JP4747955B2 - Inspection control device, inspection control method, inspection system, control program, and recording medium - Google Patents

Description

  The present invention relates to an inspection apparatus for determining whether a production object (work) is good or defective in a production line, and in particular, an inspection control apparatus, an inspection control method, an inspection system, which automatically perform various settings necessary for inspection, The present invention relates to a control program and a recording medium.

  Currently, in a production line, it is possible to automatically discriminate between good and defective workpieces using an inspection device. In such an inspection apparatus, in general, (1) information on a workpiece state is acquired, and (2) a feature amount is calculated based on a rule that defines how to extract a feature amount obtained by quantifying the workpiece state. Extraction is performed, and (3) the feature quantity is discriminated based on a preset discriminant function (threshold), thereby discriminating whether the workpiece is good or bad.

  Specifically, for example, in an inspection apparatus having an image processing function for determining whether a work is good or bad using an image obtained by photographing the work, an image is put in a certain parameter space using an image processing algorithm (rule). Mapping (extraction of feature values) is performed to determine whether the workpiece is good or bad based on a threshold value for separating good / defective products.

  Therefore, in order for an inspection device to correctly separate a workpiece into a non-defective product and a defective product, it is indispensable to select what rule to use or how to determine the threshold value. . The rules and the threshold values set in the inspection apparatus as described above are hereinafter referred to as inspection logic. That is, the inspection logic is control information for the inspection apparatus that defines a method for determining whether the inspection apparatus is good or defective.

  In other words, if the setting of the inspection logic is wrong, an inspection failure may occur in which a non-defective product is determined as a defective product or a defective product is determined as a non-defective product.

  In general, the inspection logic collects sample data (information on the state of the workpiece) and is set based on statistical data. Needless to say, the more data (number of samples) to be collected, the more accurate inspection logic or threshold value can be obtained.

  Patent Document 1 discloses an appearance inspection method and apparatus for setting a discriminant function (threshold value) from a large amount of data. Specifically, code data indicating the state of the inspection object is generated from the captured image of the inspection object, and a discriminant function is obtained based on the code data of the inspection object of the sample selected for each category.

  Patent Document 2 discloses a method and apparatus for inspecting an article that performs pass / fail determination based on a plurality of determination values, and selects and sets a determination value (threshold value) obtained with a determination result close to visual inspection. .

  Generally, at the initial stage of production when the inspection apparatus is set, since the number of samples is small, it is almost impossible to correctly set a threshold value (discriminant function / determination value) for determining good / bad. Therefore, it is necessary to increase the number of samples through repeated production, and appropriately determine the suitability of the currently set discriminant function and reset a more appropriate discriminant function.

  However, in the apparatus of Patent Document 1, it is impossible to determine whether resetting is necessary or not unless an erroneous determination (inspection failure) occurs. After all, the appropriateness of the discriminant function can only be determined by a person, especially an experienced expert.

  In addition, as disclosed in Patent Document 2, although an optimum one is selected from a plurality of judgment values, changes in the environment of the production site (consumption and replacement of production equipment, replacement of workers and skill level) In the present situation, it is necessary to reset the judgment value with the improvement of the workpiece and the change of the workpiece type.

  However, although the device of Patent Document 2 discloses that an optimum determination value is set at the time of initial setting, it is not considered to review the setting after the line is activated.

  Therefore, in order to maintain the inspection accuracy of the inspection apparatus even after line operation, Patent Document 3 discloses a stripe pattern inspection apparatus with an automatic diagnosis function that inspects the pattern width and pitch of RGB stripes (BM stripes). .

The stripe pattern inspection apparatus periodically inspects standard stripe patterns and automatically diagnoses inspection accuracy. Furthermore, the information obtained by managing the tendency of the inspection value of the actually inspected stripe pattern is fed back to the exposure table on which the BM stripe pattern is formed, and the illuminance of the exposure light source is automatically controlled. This makes it possible to maintain the inspection accuracy in the inspection environment without the need for periodic calibration, correction, and adjustment by the operator, and facilitate the maintenance and management of manufacturing quality.
JP-A-8-254501 (published on October 1, 1996) JP2001-242086 (published September 7, 2001) JP 10-40815 (published February 13, 1998)

  However, the above conventional configuration causes the following problems.

  Although the apparatus of Patent Document 3 can correct fluctuations in the inspection environment (such as the camera lens of the inspection apparatus and the illumination state), there arises a problem that it cannot cope with fluctuations in the production environment, for example, a workpiece (inspection object).

  That is, it is assumed that the standard stripe pattern is always “standard”. Therefore, if the inspection object fluctuates, the inspection accuracy cannot be maintained unless it is replaced with a standard stripe pattern that is “standard” for the new inspection object. As a result, it is necessary for a person to determine whether or not to reset the inspection logic (in this case, the standard stripe pattern) and to select an appropriate inspection logic.

  Examples of fluctuations in the production environment include, for example, changing the average width of component mounting positions due to adjustment of production equipment, and changing image characteristics (color, appearance) due to changes in part lots and vendors. It is supposed to do.

  The above-mentioned problem does not occur only in an inspection apparatus having an image processing function for determining good / bad from an image of a work using an image processing algorithm, and is based on the above-described inspection logic. If it is an inspection device that performs quality inspection, it will occur in the same way.

  The present invention has been made in view of the above-described problems, and an object of the present invention is to provide an inspection control apparatus and an inspection control method that improve inspection accuracy by dynamically setting inspection logic according to changes in the production environment. It is to realize an inspection system, a control program, and a recording medium.

  In order to solve the above-described problem, an inspection control apparatus according to the present invention uses an inspection logic that defines a method for determining whether a product is good or bad, and controls an inspection module that determines whether a target is good or bad. The inspection logic selection means for selecting the main inspection logic used by the inspection module from a plurality of inspection logics, and the final inspection indicating the result of good / bad determination of the object generated by the inspection module using the main inspection logic An evaluation means for evaluating the inspection accuracy of each inspection logic based on the result and an inspection result for evaluation indicating a determination result when the quality of the object is determined using inspection logic different from the main inspection logic; Inspection logic setting means for switching the inspection logic selected by the inspection logic selection means as the main inspection logic, and査 logic selection means, based on the evaluation result of the evaluation means to evaluate, it is characterized by selecting the main inspection logic.

  In order to solve the above problems, an inspection control method according to the present invention uses an inspection logic that defines a method for determining good or bad, and uses an inspection control in an inspection control device that controls an inspection module for determining good or bad of an object. A method for selecting a main inspection logic used by the inspection module from a plurality of inspection logics, and a result of determining whether the object is good or defective generated by the inspection module using the main inspection logic. The inspection accuracy of each inspection logic is evaluated on the basis of the final inspection result indicating the difference between the main inspection logic and the evaluation inspection result indicating the determination result when the object is determined to be good or defective using inspection logic different from the main inspection logic. An evaluation step and a switching step for switching the inspection logic selected by the inspection logic selection means as the main inspection logic. In the above selecting step, based on the evaluation result by the evaluation step, it is characterized by selecting the main inspection logic.

  According to the above configuration and method, the inspection module uses the inspection logic set as the main inspection logic by the inspection logic setting unit among the inspection logic as the control information that defines the good / bad determination method. A defect is determined, and the determination result is output as a final inspection result. The final inspection result is a final result that is adopted to identify a defective object in an actual production line. That is, the actual inspection of the object is executed by the determination method based on the main inspection logic.

  On the other hand, the inspection module determines whether the object is good or bad using inspection logic other than the main inspection logic. In this way, the inspection module obtains a determination result for each inspection logic (including the main inspection logic).

  Here, the evaluation means evaluates the discrimination result for each inspection logic. Thus, the determination result is referred to when the evaluation unit evaluates the inspection accuracy for each inspection logic, and thus can be said to be an inspection result for evaluation. Specifically, the evaluation method evaluates the inspection result for evaluation obtained for each inspection logic from the inspection module based on the final inspection result by the main inspection logic, and sets an evaluation value representing inspection accuracy for each inspection logic. To calculate.

  Next, the inspection logic selection means selects the main inspection logic used by the inspection module to output the final inspection result based on the evaluation value of each inspection logic. The selected main inspection logic is set in the inspection module by the inspection logic setting means. As described above, the inspection module uses the inspection logic set by the inspection logic setting unit as the main inspection logic to determine whether the object is good or defective and outputs the final inspection result.

  As a result, the main inspection logic of the inspection module used to detect defective products flowing in the actual production line is always selected by the inspection logic selection means based on the evaluation value. Is done. The selected main inspection logic is dynamically set by the inspection logic setting means.

  Therefore, for example, as the production environment changes, it is found that the evaluation values of other inspection logics indicate that the inspection accuracy is higher than the evaluation values of the main inspection logics already set. For example, the inspection logic selection means can select the inspection logic evaluated as having higher inspection accuracy as the main inspection logic of the inspection module. That is, the inspection logic having higher inspection accuracy can be automatically and dynamically reset for the inspection module as the main inspection logic, and the inspection accuracy can be maintained or increased.

  If the production line is in operation, changes in the production environment are inevitable. For example, consumption / failure / replacement of production facilities and inspection facilities, changes in physical environment (such as temperature and humidity), changes in the skill level of workers involved in production work, replacement of workers, and the like can be considered. If the production environment fluctuates, the already set inspection logic is not always optimal (maintaining high inspection accuracy).

  However, according to the inspection apparatus of the present invention, among the plurality of inspection logics, the optimum one can always be specified as the main inspection logic based on the evaluation value, so that the inspection is dynamically performed according to the change in the production environment. Inspection accuracy can be improved by setting logic.

  In the inspection control apparatus according to the present invention, the inspection logic defines a feature amount measuring method indicating a state of an object and a threshold value for separating the feature amount to determine good or bad, and the evaluation means Is the inspection accuracy of each inspection logic using statistical data based on the feature quantity measured for each object included in the inspection result for evaluation and the number of good and defective products of the object indicated by the final inspection result. It is preferable to evaluate by calculating an evaluation value representing.

  According to the said structure, the said evaluation means is the statistical data based on the feature-value measured for every target object contained in the said test result for evaluation, and the number of good products and the number of inferior goods of the target object which the said final test result shows. Is used to calculate an evaluation value representing the inspection accuracy of each inspection logic. Here, the feature amount is information that numerically represents the state of the object and can be separated based on a threshold value, thereby enabling the determination of the quality of the object. is there.

  In other words, the above-described statistical data includes a distribution based on a feature amount measured by another inspection logic with respect to an object determined by the main inspection logic as a non-defective product (or defective product). From such statistical data, it is possible to obtain information about how the other inspection logics have made the discrimination based on what feature quantity the object that the main inspection logic has determined to be non-defective (or defective). That is, based on the inspection result of the main inspection logic, the inspection results of other inspection logics can be analyzed and evaluated.

  Further, the evaluation means evaluates the inspection logic by calculating an evaluation value based on a difference between a non-defective product feature value obtained from the non-defective product set of the target object and a defective product feature value obtained from the defective product set. May be.

  According to the above configuration, the inspection accuracy of each inspection logic is calculated based on the difference between a non-defective product (determined by the main inspection logic) feature quantity and a defective product feature quantity measured by the inspection logic.

  A large difference between the non-defective product feature value and the defective product feature value means that the discrimination between the non-defective product and the defective product, which is the same as the main inspection logic, is executed more safely and reliably in the inspection logic. Therefore, it can be evaluated that the inspection accuracy is higher as the difference between the non-defective product feature amount and the defective product feature amount is larger.

  In addition, the method for obtaining the non-defective product (defective product) feature quantity obtained from the collection of non-defective products (defective products) of the object is not particularly limited. For example, an average value, a median value, a mode value, an upper limit value, a lower limit value, and the like of feature amounts in a set of good products (defective products) may be used as good product (defective products) feature values.

  Further, when the evaluation unit determines that the feature amount of the object is smaller than a threshold value, the evaluation unit determines that the product is a non-defective product, and the defective product lower limit value that is a lower limit value of the feature amount of the defective product set and the non-defective product set. If the evaluation value is calculated based on the difference from the upper limit value of the non-defective product, which is the upper limit value of the feature amount, and the feature value of the target object is larger than the threshold value, it is determined that the product is non-defective The inspection logic may be evaluated by calculating an evaluation value based on a difference between a non-defective product lower limit value that is a lower limit value of the non-defective product and a defective product upper limit value that is an upper limit value of the feature amount of the set of defective products.

  Thereby, the inspection accuracy of each inspection logic can be more accurately evaluated.

  In other words, when it is desired to separate the feature values of all objects into non-defective products and defective products in the same way as the final inspection result of the main inspection logic by one threshold, the distance between the non-defective product set and the defective product set based on the inspection logic feature value The larger the better, the better it should not overlap.

  More specifically, when it is determined that the target feature is smaller than the threshold value, it is determined that the target product is a non-defective product. From among the feature values measured by the set of defective products, The larger the value (difference) obtained by subtracting the maximum feature value (non-defective product upper limit value) from the feature values measured for a good product set, the safer it is possible to discriminate between good and defective as well as the final inspection result. It becomes. On the other hand, when the difference is 0 or less, the set of non-defective products and the set of non-defective products cannot be correctly separated (inspection defects such as determining non-defective products as non-defective products or determining non-defective products as non-defective products occur). It is judged that there is a possibility.

  When it is determined that the feature amount of the object is larger than the threshold value, the evaluation is higher as the value (difference) obtained by subtracting the upper limit value of the defective product from the lower limit value of the non-defective product is larger. If the difference is 0 or less, it is determined that the inspection accuracy is low as the inspection defect occurs.

  In this way, by avoiding high evaluation of the inspection logic whose difference is 0 or less, it is possible to prevent selection of inspection logic with low inspection accuracy that may cause inspection failure. As a result, the inspection accuracy of the inspection module can be improved.

  Further, it is preferable that the evaluation unit evaluates the inspection logic by calculating an evaluation value based on a quotient obtained by dividing the difference by the variance of the non-defective feature amount.

  This makes it possible to compare and evaluate inspection accuracy even when the units of feature quantities measured between different inspection logics are different from each other. For example, it is possible to appropriately evaluate the inspection logic that performs determination using an area as a feature amount and the inspection logic that performs determination using a number as a feature amount by using a common scale.

  In other words, the difference indicates the distance between the non-defective product set and the non-defective product set. The larger the distance value, the higher the evaluation that the non-defective product is safely separated. The non-defective product feature amount dispersion indicates a variation of the non-defective product set, and the larger the variation value, the less reliable the feature amount measurement procedure and the lower the evaluation. Therefore, it is possible to compare inspection logics using a common scale, assuming that the larger the quotient obtained by dividing the distance by the non-defective product feature amount, the higher the inspection accuracy.

  Alternatively, the evaluation means is at least one of a difference between the non-defective product feature value obtained from the non-defective product set of the target object and the threshold value, and a difference between the defective product feature value obtained from the defective product set and the threshold value. The inspection logic may be evaluated by calculating an evaluation value based on one.

  According to the above configuration, the evaluation value is calculated based on the distance between the non-defective product set and the defective product set, not the distance between the set and the threshold value considered to be between them. Thereby, it becomes possible to evaluate the inspection logic in which the threshold is appropriately set.

  Further, the evaluation means includes (1) a difference A between the non-defective product feature quantity and the defective product feature quantity, (2) a difference B between the defective product feature quantity and the threshold value, and (3) the non-defective product. The inspection logic may be evaluated by calculating an evaluation value based on “evaluation formula: | difference A−difference B / difference C |”, where the difference between the feature amount and the threshold is the difference C.

  According to the above configuration, the evaluation means weights the distance (difference B) between the defective product feature value and the threshold value rather than the distance (difference C) between the non-defective feature value and the threshold value. Is small (that is, when the set of defective products is closer to the threshold value), an evaluation value that particularly lowers the evaluation is calculated.

  Thereby, the evaluation of the inspection logic that discriminates a defective product as a non-defective product is calculated low, and the inconvenience that such an inspection logic is erroneously selected as the main inspection logic can be solved.

  This is based on the idea that an inspection failure (missing) that determines a defective product as a non-defective product is a more serious accident than an inspection failure that determines a good product as a defective product. There is no possibility that an inspection logic that may be missed is erroneously selected as the main inspection logic. As a result, the inspection accuracy of the inspection module can be improved.

  In the case of the above evaluation formula, the smaller the numerical value of the evaluation value, the higher the evaluation.

  Further, the evaluation means is based on the number of objects determined to be non-defective by the inspection logic in the evaluation inspection result among the objects determined to be defective by the main inspection logic in the final inspection result. Inspection logic may be evaluated.

  According to the said structure, the said evaluation means evaluates the test | inspection precision of each test | inspection logic based on the generation | occurrence | production number of the inspection defect of the above-mentioned miss. For example, the evaluation value may be calculated by, for example, evaluating an inspection logic with 0 missed cases highly, or evaluating a test logic with a larger number of missed cases.

  As a result, there is no possibility that an inspection logic that may be overlooked is erroneously selected as the main inspection logic. As a result, the inspection accuracy of the inspection module can be improved.

  The inspection logic setting unit is configured such that when the inspection logic newly selected by the inspection logic selection unit based on the evaluation result evaluated by the evaluation unit is different from the main inspection logic in use by the inspection module, Is preferably switched.

  Thus, when a proper main inspection logic is newly selected rather than the original main inspection logic currently set in the inspection module, the inspection module can perform inspection using the new main inspection logic. The main inspection logic can be switched.

  Furthermore, it is preferable that the inspection logic setting unit switches the main inspection logic when the evaluation unit evaluates inspection logic other than the main inspection logic in use by the inspection module higher than the main inspection logic.

  As a result, the inspection module can be controlled to always execute the inspection by setting the inspection logic with the highest evaluation as the main inspection logic.

  Alternatively, the inspection logic setting means monitors an evaluation value, which is included in the evaluation result and represents an inspection accuracy of the main inspection logic in use by the inspection module, and based on a differential value of the evaluation value, The switching timing may be calculated, and the main inspection logic may be switched at the calculated timing.

  According to the above configuration, the evaluation value of the main inspection logic being used by the inspection module calculated by the evaluation means is monitored, the transition of the evaluation value is recorded, and the main inspection logic is monitored based on the obtained differential value. The appropriate switching timing is calculated.

  Thereby, it is possible to detect the moment of decrease in the evaluation value of the main inspection logic and predict in advance the time point at which the main inspection logic is not optimal. Therefore, for example, before the evaluation evaluation is reversed, it is possible to switch to the inspection logic of the next point (or the highest rate of increase in evaluation value) in advance.

  As a result, it is possible to prevent an inspection failure and improve the inspection accuracy of the inspection module.

  In addition to the above-described configuration, the inspection control apparatus of the present invention includes a display control unit that controls the display unit to display the inspection result, and a user who performs an inspection on the object to be inspected that is the same as the displayed inspection result. Result input control means for receiving an input of an inspection result, and inspection result correction means for correcting the inspection result displayed on the display unit by the display control means based on the user inspection result received by the result input control means. It is preferable.

  According to the said structure, a display control means controls a display part, and displays the test result which shows the discrimination | determination result of the quality of the target object which the test | inspection module performed using the test logic. Thereby, the user can visually confirm the inspection result of the inspection module.

  And the said result input control means receives the input of the user test result as a result obtained when the user test | inspected with respect to the said target object.

  Finally, the inspection result correcting unit corrects the inspection result based on the user inspection result when the inspection result displayed on the display unit by the display control unit is different from the user inspection result received by the result input control unit. To do.

  Thereby, even when an inspection result different from the result of the inspection performed by the user is output from the inspection module, the user can correct the inspection result by reflecting the inspection result performed by the user. Since each inspection logic is evaluated based on the corrected inspection result, it is possible to more accurately select and set the proper main inspection logic.

  In addition to the above-described configuration, the inspection control apparatus of the present invention preferably includes a threshold adjustment unit that calculates the inspection logic threshold using statistical data of each inspection logic and determines the inspection logic threshold.

  As a result, the threshold defined by the inspection logic can always be kept optimal according to the inspection result based on the inspection logic. Therefore, even if the production environment varies, the inspection accuracy of the inspection logic can be improved.

  In order to solve the above problems, an inspection system according to the present invention uses an inspection logic that defines a method for determining good or bad, and uses an inspection module for determining good or bad of an object, and a final result indicating a good or bad determination result. And the above-described inspection control device that selects the main inspection logic used by the inspection module to generate the inspection result.

  By using the inspection logic, the inspection logic is dynamically set according to the change in the production environment, including the inspection module for determining whether the object is good or bad and the inspection control device for controlling the inspection module. An inspection system that improves inspection accuracy can be constructed.

  The inspection control apparatus may be realized by a computer. In this case, a control program for an inspection control apparatus that causes the inspection control apparatus to be realized by a computer by operating the computer as each of the means, and A computer-readable recording medium on which is recorded also falls within the scope of the present invention.

  As described above, the inspection control apparatus according to the present invention uses inspection logic selection means for selecting a main inspection logic used by the inspection module from a plurality of inspection logics, and the inspection module uses the main inspection logic. Based on the generated final inspection result indicating the determination result of the target object and the evaluation result indicating the determination result when determining the good or defective object using the inspection logic different from the main inspection logic Evaluation means for evaluating the inspection accuracy of each inspection logic, and inspection logic setting means for switching the inspection logic selected by the inspection logic selection means as main inspection logic, and the inspection logic selection means includes the evaluation means The main inspection logic is selected based on the evaluated evaluation result.

  As described above, the inspection system according to the present invention uses the inspection logic that defines the good / bad determination method, and uses the inspection module for determining good / bad of the object, and the final inspection result indicating the good / bad determination result. Including the inspection control device described above for selecting the main inspection logic used by the inspection module to generate.

  As described above, in the inspection control method according to the present invention, the main inspection logic used by the inspection module is selected from a plurality of inspection logics, and the inspection module is generated using the main inspection logic. Based on the final inspection result indicating the determination result of the quality of the object and the inspection result for evaluation indicating the determination result when the quality of the object is determined using the inspection logic different from the main inspection logic, An evaluation step for evaluating the inspection accuracy of each inspection logic, and a switching step for switching the inspection logic selected by the inspection logic selection means as the main inspection logic. In the selection step, based on the evaluation result by the evaluation step, Select the main inspection logic.

  Therefore, it is possible to improve the inspection accuracy by dynamically setting the inspection logic according to changes in the production environment.

<Embodiment 1>
An embodiment of the present invention is described below with reference to the drawings. In the present embodiment, as an example, each of the workpiece images obtained by photographing a workpiece (object) is extracted by using an image processing algorithm, and the extracted feature amount is determined based on a threshold value. An inspection control apparatus applied to an inspection system for determining whether a workpiece is good or defective will be described.

[Outline of inspection system]
FIG. 2 is a block diagram showing a schematic configuration of an inspection system to which the inspection control device according to the embodiment of the present invention is applied. As shown in FIG. 2, the inspection system 100 includes a control unit 1, a communication unit 2, an operation unit 3, a display unit 4, and a recording unit 40. The imaging unit C that shoots the workpiece W flowing through the production line and generates a workpiece image and the inspection system 100 are connected via a wireless or wired communication network realized by the Internet, a LAN (local area network), or the like. It is connected.

  The communication unit 2 receives the work image from the imaging unit C. The work image received by the communication unit 2 is supplied to the inspection module 20 via the input / output control unit 10 of the control unit 1.

  The operation unit 3 is used by a user to input an instruction signal for operating the inspection system 100. For example, a remote controller for remotely operating the inspection system 100, operation buttons provided on the inspection system 100 itself, or inspection A mouse and a keyboard connected to the system 100 are included. The instruction signal input by the user using the operation unit 3 is sent to each unit of the control unit 1 via the input / output control unit 10. As a result, the user can operate the inspection system 100.

  The display unit 4 outputs various data such as work images and inspection results recorded in the recording unit 40. For example, the display unit 4 is an LCD (liquid crystal display), a PDP (plasma display panel), or a CRT (cathode-ray). tube) Consists of a display device such as a display. Thereby, the user may visually inspect the quality of the workpiece from the workpiece image generated by the imaging unit C, or may confirm the inspection result of the quality of each workpiece flowing on the production line as necessary. It becomes possible.

  The imaging unit C may be built in the inspection system 100. In this case, it is not necessary to construct a communication network that connects the imaging unit C and the inspection system 100, and the communication unit 2 is not necessary.

  The control unit 1 performs overall control of the inspection system 100, and includes an input / output control unit 10, an inspection module 20, and an inspection control unit 30 therein. The control unit 1 reads various programs recorded in the recording unit 40, controls each unit of the inspection system of the present invention, and performs processing such as inspection or inspection control.

  The inspection module 20 of the control unit 1 determines whether the workpiece W is good or bad by analyzing the workpiece image generated by the imaging unit C based on the set inspection logic. As described above, the inspection logic is control information for the inspection module 20 that defines a good / bad discrimination method executed by the inspection module 20. In the present embodiment, the inspection logic is a combination of an image processing algorithm that defines a method for extracting a feature amount of a workpiece W from a workpiece image and a threshold value for separating the feature amount into a non-defective product and a defective product. Suppose there is. Details of the image algorithm will be described later.

  The inspection control unit 30 of the control unit 1 manages the inspection logic and sets the inspection logic used by the inspection module 20, and constitutes an inspection control device 200 for controlling the inspection module 20. The inspection control device 200 may include the operation unit 3, the display unit 4, and the input / output control unit 10 of the control unit 1.

  The recording unit 40 records a control program executed by the control unit 1, an OS program, and various data read out when the control unit 1 executes inspection or inspection control processing. It is constituted by a storage device. Examples of the various data recorded in the recording unit 40 include a work image generated by the imaging unit C and an inspection result output by the inspection module 20.

  Next, details of the control unit 1 and the recording unit 40 in the inspection system 100 capable of improving the inspection accuracy by dynamically setting the inspection logic according to the change in the production environment will be described.

[Details of inspection system]
FIG. 1 is a block diagram showing a main configuration of an inspection system to which an inspection control apparatus according to an embodiment of the present invention is applied. As shown in FIG. 1, the input / output control unit 10 of the control unit 1 includes an image input control unit 11, an operation receiving unit (result input control unit) 12, and a display control unit (display control unit) 13. . The inspection module 20 includes at least one main inspection unit 21 and at least one sub-inspection unit 22 inside. The inspection control unit 30 includes an inspection logic management unit 31, a statistical processing unit 32, an inspection logic evaluation unit (evaluation unit) 33, and an inspection logic selection unit (inspection logic selection unit / inspection logic setting unit) 34 therein. . The recording unit 40 includes a work information recording unit 41, an inspection result recording unit 42, a statistical data recording unit 43, an evaluation value recording unit 44, and an inspection logic recording unit 45.

  The image input control unit 11 receives the work image generated by the imaging unit C via the communication unit 2. The work image received by the image input control unit 11 is recorded in the work information recording unit 41 of the recording unit 40.

  The operation receiving unit 12 receives an instruction signal input by the user using the operation unit 3. The instruction signal received by the operation receiving unit 12 is supplied to the control unit 1 and transmitted to each unit of the control unit 1 that executes the process instructed by the instruction signal.

  The display control unit 13 controls the display unit 4 to display various data recorded in the recording unit 40. Examples of the various data include a work image recorded in the work information recording unit 41, an inspection result output by the inspection module 20 recorded in the inspection result recording unit 42, and statistical data recorded in the statistical data recording unit 43. And the evaluation value of the inspection logic recorded in the evaluation value recording unit 44, the inspection logic recorded in the inspection logic recording unit 45, and the like. Details of various data recorded in each recording unit will be described later.

  By operating the operation unit 3, the user can input an instruction to the display control unit 13 to display data to be confirmed, and can browse desired data. Further, the display control unit 13 may perform control to display a GUI (graphical user interface) screen on the display unit 4 so that displayed data can be changed or corrected by an operation.

  The inspection module 20 determines whether the workpiece is good or bad by using the inspection logic recorded in the inspection logic recording unit 45. More specifically, (1) a work image obtained by photographing a work is acquired, (2) a feature amount of the work is extracted based on an image algorithm included in the inspection logic, and (3) a threshold value included in the inspection logic. Based on the above, the feature amount is separated to determine whether the workpiece is good or bad. The inspection module 20 includes a main inspection unit 21 and a sub inspection unit 22.

  The main inspection unit 21 inspects the workpiece using the main inspection logic L1 for outputting the final inspection result indicating the good / bad determination result. The inspection result output by the main inspection unit 21 is adopted as the final inspection result for determining the non-defective product / defective product for each workpiece of the production line.

  Therefore, the work specified as a defective product by the inspection result output by the main inspection unit 21 is excluded from the production line, or an alarm is output to notify the user of the presence of the defective product, so that it is determined as a defective product. Appropriate measures are taken to ensure that the workpiece is not accidentally discharged as a good final product.

  The sub-inspection unit 22 inspects the workpiece using the inspection logic L2 (L3...) Other than the main inspection logic among the plurality of inspection logics recorded in the inspection logic recording unit 45. . The inspection result output from the sub-inspection unit 22 is used by the inspection control unit 30 to evaluate the inspection accuracy of the inspection logic.

  The inspection results output from the main inspection unit 21 and the sub inspection unit 22 are recorded in the inspection result recording unit 42. In addition, each inspection result output by each inspection unit based on each inspection logic, including the inspection result output by the main inspection unit 21, is required from the inspection result recording unit 42 by the inspection control unit 30 as an inspection result for evaluation. Is read in response to.

  Therefore, in the following, the inspection results output by the inspection module 20 for the evaluation of the inspection logic are the inspection results for evaluation, and the inspection results actually used for inspecting each work on the production line for non-defective / defective products are finalized. Distinguish it by calling it a test result. When it is not necessary to distinguish between them, it is simply referred to as an inspection result. Therefore, the inspection result output by the main inspection unit 21 is read by the inspection control device 200 as an evaluation inspection result, and is adopted and output as a final inspection result.

  In the present embodiment, the inspection logic and the inspection unit have a one-to-one correspondence, and one main inspection unit 21 that outputs the final inspection result is a plurality of inspection logics recorded in the inspection logic recording unit 45. It is assumed that the final inspection result is output using the main inspection logic L1 (inspection logic A in FIG. 1) that is always set to one. Each sub-inspection unit 22 (sub-inspection unit 22a, sub-inspection unit 22b...) Has inspection logics L2, L3... (In FIG. 1, inspection logics B, C,.・ ・) Shall be used to output each evaluation test result.

  The configuration of the inspection module 20 is not limited to the above. There may be at least one main inspection unit 21 and at least one sub inspection unit 22, and a plurality of main inspection units 21 may be provided, or one sub inspection unit 22 may be provided. When a plurality of main inspection units 21 are provided, a rule (composite determination logical expression) that defines how to obtain the final inspection result from the inspection result output by each main inspection unit 21 is combined with each main inspection unit 21. It may be recorded in association with each (combination of main inspection logic).

  The inspection control unit 30 manages the inspection logic and selects the main inspection logic based on the evaluation of each inspection logic used by the inspection module 20. More specifically, (1) the statistical processing unit 32 reads out the inspection result for each inspection logic, performs statistical processing based on the final inspection result of the main inspection logic, and (2) the inspection logic evaluation unit 33 Based on the statistical data output from the statistical processing unit 32, an evaluation value indicating the inspection accuracy of each inspection logic is calculated. (3) The inspection logic selection unit 34 should be the main inspection logic based on the evaluation value. Select the optimal inspection logic. The inspection logic selection unit 34 sets the selected inspection logic as the main inspection logic in the main inspection unit 21 of the inspection module 20.

  The inspection logic management unit 31 manages the inspection logic recorded in the inspection logic recording unit 45. Specifically, the inspection logic is newly added, updated, or deleted according to the user's instruction, or the inspection logic satisfying the condition is extracted from the inspection logic recording unit 45 according to the condition input by the user and the user The inspection logic selection unit 34 is controlled so as to set the inspection logic designated by the user as the main inspection logic in the main inspection unit 21.

  The workpiece information recording unit 41 records a workpiece image generated by the imaging unit C photographing a workpiece flowing through the production line. In this embodiment, a work ID for individually identifying a work is given to the work flowing through the production line. It is assumed that the work image recording unit 41 records the work image in association with the work ID.

  The inspection result recording unit 42 includes a result (final inspection result) that the main inspection unit 21 has determined whether the workpiece image is good or bad using the set main inspection logic, and the sub-inspection unit 22 The result (inspection result for evaluation) is recorded by determining whether the workpiece is good or bad by using another inspection logic set for each.

  The inspection result recording unit 42 records the final inspection result of good or bad determined by the main inspection unit 21 in association with the work ID. In order to clarify which main inspection logic is used as the result of the determination, a logic ID assigned to each inspection logic is also recorded in association with each other.

  Moreover, the inspection results for good / bad evaluation determined by the main inspection unit 21 and each sub-inspection unit 22 for each combination of the work ID and the logic ID of the inspection logic set in each inspection unit in the inspection module 20 Is recorded. Further, in the inspection result for evaluation, not only the good / bad discrimination result but also the feature quantity extracted from the work image according to the image processing algorithm of the inspection logic is recorded in association with the above combination.

  The statistical data recording unit 43 records the statistical data created by the statistical processing unit 32 performing statistical processing based on the evaluation test result for each inspection logic of each inspection unit. Specifically, for example, the upper limit and / or lower limit of the feature amount in a collection of non-defective products (or defective products) by a certain inspection logic is recorded for each inspection logic. In addition, the feature amount of a work in a certain inspection logic is divided for each section, and the frequency obtained for each section is recorded.

  These statistical data are processed, for example, as a histogram in order to present the test result for evaluation to the user.

  Note that the statistical data generated by the statistical processing unit 32 is not limited to the above example, and any data (average value, median value, variance, upper limit value and lower limit value) that is statistically determined from the test results for evaluation of each test logic. Or a difference with a value).

  The evaluation value recording unit 44 records the evaluation value calculated for each inspection logic by the inspection logic evaluation unit 33 from the statistical data. The evaluation value calculated by the inspection logic evaluation unit 33 based on the inspection result of the main inspection logic is recorded in association with the logic ID.

  The inspection logic recording unit 45 records inspection logic used by each inspection unit of the inspection module 20 for inspection of a workpiece. The inspection logic is managed by a logic ID, and an image processing algorithm and a threshold value are recorded in association with the logic ID.

  The image processing algorithm defines a method by which the inspection module 20 extracts a workpiece feature amount from a workpiece image acquired from the imaging unit C. For example, the image processing algorithm includes a procedure for narrowing down a specific inspection target region from the work image, a calculation formula for calculating a numerical value as a feature amount from the work image, and the like.

  The image processing algorithm is provided with parameters for more specifically defining the above calculation formulas and procedures.

  Therefore, by setting various actual values for the parameters of the image processing algorithm and the above threshold values, it is possible to complete inspection logic consisting of various image processing algorithms and threshold values from a single inspection logic template. It becomes.

  Hereinafter, the inspection logic will be described with a specific example.

(Image processing algorithm 1-2 value centroid area)
The “binary centroid area” image processing algorithm is an algorithm that binarizes a work image and extracts a feature amount based on the area of white or black pixels.

  FIG. 3 is a diagram showing a work image that has been image-processed based on a binary centroid area image processing algorithm, and is a photograph of a PTP (Press Through Package) sheet that wraps tablets and capsules as the work. . FIG. 3A shows a sample of non-defective products (in a state where six tablets are packed in a PTP sheet without excess or deficiency). FIG.3 (b) has shown the sample of the inferior goods (state with a tablet shortage).

  First, a work image is acquired by photographing a PTP sheet (work) packed with six white tablets. In the binary centroid area image processing algorithm, a procedure for converting an image with 256 gray levels taken from the imaging unit C into white or black pixels according to the density difference of the image is defined. Image processing is performed so that only the portion in which the white tablet is packed is white. In the binary centroid area image processing algorithm, a procedure for measuring the area (feature amount) of the white pixel portion from the binarized image is defined.

  As described above, when the area of the white pixel portion as the feature amount is measured, it can be compared with the threshold corresponding to the image processing algorithm in the inspection logic, and as shown in FIG. When the area is equal to or smaller than the threshold value, it is possible to detect the shortage of tablets.

  In addition, if an image algorithm for specifying the white pixel portion and an image processing algorithm for checking the position by grasping the arrangement of the white pixel portion are used, it is possible to check whether or not the position of the center of gravity is correct. .

  In addition, when there are fluctuations in the tablets packed in the PTP sheet to be produced and the color of the tablets is another color other than white, parameters such as what color to focus on and how to binarize are changed. Just do it. Alternatively, if the tablet size varies, the area threshold value may be adjusted.

(Image processing algorithm 2-scratch stain)
The “scratch stain” image processing algorithm is an algorithm for specifying a measurement area in a work image and extracting a feature amount based on a change in density in the measurement area. Thereby, it becomes possible to detect the defect of a workpiece | work.

  FIG. 4 is a diagram showing a work image that has been subjected to image processing based on a flawed dirt image processing algorithm.

  In the flaw stain image processing algorithm, a procedure for specifying the measurement region R of the acquired workpiece image and a procedure for extracting variation in density (brightness) in the measurement region as a feature amount are defined.

  Then, the presence or absence of a defective portion B such as a scratch or a stain can be inspected against the brightness variation threshold.

  Such an image processing algorithm is suitable for visual inspection such as scratches and dirt on a plain workpiece, or missing parts or burrs.

  Also in the flaw stain image processing algorithm, by appropriately adjusting the parameters (values that specify how to determine the measurement area and density variation) according to the nature of the defect to be detected (color, shape, size, etc.) Various image processing algorithms can be provided.

  Note that the image processing algorithm is not limited to the above-described example, and any information may be used as long as it defines a procedure for extracting a feature amount by performing image processing from a work image.

[Outline of inspection system flow]
Next, based on FIG. 5, an outline of the flow of processing of inspection and inspection control performed by each unit of the inspection module 20 and the inspection control unit 30 described above will be described. FIG. 5 is a flowchart showing a processing flow of the inspection module 20 and the inspection control unit 30 in the inspection system 100.

(1) Initial setting In the initial setting step S1, the main inspection logic used for actual good / bad determination is selected from the inspection logic recording unit 45 and set in the main inspection unit 21 before the inspection logic management unit 31 operates the production line. To do.

  Specifically, the user inputs a sample image for initial teaching and inspection contents, and the inspection logic management unit 31 extracts inspection logic candidates. Then, the user evaluates the inspection logic candidates, and the most suitable one is set as the main inspection logic, and the others are set as the sub inspection logic. The user selects (or creates) the main inspection logic.

(2) Inspection In the inspection step S2, inspection is performed by each inspection logic in which the main inspection unit 21 and the sub inspection unit 22 of the inspection module 20 are set, and a non-defective product / defective product is determined. The feature amount and the discrimination result based on the inspection logic are stored in the inspection result recording unit 42 as the inspection result.

(3) Statistical Processing In the statistical processing step S4, the statistical processing unit 32 reads each inspection result for each inspection logic from the inspection result recording unit 42 and performs statistical processing.

(4) Inspection Logic Evaluation In the evaluation step S5, the inspection logic evaluation unit 33 calculates an evaluation value for each inspection logic based on the statistical data obtained in S4.

(5) Inspection Logic Selection The inspection logic selection unit 34 selects an optimal inspection logic as the main inspection logic. Specifically, the evaluation values of the respective inspection logics are compared, and it is determined whether or not the current main inspection logic is optimal compared with other inspection logics (S6). If it is determined that the current main inspection logic is not optimal among the candidates (NO in S6), the inspection logic determined to be optimal among the candidates is selected as a new main inspection logic and set in the main inspection unit 21. .

  As shown in step S3 of FIG. 5, a step of determining whether or not it is time to execute a process for evaluating the inspection result may be inserted. The timing monitoring unit 35 shown in FIG. 1 monitors every state of the inspection system 100, and when a predetermined state is detected, the timing monitoring unit 35 performs a process for evaluating the inspection logic in each unit of the inspection control unit 30 (here, statistical The processing unit 32) is controlled to start.

  Specifically, for example, when the execution time of a series of processes for inspection logic evaluation is set to be evaluated once every hour, the timing monitoring unit 35 performs the last process Is detected, and the statistical processing unit 32 is controlled to start processing.

  Note that the timing for executing the evaluation process is not limited to the above-described example (a configuration in which the evaluation is performed every predetermined time). For example, the inspection module 20 may be configured to evaluate each time one workpiece is inspected, or may be configured to perform evaluation every time a predetermined number of workpieces are inspected.

  Hereinafter, the flow of inspection and inspection control processing in the inspection system 100 shown in FIG. 1 will be described in more detail.

[Details of inspection system flow]
(1) Initial setting Initial setting refers to an operation for selecting a main inspection logic for inspecting the quality of a workpiece to be produced on the production line and setting it in the main inspection unit 21 before operating the production line. That is. The inspection logic management unit 31 executes setting of the main inspection logic. The inspection logic management unit 31 selects a main inspection logic and sets it in the main inspection unit 21 according to conditions and instructions input from the user. Alternatively, the inspection logic selected by the inspection logic selection unit 34 may be set as the main inspection logic.

  FIG. 6 is a flowchart showing a flow of initial setting processing in the inspection system 100.

  First, the inspection logic management unit 31 receives an input of a sample image via the input / output control unit 10 and acquires a sample image of a workpiece (S101). The sample image is a photograph of the same workpiece that flows through the actual production line. Alternatively, a work image obtained in the past and obtained when inspected may be used as a sample image. In any case, the sample image is prepared in the same manner as the work image used for the inspection, and is handled as a sample of the work image of the work to be actually produced.

  It is preferable to prepare as many sample images as possible and any patterns that can be assumed (particularly defective patterns). In this way, highly accurate teaching can be performed, and the inspection can be executed using the highly accurate main inspection logic even in the initial operation stage where the production line tends to become unstable.

  Next, information about the inspection to be carried out in the production line that operates is acquired (S102). Specifically, it is conceivable to accept what the user has entered the content of the inspection what kind of inspection the user wants to perform. The contents of the inspection include a direction inspection for inspecting the direction of the workpiece, an inspection for the presence or absence of a workpiece, the presence or absence of a part that should be assembled in the relevant process, an inspection to check for scratches and dirt on the workpiece, etc. It is done.

  For example, when the user inputs an inspection of dirt, the inspection logic management unit 31 extracts inspection logic based on an image algorithm suitable for detecting dirt from the inspection logic recording unit 45.

  Alternatively, it may be possible to accept what part (region) of the image the user designates which of the image algorithms should be analyzed. For example, when it is desired to inspect an edge portion of a work, if the user designates an area so that the edge portion of the work is covered as an area to be inspected, the inspection logic management unit 31 analyzes the area. Such inspection logic can be selected or added.

  Next, information indicating whether the input one or a plurality of sample images is a good sample or a defective sample is acquired for each sample image (teaching) (S103). In addition, when inputting a sample image in S101, you may input including the information which shows the discrimination | determination result of the quality in the said sample image.

  If the inspection logic management unit 31 determines that the input sample image is a sample of a defective product (NO in S104), the defective portion of the sample image such as which region of the sample image is a defective portion. May be acquired from the user (S105). Thus, by acquiring a lot of information about defective products, a more appropriate inspection logic can be selected and created.

  Subsequently, based on the information acquired in each step described above, the inspection logic management unit 31 infers an image algorithm by referring to a knowledge base or a past case in some cases (S106). Then, a test logic template is generated from the inferred image processing algorithm and recorded in the test logic recording unit 45 (S107). The test logic model recorded in the test logic recording unit 45 may be displayed on the display unit 4 via the display control unit 13 (FIG. 1) (S108). This makes it possible for the user to check what type of inspection logic has been completed based on the information entered by the user, and to set the feature value and threshold in association with this template. The inspection logic can be completed just by making fine adjustments.

  When the user who has confirmed the inspection logic template has added or changed a part of the inspection logic (YES in S109), the inspection logic management unit 31 sends the instruction signal via the operation reception unit 12. In response to the instruction signal, a new test logic template is added or a part thereof is corrected (S110).

  Next, for each created inspection logic, a feature amount and a threshold value for separating good / bad workpieces are calculated based on information acquired by processing such as teaching (S110). As a result, it is possible to complete a plurality of inspection logics associated with all feature quantities and thresholds from the inspection logic template and record them in the inspection logic recording unit 45. Here, when it is desired to discriminate between good and bad using two or more feature quantities (when composite judgment is executed), a logical expression for the composite judgment may also be created.

  Finally, the inspection logic selection unit 34 selects the inspection logic with the highest evaluation as the initial main inspection logic based on the evaluation performed on each inspection logic (main inspection logic candidate) created in S111. (S112). The inspection logic management unit 31 sets the main inspection logic selected by the inspection logic selection unit 34, and the initial setting is completed.

  The evaluation of each inspection logic for the initial setting may be performed by a method similar to the evaluation value calculation procedure executed by the inspection logic evaluation unit 33 after the production line is operated. The difference from the evaluation method in actual operation is that the evaluation is performed on the hypothetical inspection result on the assumption that the input sample image is inspected, not the evaluation based on the work image flowing on the actual production line. . Details of the inspection logic evaluation method will be described later.

  In the above description, the inspection logic selection unit 34 selects the inspection logic with high evaluation as the initial main inspection logic. However, the present invention is not limited to this. For example, based on the feature amount calculated in S111, the histogram created by the statistical processing unit 32 is presented to the user for each inspection logic, and the user confirms the histogram and is suitable for the initial main inspection logic. It is also possible to select a configuration.

  The initial main inspection logic setting before the production line operation does not have to be performed by using each unit (FIG. 1) of the inspection control device 200. The main inspection in which the user determines the optimum at the initial time point from past experience. The initial setting may be completed by selecting a logic from candidate inspection logics or creating a new inspection logic (inputting inspection contents, feature amounts, threshold values, etc.). Alternatively, the configuration may be such that the user selects the main inspection logic from the inspection logic completed in advance and sets it in the main inspection unit 21.

  The above is the initial main inspection logic setting procedure to be set before the production line is operated. The main inspection logic to be set before the production line is operated is evaluated after the actual production line is operated even if the correct main inspection logic is selected by the detailed evaluation (or at the discretion of the expert) based on past experience. However, it is almost impossible for the initial main inspection logic to remain the most suitable main inspection logic. This is because the above-described various production environment fluctuations are assumed, and the optimum main inspection logic always changes in accordance with the fluctuations.

  Therefore, in the following, in order to maintain high inspection accuracy by selecting an optimal main inspection logic in accordance with changes in the production environment, in the inspection performed by the inspection module 20 on the production line in operation after the initial setting. The process flow of the inspection control apparatus 200 will be described in more detail.

  In order to evaluate and dynamically reset the optimal inspection logic while the production line is in operation, the inspection control apparatus 200 outputs an evaluation that is output as a result of the inspection that the inspection module 20 has actually performed on the workpiece flowing through the line. Use inspection results. First, the flow of the inspection process in the inspection module 20 and the inspection result recorded in the inspection result recording unit 42 will be described in detail.

(2) Inspection FIG. 7 is a flowchart showing a flow of inspection processing in the inspection system 100.

  First, the imaging unit C receives an imaging command signal from a user's operation or from the inspection system 100 at a timing when a workpiece is carried into a line for inspection (S201). Receiving the imaging command signal, the imaging unit C captures the loaded workpiece and generates a workpiece image (S202). At this time, the work ID of the photographed work may be embedded in the data of the work image. The image input control unit 11 (FIG. 1) that has received the work image from the imaging unit C records the received work image in the work information recording unit 41 in association with the work ID (S203).

  Subsequently, when a new work image is acquired via the image input control unit 11 (YES in S204), each inspection unit (main inspection unit 21, sub-inspection unit 22) of the inspection module 20 is set in advance. The inspection of the workpiece is executed using the inspection logic.

  For example, the coordinate system of the work image is corrected based on the method of specifying the center of gravity position specified by the image processing algorithm included in the inspection logic (S205), and the inspection target region is also determined based on the method of specifying the measurement region specified by the algorithm. Is specified (S206). Then, based on the image processing algorithm, analysis (image processing) of the identified region is executed (S207), and a feature amount is measured based on the analysis result (S208). For example, it is assumed that the grayscale image is binarized and the area of a pixel of a predetermined color is measured, or the variation in brightness of a specified area is measured.

  Next, the feature amount extracted from the work image based on the image processing algorithm in S205 to S208 is compared with a threshold value included in the inspection logic (S209). If it is determined that the feature amount exceeds the threshold (YES in S209), the workpiece of the workpiece image is determined as a non-defective product (S210). On the other hand, when it is determined that the value is equal to or less than the threshold value (NO in S209), the workpiece is determined as a defective product (S211). Then, the inspection result and the extracted feature amount are recorded in the inspection result recording unit 42 in association with the work ID (S212).

  In this way, when the inspection result for one inspection logic is recorded, the same processing is performed for the remaining inspection logics (if YES in S213, the process returns to S205), and the inspection results for all inspection logics are obtained. When the recording is completed (NO in S213), the inspection result based on the main inspection logic is output as the final inspection result (S214), and the inspection is terminated.

  In the above description, when the feature amount exceeds the threshold value in S209, the workpiece is determined to be a non-defective product. However, the present invention is not limited to this. For example, when the feature amount is large, the workpiece is defective. Further, it can be configured to discriminate it as a non-defective product when it falls below the threshold.

  FIG. 8 is a diagram illustrating an example of inspection results recorded in the inspection result recording unit 42. In the present embodiment, as shown in FIG. 8, the inspection result data recorded in the inspection result recording unit 42 is used by the table 81 for recording the final inspection result by the main inspection logic and each inspection unit of the inspection module 20. It is comprised from the table 82-84 ... which records the test result for evaluation of each test | inspection logic. The tables 82, 83, and 84 are tables of evaluation inspection results output based on the inspection logics A, B, and C recorded in the inspection logic recording unit 45, respectively.

  In the table 81 for recording the final inspection result based on the main inspection logic, at least the work ID (C11), the good / bad discrimination result for each work (that is, the final inspection result (C12)) associated therewith, As the final inspection result, the logic ID (S13) of the main inspection logic indicating which inspection logic is the result obtained as the main inspection logic is recorded.

  The data structure of the table 81 is not limited to the above. As shown in FIG. Information such as (C14), inspection date (C15), and work image (C16) used in the inspection may be recorded in association with each other. Further, the logic ID (C17) of the sub inspection logic used by the sub inspection unit 22 when the inspection of the workpiece is executed in parallel may be recorded in association with each other. As a result, it is possible to link the final inspection result and the inspection inspection result with another inspection logic, and the inspection result can be compared between the main inspection logic and the inspection logic.

  In addition, when there are a plurality of adopted main inspection logics (a plurality of main inspection units 21), it is preferable to record a composite determination logical expression (C18). Thereby, it is possible to link the inspection results of each of the plurality of main inspection logics, and it is possible to obtain one final inspection result (C12) from the inspection results of each main inspection logic aggregated based on the logical expression. . In the example shown in FIG. 8, two main inspection logics, inspection logic A and inspection logic B, are set, and the logical expression stipulates that the final inspection result is taken from the logical product of the above inspection logics. In both the inspection logic A and the main inspection logic B, only a work that has been determined to be non-defective is finally determined to be non-defective.

  Each of the tables 82 to 84... Records at least a work ID (C21), a feature amount (C22), and an evaluation test result (C23) associated therewith.

  The data structures of the tables 82 to 84... Are not limited to the above, and parameters (C24), threshold values (C25), etc. in the image processing algorithm of the inspection logic are recorded for each work as necessary as shown in FIG. You may keep it. In addition, since these pieces of information are recorded in the inspection logic recording unit 45 for each logic ID of the inspection logic, the logic ID of the inspection logic is recorded instead of C24 and C25, and A data structure for performing association may be used.

  As described above, the inspection results of all the inspection logics recorded in the inspection result recording unit 42 are read as necessary in the next statistical process as the inspection results for evaluation.

(3) Statistical Processing FIG. 9 is a flowchart showing the flow of statistical processing in the inspection system 100. The statistical processing unit 32 (FIG. 1) starts statistical processing based on the control of the timing monitoring unit 35 when it is time to evaluate each inspection logic.

  The statistical processing unit 32 acquires the test result for evaluation (table 82 shown in FIG. 8) of the test logic to be evaluated (for example, test logic A) from the test result recording unit 42 (S301). Here, it is determined whether or not statistical processing of the acquired table 82 (inspection logic A) has been performed so far, and when statistical processing is performed for the first time (YES in S302), statistical data is prepared. More specifically, in R <b> 1 to R <b> 3 of the table 82, a non-defective section and a defective section are set based on the feature amount and threshold value for each workpiece, and the feature amount is further classified for each predetermined section. (S303).

  Subsequently, the inspection result of the unadded work (for example, the inspection result “OK” of the work ID 0001 in the record R1) is acquired from the inspection inspection result (table 82) (S304). The statistical processing unit 32 determines the inspection result of the acquired record, and if the inspection result is a non-defective product (YES in S305), the frequency of the record (work) is set to one for the good product section to which the feature amount belongs. to add. On the other hand, if the inspection result is a defective product (NO in S305), one work frequency is added to the corresponding defective product section (S307).

  If the inspection result of the non-addition process still remains (for example, records R2 and R3 in the table 82) (YES in S308), the process returns to S304, and the frequency addition process is repeated. On the other hand, when the additional processing is completed for the inspection results of all the workpieces in one inspection logic (NO in S308), the statistical processing of the inspection logic is ended.

  Therefore, if unprocessed inspection logic (for example, inspection logic B or inspection logic C) remains (YES in S309), the process returns to S301, and the inspection result for evaluation of unprocessed inspection logic (for example, inspection logic B). Table 83) is obtained, and the subsequent processing is repeated.

  When the statistical processing is completed and statistical data is created for all inspection logic evaluation test results (NO in S309), the above-described statistical data is recorded in the statistical data recording unit 43 (S310).

  FIG. 10 is a diagram schematically illustrating data input to and output from the statistical processing unit 32. Here, the statistical data output by the statistical processing unit 32 and recorded in the statistical data recording unit 43 is represented by a histogram. In the example illustrated in FIG. 10, statistical processing is performed based on the inspection result record 85 (including the inspection result and feature amount of each inspection logic and the final inspection result) for the workpiece X recorded in the inspection result recording unit 42. The unit 32 executes statistical processing and records the output statistical data 90 in the statistical data recording unit 43.

  The statistical data 90 includes histograms 91 to 93... Created for each inspection logic, and the histograms 91 to 93 represent the statistically processed inspection results of the inspection logics A to C, respectively.

  In each histogram, the vertical axis represents the frequency, and the horizontal axis represents the feature amount. A block belonging to a section below the threshold 94 with a threshold 94 as a boundary is a non-defective product, and a block belonging to a section exceeding the threshold 94 is a defective product. This indicates that each inspection logic has been determined.

  Further, the color coding of the bar graph indicating the frequency of the work indicates the classification of the non-defective product and the defective product based on the final inspection result 86. Therefore, a bar graph with halftone dots is determined as good by the main inspection logic (in this case, inspection logic A), and a bar graph with diagonal lines is determined as defective by the main inspection logic. Is shown.

  Here, when the inspection result record 85 is newly added, the statistical processing unit 32 counts the frequency of the work X of the inspection result record 85 in an appropriate section of each histogram based on the inspection result and feature amount of each inspection logic. 1 is added. More specifically, in the histogram 91 of the inspection logic A, a block of frequency 1 with the final inspection result 86 being non-defective (halftone dot) is added to the section of the inspection logic A feature amount 10. In the inspection logic B histogram 92, a halftone dot block is added to the section of the inspection logic B feature 20. In the histogram 93 of the inspection logic C, a halftone dot frequency block is added to the section of the inspection logic C feature amount 30.

  As described above, the statistical data generated by the statistical processing unit 32 and recorded in the statistical data recording unit 43 is read by the inspection logic evaluation unit 33 as necessary in the next inspection logic evaluation process.

  Note that the statistical data does not necessarily have to be recorded as a histogram as shown in FIG. 10, and any statistical data can be extracted as long as the inspection logic evaluation unit 33 can extract information necessary for executing the inspection logic evaluation. Such a data structure may be used. For example, as shown in FIG. 11, various data (upper / lower limit values, average values, variances, median values, mode values, etc. of feature values for each good / bad product in the final inspection result) are associated with each inspection logic. Statistical data having a table structure may be recorded in the statistical data recording unit 43.

  However, when the statistical data is displayed on the display unit 4 and presented to the user, it is preferably output as a histogram. If a histogram as shown in FIG. 10 is presented, the user discriminates how each inspection logic determines whether the work is good or bad, and how the inspection result is different from the final inspection result. It can be easily visually confirmed whether or not.

  Here, in the above-described example, the color coding of the histogram bar graph is performed based on the final inspection result 86, but is not limited thereto. When a histogram is generated for each inspection logic, color classification may be performed according to the inspection result of each inspection logic. Thereby, the user can confirm at a glance the frequency of non-defective / defective products with each inspection logic. In the case of a configuration in which the final inspection result 86 is determined by the composite determination, statistical data may be created using the feature amounts of a plurality of main inspection logics employed as shown in FIG.

  In addition, when the statistical processing unit 32 determines that the inspection results of good and bad do not match in all inspection logics, the statistical processing unit 32 is configured not to add such an inspection result record 85 as statistical data. It is also possible.

(4) Inspection Logic Evaluation FIG. 13 is a flowchart showing the flow of inspection logic evaluation processing in the inspection system 100. When the statistical processing unit 32 records the statistical data in the statistical data recording unit 43, the inspection logic evaluation unit 33 (FIG. 1) starts the evaluation process using the statistical data.

  The inspection logic evaluation unit 33 acquires the statistical data of each inspection logic from the statistical data recording unit 43 (S401), and the evaluation unit used to calculate the evaluation value from the statistical data is a recording unit that records the evaluation equation (For example, the evaluation formula may be recorded in the evaluation value recording unit 44) (S402). Incidentally, if the evaluation formula is more recorded, the user may select each time, it may be set whether to adopt the advance which evaluation formula. Moreover, it is good also as a structure which can input directly the evaluation formula which a user employ | adopts.

  Next, the inspection logic evaluation unit 33 calculates an evaluation value of each inspection logic from the statistical data according to the acquired evaluation formula (S403). Details of the evaluation value calculation process will be described later. After completing the calculation of the evaluation values for all the inspection logics, the inspection logic evaluation unit 33 records the calculated evaluation values in the evaluation value recording unit 44 for each inspection logic (S404).

  FIG. 14A is a diagram illustrating an example of the evaluation value of the inspection logic recorded in the evaluation value recording unit 44. As shown in FIG. 14A, the evaluation value calculated by the inspection logic evaluation unit 33 for each inspection logic is recorded in association with the logic ID. Here, when a plurality of evaluation expressions are recorded in the inspection system 100, information (for example, evaluation expression ID) for specifying the evaluation expression is also included so that it is possible to determine which evaluation expression is used to determine the evaluation value. It is preferable to record in association.

  FIG. 14B is a diagram illustrating an example of an evaluation formula used for evaluation value calculation. As shown in FIG. 14B, each evaluation formula is managed by identification information (evaluation formula ID) for uniquely identifying the evaluation formula. Thus, since the evaluation value and the evaluation equation can be linked by the evaluation equation ID, for example, the evaluation value of each inspection logic shown in FIG. It can be seen that the value is calculated by “Lower limit value—Upper limit value of non-defective product”.

  In the above description, the evaluation formula employed first is selected, and the evaluation value is calculated based on the selected evaluation formula. However, the present invention is not limited to this. For example, each evaluation value is calculated based on all the recorded evaluation formulas, and a final evaluation value is selected (or one evaluation value is finally calculated from a plurality of evaluation values). Also good.

  Next, details of the evaluation value calculation process shown in S403 of FIG. 13 will be described by taking as an example the case where the evaluation expression of “Expression 1” of FIG. 14B is adopted.

  First, the inspection logic evaluation unit 33 acquires A: non-defective upper limit value (feature amount) from statistical data (for example, FIG. 11) for one inspection logic (S41). Next, B: The lower limit value of the defective product is acquired (S42). Finally, according to the evaluation formula (formula 1: lower limit value of defective product−upper limit value of non-defective product), a difference (in this case, B−A) of the acquired values is calculated and used as an evaluation value (S43). Here, “non-defective product” and “defective product” indicate “good product” and “defective product” indicated by the final inspection result by the main inspection logic.

  According to the above-described evaluation value calculation method, when a feature amount is large, a defective product is obtained. Therefore, it is possible to highly evaluate an inspection logic in which a good product and a defective product are correctly separated based on the feature amount. In other words, the longer the distance between the minimum feature value of a defective product and the maximum feature value of a non-defective product (the above-mentioned “BA”), the inspection logic can safely and correctly separate the non-defective product and the defective product. Can be considered.

  Contrary to the above-described example, when the feature amount is small, it becomes a defective product, and “Formula 4: Lower limit value of non-defective product—Upper limit value of defective product” in FIG. What is necessary is just to obtain | require the distance of this feature-value and the largest feature-value of inferior goods, and to output it as an evaluation value.

  Here, the method of calculating the evaluation value is not limited to the above. In the inspection system 100 of the present invention, any evaluation method can be adopted using various pieces of information of statistical data.

  FIGS. 15A to 15F are diagrams illustrating other examples of the evaluation value calculation method using statistical data (histograms).

  The example shown in FIG. 15A shows a method of calculating the evaluation value based on the distance d between the non-defective product set and the defective product set, as in the above-described example. As the value of the distance d is larger, it means that the non-defective product and the defective product are correctly and safely separated, and the evaluation becomes higher.

  The example shown in FIG. 15B shows a method of calculating an evaluation value based on the distance d between the average value Gμ of feature values in a set of non-defective products and the average value Bμ of sets of defective products. As the value of the distance d is larger, it means that the non-defective product and the defective product are correctly and safely separated, and the evaluation becomes higher.

  The example shown in FIG. 15C shows a method for calculating the evaluation value based on the variance Gσ of the feature quantity of the non-defective product and / or the variance Bσ of the feature quantity of the defective product. A large variance means that the measured feature values vary within a set of non-defective products (defective products), and the reliability of the measurement procedure for the feature values is low. Therefore, the evaluation logic is higher as the inspection logic has a smaller variance value.

  The example shown in FIG. 15D shows a method of calculating an evaluation value based on a differential value of the distance d between the non-defective product set and the defective product set (speed at which the distance d narrows). As described above, the greater the distance d, the higher the evaluation, and the greater the differential value of the distance d (the greater the momentum at which the distance d is narrowed), the lower the evaluation. That is, as shown in FIG. 15 (d), the smaller the difference between the distance d1 and the distance d2, the higher the evaluation.

  The example shown in FIG. 15E is a method for calculating an evaluation value based on a difference between a distance d1 between a non-defective product set and an inspection logic threshold value and a distance d2 between a defective product set and an inspection logic threshold value. Show. When the difference between the distance d1 and the distance d2 is large, only one set of non-defective products and defective products is close to the threshold value. If only one set is approaching the threshold value, for example, if the set of defective products is near the threshold value, it means that there is a high possibility of an inspection failure in which the defective product is identified as a non-defective product. . Therefore, it is ideal that the distance d1 and the distance d2 are equal, and the smaller the difference between the distance d1 and the distance d2, the higher the evaluation.

  Furthermore, when the distance d2 (the distance between the defective product and the threshold value) is weighted and the distance d2 is smaller (that is, when the set of defective products is closer to the threshold value), the evaluation is particularly lowered. The inspection logic evaluation unit 33 may be configured to calculate such an evaluation value. This is based on the idea that an inspection failure (missing) that determines a defective product as a non-defective product is a more serious accident than an inspection failure that determines a good product as a defective product. As a result, it is possible to eliminate the inconvenience that an inspection logic that is overlooked is erroneously selected as the main inspection logic.

  When a specific example of the evaluation formula is shown, “Formula 3: | (Lower limit value of defective product−Upper limit value of non-defective product) − (Lower limit value of defective product−Threshold value) / (Threshold value−Upper limit value of non-defective product)” in FIG. | ". In this case, the smaller the evaluation value, the higher the evaluation.

  Alternatively, in another evaluation method, the evaluation value may be calculated based on the number of workpieces that the main inspection logic has determined to be defective (the number of missed workpieces) that the inspection logic has determined to be non-defective. Good. In this case, the evaluation becomes lower as the number of missed items increases.

  FIG. 16 is a graph showing the transition of the evaluation value according to the threshold position. The vertical axis represents the evaluation value calculated based on Equation 3 (the smaller the evaluation is, the higher the evaluation is), and the horizontal axis represents the threshold position. The position of the threshold means which value the threshold value indicates between the non-defective product upper limit value P and the defective product lower limit value Q in FIG. That is, the distance d1 from the acceptable product upper limit P to the threshold and the distance d2 from the threshold to the defective product lower limit Q are shown.

  The left end 0 of the horizontal axis indicates the non-defective product upper limit value P, and the right end 0.9 indicates the defective product lower limit value Q. The closer the threshold approaches the defective product lower limit value Q, the closer to the defective product (the more it goes to the right), It can be seen that the evaluation value increases rapidly and the evaluation decreases. The degree to which the evaluation value increases is greater than the degree at which the threshold approaches the non-defective product upper limit P. Thereby, it is possible to prevent an inspection logic that may be overlooked from being erroneously selected as the main inspection logic.

  The example shown in FIG. 15 (f) shows a method for calculating an evaluation value based on the degree of similarity compared with model statistical data, which indicates an ideal state of statistical data. For example, when the statistical data is a histogram, a model histogram 95 as model statistical data is prepared in advance, the shape of the model histogram 95 is compared with the actual histogram, pattern matching is performed, and the similarity is calculated. The similarity between the shape and the model histogram 95 means that the inspection logic can output an ideal inspection result, and the evaluation becomes higher.

  As described above, the evaluation value calculated by the inspection logic evaluation unit 33 and recorded in the evaluation value recording unit 44 is read by the inspection logic selection unit 34 as necessary in the next inspection logic selection process.

(5) Inspection Logic Selection FIG. 17 is a flowchart showing a flow of inspection logic selection processing in the inspection system 100. When the inspection logic evaluation unit 33 records the evaluation value of each inspection logic in the evaluation value recording unit 44, the inspection logic selection unit 34 (FIG. 1) starts the inspection logic selection process using the evaluation value.

  The inspection logic selection unit 34 acquires the evaluation value of each inspection logic from the evaluation value recording unit 44 (S501). Next, based on the acquired evaluation value, the inspection logic with the highest evaluation (optimal as the main inspection logic) is specified (S502).

  Subsequently, the inspection logic selection unit 34 matches the specific inspection logic specified in S502 with the main inspection logic (for example, inspection logic A (FIG. 1)) currently set in the main inspection unit 21 (FIG. 1). It is determined whether or not to perform (S503). If it is determined that the specific inspection logic is the inspection logic A, the inspection logic selection unit 34 adopts the inspection logic A as it is because it is not necessary to switch the inspection logic (S504). On the other hand, if the inspection logic selection unit 34 determines that the specific inspection logic is different from the inspection logic A, the inspection logic selection unit 34 newly selects the specific inspection logic as the main inspection logic (S505), and selects the main inspection logic L1 of the main inspection unit 21. Then, the inspection logic A is switched to the specific inspection logic (S506).

<Embodiment 2>
As described in the above-described embodiment, according to the configuration of the inspection control apparatus 200 (FIG. 1) and the inspection control method (S4 to S7 in FIG. 5) of the present invention, the latest inspection results are evaluated and always optimal. The inspection system 100 (inspection module 20) can be inspected by setting the inspection logic dynamically in accordance with changes in the production environment. It becomes possible to improve.

  Next, using a specific example, the inspection logic switching process and the effect realized by the inspection control apparatus 200 of the present invention will be described in more detail. In the present embodiment, the inspection system 100 of the present invention is an inspection system that inspects dirt generated on white paper as a workpiece. The inspection logic recording unit 45 of the inspection system 100 includes two inspection logics, that is, a first inspection logic that employs the above-described binary centroid area image processing algorithm and threshold = 300 (area), and a flaw stain image processing algorithm. And the second inspection logic adopting threshold = 150 (defect degree) are recorded in advance.

  FIG. 18 is a diagram showing a sample image of a work used by the inspection system 100 in the dirt inspection. As illustrated in FIG. 18, the imaging unit C captures an imaging range R of the workpiece 50 flowing through the production line, and generates a workpiece image 51 (52). The generated work image 51 (52) is recorded in the work information recording unit 41 via the image input control unit 11. The work image 51 shows a photograph of a non-defective work, and the work image 52 shows a work that is a defective product (which should be detected as a stain). In this embodiment, a non-defective sample image such as a work image 51 and a defective sample image such as a work image 52 are input in advance in order for the inspection logic management unit 31 to perform initial setting.

  In the present embodiment, as shown in FIG. 19, as an example of fluctuations in the production environment, as the number of production increases, the production line becomes stable (improvement of worker skill, production facilities have become stable, etc. It is assumed that a situation occurs in which the size of the dirt adhering to the work (blank paper) gradually decreases.

[Selection of main inspection logic at initial setting]
FIG. 20 is a diagram illustrating an initial setting procedure performed by the inspection logic management unit 31. Before operating the production line, the inspection logic management unit 31 of the inspection control device 200 first sets the inspection logic to be set in each inspection unit of the inspection module 20. In this embodiment, the evaluation of the inspection logic for initial setting is the same as the evaluation method after the production line is operated. That is, the inspection logic evaluation unit 33 evaluates each inspection logic using the sample image, and the inspection logic selection unit 34 selects the main inspection logic. The inspection logic management unit 31 sets an initial main inspection logic in the main inspection unit 21 in accordance with the selection of the inspection logic selection unit 34.

  As shown in FIG. 20, the inspection logic evaluation unit 33, the statistical data 53 of the inspection result of the first inspection logic, the statistical data 54 of the inspection result of the second inspection logic for the acquired sample image of the workpiece (blank), To get. Then, as an evaluation value calculation method, “Formula 2: (Lower limit value of defective product−Upper limit value of non-defective product) / (Upper limit value of good product−Lower limit value of non-defective product)” shown in FIG. An evaluation value is calculated for each of the two inspection logics.

When calculating the evaluation value based on Equation 2,
Evaluation value of first inspection logic: (764-23) / (23-1) = 33.681818
Evaluation value of second inspection logic: (255-50) / (50-28) = 9.3318181
Thus, the inspection logic selection unit 34 selects the first inspection logic (including the binary centroid area image processing algorithm) with high evaluation as the initial main inspection logic.

  Here, “lower limit value of defective product−upper limit value of non-defective product” in the above formula 2 indicates the distance between the non-defective product set and the non-defective product set, and the evaluation increases as the distance value increases.

  On the other hand, “the upper limit value of the non-defective product−the lower limit value of the non-defective product” indicates the variation of the non-defective product. When the non-defective product variation is large, it is considered that the parameters are not appropriate in the image processing algorithm, the feature quantity measurement procedure is not appropriate, and the like. Therefore, the evaluation becomes lower as the variation value is larger.

  As described above, if the evaluation value obtained by dividing the distance by the variation is used, even inspection logic in which feature quantities of different units are measured by an entirely different image processing algorithm as in the above example can be easily obtained. It becomes possible to compare and evaluate inspection accuracy.

[Changes in production quantity and inspection result statistics after production line operation]
FIG. 21 is a diagram showing the number of productions and the transition of statistics of inspection results by the first inspection logic employing the binary centroid area image processing algorithm.

  As shown in FIG. 21, as the number of production increases to 50, 100,..., The number of workpieces that are determined to be defective increases, and the lower limit value (defective product lower limit value Q) of the measured feature value decreases. Then, it approaches the threshold (and the set of good products). When the production number reaches 150, the lower limit value Q of the defective product reaches the threshold value. If the first inspection logic continues to be adopted at this time, when 200 workpieces are produced, the lower limit value Q of the defective product divides the threshold value, and therefore an inspection that erroneously determines that the defective product is a non-defective product. Defects will occur.

  FIG. 22 is a diagram showing the number of productions and the transition of statistics of inspection results by the second inspection logic employing the flawed dirt image processing algorithm.

  As shown in FIG. 22, although the variation in the set of non-defective products is larger than that in the first inspection logic, the distance between the defective product lower limit Q and the threshold value is stable even when the number of production increases. Therefore, it can be seen that the inspection accuracy of the second inspection logic is stable regardless of the increase in the number of production.

[Changes in evaluation values]
FIG. 23 is a graph showing the transition of the evaluation value of each inspection logic as the number of production increases. The vertical axis represents the evaluation value, and this evaluation value is calculated according to the above-described equation 2. The horizontal axis indicates the number of products produced. As shown in FIG. 21, in the first inspection logic in which the distance between the defective product set and the non-defective product set decreases as the production number increases, the evaluation value decreases as the production number increases (the evaluation decreases). ). On the other hand, as shown in FIG. 22, in the second inspection logic in which the distance from the set of non-defective products and the threshold is stable regardless of the increase in the number of products produced, the evaluation value is almost constant.

  From the graph of FIG. 23, it can be seen that in order for the inspection system 100 to maintain a constant inspection accuracy, the main inspection logic must be switched from the first inspection logic to the second inspection logic at time T2. .

  The inspection logic selection unit 34 of the inspection control device 200 detects that the evaluation value is reversed based on the evaluation value calculated by the inspection logic evaluation unit 33, and changes the main inspection logic from the first inspection logic to the second inspection logic. And switch.

  Note that the timing of main inspection logic switching is not limited to the point in time when the evaluation value such as T2 is reversed. For example, the transition of the evaluation value for a certain period (for example, from T0 to T1 in FIG. 23) is recorded, and the evaluation value is reversed by referring to the differential value (evaluation momentum of the evaluation value) of each evaluation value. It is also possible to configure the inspection logic selection unit 34 to predict the time point and calculate the appropriate switching timing. Alternatively, based on the differential value of the evaluation value of the main inspection logic, the inspection logic selection unit 34 is configured to start the process of switching to an appropriate main inspection logic when the evaluation value drops at a certain moment or more. You can also.

  According to the above configuration, the inspection logic evaluation unit 33 uses the first inspection logic as the main inspection logic and the reference based on the final inspection result by the main inspection logic based on a predetermined evaluation expression (for example, Expression 2). An evaluation value of the second inspection logic as the sub inspection logic is calculated. The evaluation value is reversed when the production number reaches 150 pieces, and the inspection logic selection unit 34 that detects the evaluation value, as shown in FIG. 24, the second inspection logic is optimal when the production number reaches 150 pieces. The main inspection logic is switched from the first inspection logic to the second inspection logic.

  As a result, it is possible to avoid defective inspection. In other words, if the first inspection logic continues to be adopted without switching the main inspection logic even after the time when 150 pieces have been reached, an overlooked inspection defect will occur when the production number reaches 200 pieces. It was. However, by switching the main inspection logic to the second inspection logic, the inspection module 20 of the inspection system 100 can correctly determine good or bad even when the production number reaches 200 (FIG. 24). .

  The inspection control apparatus 200 of the present invention is not limited to the above-described embodiments. The inspection control apparatus 200 may further include a threshold adjustment unit (threshold adjustment unit) 36 and / or a statistical data correction unit (inspection result correction unit) 37 inside the inspection control unit 30 (FIG. 1). Good.

  FIG. 25 is a block diagram showing a detailed configuration of the inspection control unit 30 in the inspection control apparatus 200 of the present invention. Although not shown here, it is assumed that the inspection control unit 30 includes each unit of the inspection control unit 30 illustrated in FIG. The description of those functions will not be repeated.

<Threshold adjustment function>
The threshold adjustment unit 36 calculates and adjusts the optimum threshold of the inspection logic based on the statistical data for each inspection logic recorded in the statistical data recording unit 43. When statistical data of a certain inspection logic is newly created or updated by the statistical processing unit 32, the threshold adjustment unit 36 reads the statistical data from the statistical data recording unit 43 and calculates an optimum threshold value.

  Here, when the calculated threshold value is different from the current threshold value of the inspection logic, the calculated new threshold value is newly set as the threshold value of the inspection logic, and the inspection recorded in the inspection logic recording unit 45 Update the logic threshold.

  At this time, a new inspection logic in which only the threshold value is changed from the original inspection logic is generated without updating the original inspection logic recorded in the inspection logic recording unit 45 and added to the inspection logic recording unit 45. You may do it.

  As a result, the inspection logic can always be kept in an optimum state, and the inspection accuracy can be improved. In particular, there is no need to change the image processing algorithm itself, and in the case where the inspection accuracy of the inspection logic is improved only by fine adjustment of the threshold value, the process until obtaining the optimal inspection logic can be shortened, which is effective.

  The method for calculating the optimum threshold value is not particularly limited. For example, an intermediate value between the upper limit value of the non-defective product and the lower limit value of the non-defective product is calculated as the optimum threshold value. It is possible to do.

  FIGS. 26A and 26B are diagrams showing the contents of the threshold adjustment process of the threshold adjustment unit 36. FIG. The histogram shown in FIG. 26A shows a state in which the inspection result 55 of a certain work is output based on a certain inspection logic, and the frequency of a certain section is added by one based on the feature amount of the work. Here, H1 represents the threshold value of the inspection logic.

  Here, when the inspection result 55 is added to the section B1, the distance D1 between the non-defective product set and the defective product set remains the same as D1. Therefore, the threshold adjustment unit 36 recognizes that the intermediate value does not change and does not update the threshold.

  On the other hand, when the inspection result 55 is added to the section B2, the distance D1 changes to the distance D2. Therefore, the threshold value adjustment unit 36 calculates a new intermediate value (H2) as shown in FIG. 26 (b) in accordance with the fluctuation of the upper limit value of the non-defective product, and sets the value of H2 as a new threshold value. Update inspection logic.

  In the above description, the timing for recalculating and resetting the threshold is the timing at which the workpiece inspection result is output. However, the timing is not limited to this. For example, the threshold value is recalculated when the upper limit (lower limit) of the set of non-defective products (defective products) fluctuates as in the case where the frequency is added to the section B2 in FIG. Also good. Alternatively, it may be recalculated when the intermediate value, average value, mode value, etc. of the non-defective product set change, or may be recalculated every time the inspection result is output a predetermined number of times.

  Furthermore, in the above description, when the threshold adjustment unit 36 calculates a new threshold, the threshold is automatically updated. However, the present invention is not limited to this, and notifies the user that the intermediate value has changed. An inquiry may be made as to whether or not to update to a new threshold, and the threshold may be updated after receiving a signal instructing the update.

<Inspection defect correction function>
The statistical data correction unit 37 corrects the statistical data of each inspection logic recorded in the statistical data recording unit 43. Specifically, a test result input by the user using the operation unit 3 (FIG. 1) is received via the operation receiving unit 12, and the test result is reflected in the statistical data. Alternatively, statistical data different from the user's inspection result is corrected in accordance with the user's inspection result as compared with the inspection result when the user recorded in advance is visually inspected.

  In this case, it is preferable to configure the statistical data correction unit 37 so that the statistical data to be corrected is corrected offline (locked so that it cannot be read). Thereby, since the evaluation by the inspection logic evaluation unit 33 can be received in a state after the correction is completed, the correction content can be correctly reflected in the evaluation process.

  FIG. 27 is a diagram illustrating the contents of the statistical data correction process of the statistical data correction unit 37. It is assumed that H1 is a threshold value of the inspection logic. The display control unit 13 presents the histogram 61 to the user via the display unit 4. Thereby, the user visually confirms the inspection result 56 determined to be defective in a certain work, and confirms that it is different from the inspection result of his own visual inspection determined to be a non-defective product. The statistical data correction unit 37 receives an instruction to change the inspection result 56 from a defective product to a non-defective product inspection result via the operation receiving unit 12 by the user, and the statistical data recorded in the statistical data recording unit 43 is in accordance with the instruction. Correction (histogram 62).

  Here, when the upper limit value of the non-defective product is changed, the above-described threshold value adjustment unit 36 recalculates the threshold value (histogram 63), and executes the process of updating the threshold value of the inspection logic from H1 to H2. Good (histogram 64).

  After the statistical data of each inspection logic is corrected and the threshold value is updated by the method described above, the inspection logic evaluation unit 33 calculates an evaluation value, thereby performing a more accurate evaluation, and more appropriate main inspection logic. Can be selected.

  In each of the above-described embodiments, the case where the inspection system 100 including the inspection control apparatus 200 and the inspection module 20 of the present invention is constructed by one computer has been described. However, the inspection system of the present invention is not limited to the above configuration. For example, the inspection module 20 and the inspection control device 200 may be realized by separate computers, and an inspection system in which the inspection module 20 and the inspection control device 200 are connected so as to communicate with each other may be constructed.

<Embodiment 3>
FIG. 28 is a diagram showing a schematic configuration of an inspection system according to another embodiment of the present invention. As illustrated in FIG. 28, the inspection system 300 performs inspection based on at least one inspection device 220 including the main inspection unit 21 that performs inspection based on the main inspection logic, and inspection logic other than the main inspection logic. It includes at least one inspection device 231 (inspection device 232) provided with the sub-inspection unit 22, and an inspection control device 200 for setting inspection logic of each inspection device. The inspection logic and the inspection device have a one-to-one correspondence, and each inspection device constitutes the inspection module 20.

  Note that the reference numerals given to the respective constituent elements in FIG. 28 correspond to the reference numerals given to the respective constituent elements in FIG. 1, and the same reference numerals indicate the same constituent elements. Therefore, the description about the component already demonstrated by each above-mentioned embodiment is not repeated.

  Each inspection apparatus performs an inspection using the inspection logic set in itself, and records the inspection result in the inspection result recording unit 42. The inspection control apparatus 200 acquires the inspection result of the inspection result recording unit 42, executes statistical processing and evaluation processing, and selects an optimal inspection logic. The selected inspection logic is set for the inspection apparatus 220 by the inspection logic selection unit 34 via the communication network. Or it is good also as a structure which the test | inspection apparatus 220 receives the instruction | indication of the inspection logic switch which the inspection logic selection part 34 transmits, and switches own inspection logic based on an instruction | indication.

<Embodiment 4>
FIG. 29 is a diagram showing a schematic configuration of an inspection system according to another embodiment of the present invention. As shown in FIG. 29, the inspection system 400 includes an inspection device 240 and an inspection control device 200 that sets the inspection logic of the inspection device 240. The inspection apparatus 240 includes at least one main inspection unit 21 that performs inspection based on main inspection logic and at least one sub-inspection unit 22 that performs inspection based on inspection logic other than the main inspection logic. Each inspection unit constitutes an inspection module 20. The inspection logic and the inspection unit in the inspection device 240 have a one-to-one correspondence, and the point that the inspection module 20 is realized by one computer is different from the inspection system 300 described above.

  Each inspection unit of the inspection apparatus 240 performs inspection using the inspection logic set in itself, and records each inspection result in the inspection result recording unit 42. The inspection control device 200 selects an optimal inspection logic based on the calculated evaluation value, and sets the inspection logic of each inspection unit. Alternatively, the inspection device 240 may receive an instruction for switching the inspection logic transmitted by the inspection control device 200, and the inspection logic of each inspection unit of the inspection module 20 may be switched based on the instruction.

  FIG. 30 is a flowchart showing the flow of processing of the inspection module and the inspection control unit in the inspection system 300 or the inspection system 400. As shown in FIG. 30, the inspection process is executed by the inspection module 20, and the evaluation process is executed by the inspection control device 200. The inspection module 20 transmits the inspection result to the inspection control device 200 (S12), and the inspection control device 200 determines the timing to be evaluated (S13), and evaluates the inspection accuracy of each inspection logic from the inspection result (S13). S14, S15). When it is determined that the current inspection logic of the inspection module 20 is not appropriate (NO in S16), the inspection control device 200 transmits an instruction signal to the inspection module 20 to switch the main inspection logic (S17). In response to the instruction signal transmitted in S17, the inspection module 20 switches the inspection logic (S19) and continues the inspection process when an instruction for switching the main inspection logic is received (YES in S18).

  The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope shown in the claims, and embodiments obtained by appropriately combining technical means disclosed in different embodiments. Is also included in the technical scope of the present invention.

  Finally, each block of the inspection control apparatus 200, in particular, the statistical processing unit 32, the inspection logic evaluation unit 33, and the inspection logic selection unit 34 may be configured by hardware logic, and uses a CPU as follows. It may be realized by software.

  That is, the inspection control device 200 includes a CPU (central processing unit) that executes instructions of a control program that realizes each function, a ROM (read only memory) that stores the program, and a RAM (random access memory) that expands the program. And a storage device (recording medium) such as a memory for storing the program and various data. An object of the present invention is a recording medium on which a program code (execution format program, intermediate code program, source program) of a control program of the inspection control apparatus 200, which is software that realizes the functions described above, is recorded so as to be readable by a computer. This can also be achieved by supplying the inspection control apparatus 200 and reading and executing the program code recorded on the recording medium by the computer (or CPU or MPU).

  Examples of the recording medium include a tape system such as a magnetic tape and a cassette tape, a magnetic disk such as a floppy (registered trademark) disk / hard disk, and an optical disk such as a CD-ROM / MO / MD / DVD / CD-R. Card system such as IC card, IC card (including memory card) / optical card, or semiconductor memory system such as mask ROM / EPROM / EEPROM / flash ROM.

  Further, the inspection control apparatus 200 may be configured to be connectable to a communication network, and the program code may be supplied via the communication network. The communication network is not particularly limited. For example, the Internet, intranet, extranet, LAN, ISDN, VAN, CATV communication network, virtual private network, telephone line network, mobile communication network, satellite communication. A net or the like is available. Further, the transmission medium constituting the communication network is not particularly limited. For example, even in the case of wired such as IEEE 1394, USB, power line carrier, cable TV line, telephone line, ADSL line, etc., infrared rays such as IrDA and remote control, Bluetooth ( (Registered trademark), 802.11 wireless, HDR, mobile phone network, satellite line, terrestrial digital network, and the like can also be used. The present invention can also be realized in the form of a computer data signal embedded in a carrier wave in which the program code is embodied by electronic transmission.

  Since the inspection control apparatus of the present invention can dynamically set inspection logic in accordance with changes in the production environment, it can be applied to an inspection system that automatically determines whether a workpiece is good or defective on the production line.

It is a block diagram which shows the principal part structure of the test | inspection system to which the test | inspection control apparatus concerning embodiment of this invention is applied. It is a block diagram showing a schematic structure of an inspection system to which an inspection control device concerning an embodiment of the present invention is applied. It is a figure which shows the workpiece | work image image-processed based on the binary gravity center area image processing algorithm. It is a figure which shows the workpiece | work image image-processed based on the flaw dirt image processing algorithm. It is a flowchart which shows the flow of a process of a test | inspection module and a test | inspection control part in a test | inspection system. It is a flowchart which shows the flow of the process of initial setting in a test | inspection system. It is a flowchart which shows the flow of the process of an inspection in an inspection system. It is a figure which shows the example of the test result recorded on a test result recording part. It is a flowchart which shows the flow of a statistical process in a test | inspection system. It is a figure which shows typically the data input / output to a statistical processing part. It is a figure which shows the other example of the statistical data recorded on a statistical data recording part. It is a figure which shows the example of the statistical data in the case of the composite determination recorded on a statistical data recording part. It is a flowchart which shows the flow of an inspection logic evaluation process in an inspection system. (A) is a figure which shows the example of the evaluation value of the test | inspection logic recorded on an evaluation value recording part, (b) shows the example of the evaluation formula which a test | inspection logic evaluation part refers when calculating an evaluation value. FIG. (A)-(f) is a figure which shows the other example of an evaluation value calculation method by statistical data (histogram). It is a graph which shows transition of the evaluation value by the position of a threshold. It is a flowchart which shows the flow of a test | inspection logic selection process in a test | inspection system. It is a figure which shows the sample image of the workpiece | work which an inspection system uses by a stain | pollution | contamination inspection. It is a figure which shows the example of the fluctuation | variation of a production environment which occurs in a dirt inspection. It is a figure which shows the procedure of the initial setting which an inspection logic management part performs. It is a figure which shows transition of the statistics of the inspection result by the inspection logic which employ | adopted the production number and the binary gravity center area image processing algorithm. It is a figure which shows the transition of the statistics of the inspection result by the inspection logic which employ | adopted the number of production and a crack dirt image processing algorithm. It is a graph which shows transition of the evaluation value of each inspection logic accompanying the increase in the number of production. It is a figure which shows the result of the inspection logic switch which an inspection logic selection part performs. It is a block diagram which shows the detailed structure of the test | inspection control part in the test | inspection control apparatus of this invention. (A) And (b) is a figure which shows the content of the threshold value adjustment process of a threshold value adjustment part. It is a figure which shows the content of the statistical data correction process of a statistical data correction part. It is a block diagram which shows schematic structure of the test | inspection system concerning other embodiment of this invention. It is a block diagram which shows schematic structure of the test | inspection system concerning other embodiment of this invention. It is a flowchart which shows the flow of a process of a test | inspection module and a test | inspection control part in the test | inspection system concerning other embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Control part 2 Communication part 3 Operation part 4 Display part 10 Input / output control part 11 Image input control part 12 Operation reception part (result input control means)
13 Display control unit (display control means)
20 inspection module 21 main inspection unit 22 sub inspection unit 30 inspection control unit 31 inspection logic management unit 32 statistical processing unit 33 inspection logic evaluation unit (evaluation means)
34 Inspection logic selection section (inspection logic selection means / inspection logic setting means)
35 Timing monitoring unit 36 Threshold adjustment unit (threshold adjustment means)
37 Statistical data correction section (test result correction means)
40 recording unit 41 work information recording unit 42 inspection result recording unit 43 statistical data recording unit 44 evaluation value recording unit 45 inspection logic recording unit 100 inspection system 200 inspection control device 300 inspection system 400 inspection system

Claims (17)

In the inspection control device that controls the inspection module that determines the quality of the object, using the inspection logic that defines the method for determining good or bad,
Inspection logic selection means for selecting a main inspection logic used by the inspection module from a plurality of inspection logics;
When the inspection module determines the quality of the object using the inspection logic different from the final inspection result generated by using the main inspection logic and indicating the determination result of the quality of the object. Evaluation means for evaluating the inspection accuracy of each inspection logic based on the evaluation inspection result indicating the discrimination result;
Inspection logic setting means for switching the inspection logic selected by the inspection logic selection means as the main inspection logic,
The inspection control device, wherein the inspection logic selection means selects the main inspection logic based on the evaluation result evaluated by the evaluation means. The inspection logic defines a method for measuring a feature amount indicating a state of an object, and a threshold value for separating the feature amount to determine good or bad,
The evaluation means is
Express the inspection accuracy of each inspection logic using statistical data based on the feature quantity measured for each object included in the evaluation inspection result and the number of good and defective objects of the object indicated by the final inspection result The inspection control apparatus according to claim 1, wherein the evaluation is performed by calculating an evaluation value. The evaluation means is
The evaluation is performed by calculating an evaluation value based on a difference between a non-defective feature amount obtained from the non-defective product set of the objects and a defective product feature amount obtained from the defective product set. Inspection control device. The evaluation means is
When it is determined that the feature amount of the target object is smaller than a threshold value, it is determined as a non-defective product lower limit value which is a lower limit value of the feature amount of the defective product set and an upper limit value of the feature value of the non-defective product set. Calculate the evaluation value based on the difference from a certain good upper limit value,
When it is determined that the feature value of the target object is larger than the threshold value, the non-defective product lower limit value that is the lower limit value of the non-defective product set and the upper limit value of the feature value of the defective product set 4. The inspection control apparatus according to claim 3, wherein the evaluation is performed by calculating an evaluation value based on a difference from a certain defective product upper limit value. The evaluation means is
The inspection control apparatus according to claim 3 or 4, wherein the evaluation is performed by calculating an evaluation value based on a quotient obtained by dividing the difference by the variance of the non-defective feature amount. The evaluation means is
The evaluation value is calculated based on at least one of the difference between the non-defective product feature value obtained from the non-defective product set of the target object and the threshold value, and the difference between the defective product feature value obtained from the defective product set and the threshold value. The inspection control apparatus according to claim 2, wherein the evaluation is performed by performing the evaluation. The evaluation means is
The difference A between the non-defective product feature amount and the defective product feature amount
The difference between the defective product feature amount and the threshold is the difference B
The difference C between the non-defective feature amount and the threshold value is the difference C
As the evaluation formula,
| Difference A-Difference B / Difference C |
The inspection control apparatus according to claim 6, wherein the evaluation is performed by calculating an evaluation value based on the evaluation. The evaluation means is
Of the objects identified as defective by the main inspection logic in the final inspection result, evaluating the inspection logic based on the number of objects determined as non-defective by the inspection logic in the evaluation inspection result. The inspection control apparatus according to claim 1, wherein The inspection logic setting means is
When the inspection logic newly selected by the inspection logic selection unit based on the evaluation result evaluated by the evaluation unit is different from the main inspection logic in use by the inspection module,
The inspection control apparatus according to claim 1, wherein main inspection logic is switched. The inspection logic setting means is
10. The inspection control apparatus according to claim 9, wherein the evaluation unit switches the main inspection logic when the inspection logic other than the main inspection logic being used by the inspection module is evaluated higher than the main inspection logic. The inspection logic setting means is
The evaluation value included in the evaluation result is monitored for an evaluation value representing the inspection accuracy of the main inspection logic in use by the inspection module, and the timing for switching the main inspection logic is calculated based on the differential value of the evaluation value. The inspection control apparatus according to claim 9, wherein the main inspection logic is switched at a timing. Display control means for controlling the display unit to display the inspection results;
A result input control means for receiving an input of a user inspection result of an inspection by a user for the same inspection target object as the displayed inspection result;
2. The inspection according to claim 1, further comprising: an inspection result correcting unit that corrects the inspection result displayed on the display unit by the display control unit based on the user inspection result received by the result input control unit. Control device.   The inspection control apparatus according to claim 2, further comprising a threshold adjustment unit that calculates a threshold value of the inspection logic using statistical data of each inspection logic and determines the threshold value of the inspection logic. An inspection module that determines the quality of an object using inspection logic that defines a method for determining good or bad, and
The inspection control device according to any one of claims 1 to 13, which selects a main inspection logic used by the inspection module to generate a final inspection result indicating a good / bad determination result. Inspection system. An inspection control method in an inspection control apparatus for controlling an inspection module for determining whether a target is good or bad, using an inspection logic that defines a method for determining good or bad,
A selection step of selecting a main inspection logic used by the inspection module from a plurality of inspection logics;
When the inspection module determines the quality of the object using the inspection logic different from the final inspection result generated by using the main inspection logic and indicating the determination result of the quality of the object. An evaluation step for evaluating the inspection accuracy of each inspection logic based on the inspection result for evaluation indicating the discrimination result;
A switching step of switching the inspection logic selected by the inspection logic selection means as the main inspection logic,
In the selection step, the main control logic is selected based on the evaluation result in the evaluation step.   A control program for an inspection control apparatus for causing a computer to function as each means of the inspection control apparatus according to any one of claims 1 to 13.   The computer-readable recording medium which recorded the control program of the test | inspection control apparatus of Claim 16.