JP5798371B2 - How to create a fiducial mark model template - Google Patents

Description

  The present invention relates to an electronic circuit component mounting machine for mounting an electronic circuit component on a circuit substrate such as a printed circuit board, a screen printing machine for printing cream solder for soldering the electronic circuit component on the circuit substrate, and an electronic circuit component. For circuit substrate working machines that perform prescribed work on circuit substrates, such as adhesive application machines that apply adhesives to circuit substrates, electronic circuit inspection machines that inspect electronic circuits, etc. On the other hand, the present invention relates to a technique for meeting the demands that have become increasingly severe in recent years.

In Patent Document 1 below, a reference mark provided on an object such as a circuit substrate is imaged by an imaging device, and an image acquired by the imaging is compared with a plurality of model templates stored in advance in a storage unit. By doing so, it is described that the model template corresponding to the read image is selected. The reference mark is recognized by template matching between the selected model template and an image acquired by the imaging device from the sequentially loaded object, and the position of the reference mark is detected. Instead of storing the model template itself in the template memory, only the feature data of the model template may be stored and the model template may be created from the feature data. If the size of the model template is different from the actual image, Use an adjusted model template size as a model template, and if there is no model template suitable for the template memory, newly register image data acquired by the imaging device as optional data Is also described as a model template.
However, in recent years, there has been an increasing demand for miniaturization of electronic circuits, and there is no description of measures to be taken when there are a plurality of items that are likely to be reference marks in the screen of the imaging apparatus.

  Further, as the demand for electronic circuit miniaturization becomes stronger, miniaturization of circuit substrates and electronic circuit components has progressed, and it has become difficult to properly attach electronic circuit components to circuit substrates. On the other hand, in Patent Documents 2 and 3 below, an electronic circuit component is picked up and taken out from a component supply device by a suction nozzle, and a state where the electronic circuit component is attached to a circuit substrate is picked up by an image pickup device and stored as a moving image. It is described that when a mounting mistake occurs, it is useful for investigating the cause. However, storage of moving image data requires a large-capacity storage device, and it is not easy to select data useful for elucidating the cause of the mounting error from a large amount of moving image data.

  Furthermore, if the cause of mounting errors can be clarified and the control program is changed in order to prevent it, the new control program is tested to determine whether it is appropriate before the production site. However, the problems of the new control program became clear after the production of electronic circuits was started, or the problems that had existed before by adopting the new control program May manifest. Conventionally, measures have been taken after the fact has been clarified, which may interfere with the production of electronic circuits.

JP 2003-317097 A JP 2008-098411 A JP 2008-103426 A

  The present invention has been made as an object to solve at least one of the above problems in a circuit substrate working machine, and the invention (hereinafter referred to as "hereinafter," The claimable invention includes “the invention of the present application” which is the invention described in the claims, but the subordinate concept invention of the present invention, the superordinate conception of the present invention, or the invention of another concept. Some of the embodiments are described and explained below. As with the claims, each aspect is divided into sections, each section is numbered, and is described in a form that cites the numbers of other sections as necessary. This is for the purpose of facilitating the understanding of the claimable invention, and is not intended to limit the combinations of the constituent elements constituting the claimable invention to those described in the following sections. In other words, the claimable invention should be construed in consideration of the description accompanying each section, the description of the embodiments, the prior art, and the like. The added aspect and the aspect in which the constituent elements are deleted from the aspect of each item can be an aspect of the claimable invention.

Aspects of the Invention

  In each of the following terms, (1) corresponds to claim 1, (2) corresponds to claim 2, (4) corresponds to claim 3, (5) corresponds to claim 4, (6) corresponds to claim 5, (7) corresponds to claim 6, (8) corresponds to claim 7, and (9) corresponds to claim 8.

(1) an imaging step of imaging the reference mark formation scheduled position of the circuit base material and its periphery by an imaging device;
A reference mark candidate extraction step of extracting a reference mark candidate that is an image that is highly likely to be an image of a reference mark from images obtained by performing the imaging step;
A mark type determination step for determining which of the plurality of types of reference marks that are extracted from the extracted reference mark candidates;
And a template creation step for creating a model template of a reference mark whose type has been clarified through the execution of the mark type discrimination step, and the reference mark candidate extraction step is a hierarchical Adaboost using Haar-like features (AdaBoost) A method of creating a reference mark model template, which is a step of extracting reference mark candidates using a detector.
(2) The method according to (1), including a dimension measuring step of measuring a dimension of an image whose type has been found by performing the mark type determining step, and creating a model template of the measured dimension in the template creating step. How to create a standard mark model template.
(3) In the reference mark candidate extraction step, a scan window having a predetermined size is set in a processed image that is an image acquired in the imaging step, and an image in the scan window is the reference mark candidate. The step (1) or (2) includes a step of determining whether or not the image is a single image by moving the scan window by a predetermined amount along a predetermined path. The reference mark model template creation method described in the section.
(4) Each time the reference mark candidate extracting step sets the scan window to a size suitable for detection of the smallest one of those initially scheduled as the reference mark, and performs one scan. The reference mark model template creation method according to item (3), further including a step of enlarging the scan window by a set amount.
If there is only one type of reference mark planned, the scan window may be set to a size suitable for extracting the reference mark of that one type. When the reference mark is scheduled, it is desirable to change the size of the scan window into a plurality of types and perform scanning for each size of the scan window.
(5) The model template creation method according to any one of (1) to (4), wherein the mark type discrimination step includes a step of discriminating a type of reference mark candidate using a neural network.
(6) As the neural network, an input layer composed of N elements × M elements, an output layer composed of a number of elements obtained by adding 1 to the predetermined number of reference marks, and between the input layer and the output layer And an intermediate layer composed of a plurality of elements provided, and the teacher data for a mark obtained by normalizing an image of the planned reference mark to N pixels × M pixels, and an image of a reference mark that is not a reference mark is N pixels × M The reference mark model template creation method according to the item (5), wherein learning is performed using non-mark teacher data normalized to pixels.
(7) When a plurality of model templates are created for one reference mark formation scheduled position by executing the reference mark candidate extraction step, the mark type determination step, and the dimension measurement step, the plurality of model templates The method of creating a reference mark model template according to any one of items (1) to (6), including a final model template determination step of determining one of them as a final model template.
(8) The final model template determination step includes:
A plurality of model template notification steps for notifying that one of the plurality of model templates should be selected and the final model template should be designated when a plurality of model templates are created for one reference mark formation scheduled position When,
The reference mark model template creation method according to item (7), further including a final model template designation input step in which an operator designates a final model template in accordance with execution of the model template multiple notification step.
(9) The step of determining the final model template includes a step of automatically determining a model template that best matches a predetermined condition among the plurality of model templates as the final model template (8) The method for creating the fiducial mark model template described in item).
The predetermined condition may be, for example, an image closest to the planned formation position of the reference mark or an image having the highest matching degree with any of a plurality of types of reference marks scheduled. it can.
(10) Notifying that if no model template has been created for one reference mark formation scheduled position by executing the reference mark candidate extraction step, the mark type discrimination step, and the template creation step The reference mark model template creation method according to any one of items (1) to (9), including a model template creation failure notification step.
In the case of this section, there is a high possibility that there is some problem, so the operator is required to search for the cause and take necessary measures.
(11) an imaging device for imaging the reference mark formation planned position of the circuit base material and its periphery;
A reference mark candidate extraction unit that extracts a reference mark candidate that is an image that is likely to be an image of a reference mark from images captured by the imaging device;
A mark type discriminating unit for discriminating which of the plurality of types of fiducial mark candidates the extracted fiducial mark candidates are;
A template creation unit that creates a model template of a reference mark whose type has been clarified by the mark type discrimination unit, and the reference mark candidate extraction unit includes a hierarchical type AdaBoost using a Haar-like feature. ) A reference mark model template creation device including a detector.
The features described in the items (2) to (10) are also applicable to the reference mark model template creation device according to this item.
(12) a component supply device for supplying electronic circuit components;
A substrate holding device for holding a circuit substrate on which electronic circuit components are to be mounted;
A mounting device that receives an electronic circuit component from the component supply device and mounts the electronic circuit component on a circuit base material held by the base material holding device; a component receiving unit that receives the electronic circuit component from the component supply device; and an electronic circuit component In an electronic circuit component mounting machine including a moving image capturing device that captures at least one moving image of mounting a component on the circuit substrate, and a method of storing a moving image captured by the moving image capturing device,
A temporary storage step of temporarily storing video data that is video data captured by the video imaging device;
An electronic circuit comprising: a normal storage step for properly storing the moving image data stored in the temporary storage step when at least one of the component reception and the component mounting is not normally performed. Video data storage method for component mounting machines.
(13) In the temporary storage step, the moving image data is stored in a buffer memory, and the moving image data stored in the buffer memory is deleted when at least one of the component reception and the component mounting is normally performed. The moving image data storage method in the electronic circuit component mounting machine according to (12), wherein, when at least one of them is not performed normally, the moving image data is transferred from the buffer memory to the data collection memory in a normal storage step.
(14) Of the at least one of the component receiving and the component mounting, the temporary storage step is executed for those satisfying a predetermined condition, and the temporary storage is performed for those not satisfying the predetermined condition The moving image data storage method according to item (12) or (13), wherein the step is not executed.
The “predetermined condition” is, for example, that (a) the type of electronic circuit component (including at least one of shape, size, and model) is predetermined, and (b) the component holder ( At least one of the kind (including at least one of the size and type) and the identification code is predetermined, and (c) at least one of the feeder and the tray ) Including at least one of the types (including at least one of the dimensions and model) and the identification code. In addition, the “predetermined condition” may not be normally performed normally even if it is set automatically or by an operator based on the fact that component reception or component mounting was not actually performed normally. It may be set automatically or by the operator based on the fact that the property is high.
(15) In order to control the circuit base material work system based on information supplied from a detection device provided in the counter circuit base material work system that performs work on the circuit base material, the circuit base material work system A method for updating software used in a production computer,
Software that is to be updated based on information supplied from the detection device by operating the actual computer with information from the detection device and old software that is software before update Circuit base work system, characterized in that a new computer software is executed by a test computer, and after confirming that there is no problem with the new software, the old software is updated to the new software. Software update method.
The execution of the new software by the test computer can be performed in parallel with the operation of the circuit substrate work system in conjunction with the operation of the production computer, or can be performed separately from the operation of the circuit substrate work system. Is possible. In the former case, a production computer and a test computer are connected in parallel to the detection device. In the latter case, information from the detection device or control device is temporarily stored in the storage device, and the stored information is stored in the storage device. And the test computer is activated based on the new software. In the former case, there is an advantage that the new software can be evaluated in parallel with the operation of the circuit board working system. In the latter case, for example, the operation is performed while the circuit board working system is stopped. Since the new software can be evaluated using the computer itself as a test computer, there is an advantage that problems after the software update can be avoided more reliably.
The software includes a program for controlling the computer and data used when the program is executed. In addition, the above “confirmation that there is no problem with the new software” can be performed by a person or automatically by a test computer using the confirmation software. It is. In the former case, for example, the operation result of the test computer is displayed on a display device or recorded on a recording sheet, and a person performs an evaluation based on the display or recording.
(16) Confirming that there is no problem with the new software by comparing the operation result of the production computer and the operation result of the test computer. Update method.
When running the new software by the test computer separately from the operation of the circuit substrate work system, the operation result of the production computer can be stored in the storage device or a hard copy indicating the operation result can be left. That's fine.
(33) The detection device includes an imaging device, the actual computer includes an image processing unit that processes data of an image captured by the imaging device, and the old software and the new software are the image processing unit. The software update method for a circuit board work system according to item (31) or (32), including image processing software used by the computer.

1 is a front view showing an external appearance of an electric circuit assembly system, which is an embodiment of the claimable invention and is an apparatus in which an embodiment method is implemented. It is a side view which shows the structure of the screen printing machine which is a component of the said electric circuit assembly system by removing a cover. It is a top view which takes out and shows the reference mark imaging system of the said screen printer. It is a top view which shows an example of the screen used for the said screen printer. It is a top view which shows an example of the printed circuit board by which cream-like solder is printed by the screen printing machine. It is a perspective view which removes a part of cover and shows the structure of the mounting module which is a component of the said electric circuit assembly system. It is a perspective view which takes out and shows the mounting apparatus in the said mounting module. It is a perspective view which shows an example of the mounting head which is a component of the said mounting apparatus. It is a perspective view which shows another example of the mounting head which is a component of the said mounting apparatus. It is a perspective view which removes a cover and shows the mounting head of another example of the above. It is a perspective view which shows the adhesive application head mounted in a mounting module selectively with a mounting head. It is side surface sectional drawing which shows the principal part of the said adhesive agent coating head. It is a block diagram which shows the control apparatus of the said screen printer. It is a block diagram which shows the control apparatus of the said mounting module. It is a figure which shows an example of the process image imaged with the reference mark imaging device. It is a flowchart which shows an example of a reference mark model template creation program. It is a figure which shows the reference mark currently scheduled to be used in the said electric circuit assembly system. A hierarchical Adaboost detector as a part of a model template creation apparatus configured by a part for storing the reference mark model template creation program and a part for executing the program of the image processing computer of the screen printing machine is conceptually shown. FIG. It is a figure which shows an example of the setting of the characteristic to the scan window in the said hierarchical type | mold Adaboost detector. It is a figure which shows the characteristic used in the said hierarchical Adaboost detector. It is a figure showing the teacher data for the bright color mark used for the learning of the said hierarchical type Adaboost detector. It is a figure showing the teacher data for dark mark used for learning of the above-mentioned hierarchical type Adaboost detector. It is a figure showing some examples of teacher data for non-marks used for learning of the above-mentioned hierarchical type Adaboost detector. Concept of a mark type discriminator as another part of a model template creation apparatus constituted by a part for storing the reference mark model template creation program and a part for executing the program of the image processing computer of the screen printing machine FIG. It is a graph which shows the kind of reference mark output to each element of the output layer of the said mark type discriminator. It is a figure for demonstrating learning of the said mark type discriminator. It is another figure for demonstrating learning of the said mark type discriminator. It is another figure for demonstrating learning of the said mark type discriminator. It is a perspective view which shows the state with which the CCD camera was attached to the mounting module of the said electric circuit assembly system in order to investigate the cause of component adsorption | suction and mounting mistake. It is a block diagram which shows the image data collection control part connected to the control apparatus of a mounting module with the attachment of the said CCD camera. It is a flowchart showing the adsorption | suction and mounting | wearing mistake image collection program stored in ROM of the image data collection computer of the said image data collection control part. It is a flowchart showing the new adsorption | suction / mounting mistake image collection program different from the said adsorption | suction / mounting mistake image collection program. FIG. 31 is a block diagram showing a state in which a test computer for testing the validity of the new adsorption / mounting mistake image collection program is connected in parallel to the image data collection computer of the image data collection control unit of FIG. 30.

  Hereinafter, an embodiment of the claimable invention will be described with reference to the drawings. In addition to the following examples, the claimable invention can be practiced in various modifications based on the knowledge of those skilled in the art, including the aspects described in the above [Aspect of the Invention] section.

  FIG. 1 shows an example of an electric circuit assembly system (hereinafter abbreviated as an assembly system). The assembly system includes a substrate supply device 2, a screen printing machine 4, a shift conveyor 6, an adhesive applicator 8, a component mounting line 10, an adhesive curing device 12, and a reflow furnace 14 arranged in a line from the upstream side. . The component mounting line 10 includes a plurality of mounting modules 16 arranged close to each other. A plurality of mounting modules 16 are arranged on a base 18 so that an arbitrary one can be pulled forward. Yes. The adhesive applicator 8 can be configured as a dedicated device, but in the present embodiment, as will be described later, a part of the mounting module 16 (one unit in the illustrated example) is used. It has been done. The substrate supply device 2, the screen printing machine 4, the shift conveyor 6, the adhesive applicator 8, the mounting module 16, the adhesive curing device 12, the reflow furnace 14 and the like each have a control device mainly composed of a control computer. These control computers are connected to each other and to the host computer 20.

  As shown in FIG. 2, the screen printing machine 4 uses a screen 26 to print cream-like solder on a printed circuit board 28, which is a kind of circuit substrate, and includes a frame 30 as a printing machine body, a printing machine. A substrate transfer device 32, a substrate holding device 34, a screen holding device 36, a squeegee device 38, a mark imaging system 40 shown in FIG. 3, and a control device 218 described later (see FIG. 13) are included. As shown in FIG. 4, the screen 26 is stretched and fixed to the screen frame 46 without slack, and is provided with a plurality of printing through holes 48 penetrating in the thickness direction. The screen 26 is also provided with a plurality of, for example, two fiducial marks 50 at two locations separated diagonally.

  As schematically shown in FIG. 5, a plurality of lands 52 are formed on the printed circuit board 28, and the plurality of through holes 48 are formed at locations corresponding to the lands 52 of the screen 26. Corresponding shape and dimensions. A plurality of, for example, two fiducial marks 54 are also provided on the printed circuit board 28 at positions corresponding to the fiducial marks 50 that are diagonally separated.

  The printing machine substrate transport device 32 is constituted by, for example, a belt conveyor which is a kind of conveyor, and supports the printed circuit board 28 in a horizontal posture at both side edges thereof, and is in the X-axis direction. Transport in a direction perpendicular to The substrate holding device 34 is provided in the middle of the conveyance path of the printing machine substrate conveyance device 32, holds the printed circuit board 28, and raises it from the belt conveyor. The screen holding device 36 is provided above the substrate holding device 34 of the frame 30. The screen 26 is positioned by the positioning device 62 with respect to the screen support base 60, fixed by the fixing device 64, and held in a horizontal posture. To do.

  The squeegee device 38 includes a pair of squeegee heads 66 as work heads, a squeegee head moving device 68 as a work head moving device, and a pair of squeegee head lifting devices 70. The squeegee head moving device 68 includes a moving member 72 and a moving member moving device 74. The moving member moving device 74 includes a squeegee moving motor 76 that is a driving source, and a feed screw mechanism 82 that is a motion converting mechanism including a ball screw 78 and a nut 80. Each of the pair of squeegee heads 66 includes a squeegee holding member 84 and a squeegee 86. The squeegee head lifting and lowering device 70 and the pair of squeegee head lifting and lowering devices 70 are provided on the moving member 72. The screen 26 is selectively brought into contact with and separated from the screen 26, and is moved along the screen 26 in the Y-axis direction perpendicular to the X-axis direction by the squeegee head moving device 68.

  As shown in FIG. 3, the mark imaging system 40 includes a reference mark imaging device 90 and an imaging device moving device 92, and as shown in FIG. 2, between the screen holding device 36 and the substrate holding device 34 in the lowered position. It is provided in the part. The fiducial mark image pickup device 90 picks up an object to be picked up by a CCD camera, is configured in the same manner as the image pickup device described in Japanese Patent Application Laid-Open No. 2007-38456, and is provided on the printed circuit board 28 and the screen 26, respectively. The reference marks 50 and 54 can be imaged. The imaging device moving device 92 includes an X-axis direction moving device 96 and a Y-axis direction moving device 98. The X-axis direction moving device 96 includes an X-axis slide 100 and an X-axis slide moving device 102 as movable members. A Y-axis direction moving device 98 is provided on the X-axis slide 100, and a Y-axis slide as a movable member. 104 and a Y-axis slide moving device 106 are included. The reference mark imaging device 90 is provided on the Y-axis slide 104 and is moved to an arbitrary position in the horizontal plane.

  Each of the plurality of mounting modules 16 can also be referred to as an electric circuit component mounting machine, and shares the mounting of electronic circuit components on the printed circuit board 28 and performs them in parallel. Although not shown, the adhesive curing device 12 includes a curing device substrate transport device, a heating device, and a control device, and heats the printed circuit board 28 to cure the applied adhesive to transfer the electronic circuit component to the printed circuit board 28. To the reflow furnace 14. The reflow furnace 14 is connected to lead wires (not shown) formed on the printed circuit board 28 by heating and melting the cream-like solder and soldering electronic circuit components to the lands 52 of the printed circuit board 28. Is.

The mounting module 16 is described in detail, for example, in Japanese Patent Application Laid-Open No. 2004-104075, and parts other than those related to the claimable invention will be briefly described.
As shown in FIG. 6, each mounting module 16 includes a module main body 120 as a main body frame, a mounting module substrate transfer device 122, a substrate holding device 124, a component supply device 126, a mounting device 128, a reference mark imaging device 130 and components. An imaging device 132 is provided.

  The mounting module board transfer device 122 includes two-lane board conveyors 154 and 156, and transfers the printed board 28 in the horizontal direction parallel to the X-axis direction in which the plurality of mounting modules 16 are arranged. The substrate holding device 124 is provided for each of the two substrate conveyors 154, 156, and includes a support member that supports the printed circuit board 28 from below and a clamp member that clamps both side edges parallel to the conveyance direction of the printed circuit board 28. The printed circuit board 28 is held in such a posture that the mounting surface on which the electronic circuit component is mounted is horizontal.

  The component supply device 126 is provided on one side of the mounting module substrate transfer device 122 and on the front side of the mounting module 16. The component supply device 126 supplies electronic circuit components by a tape feeder (hereinafter abbreviated as a feeder) 158 which is a type of component feeder as a component supply tool, for example, and includes a plurality of feeders 158 and the feeders 158. And a feeder holding device (not shown) to which is attached. The feeder holding device includes a plurality of feeder mounting portions, and a feeder 158 is detachably mounted on the feeder mounting portions. The component supply device may be a device that supplies electronic circuit components by a tray that is a kind of component supply tool.

  The mounting device 128 includes a mounting head 160 and a head moving device 162 that moves the mounting head 160. As shown in FIG. 7, the head moving device 162 includes an X-axis direction moving device 164 and a Y-axis direction moving device 166. The Y-axis direction moving device 166 includes a linear motor 170 provided on the module main body 120 across the component supply unit of the component supply device 126 and the substrate holding device 124, and a Y-axis slide 172 as a movable member is moved in the Y-axis direction. Move to any position. The X-axis direction moving device 164 is provided on the Y-axis slide 172, is moved relative to the Y-axis slide 172 in the X-axis direction, and is relatively movable in the X-axis direction with respect to each other. 176 and an X-axis slide moving device 178 for moving the slides 174 and 176 in the X-axis direction (only a moving device for moving the first X-axis slide 174 is shown in FIG. 7). Yes. Each of the two X-axis slide moving devices includes, for example, a servo motor which is a kind of electric rotary motor as a drive source, and a feed screw mechanism including a ball screw and a nut. Move to any position in the axial direction.

  The mounting head 160 is detachably mounted on the second X-axis slide 176, and the head moving device 162 allows an arbitrary movement within a mounting work area that is a moving area across the component supply unit of the component supply apparatus 126 and the substrate holding apparatus 124. Moved to position. The mounting head 160 holds electronic circuit components by a suction nozzle 186 (186a, 186b), which is a kind of component holder. The mounting head 160 holds the suction nozzle 186 and is a nozzle holder that constitutes the holder holding portion. Plural types of mounting heads 160a and 160b having different numbers are prepared, and selectively mounted on the second X-axis slide 176 according to the type of the printed circuit board 28 on which the electronic circuit components are mounted. For example, the mounting head 160a illustrated in FIG. 8 includes one nozzle holder 188a and holds one suction nozzle 186a, and the mounting head 160b illustrated in FIG. 9 includes a plurality of nozzle holders 188b, for example, three or more (in the illustrated example, 12), and a maximum of 12 suction nozzles 186b can be held.

  In the mounting head 160a, the nozzle holder 188a is a moving device provided in the head main body, which is provided in the head main body, which is provided in the head main body so as to be movable in the axial direction and in the vertical direction and rotatable about its own axis. It is moved up and down and rotated by the lifting device and the rotating device. As shown in FIG. 10, the mounting head 160b includes a rotating body 200 provided on the head main body 196 so as to be rotatable around a vertical rotation axis, and a rotating body rotation that rotates the rotating body 200 in both forward and reverse directions at an arbitrary angle. The rotary head includes the device 202. The rotary body 200 has a nozzle holder 188b at the rotational axis of the rotary body 200 at an appropriate interval on one circumference around the rotational axis, and at 12 positions equidistant from the mounting head 160b shown in the figure. The suction nozzle is provided so as to be relatively movable in a parallel direction and to be rotatable about its own axis, and at each protruding end from the rotating body 200 (the lower end when the mounting head 160 is attached to the second X-axis slide 176). 186b is detachably held.

  The twelve nozzle holders 188b are swung around the rotation axis of the rotating body 200 by the rotation of the rotating body 200, and are sequentially moved to the component suction mounting position which is one of the twelve stop positions. It is moved up and down by a lifting device 204 which is a moving device provided at a position corresponding to the component suction mounting position 196. The nozzle holder 188b is further rotated around its own axis by a holder rotating device 206 provided in the head body 196. As shown in FIG. 7, the reference mark imaging device 130 is mounted on the second X-axis slide 176 and is moved together with the mounting head 160 by the head moving device 162.

  In the mounting module 16 configured as described above, the mounting head 160 is attachable to and detachable from the second X-axis slide 176, and can be replaced with any mounting head 160. The agent application head 210 is also replaceable. The adhesive application head 210 is provided with a syringe 212 as shown in FIG. A piston 214 is fitted in the syringe 212 so as to be airtight and slidable. When compressed air is supplied into the syringe 212, the piston 214 is lowered, and an adhesive is discharged through the discharge pipe 216 by a predetermined amount. And applied to the printed circuit board 28. Although not shown, the adhesive application head 210 is a moving device that moves the syringe 212 in the axial direction thereof, and includes an elevating device that moves the syringe 212 up and down, and moves the syringe 212 closer to and away from the circuit board. The mounting module 16 becomes an adhesive coating module by mounting the adhesive coating head 210 instead of the mounting head 160. This adhesive application module is the adhesive application machine 8.

  As is apparent from the above description, the printing press substrate transport device 32 is composed of a one-lane substrate conveyor, while the mounting module substrate transport device 122 is composed of two-lane substrate conveyors 154 and 156. For this purpose, the shift conveyor 6 is provided between the substrate transfer devices 32 and 122. Although not shown in the figure, the shift conveyor 6 is a movable conveyor that is a movable belt conveyor, a position that moves the movable conveyor in the Y-axis direction, and is connected to the substrate conveyor of one lane that is the printing machine substrate transfer device 32. The mounting module substrate transfer device 122 includes a movable conveyor shift device that selectively positions each of the two lanes of the substrate conveyors 154 and 156 in continuous connection with each other.

The screen printing machine 4, the adhesive applicator 8, the mounting module 16, the adhesive curing device 12, the reflow furnace 14 and the like each have a control device mainly composed of a control computer, but a screen printing machine deeply related to the present invention. 4 and the control device 220 of the mounting module 16 will be described representatively.
As shown in FIG. 13, the control computer 222 constituting the main body of the control device 218 includes a CPU 224, a ROM 226, a RAM 228, an I / O port 230, and a bus 232 for connecting them, and each drive is connected to the I / O port 230. Via the circuit 240, the printing press substrate transport device 32, substrate holding device 34, positioning device 62, fixing device 64, squeegee head moving device 68, squeegee head lifting device 70, etc. Actuator) is connected. Further, the host computer 20 and the reference mark imaging device 90 are connected to the I / O port 230. The fiducial mark imaging device 90 is further connected to an image processing computer 244, and the image processing computer 244 is also connected to the I / O port 230. Also connected to the I / O port 230 is a display 246 that functions as a display device for the control computer 222 and the image processing computer 244.

  As shown in FIG. 14, the control computer 248 constituting the main body of the controller 220 of the mounting module 16 includes a CPU 250, a ROM 252, a RAM 254, an I / O port 256 and a bus 260 for connecting them, and an I / O port 256. In addition to the linear motor 170, the X-axis slide moving device 178, the rotating body rotating device 202, the lifting device 204, the holder rotating device 206, the mounting module substrate transport device 122, and the shift conveyor via the drive circuits 270. 6 etc. (strictly, each actuator serving as a driving source thereof) is connected. The I / O port 256 is further connected to a feeder control computer 272 provided in each of the host computer 20 and the feeder 158, the reference mark imaging device 130, and the component imaging device 132. The reference mark imaging device 130 and the component imaging device 132 are further connected to an image processing computer 280, and the image processing computer 280 is also connected to the I / O port 256. The image processing computer 280 includes a CPU 282, a ROM 284, a RAM 286, an I / O port 288, and a bus 290.

  The ROM 226 stores a control program for controlling the screen printer 4, and the RAM 228 stores various control data. The CPU 224 controls the screen printer 4 using the control data. With this control, the printed board 28 is carried into a predetermined position in the screen printer 4 by the printing machine board conveying device 32 and held by the board holding device 34. The held fiducial mark 54 of the printed circuit board 28 and the fiducial mark 50 of the screen 26 are imaged by the fiducial mark imaging device 90, and the relative position error between the printed circuit board 28 and the screen 26 is acquired based on the imaging result. After the relative position error is eliminated by the control of the positioning device 62 and the fixing device 64, the printed circuit board 28 is brought into contact with the screen 26 by raising the substrate holding device 34, and the squeegee head 66 is moved along the screen 26. And printing is performed.

  On the other hand, the ROM 252 of the control computer 248 stores a control program for controlling the mounting module 16, and the RAM 254 stores a sequence list supplied from the host computer 20, and the CPU 250 stores the sequence list. The mounting module 16 is controlled by executing the control program using the RAM 254 according to the above. With this control, the printed circuit board 28 is carried into a predetermined position in the mounting module 16 by one of the substrate conveyors 154 and 156 and is held by the substrate holding device 124. The fiducial mark 54 of the held printed circuit board 28 is imaged by the fiducial mark imaging device 130, and the retained position error of the printed circuit board 28 in the substrate retaining device 124 is acquired based on the imaging result. On the other hand, the mounting head 160 of the mounting device 128 is moved to the component supply unit of the predetermined feeder 158 of the component supply device 126, a predetermined electronic circuit component is taken out by the predetermined suction nozzle 186, and the electronic circuit component is held by the substrate. While being moved upward of the device 124, an image is picked up by the component image pickup device 132, whereby an electronic circuit component holding position error by the suction nozzle 186 is acquired. Then, after correcting the rotation axis position and the rotation position of the mounting head 160 for eliminating the holding position error of the printed board 28 and the holding position error of the electronic circuit parts, the electronic circuit parts are printed on the printed board. At 28 predetermined positions.

The present embodiment has an important feature in the detection of the fiducial marks 50 and 54 by the fiducial mark imaging devices 90 and 130 in the above control, and this feature will be described in detail below. Since detection of the reference mark 50 is substantially the same as detection of the reference mark 54, detection of the reference mark 54 will be described as a representative.
The detection of the reference mark 54 is performed in the mounting module 16 that functions as the screen printing machine 4 and the adhesive applicator 8 and the mounting module 16 that constitutes the component mounting line 10, but these detections are substantially the same. Therefore, a description thereof will be omitted, and the detection of the reference mark 54 in the screen printing machine 4 will be representatively described below.

  The detection of the reference mark 54 is performed by template matching between the image of the reference mark 54 acquired by the reference mark imaging device 90 and a model template stored in the RAM 252 in advance. Conventionally, the creation of the model template has been performed offline by an operator. However, in this embodiment, the model is automatically modeled based on the image data of the reference mark 54 obtained by imaging by the reference mark imaging device 90. A template is created. Hereinafter, creation of the model template will be described. This creation can be performed in the control computers 222 and 248, the image processing computer 280, the host computer 20, or a personal computer connected to them, but in this embodiment, the screen printer 4 is used. This is performed in the image processing computer 244. The imaging data of the reference mark imaging device 90 is sent to the image processing computer 244, and a model template is created based on the image processing result.

  If the reference mark 54 is provided in a state of being sufficiently isolated from the land 52 of the printed circuit board 28, the printed circuit such as the lead, and the through hole, it is easy to detect the reference mark 54. However, in recent years, the demand for downsizing of the printed circuit board 28 has become stronger, and it is necessary to allow a certain amount of error in the holding position of the printed circuit board 28 by the substrate holding device 34. Since the imaging surface is often made larger, there are many cases where there is something other than the reference mark 54 in the captured image. Therefore, after the printed circuit board 28 is carried into the screen printer 4 and held by the board holding device 34, the reference mark imaging is performed on the basis of the coordinate data of the reference mark 54 in the control data in order to create a model template. When the apparatus 90 is moved to a position where it should be opposed to the reference mark 54 and an image is taken, the acquired image (displayed on the display 246 of the screen printing machine 4) is displayed as illustrated in FIG. An image other than the reference mark 54, such as a part of the printed circuit 302, may be present in the processed image 292) together with the image of the reference mark 54. In this embodiment, even under such circumstances, the reference mark 54 is obtained by executing the model creation program stored in the ROM 294 of the image processing computer 244 using the RAM 296 by the CPU 295 even in such a situation. Are automatically extracted or detected, and a model template is created based on the detection result.

  First, in S1, a scan window 304 is set as shown in FIG. The scan window 304 is initially set at a predetermined position (upper left corner in the illustrated example) of the processed image 292 with a larger set amount than the smallest of the planned reference marks 54. Then, in S2, it is checked whether or not the reference mark candidate has been extracted, that is, whether or not there is something that seems to be the reference mark in the scan window 304. A determination is made. The extraction of the reference mark candidates and the determination of the mark type will be described in detail later.

  Various reference marks 54 of the printed circuit board 28 can be used. Here, the circular, square, rhombus, inverted regular triangle, equilateral triangle, double square upper left, double square upper right and cross shown in FIG. It is assumed that the use of fiducial marks 54 having the eight types of shapes is planned. Each of the eight types of reference marks 54 includes a light color mark that exists in a dark ground and a dark color mark that exists in a light ground, for a total of 16 reference marks. The reference marks 54 that are scheduled to be used and registered in advance and are scheduled to be used are referred to as standard marks.

  Subsequent to the determination of the reference mark type, it is determined in S4 whether the reference mark candidate is one of the standard marks shown in FIG. 17, and if it is a standard mark, a diagram corresponding to each mark type is displayed in S5. Measurement of dimensions shown in FIG. 17 is performed. Whether or not the dimension measurement is successful is determined in S6. If the measurement is successful, the mark type and the dimension are stored in S7, and then S8 and subsequent steps are executed. On the other hand, when the determination results of S2, S4, and S6 are NO (mostly in this case), S3, S5, and S7 are bypassed, and S8 and subsequent steps are executed.

  In S8, it is determined whether or not the update of the position of the scan window 304 has been completed, that is, whether or not the scan window 304 has been updated to the last position of the processed image 292. If the determination result is NO, in S9 The position of the scan window 304 is changed by a predetermined amount (for example, to the right by one pixel or down by one line and to the left end) along the path indicated by the arrow in FIG. If the determination result is YES, it is determined in S10 whether or not the size of the scan window 304 has reached a predetermined maximum value. If the determination result is NO, the scan window 304 is set to a predetermined amount in S11. In addition to being enlarged (in this embodiment 1.1 times), it is returned to its initial position. After execution of S9 or S11, program execution returns to S2.

  If the scan window 304 is increased to a predetermined maximum size at the time of the determination at S10, it is determined at S12 whether or not a reference mark candidate has been extracted. If it has been extracted, it is extracted at S13. It is determined whether or not there are a plurality of reference marks. If there are a plurality of reference mark candidates, one of the plurality of reference mark candidates is selected as the reference mark 54 in S14. This selection may be performed by an operator or may be performed automatically. In the former case, for example, a plurality of reference mark candidates may be displayed on the display 246 together with a command to select one of them, and may be selected by the operator according to the command. In the latter case, for example, the closest reference mark position among the plurality of reference mark candidates (in the present embodiment, the closest to the center of the processed image 292), the most undistorted shape, etc. What satisfies the predetermined condition may be selected as the reference mark 54.

  If the reference mark candidate is not extracted at the time of the determination in S12 and the determination result is NO, the display 246 is notified in S15, and accordingly, the operator takes a necessary action in S16. Done. If the determination result in S12 is NO, even though the reference mark imaging device 90 has moved to a position where the reference mark 54 should be imaged and has taken an image, the image that appears to be a reference mark is also taken. Since such a situation is unscheduled, the cause, for example, the printed board 28 held in the board holding device 34 is different from the planned one, or the control data The existence of the fact that the coordinates of the reference mark 54 inside and the actual position of the reference mark 54 do not match, or the failure of the reference mark imaging device 90 or the imaging device moving device 92 is investigated, and the unexpected situation is resolved. The actions necessary to do it are taken.

Now, details of the reference mark candidate extraction performed in S2 will be described. In this embodiment, this extraction is performed using “(AdaBoost detector using Haar-like feature)”, which is known as a “face detection” technique in a camera or the like. 18, in the hierarchical Adaboost detector, whether the image in the scan window 304 is a reference mark image (hereinafter abbreviated as a mark image) or an image that is not a mark image (hereinafter referred to as a mark image). And a plurality of discriminators 310 connected in series in which the mark image is allowed to pass and the non-mark image is not allowed to pass. If the pass rate is Dr (0 <Dr <1) and the pass rate of the non-marked image is Fp (0 <Fp <1), the pass rate of the mark image and the non-marked image of the n-th discriminator 310 Haso Respectively (Dr) ^n, a (Fp) ^n. Thus, for example, sets the passing rate Dr of the mark image to a value sufficiently close to 1.0, a relatively small value the passage rate Fp of unmarked images ( For example, if it is set to 0.5), it is possible to obtain a hierarchical Adaboost detector that passes most of the mark image and does not pass most of the non-mark image.

Each of the classifiers 310 is a classifier using a Haar-like feature. The Haarlike feature is a rectangular region having a white region and a black region in a unique pattern, and is set with a unique pattern, position, and size in the scan window 304 as illustrated in FIG. The quantity f is expressed by equation (1).

Here, I (i) and I (j) are the gray values of the pixels belonging to the white region R _w and the black region R _b , respectively, and ω _w and ω _b are the areas of the white region R _w and the black region R _b , respectively. It is a coefficient for weighting the total value of each region as if they were equal.
A function having a function value of 1 if the feature value f is greater than or equal to the threshold value and a function value of 0 if the feature value f is less than the threshold value is referred to as a weak hypothesis. A function value 1 indicates that the image in the scan window 304 is a mark image candidate, and a function value 0 indicates that the image in the scan window 304 is not a mark image candidate.

  A plurality of feature patterns are usually selected. In this embodiment, eight types of features (a) to (h) shown in FIG. 20 are selected. Further, as described above, in this embodiment, eight types of white marks and black marks are planned as standard marks, and the image in the scan window 304 is a total of 16 types in the discriminator 310 at each stage. If there is a possibility that it is one of the standard marks, it is determined as a mark candidate image, and if there is no possibility that it is any mark, it is determined as a non-mark image. One set of one feature (a combination of pattern, size and position in the scan window 304) and a threshold corresponds to one weak hypothesis. In this embodiment, each stage of the hierarchical Adaboost detector When each of the plurality of weak hypotheses includes a plurality of weak hypotheses and the function value of the plurality of weak hypotheses is 1 more than that of 0, the image in the scan window 304 is determined to be a mark image candidate at each stage. It is supposed to be. Further, the number of weak hypotheses in each stage of the hierarchical Adaboost detector is made smaller in the upstream stage so that the image of the scan window 304 that is not a mark candidate is quickly eliminated.

  In order for the hierarchical Adaboost detector to be able to appropriately extract a mark image candidate, it is necessary that the number of stages of the classifier 310 and the weak hypothesis at each stage be appropriate. In the example, an experiment was performed using the eight types of bright mark teacher data shown in FIG. 21, the eight types of dark mark teacher data shown in FIG. 22, and the non-mark teacher data exemplified in FIG. Thus, (a) the pattern and size of the feature at each stage, the position in the scan window 304, (b) the set of threshold values, and the number of stages were determined. Specifically, the characteristics of each pattern are set while changing the size and position in an area of N × N (N is a natural number, and the larger the standard mark, the larger the number is desirable), The ratio of determining whether each of the teacher data is marked or non-marked and determining that the non-marked image is a marked image is equal to or less than the first setting percentage, and the rate of erroneously determining that the marked image is a non-marked image is further Searching for weak hypotheses that are smaller than the second second setting percentage, and those that have a misjudgment rate of the judgment result by a combination of the respective numbers of the weak hypotheses being less than or equal to a sufficiently satisfactory value, were used as temporary discriminators. Then, the detection experiment with the provisional hierarchical Adaboost detector composed of these provisional discriminators and the addition of the erroneously determined image to the teacher data were repeated several times to improve the detection rate and the robustness. .

The feature amount f is calculated by using an integral image represented by equation (2), and the calculation result is stored in a storage area of a computer and repeatedly used to shorten the calculation time.

In addition, as the teacher data, a part of the actual printed circuit board 28 captured by the reference mark imaging device 130 is used. Normalized to the same size of N × N pixels.

Next, the determination of the reference mark candidate type will be described. This type discrimination is performed by a mark type discriminator using a neural network. Since the neural network itself is a known one, the description will be simplified.
As shown in FIG. 24, the mark type discriminator 320 of this embodiment is a three-layer neural network having an intermediate layer 326 between an input layer 322 and an output layer 324, and the input layer 322 includes the mark image. Are extracted (cut out) by N × N elements corresponding to an image normalized to N × N pixels, and the output layer 324 includes the non-standard mark 1 in the 16 types of standard marks shown in FIG. The intermediate layer 326 is composed of 17 elements corresponding to 17 types including the types, and the intermediate layer 326 is 2 ^Q (Q is preferably a natural number of 5 or more).

  The mark type discriminator 320 configured as described above was trained through a large number of bright mark teacher data and dark mark teacher data. As a specific learning method, as shown in FIG. 26, the same teacher data is repeatedly learned P times, the next teacher data is learned, and the order of the teacher data to be learned is randomly determined for each learning. Replaced. Further, when learning each teacher data, as shown in FIG. 27, an error was added by translating the image by ± Δx, ± Δy or rotating it by ± Δθ. Further, as shown in FIG. 28, an image obtained by horizontally inverting, vertically inverting and horizontally / vertically inverting the original image was also learned at the same time.

A lot of test data was passed through the provisional mark type discriminator 280 obtained by the learning, and a discrimination experiment was performed. In this determination experiment, changing the number 2 ^Q slope U0 and the intermediate layer of the sigmoid function to a plurality of types was selected appropriate in terms of both the convergence degree and average error ratio learning curve.

As is apparent from the above description, the mark type discriminator 320 of the present embodiment eventually has an N × N element input layer 322, 2 ^Q (Q is a natural number of 5 or more) element intermediate layer 326, and 17 elements. Output layer 324, and the slope of the sigmoid function is U0. The mark type discriminator 320 is not completely free of errors. However, as a result of visual confirmation, the image in which an error has occurred is only one that may be erroneous even if it is judged by a person, and can sufficiently withstand practical use. It was confirmed that it was.

  Model template data obtained by executing the fiducial mark model template creation program described above is stored in the RAM 252 of the image processing computer 244 itself of the screen printing machine 4 and via the host computer 20 or directly. It is supplied to the coating machine 8 and the control computer 248 of each mounting module 16 and stored in the RAM 286 of those image processing computers 280. Then, after the printed board 28 is carried into the adhesive applicator 8 or each mounting module 16 and held by the board holding device 124, it is used to detect a holding position error of the printed board 28.

After the above processing is performed on the printed circuit board 28 that is first carried into the screen printer 4, the image processing computer 244 of the screen printer 4 performs the template matching process using the created model template. The relative position of the reference mark 54 with respect to the reference mark imaging device 90 is acquired.
Further, the relative position of the reference mark 50 on the screen 26 with respect to the reference mark imaging device 90 is acquired in the same manner. At this time, if the reference mark 50 and the reference mark 54 are made common, the model template of the reference mark 54 can be used as the model template of the reference mark 50, but the two reference marks 50 and 54 are not the same. In this case, it is necessary to create a model template of the reference mark 50 in the same manner as in the case of the reference mark 54, and use it to acquire the relative position of the reference mark 50 with respect to the reference mark imaging device 90.

  In any case, the relative position shift of the reference marks 50 and 54 can be acquired based on the data of the relative position of the reference marks 50 and 54 with respect to the reference mark imaging device 90, and this relative position shift can be obtained from the positioning device 62. After being eliminated by the control of the fixing device 64, the printed circuit board 28 is brought into contact with the screen 26, and the first printing is performed. In this embodiment, the creation of the model template for the fiducial marks 50 and 54 is performed using the printed circuit board 28 to be a product.

  However, this is not indispensable. After the model template is created using the printed circuit board 28 dedicated to creating the model template, the printed circuit board 28 that is supposed to be an actual product is carried into the screen printing machine 4, and It is also possible to perform position detection once. This method is particularly useful when the image printing computer 280 of the adhesive applicator 8 or the mounting module 16 is to create a model template because the screen printing machine 4 does not have a model template creation function. It is valid. Of course, if the printing machine substrate transfer device 32, the shift conveyor 6 and the mounting module substrate transfer device 122 are capable of transferring the printed circuit board 28 in the reverse direction, the model template is created by the adhesive applicator 8 or After making the image processing computer 280 of the mounting module 16 perform, it is also possible to return the printed circuit board 28 used to create the model template to the screen printer 4 and use it as a printed circuit board 28 for production.

  In addition, in the above-described embodiment, the conclusion of the classifier 310 at each stage is determined by the majority of the identification results of a plurality of weak hypotheses included in each stage, but this is not essential. For example, when all of the conclusions of the plurality of weak hypotheses are 1, it is possible to determine that the image in the scan window 304 is a mark image candidate. It is also possible to multiply each identification result of a plurality of weak hypotheses by an evaluation value of creditworthiness and determine the conclusion of each stage by comparing the sum and the threshold value.

  In the assembly system of the present embodiment, the reference marks on the printed circuit board 28 and the screen 26 are used in the screen printing machine 4, the adhesive applicator 8, the mounting module 16 and the like using the model template created as described above. The positions 54 and 50 are detected, and based on the positions of the reference marks, the displacement of the printed circuit board 28 and the screen 26 from the predetermined position is detected, and the printing, adhesive application, and electronic circuit are canceled while the positional deviation is eliminated. Electronic circuit production work such as mounting of parts is performed. In the screen printing machine 4, the fixing of the screen frame 46 by the fixing device 64 is released, and then the screen frame is moved by the screen positioning device 62, and the relative position error of the screen 26 held by the screen frame 46 with respect to the printed circuit board 28. In the adhesive applicator 8 and the mounting module 16, the adhesive application head 210 and the mounting head 160 are moved to positions where the positioning error of the printed circuit board 28 is eliminated by the control of the head moving device 162. It is done.

  The above-mentioned electronic circuit production work, that is, work such as printing, adhesive application, and electronic circuit component mounting is not performed as scheduled (abbreviated as occurrence of defective work), or is performed as planned, but with higher work accuracy and work In some cases, improvement in efficiency is desired (abbreviated as performance improvement). In order to deal with these problems, the present assembly system includes a device for monitoring work execution status and updating software. Hereinafter, these points will be described.

First, regarding the monitoring of the work situation, the component mounting line 10 will be described as an example.
In the case where a component mounting error as a kind of work error occurs in the mounting module 16 constituting the component mounting line 10 and it is necessary to investigate the cause thereof, as shown in FIG. The CCD camera 352 held in the above is mounted. The CCD camera 352 captures an image once every fixed minute time, acquires image data of a set of a plurality of still images (which may be referred to as moving images) obtained by the plurality of times of image capturing. The circuit component 354 is monitored for mistakes in suction from the feeder 158 and mistakes in mounting on the printed circuit board 28. In many cases, a component mounting error occurs in a small electronic circuit component. Therefore, here, a component mounting error by the nozzle holder 188b will be described. However, a component mounting by the nozzle holder 188a having one suction nozzle 186a is described. If a mistake needs to be monitored, it can be monitored in that case as well.

  The CCD camera 352 can be always attached to the head main body 196 to monitor the status of component mounting work. However, in this embodiment, in order to avoid an increase in the inertial load on the head moving device 162, it is usually removed and attached as necessary. In addition, the image pickup by the CCD camera 352 can be performed continuously during the operation of the mounting module 16, but in this embodiment, information that is particularly effective for pursuing the cause of the component mounting error is obtained. This is performed only from the start of the lowering of the nozzle holder 188b to the end of the upward movement.

  That is, as the CCD camera 352 is attached, an image data collection controller 360 is connected to the controller 220 as shown in FIG. The image data collection control unit 360 includes a frequency divider 364 that divides the pulse signal supplied from the drive circuit 270 to the stepping motor 362 that is a drive source of the lifting device 204, and one pulse is output from the frequency divider 364. Each time it is output, that is, every time the nozzle holder 188b is moved down or raised by a minute amount, the CCD camera 352 performs one imaging, and the image data acquired by the imaging is stored in the buffer memory 366.

  For this purpose, the image data collection control unit 360 includes an image data collection computer 368 together with the frequency divider 364 and the buffer memory 366. The image data collection computer 368 includes a CPU 372, a ROM 374, a RAM 376, and an I / O port 378. The ROM 374 stores a suction / mounting error image data collection program represented by the flowchart of FIG. It is executed while using it.

  First, after initial setting such as setting the count value N to 0 in S21 (representing step 21, but simply representing S21 for the sake of simplification), the control is performed in S22. Next data sent from the computer 248 relating to suction or mounting, that is, at the time of suction, identification codes of the feeder 158 and the suction nozzle 186 (not a code indicating the type, but specific to each feeder or suction nozzle) ) And data indicating the type of electronic circuit component, and when mounted, the identification code of the suction nozzle 186 and data indicating the type of electronic circuit component are acquired and stored in the RAM 376. Thereafter, S23 is repeatedly executed to wait for the frequency divider 364 to output a frequency-divided pulse. The divided pulse is issued from the frequency divider 364 every time the nozzle holder 188b is lowered or raised by a minute amount, and accordingly the determination in S23 is YES, and in S24, an imaging command is issued to the CCD camera 352, and the count value N is increased by one. Thereafter, S25 is repeatedly executed, and the elapse of time T1 is waited, and during that time, imaging by the CCD camera 352 is performed. Subsequently, the image data acquired by the CCD camera 352 in S26, that is, image data for one image is stored in the buffer memory 366 in association with the count value N and the data relating to suction or mounting acquired in S22. In S27, it is determined whether or not the count value N has reached the set value N2. Initially, the determination result is NO and program execution returns to S23. By repeating the above execution, data of a plurality of images indicating the adsorption or mounting status at minute time intervals is collected in the buffer memory 366.

  Eventually, the determination result of S27 becomes YES, and the image data when the count value N is N1 and the image data when N2 is N2 are compared in S28. The count value N1 and the count value N2 indicate that in the image acquired by the CCD camera 352, the suction nozzle 186 that is descending and the suction nozzle 186 that is ascending are in the same position, and the suction nozzle 186 has an electron. If the circuit component 354 is attracted, the count value corresponding to the time when the entire image of the electronic circuit component 354 should be included is selected. Therefore, when the electronic circuit component 354 is sucked, the image corresponding to the count value N1 includes the image of the suction nozzle 186 that does not suck the electronic circuit component 354, and the image corresponding to the count value N2 Should include an image of the suction nozzle 186 that sucks the electronic circuit component 354, and the images of the suction nozzle 186 in the two images are the same. When the electronic circuit component 354 is mounted, the image corresponding to the count value N1 includes the image of the suction nozzle 186 that sucks the electronic circuit component 354, and the image corresponding to the count value N2 Should include an image of the suction nozzle 186 that does not suck the electronic circuit component 354, and the images of the suction nozzle 186 in the two images are the same.

  Therefore, if the data of the two images is binarized and the absolute value of the sum of the differences between the binarized data of the corresponding pixels is obtained, the electronic circuit component 354 is picked up or attached when the electronic circuit component 354 is sucked or mounted. If the absolute value of the sum exceeds the set value and is not performed, it should be below the set value. Based on this fact, in S28, the image corresponding to the count value N1 and the image corresponding to the count value N2 are compared, that is, the difference between the binarized data of the pixels corresponding to each other in both images is acquired. In S29, it is determined whether or not the image data match each other, that is, whether or not the absolute value of the difference is equal to or less than a threshold value set in advance to a relatively small value. If the determination result is YES, the image data stored in the buffer memory 366 is transferred to the external memory 380 as a data collection memory in S30. Transferred. At the time of this transfer, the suction attachment related data acquired in S22 is sent together. The “external memory 380” means a storage device having a large storage capacity provided outside the image data collection control unit 360.

  On the other hand, if the determination result in S29 is NO, it is determined in S31 whether or not the suction posture of the electronic circuit component 354 by the suction nozzle 186 is normal. The simple determination in S29 is a simple determination as to whether or not suction or mounting has been performed, and does not determine whether or not the suction posture of the electronic circuit component 354 by the suction nozzle 186 is normal. The control computer 248 is inquired of the determination result of “whether the suction posture of the electronic circuit component 354 by the suction nozzle 186 is normal” performed in the image processing computer 280 based on the imaging result of 132. The buffer memory 366 is cleared in S32 only when the suction posture of 354 is normal and the determination result in S31 is YES, and the image data temporarily stored therein is deleted. However, when the suction posture of the electronic circuit component 354 by the suction nozzle 186 is not normal, it cannot be said that the suction is normally performed, and S30 is executed and the image data in the buffer memory 366 is transferred to the external memory 380. The

  As described above, in this embodiment, the state of suction and mounting of the electronic circuit component 354 by the suction nozzle 186 is acquired as a set of still images at every minute time (this set can be called a moving image), and is temporarily stored in the buffer memory 366. Then, only the image data when the suction or mounting is not normally performed is automatically stored in the external memory 380. Therefore, there is an advantage that the storage capacity of the external memory 380 can be relatively small and only the image data useful for investigating the cause is accumulated, so that the cause can be easily investigated. In addition, the image data stored in the external memory 380 is associated with identification codes of the feeder 158 and the suction nozzle 186 and data indicating the type of electronic circuit component if it is at the time of suction. In this case, since the identification code of the suction nozzle 186 is associated with the data representing the type of electronic circuit component, the cause can be further easily determined.

In this embodiment, the image processing performed in S28 is simplified, and the image processing result by the image processing computer 280 is used to determine whether or not the suction posture of the electronic circuit component 354 by the suction nozzle 186 is normal. Has been. Therefore, when transfer of image data from the buffer memory 366 to the external memory 380 is required, the transfer is started early, while clearing the buffer memory 366 confirms that the suction posture of the electronic circuit component 354 is normal. Therefore, there is an advantage that it is possible to surely avoid that necessary image data is cleared.
However, based on the image data acquired by the CCD camera 352, it is possible to determine whether or not the suction posture of the electronic circuit component 354 by the suction nozzle 186 is normal. May be provided with a dedicated image processing computer, or the image processing computer 280 may be supplied with image data for processing.

  In the above-described embodiment, the state of suction and mounting of all electronic circuit components is imaged. However, in S22, identification codes and electronic circuits of the feeder 158 and suction nozzle 186 involved in suction or mounting are used. After the data representing the type of part is acquired, before the execution of S23, in S22a indicated by a two-dot chain line in FIG. 31, the data matches the designated data stored in the RAM 376 as an object to be imaged in advance. If it is determined whether or not to match the specified data, it is possible to end the execution of one suction / mounting image data collection program without executing S23 and subsequent steps. . In this way, only when the conditions are the same as the actual occurrence of suction or mounting mistakes, or only when it is known from past results that suction or mounting errors are likely to occur. Can be collected.

  Next, a device for updating software will be described. This device can be used at the time of updating all software used in the assembly system. Here, as an example, the image data stored in the buffer memory 366 in the suction / mounting error image data collection program is described here. As an example, a case where the part of determination as to whether or not to transfer to the external memory 380 is changed to determination using the same image processing as in the image processing computer 280 will be described.

  The image processing in the image processing computer 280 is mainly intended to acquire the holding position error of the electronic circuit component 354 by the suction nozzle 186. Here, the image processing is performed by the method described in JP-A-8-180191. It is assumed that the holding position error of the electronic circuit component 354 held by the nozzle 186 is acquired. According to this method, it is possible to determine whether or not the electronic circuit component 354 is held in a normal posture as well as whether or not the electronic circuit component 354 is held by the suction nozzle 186.

  When it is necessary to update the suction / mounting error image data collection program, first, as shown in FIG. 33, the image data collection computer 368 and the frequency divider 364 of the image data collection control unit 360 are connected. In parallel, a test computer 400 including a CPU 392, a ROM 394, a RAM 396, a buffer memory 397, and an I / O port 398 is connected. The external memory 402 is connected to the I / O port 398 of the test computer 400 and the I / O port 256 of the control computer 248 is also connected. The ROM 394 stores an update program, that is, a new suction / mounting image data collection program obtained by changing S28 and subsequent steps in the flowchart shown in FIG. 31 to those shown in FIG.

  Only the changed part will be described below. If the determination result in S27 is YES, in S41, it is determined whether or not the currently performed lowering operation of the suction nozzle 186 is for sucking the electronic circuit component 354. This determination is made based on the data relating to suction or mounting acquired in S22. If the determination result in S41 is YES, image processing based on the image data stored in the buffer memory 397 in association with the count value N2 is performed in S42. Based on the processing result, the electronic circuit component is processed in S43. It is determined whether or not 354 is held in the suction nozzle 186 in a normal posture. If the determination result is YES, the buffer memory 397 is cleared in S44. On the other hand, if the determination result is NO, in S45, the image data stored in the buffer memory 397 is transferred to the external memory 402 in association with the data relating to suction acquired in S22.

  As described above, at the time of suction, the holding state of the electronic circuit component 354 is checked only while the suction nozzle 186 is raised, and is not checked while the suction nozzle 186 is lowered. This is because it is guaranteed that the suction operation is not performed while the electronic circuit component 354 is held. However, if it is not guaranteed, or it may be guaranteed, it may be checked whether the electronic circuit component 354 is held while the suction nozzle 186 is lowered. .

  On the other hand, if the determination result in S41 is NO, that is, when the electronic circuit component 354 is mounted, image processing based on the image data stored in the buffer memory 397 in association with the count value N1 is performed in S46. Based on the processing result, it is determined in S47 whether or not the electronic circuit component 354 is held by the suction nozzle 186 in a normal posture. Since this determination has already been performed once in S43, it can be omitted. However, the electronic circuit component 354 is not necessarily dropped during the movement of the suction nozzle 186. Therefore, the above determination is made. If the determination result in S47 is NO, S45 is executed, and the image data stored in the buffer memory 397 is transferred to the external memory 402.

  On the other hand, if the determination result in S47 is YES, image processing based on the image data stored in the buffer memory 397 in association with the count value N2 is performed in S48, and based on the processing result, in S48. It is determined whether or not the electronic circuit component 354 is no longer held by the suction nozzle 186. If the determination result is YES, it is assumed that the electronic circuit component 354 is mounted on the printed circuit board 28, S44 is executed, and the buffer memory 397 is cleared. On the other hand, if the determination result is NO, assuming that the electronic circuit component 354 has been brought home, the image data stored in the buffer memory 397 is associated with the data relating to the mounting acquired in S22 in S45, and the external memory 402 Forwarded to

Therefore, in a state where the test computer 400 is connected in parallel with the image data collection computer 368, the assembly system including the mounting module 16 performs normal assembly work, and the image data connected to the mounting module 16 at that time By comparing the operations of the collection computer 368 and the test computer 400, it is possible to test whether or not the new adsorption / mounting error image data collection program operates as scheduled. In particular, in the present embodiment, by comparing the image data stored in the external memory 380 and the external memory 402, it is possible to easily determine whether the new suction / mounting error image data collection program functions as planned. Can be evaluated. For example, if the purpose of the test is to increase the reliability of picking up / mounting error image data collection, confirm that there are no new problems due to program changes and that the reliability has increased as planned. It can be done.
In this embodiment, this confirmation is performed by a person. The external memory 380 and the external memory 402 are connected to a computer storing a comparison program for comparing image data, and the comparison program is executed. Thus, the above confirmation can be automatically performed.

  The present embodiment tests a new suction / mounting image data collection program that is not for directly controlling the operation of the mounting module 16, but for directly controlling the operation of the mounting module 16. It is also possible to test the validity of the new control program. Even in this case, the new control program is not stored in the ROM 252 of the control computer 248 of the mounting module 16, but is stored in the ROM of the test computer connected in parallel with the control computer 248, and the control computer 248 is stored in the control program before the update. To control the mounting module 16. Along with this control, a control command issued from the test computer is stored in the external memory, and later, the validity of the control command is confirmed. At that time, although it is not indispensable, if the control command issued from the control computer 248 that operates according to the control program before the update is stored in the external memory, the validity of the new control program can be confirmed more easily and can be automated. become.

  As mentioned above, although one Example of this invention was described in detail, this is only an illustration and this invention is based on the knowledge of those skilled in the art including the aspect described in the above-mentioned [Aspect of invention]. The present invention can be implemented in variously modified forms.

  2: substrate supply device 4: screen printer 6: shift conveyor 8: adhesive applicator 10: component mounting line 12: adhesive curing device 14: reflow furnace 16: mounting module 18: base 20: host computer 26: screen 28: Printed circuit board 30: Frame 32: Printing machine board transfer device 34: Board holding device 36: Screen holding device 38: Squeegee device 40: Mark imaging system 42: Printing machine control device 46: Screen frame 48: Through hole 50: Reference Mark 52: Land 54: Reference mark 60: Screen support base 62: Positioning device 64: Fixing device 66: Squeegee head 68: Squeegee head moving device 70: Squeegee head lifting device 72: Moving member 74: Moving Member moving device 76: Squeegee moving motor 78: Ball screw 80: Nut 82: Feed screw mechanism 84: Squeegee holding member 86: Squeegee 90: Reference mark imaging device 92: Imaging device moving device 96: X-axis direction moving device 98: Y-axis direction moving device 100: X-axis slide 102: X-slide moving device 104: Y-axis slide 106: Y-axis slide moving device 120: Module main body 122: Mounting module substrate transfer device 124: Board holding device 126: Component supply device 128 : Mounting device 130: Reference mark imaging device 132: Component imaging device 154, 156: Substrate conveyor 158: Tape feeder (feeder) 160a, b: Mounting head 162: Head moving device 164: X-axis direction moving device 66: Y-axis direction moving device 170: Linear motor 172: Y-axis slide 174: First X-axis slide 176: Second X-axis slide 178: X-axis slide moving device 186 (186a, 186b): Suction nozzle 188 (188a, 188b) : Nozzle holder 190: Suction tube 192: Background forming plate 196: Head body 200: Rotating body 202: Rotating body rotating device 204: Lifting device 206: Holder rotating device 210: Adhesive application head 212: Syringe 214: Piston 218: Control Device 220: Control device 222: Control computer 240: Drive circuit 244: Image processing computer 246: Display 248: Control computer 270: Drive circuit 272: Feeder control computer 80: image processing computer 300: processed image 302: printed circuit 304: scan window 310: discriminator 320: mark type discriminator 322: input layer 324: output layer 326: intermediate layer 350: bracket 352: CCD camera 354: electronic circuit Component 360: Image data collection control unit 362: Step motor 364: Frequency divider 366: Buffer memory 368: Image data collection computer 380: External memory 397: Buffer memory 400: Test computer 402: External memory

Claims (8)

An imaging step of imaging the reference mark formation planned position of the circuit base material and its periphery by an imaging device;
A reference mark candidate extraction step of extracting a reference mark candidate that is an image that is highly likely to be an image of a reference mark from images obtained by performing the imaging step;
A mark type determination step for determining which of the plurality of types of reference marks that are extracted from the extracted reference mark candidates;
And a template creation step of creating a model template of a reference mark whose type has been clarified by performing the mark type discrimination step, and the reference mark candidate extraction step uses a hierarchical Adaboost detector using a Haar-like feature. A reference mark model template creation method, characterized in that it is a step of extracting reference mark candidates. 2. The reference mark model according to claim 1 , further comprising a dimension measuring step of measuring a dimension of an image whose type has been clarified by performing the mark type determining step, and creating a model template of the measured dimension in the template creating step. Template creation method.   The reference mark candidate extraction step sets a scan window having a size suitable for detection of the smallest one of those scheduled as the reference mark in the processed image that is an image acquired in the imaging step, Whether or not the image in the scan window is the image of the reference mark candidate is determined by moving the scan window along a predetermined path by a predetermined amount and performing one scan. 3. The reference mark model template creation method according to claim 1, further comprising a step of performing a plurality of scans by enlarging the scan window by a set amount each time the one scan is performed. The mark type determination process, the reference mark model templates A method as claimed in any one of claims 1 to 3 comprising the step of determining the type of reference mark candidate by using a neural network.   The neural network is provided between an input layer composed of N elements × M elements, an output layer composed of a number of elements obtained by adding 1 to the predetermined number of reference marks, and between the input layer and the output layer. An intermediate layer composed of a plurality of elements, and the mark teacher data obtained by normalizing the image of the planned reference mark to N pixels × M pixels, and the image of a non-reference mark image normalized to N pixels × M pixels The reference mark model template creation method according to claim 4, wherein learning using non-mark teacher data that has been made into a non-mark is used.   When a plurality of model templates are created for one reference mark formation scheduled position by executing the reference mark candidate extraction step, the mark type determination step, and the template creation step, one of the plurality of model templates. 6. The reference mark model template creation method according to claim 1, further comprising a final model template determination step of determining one as a final model template. The final model template determination step includes
A plurality of model template notification steps for notifying that one of the plurality of model templates should be selected and the final model template should be designated when a plurality of model templates are created for one reference mark formation scheduled position When,
The reference mark model template creation method according to claim 6, further comprising: a final model template designation input step in which an operator designates a final model template in accordance with execution of the model template multiple notification step.   The reference mark according to claim 6 or 7, wherein the final model template determining step includes a step of automatically determining a model template that best matches a predetermined condition among the plurality of model templates as a final model template. Model template creation method.