smart camera is compared with the cumulative registered previous mismatch block that satisfies the preset similarity degree range, , ≪ / RTI >
When the cumulative registered mismatch block satisfying the pre-set similarity degree range does not exist in the current border region image, the judgment result input by the administrator through the management terminal is provided as the examination result of the current border region image A method of inspecting an object to be inspected using a learning type smart camera. 5. The method of claim 4,
In step 17, if the result of the mismatch block determination of the predetermined similarity degree ranges for the plurality of mismatch blocks for the corresponding image is determined to be different from each other, And providing a result of the determination as a result of the inspection. 5. The method of claim 4,
In step 17, the smart camera provides a check result of "bad" when there is a mismatch block extracted from the current image among the cumulatively registered mismatch blocks and a mismatch block whose determination result of the mismatch block is greater than the reference similarity degree is & Wherein the inspection target object is inspected using a learning type smart camera.
KR1020150111121A 2015-08-06 2015-08-06 Method for inspecting of product using learning type smart camera KR101643713B1 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
KR1020150111121A KR101643713B1 (en) 2015-08-06 2015-08-06 Method for inspecting of product using learning type smart camera

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
KR1020150111121A KR101643713B1 (en) 2015-08-06 2015-08-06 Method for inspecting of product using learning type smart camera

Publications (1)

Publication Number Publication Date
KR101643713B1 true KR101643713B1 (en) 2016-08-11

Family

ID=56714266

Family Applications (1)

Application Number Title Priority Date Filing Date
KR1020150111121A KR101643713B1 (en) 2015-08-06 2015-08-06 Method for inspecting of product using learning type smart camera

Country Status (1)

Country Link
KR (1) KR101643713B1 (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018074622A1 (en) * 2016-10-19 2018-04-26 (주)코글릭스 Inspection method and apparatus
KR20190037889A (en) * 2017-09-29 2019-04-08 주식회사 엘지씨엔에스 Method and system for detecting defect in produnt
US10677681B2 (en) 2016-10-19 2020-06-09 Coglix Co. Ltd. Inspection method and apparatus for testing cap sealing of container that involves calculating surface temperature of temperature sensor and comparing difference between measured temperature and the calculated temperature with threshold for correcting thermal image data
WO2021217467A1 (en) * 2020-04-28 2021-11-04 华为技术有限公司 Method and apparatus for testing intelligent camera
KR20230116537A (en) 2022-01-28 2023-08-04 경운대학교 산학협력단 Defective product sorting device and sorting method based on deep learning technology

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040054893A1 (en) 2002-09-18 2004-03-18 Anthony Ellis Method and system for a file encryption and monitoring system
KR20060133271A (en) * 2005-06-20 2006-12-26 주식회사 세영 The method and an apparatus for inspecting harness by edge-detection
KR20080101367A (en) * 2007-05-17 2008-11-21 주식회사 에이디피엔지니어링 Apparatus for inspection stamp and method for forming a nano-pattern
KR101043236B1 (en) * 2010-11-17 2011-06-22 표준정보기술 주식회사 Apparatus and method for inspecting appearance of led chip
US20150022249A1 (en) 2013-07-17 2015-01-22 Taiwan Semiconductor Manufacturing Co., Ltd. Method and apparatus for generating a ramp signal

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040054893A1 (en) 2002-09-18 2004-03-18 Anthony Ellis Method and system for a file encryption and monitoring system
KR20060133271A (en) * 2005-06-20 2006-12-26 주식회사 세영 The method and an apparatus for inspecting harness by edge-detection
KR20080101367A (en) * 2007-05-17 2008-11-21 주식회사 에이디피엔지니어링 Apparatus for inspection stamp and method for forming a nano-pattern
KR101043236B1 (en) * 2010-11-17 2011-06-22 표준정보기술 주식회사 Apparatus and method for inspecting appearance of led chip
US20150022249A1 (en) 2013-07-17 2015-01-22 Taiwan Semiconductor Manufacturing Co., Ltd. Method and apparatus for generating a ramp signal

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018074622A1 (en) * 2016-10-19 2018-04-26 (주)코글릭스 Inspection method and apparatus
US10677681B2 (en) 2016-10-19 2020-06-09 Coglix Co. Ltd. Inspection method and apparatus for testing cap sealing of container that involves calculating surface temperature of temperature sensor and comparing difference between measured temperature and the calculated temperature with threshold for correcting thermal image data
US10853933B2 (en) 2016-10-19 2020-12-01 Coglix Co. Ltd. Inspection method and apparatus
US11599989B2 (en) 2016-10-19 2023-03-07 Coglix Co. Ltd. Inspection method and apparatus
KR20190037889A (en) * 2017-09-29 2019-04-08 주식회사 엘지씨엔에스 Method and system for detecting defect in produnt
KR101995396B1 (en) 2017-09-29 2019-09-30 주식회사 엘지씨엔에스 Method and system for detecting defect in produnt
WO2021217467A1 (en) * 2020-04-28 2021-11-04 华为技术有限公司 Method and apparatus for testing intelligent camera
KR20230116537A (en) 2022-01-28 2023-08-04 경운대학교 산학협력단 Defective product sorting device and sorting method based on deep learning technology

Similar Documents

Publication Publication Date Title
KR101643713B1 (en) Method for inspecting of product using learning type smart camera
CN106920219B (en) Article defect detection method, image processing system and computer readable recording medium
KR20220041212A (en) PCB maintenance system and maintenance method based on false point defect detection
CN102253048A (en) Machine vision detection method and system for detection of various products
KR100598381B1 (en) in-line typed apparatus for auto wafer-defect classification and control method thereof
JPH04270950A (en) Visual inspection apparatus
CN112233994B (en) Chip gold wire detection method
CN113723325A (en) Tool defect detection system for prefabricated parts
CN113269769A (en) Defect detection method, system and equipment based on image registration
CN113284094A (en) Method, device, storage medium and equipment for acquiring defect information of glass substrate
CN111610190B (en) Appearance inspection device for household appliances
US8355546B2 (en) Method for identifying fingerprint image
JP2002022671A (en) Apparatus and method for inspecting inner wall surface of cylinder
KR20160084212A (en) Apparatus for inspecting defect and method thereof
CN114280073A (en) Tile defect detection device adopting laser and camera composite detection mode and classification method
US6920241B1 (en) System and method for bundled location and regional inspection
CN116091506B (en) Machine vision defect quality inspection method based on YOLOV5
TWI603077B (en) Map matching device, comparison method and comparison program
CN116660281A (en) Defect detection method and device based on machine vision and electronic equipment
CN113723841A (en) Online detection method for tool missing in assembly type prefabricated part
JP2019178928A (en) Inspection device and inspection method
CN111412941A (en) Method and device for detecting mounting quality
CN106934791B (en) Visual detection method and control equipment for FPC (Flexible printed Circuit) on mobile terminal
JPH10141925A (en) Appearance inspection device
CN117309766B (en) Visual detection method and system for product surface defects

Legal Events

Date Code Title Description
E701 Decision to grant or registration of patent right
GRNT Written decision to grant
FPAY Annual fee payment

Payment date: 20190716

Year of fee payment: 4