JPH03154173A - Defective deciding image processor - Google Patents

Abstract

PURPOSE:To execute the inspection by a quick and simple processing by deriving a decision value or a reference value by a statistical processing from the number of picture elements shown by a non-defective work, comparing these values and the number of picture elements shown by an inspection work and deciding whether a parts omission exists or not. CONSTITUTION:Plural works confirmed to be a non-defective separately are photographed in advance by television cameras TV1 - TV4, and by binarizing an image signal by an image processor 20, the number of picture elements of the non- defective work is counted, and by bringing the count value to statistical processing, an average value mu and a standard deviation sigma are derived, and mu+ or -3sigma is stored as a decision value. Subsequently, an inspection work is photographed by the same television camera, and by binarizing its image signal, the number of picture elements of the inspection work is counted, and when the number of picture elements of the inspection work deviates from the decision value, it is decided that the inspection work is defective, and in the case of a non-defective, the count value of the picture element of its inspection work is used for calculating the average value mu and the standard deviation sigma and the accuracy of the statistical processing is enhanced. In such a way, an omission inspection can be executed simply, quickly and automatically.

Description

DETAILED DESCRIPTION OF THE INVENTION Industrial Application The present invention relates to an image processing device for determining defective products, and is suitable for use in inspecting the quality of products with complex shapes, particularly for inspecting defective parts that are missing.

B. Summary of the Invention The defective product determination image processing device of the first invention photographs a plurality of workpieces that have been separately confirmed to be non-defective products with a television camera, and binarizes the image signals of the television camera to identify the non-defective workpieces. Count the number of pixels, statistically process the count value to find the average value μ and standard deviation σ, store μ ± 3σ as the judgment value, then photograph the inspection work with the same TV camera, and record the image. The signal is binarized and the number of pixels of the inspection work is counted. If the number of pixels of the inspection work is outside the judgment value, the inspection work is determined to be defective, and if it is non-defective, the pixel count value of the inspection work is determined. the mean value μ and standard deviation σ
It is used to increase the accuracy of statistical processing.

The defective product determination image processing device of the second invention photographs a plurality of workpieces that have been separately confirmed to be non-defective products using a TV camera, binarizes the image signals of the TV camera, and counts the number of pixels of the non-defective workpieces. Then, statistically process the count values to find the average value, store the average value as a reference value, then photograph the workpiece to be inspected with the same TV camera, binarize the image signal, and calculate the number of pixels of the workpiece to be inspected. If there is a difference of more than a set value between the reference value and the number of pixels of the inspection work, the inspection work is determined to be defective, and if the inspection work is good, the pixel count value of the inspection work is calculated as an average value. It is used to improve the accuracy of statistical processing.

C6 Prior Art One typical example of a product with a complex shape is automobile wiring eights.

A wide variety of electrical components are used in automobiles, and there are many electrical components with relatively low current capacity that perform lighting, signals, control, warnings, charging, meters, window cleaning, etc.

Wiring eightness is a pre-assembled form in which the wiring connecting these electrical components can be assembled smoothly on an automobile assembly line. A wiring harness is constructed by cutting a large number of electric wires to the required length, gathering them together using vinyl tape, and attaching harness parts (tubes, clampers, grommets, protectors, taping, etc.) to the wires.

In the normal manufacturing method for wiring/eighthness, the bundled wires are set on a drawing board (a board with manufacturing drawings pasted on it), and the drawing board is sent along the production line to determine the required number of bundled wires. It is made by attaching the necessary 8-ness parts to the parts.

The wiring harness made in this way is inspected to see if all necessary parts are attached, and if any parts are missing, it is judged to be defective.

Conventionally, human visual inspection has been performed as a method for inspecting parts for missing parts, and the introduction of visual inspection using image processing is being considered.

A conventional visual inspection using image processing will be explained with reference to FIG.

As shown in FIG. 8, the wiring eightness 1 is set on a drawing board 2, and the drawing board 2 is fed in the direction of the arrow. 1a is a converging wire, and 1b is a harness component. Then, the wiring/eightness 1 and the drawing board 2 are photographed with an industrial television camera 3, and an image signal is sent to the image processing device 4. The image processing device 4 has a built-in television camera interface and a microcomputer, and divides the surface of the drawing board 2 into a plurality of photographing areas corresponding to the photographic field of view of the television camera 3, for example, divided by the dotted chain line in the figure. 4 shooting areas I,
The image processing device 4 captures the image signal of the television camera 3 at each timing when I, If, and N match the photographic field of view of the television camera 3.

The image processing device 4 has a terminal 5 (keyboard, CRT
), an image showing the position where each harness component is normally occupied and the regular shape of each harness component is registered in correspondence with each imaging region. Further, a monitor 6 is connected to the image processing device 4.

The image processing device 4 then processes each harness component 1 of the wiring/eightness 1 to be inspected based on the captured image signal.
By detecting the position and shape of each harness component 1b and comparing the detected actual position and shape with the registered image, it is determined whether each harness component 1b is correctly attached.

D. Problems to be Solved by the Invention By the way, among the methods for inspecting parts missing in wiring and eight-ness, visual inspection has poor inspection accuracy and causes severe worker fatigue.

On the other hand, in visual inspection using conventional image processing, all eight
Since the regular position and shape of the nest part 1b must be registered as an image, it is necessary to register a huge amount of data, and it takes time to register. In addition, it takes time to inspect the position and shape of the eight-ness part 1b of the wiring/eightness part 1 to be inspected based on the image signal.
Furthermore, it is difficult to accurately detect the position and shape. For these reasons, testing cannot be carried out easily or quickly.

In view of the above-mentioned prior art, it is an object of the present invention to provide a defective product determination image processing device that can automatically and accurately perform a defect inspection simply, quickly, and without visual inspection.

E. Means for Solving the Problems The defective product determination image processing device according to the first invention includes a television camera for photographing a workpiece, a device for switching and outputting a teaching command and an inspection command, and an image signal input from the television camera. and an image processing device that determines whether or not the workpiece is defective, the image processing device including a binarization processing unit that divides the image signal of the workpiece from the television camera into pixels of the back shadow and pixels of the workpiece. , a counting unit that counts the number of pixels of the workpiece, a memory that stores judgment values, and a judgment value that stores the count value of the pixels of the workpiece obtained by the counting unit in the memory when an inspection command is output. A determination unit that compares the count value and determines that the workpiece is good if it is within the judgment value, and determines that the workpiece is defective if it is outside the judgment value, and outputs an inspection command and determines that the workpiece is good. In this case, and when a teaching command is output, the count value of the pixels of the workpiece obtained by the counting section is input, the average value μ and standard deviation σ are calculated, and μ±3σ is output to the memory as the judgment value. It has a statistical processing section.

Further, the defective product determination image processing device according to the second invention includes a television camera that photographs the workpiece, a device that switches and outputs teaching commands and inspection commands, and an image processing device that inputs image signals from the television camera and determines whether the workpiece is defective or not. an image processing device that determines the number of pixels of the workpiece; the image processing device includes a binarization processing unit that divides the image signal of the workpiece from the television camera into pixels of the back shadow and pixels of the workpiece; and a binarization processing unit that counts the number of pixels of the workpiece. When the inspection command has been output, the count unit that stores the reference value and the count value of the pixel of the workpiece obtained by the count unit are compared with the reference value stored in the memory when the inspection command is output, and the difference is less than or equal to the set value. A determination unit that determines that the workpiece is good if the value is greater than a set value, and that the workpiece is defective if it is greater than a set value, and a teaching command that outputs an inspection command and determines that the workpiece is good. is output, the statistical processing unit inputs the count value of the pixels of the workpiece obtained by the counting unit, calculates the average value, and outputs the average value to the memory as a reference value. be.

F. Function: Since the area of a good workpiece and a defective workpiece with missing parts is different, if they are photographed with the same television camera and the same optic kidney, the number of pixels of the workpiece obtained by binarizing the image signal will exceed the error range. It makes a difference.

Therefore, in the device of the first invention, as a teaching step, multiple workpieces that have been separately confirmed to be good are photographed in advance with a television camera, the image signals are binarized, the number of pixels of the good workpieces is counted, and the number of pixels of the good workpieces is counted. The values are statistically processed to obtain the average value μ and standard deviation σ, and μ±3σ is stored as a judgment value. Next, during inspection, the work to be inspected is photographed using the same television camera, the image signal is binarized, and the number of pixels of the work to be inspected is counted. The number of pixels of the inspection workpiece is compared with a determination value, and if it deviates from the determination value, the inspection workpiece is determined to be defective. Also, if the inspected workpiece is good,
The count value of that pixel is used to calculate the average value μ and standard deviation σ.

In this way, by statistically processing the count value of only good workpieces and setting the judgment value μ±3σ, it is possible to compare and There is no need to prepare samples of defective products (this is advantageous. For example, when there are many inspection points such as the appearance inspection of a wire harness, it is troublesome to intentionally cause a defective state at each inspection point. Moreover, it is wasteful.Furthermore, since the defective sample is measured, the number of samples is large, and it takes a long time to determine the judgment value.

Furthermore, as the number of inspections increases, the number of non-defective workpieces used to calculate the average value μ and standard deviation σ increases, so the judgment μ ± 3σ
is highly accurate by absorbing variations in the workpiece, etc., and stable judgments can be made. For example, even if the lighting gradually becomes darker, the determination will be stable.

On the other hand, in the device of the second invention, first, as a teaching,
M number of non-defective workpieces are photographed in advance with a TV camera, the image signals are binarized, and the number of pixels of the non-defective workpieces is counted.
The count values are statistically processed to obtain an average value, and the average value is stored as a reference value. Next, during inspection, the work to be inspected is photographed using the same television camera, the image signal is binarized, and the number of pixels of the work to be inspected is counted. The reference value and the number of pixels of the inspection work are compared, and if the difference is greater than a set value, the inspection work is determined to be defective. In addition, if the inspection workpiece is a non-defective item, the count value of that pixel is used to calculate the average value.

In this case as well, since the reference value is determined by statistical processing, variations in non-defective workpieces have little effect on the determination results. Further, since the allowable range of the difference is determined by the set value, the determination accuracy can be freely determined based on the set value.

Further, as in the first invention, as the number of inspections increases, the number of non-defective workpieces used to calculate the average value increases, so the reference value becomes highly accurate by absorbing variations in the workpieces, etc., and stable determination can be made.

G. Examples Hereinafter, the present invention will be explained in detail based on the drawings. Fig. 1 is a block diagram showing the system system of the embodiment of the present invention, Fig. 2 is a plan view showing the conveyance line of the drawing board, Fig. 3 is a block diagram of the drawing board, and Fig. 4 shows the timing of operation. 5 is a flowchart of teaching, FIG. 6 is a flowchart of statistical processing, and FIG. 7 is a flowchart of inspection.

As shown in FIG. 3, the drawing board 10 is installed obliquely on a truck 11, and the truck 11 runs along a guide rail 12 by a power source (not shown). The drawing board 10 is white. 10a is a support member for the harness, and 11a is a wheel.

Therefore, as shown in FIG. 2, the drawing board 10 is sent in the direction of the arrow B and circulates through the assembly line 13 and the inspection line 15 at a line speed V.

Then, while the drawing board 10 moves along the assembly line 13, a wiring harness 14 (see FIG. 2) is assembled on the drawing board 10. 14a is, for example, a black converging wire, 14
b is a harness component in a color other than white, such as blue or red.

The assembled wiring harness 14 is inspected for missing parts while moving along the inspection line 15. A specific explanation of the test will be given later. The wiring eights that have passed the inspection are removed from the drawing board 10, and the drawing board 10 is returned to the assembly line 13 again. Wiring harnesses determined to be defective are sent to the assembly line 13 while attached to the drawing board 10.
It will be returned to and revised.

Also, on the drawing board 10, as a pre-processing (teaching) for inspection, a good workpiece, that is, a required harness part 14 is shown.
The wiring harness G, which has already been visually confirmed to be correctly attached to the converging wires 14a, is attached. For statistical processing, a plurality of different good workpieces G are used.

In this embodiment, four industrial television cameras TVI are used to increase the resolution according to the size of the workpiece.

TV2. TV3. Iteiro for TV4. As shown in FIGS. 1 to 3, a support member 16 is installed obliquely above the drawing board 10 in the inspection line 15, and each television camera TVI, TV2. TV3. Install TV4,
It is directly facing the figure [10]. The imaging field of each television camera is adjusted according to the inspection accuracy, but since it is set to be significantly narrower than the drawing board 10, the drawing board 10 can be moved vertically by shifting it in the direction C perpendicular to the drawing direction. I try to cover it. Also, so as to cover the drawing board 10 laterally,
Marks 17 are provided on the drawing board 10 at intervals in the horizontal direction, and the image processing device 20 captures the image signal of the television camera every time the marks 17 are detected by a photoelectric sensor or the like provided in the inspection line 15. be. Furthermore, four television cameras TVI are installed so that image processing is performed for each television camera in turn.
~ TV4 is shifted in the drawing board traveling direction B, and a sensor is provided corresponding to each TV camera, and four sensors PH1, PH2,
PH3 and PH4 are also shifted in the same way. Further, a limit switch LS is provided in the inspection line 15 to detect when the drawing board 10 reaches the inspection line 15 and give a start command to the image processing device 20.

As a result of the above, the numbers on the diagram board 10 are ■, ■, ■, ■.

■... As shown in FIG. 4, each time the sensors PH1 to PH4 sequentially detect the mark 17, the photographing areas are divided into ■, ■, ■.

Image processing for each imaging area is performed in the order of (1), (2), and so on.

Note that 18 is an illumination light source such as a fluorescent lamp, and BS is an emergency stop button switch. Further, the inspection line 15 is covered with a blackout curtain 19 to block external light.

As shown in FIG. 1, the image processing device 20 receives analog image signals 3, S2,
33. S4 is sent, and mark detection signals Ml, M2 . M3. M4, a start command signal 21 from the limit switch LS, an emergency stop signal 22 from the button switch BS, and a command signal 23 from the switch SW: the teaching command and the inspection command are switched and sent (respectively).

The image processing device 20 has a binarization processing section, a counting section, a statistical processing section, a memory, and a determination section, and determines the wiring of a good product as shown in FIGS. 5 and 6 by operating a switch SW.・Teaching based on the 8-nes G and inspection of the wiring harness 14 to be inspected shown in FIG. 7 are performed.

Each functional section of the image processing device 20 is realized by a microcomputer and appropriate hardware and software.

The binarization processing unit converts the input image signals 81 to S4 into a black pixel signal indicating the wiring/eighthness G, 14 and a white pixel signal indicating the drawing board 10 each time the marks of the sensors P) Il to PH4 are detected. It is divided, binarized, and sent to the counting section. Note that, in order to shorten the processing time, immediately after capturing the image signal of one television camera, preparations are made to capture the image signal of the next television camera. In addition, in order to adjust the area to be binarized, targets 24 are placed at intervals in the horizontal direction on the drawing board 10.
is provided, and correction is performed so that only the image signal in a predetermined area with the target 24 as a reference is binarized.

The counting section calculates the number of pixels of the black pixel signal in the binarized image signal for each imaging area (■), (2).

(2) Count each time, and send the count value to the statistical processing unit, memory, and determination unit.

When a teaching command is given from the switch SW, the memory saves the count value from the counting section to an address corresponding to the imaging area for statistical processing, and the switch SW outputs an inspection command. In this case, if the determining section determines that the inspection work is a good product, the count value from the counting section is saved in an address corresponding to the imaging area for statistical processing. Furthermore, the memory saves the calculation results of the statistical processing section at an address corresponding to the imaging area.

As an embodiment of the first invention, the statistical processing section receives a teaching command from the switch SW, and the statistical processing section receives a teaching command from the switch SW.
Even when the W outputs an inspection command or the judgment unit judges the inspection workpiece to be non-defective, the count value is input from the memory for each imaging area, and the average value μ and standard deviation shown in Fig. 6 are calculated. (Dispersion) From the calculation of σ, a judgment value μ±3σ is calculated and stored in the memory. However, n is the number of samples, X, , is n
Count value of the th sample, μ. is the previous average value, μ
. 1 is the current average value, and for the first time, the count value is used as the average value. Also, σ: is the previous variance, and σ: is the current variance.

In the second invention, the statistical processing section only calculates the average value μ shown in FIG. 6, and stores μ in the memory as a reference value.

As an embodiment of the first invention, the determination section compares the count value from the counting section with the determination value μ±3σ stored in the memory for each imaging region when an inspection command is given from the switch SW, and As a result, if any region deviates from the determination value, it is determined that the eight-ness component is missing or redundant, and a determination signal 25 indicating "defective" is output. Furthermore, if the count value is within the judgment value μ±3σ,
A signal indicating "good product" is given to the memory and statistical processing section.

In the second invention, when receiving an inspection command from the switch SW, the determining section compares the count value from the counting section and the reference value μ stored in the memory for each imaging region, and as a result of the comparison, If the difference between the two in any area is larger than the set value (this is determined in advance according to variations in the size of the harness parts, judgment accuracy, etc.), it is determined that the harness parts are missing or are attached in excess. If the difference is less than or equal to the set value, the determining section supplies a signal indicating "good product" to the memory and statistical processing section.

For this reason, in teaching, we use multiple samples of good quality products and mark the wiring of good quality products on each drawing board 10.
By setting the harness G and sequentially sending it to the inspection line 15, the image processing device 20 captures the image signals of the television cameras TVI to TV4 every time the marks of the sensors PHI to PH4 are detected, and counts the number of black pixels. The count value is taken as the area of each imaging region of the wiring harness G, and the judgment value μ±3σ or the reference value μ is taught.

Next, in the inspection, the image processing device 20 sends the wiring harness 14 assembled on the drawing board 10 to the inspection line 15 without changing the field of view of each of the television cameras TVI to TV4. Every time a mark is detected, the image signal is captured and the number of black pixels is counted, and the judgment value μ of the number of black pixels is taught for each shooting area.
It is determined whether or not it is defective by comparing it with ±3σ or the reference value μ.

In the case of a non-defective product, the count value of the number of black pixels of the workpiece is taken in as data for statistical processing to determine the judgment value μ±3σ or the reference value μ.

If it is defective, the worker will repair it.

Note that the image processing device 20 has a switch SW on the keyboard.
A personal computer 26 having an image processing device 2 is connected to the image processing device 2.
Used for 0 operation.

Further, when it is determined that a component is missing, the determination result is displayed on the CR7 display device 27. Further, a monitor 28 is connected to the image processing device 20 for adjusting image processing, and images taken by each of the television cameras TVI to TV4 are displayed on the monitor 28.
In addition to video recording equipment (VTR)
29 so that the details of the inspection can be investigated later. In FIG. 1, 30 is a power supply unit which includes a camera power supply and a high frequency power supply for fluorescent lamps.

After all, in this embodiment, a plurality of good quality wiring harnesses G with no missing parts are used in advance, and a judgment value or reference value is obtained by statistical processing from the number of black pixels of each good quality product, and these values are used as the inspection target. The presence or absence of missing parts is determined by a simple area comparison using binarization, in which the number of black pixels representing the wiring harness 14 is compared. Therefore, the determination accuracy is high. In addition, the count value of the number of black pixels of a workpiece that is determined to be good through inspection is also incorporated into the statistical processing to determine the judgment value or reference value, so as the number of inspections increases, product variations are absorbed and the judgment value or reference value increases. Accuracy becomes higher and judgment becomes more stable. For example, even if the lighting gradually becomes darker, the determination will be stable. Furthermore, since the image processing device 20 only performs simple processing such as binarization, counting, statistical processing, and comparison, the processing speed is extremely fast, so the moving speed of the drawing board 10 can be increased, and the inspection efficiency can be increased. Tests can be performed at online speeds.

Note that the present invention can be applied not only to wiring harnesses but also to inspecting other inspection workpieces for missing parts.

In this case, no matter how many shapes there are in the inspection work or even if it is difficult to detect individual parts, there is no problem at all because the missing parts are determined by image comparison. Further, the field of view of the television camera may be adjusted according to the inspection accuracy, and therefore one to many television cameras may be used depending on the inspection accuracy and the size of the workpiece.

■ Effects of the Invention As specifically explained above in conjunction with the embodiments, according to the present invention, the image signal obtained by photographing the workpiece with a television camera is binarized, and the judgment is made by statistical processing based on the number of pixels indicated by a good workpiece. Values or reference values are determined, and these values are compared with the number of pixels indicated by the inspection workpiece to determine whether there is a missing part. Since the judgment value or reference value is determined by statistical processing in this way, the error is small. Furthermore, since the determination is based on a simple area comparison, there is no need to teach the position, dimensions, etc. of the workpiece parts, and inspection can be performed through quick and simple processing. Furthermore, when the type of work to be inspected is changed, it is sufficient to simply designate the type of work that has been confirmed to be non-defective as the non-defective work, and there is no need to change the inspection program.

[Brief explanation of the drawing]

FIG. 1 is a system diagram showing a system system according to an embodiment of the present invention, FIG. 2 is a plan view showing a drawing board conveyance line, FIG. 3 is a configuration diagram showing the drawing board conveying state, and FIG. 4 is a diagram showing the operation timing, Figure 5 is a teaching flowchart,
FIG. 6 is a flowchart of statistical processing, FIG. 7 is a flowchart of inspection, and FIG. 8 is a system diagram showing a conventional technique. In the drawing, 10 is a drawing board, 14 is a wiring harness as an inspection work, G
is the wiring harness as a good workpiece, SW is the teaching and inspection command changeover switch, TVI, T
V2. TV3. TV4 (15Lk' force 15.20 is an image processing device. Fig. 3 Structure of drawing board conveyance Funeral diagram 6 Fig. Flowchart of statistical processing Q Two funeral diagrams - Ching flow Funeral diagram Prior art)

Claims (2)

[Claims] (1) Equipped with a television camera that photographs the workpiece, a device that switches and outputs teaching commands and inspection commands, and an image processing device that inputs image signals from the television camera and determines whether or not the workpiece is defective. , the image processing device includes a binarization processing unit that divides the image signal of the workpiece from the television camera into pixels of the back shadow and pixels of the workpiece, a counting unit that counts the number of pixels of the workpiece, and a determination value that is stored. When the memory and the inspection command are output, the count value of the pixels of the workpiece obtained by the counting section is compared with the judgment value stored in the memory, and if the count value is within the judgment value, the workpiece is determined to be good. There is a judgment unit that determines that the workpiece is defective if the judgment value is exceeded, and a count unit that determines the workpiece is defective if the inspection command is output and the workpiece is determined to be non-defective. A defective product determination image processing device comprising: a statistical processing unit that inputs count values of pixels of a workpiece, calculates an average value μ and standard deviation σ, and outputs μ±3σ to a memory as a determination value. . (2) Equipped with a television camera that photographs the workpiece, a device that switches and outputs teaching commands and inspection commands, and an image processing device that inputs image signals from the television camera and determines whether or not the workpiece is defective. , the image processing device includes a binarization processing unit that divides the image signal of the workpiece from the television camera into pixels of the back shadow and pixels of the workpiece, a counting unit that counts the number of pixels of the workpiece, and a reference value. Compares the reference value stored in the memory when an inspection command is output to the memory and the count value of the pixel of the workpiece obtained by the counting section, and determines that the workpiece is good if the difference is less than the set value. There is a determination section that determines that the workpiece is defective if it is larger than a set value, and a counting section that determines that the workpiece is defective if the value is larger than the set value, and a counting section that determines that the workpiece is defective if the inspection command is output and the workpiece is determined to be good. What is claimed is: 1. A defective product determination image processing device comprising: a statistical processing unit that inputs count values of pixels of a workpiece obtained in step 1, calculates an average value, and outputs the average value to a memory as a reference value.