JP2019133426A - Control device, abnormality detection system, control method for control device, and control program - Google Patents

Abstract

To provide a control device, an abnormality detection system, a control method for the control device, and a control program for detecting a workpiece abnormality caused by a workpiece putting-into loss.SOLUTION: A PC (64) includes a control unit (643) that acquires, when a workpiece is put into a conveyance path at a normal posture and a time interval, a photographed image obtained by photographing a specific position, which is a position where the workpiece is scheduled to reach on the conveyance path, at a specific angle; an analysis unit (643A) that analyzes, by collating an image in at least a part of an object region of the photographed image with a partial image including a characteristic contour portion of the workpiece, similarity between an image in the object region and the partial image; and a workpiece presence/absence determination unit (643B) that determines whether the workpiece is included in the object region according to the similarity.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a control device that detects an abnormality of a workpiece based on a photographed image.

  Conventionally, various production lines for executing a heat treatment process using a heating apparatus such as a heating furnace are known. Since the heating temperature of the heating furnace cannot be easily changed, in the heat treatment process, it is common to operate the heating furnace at a constant set temperature and to feed the workpiece into the heating furnace at a constant speed and at regular intervals. is there.

  In such a heat treatment process, there is a case where the workpiece loading interval is shifted for some reason or the workpiece is not loaded. In this case, not only production loss occurs, but there is a possibility that the quality of the workpiece processed subsequently may be adversely affected. This problem occurs because the heat balance in the heating furnace, which is maintained by loading workpieces at normal intervals, changes due to abnormalities in the loading intervals.

  In addition, when a human touches the workpiece before heat treatment, or when the workpiece touches any protrusion during transportation, the workpiece is put into the heating furnace in a different orientation, shape, or position than usual due to some trouble. It may be done. In this case, the degree of temperature rise of each workpiece in the heating furnace changes, which may adversely affect the quality of the workpiece. This problem occurs because the manner in which hot air strikes the workpiece or the manner in which infrared radiation strikes the workpiece changes due to a change in the orientation, shape, or position of the workpiece.

  In order to detect the occurrence of such a problem, there is a need for a technique for monitoring whether or not an abnormality has occurred in the work put into the heating device.

Japanese Utility Model Publication No. 4-82756 JP-A-10-286730 JP 2001-2224011 A Japanese Patent Laid-Open No. 11-098491 JP 2001-076268 A JP 7-192186 A JP-A-7-167787

  Patent Documents 1 and 2 disclose a technique for identifying a product conveyed by a conveyor. Patent Document 1 discloses a product identification device that identifies a product by reading a barcode attached to the product being conveyed. However, since the technique described in Patent Document 1 only reads the barcode, the posture (direction, shape, position, etc.) of the product cannot be determined. Therefore, it cannot be determined whether or not the product is being conveyed in a normal posture. Also, it cannot be determined whether the product has passed through the specific position at normal time intervals.

  In Patent Document 2, a sensor detects that a pallet with a workpiece flowing on a production line has arrived, and measures time information at the time of detection. In the technique described in Patent Document 2, since the workpiece image is not acquired, it cannot be determined whether or not the posture of the workpiece is normal.

  Furthermore, Patent Documents 3 to 7 disclose techniques related to a method and apparatus for detecting an abnormality by analyzing a captured image. These patent documents disclose techniques related to automation such as detection of an abnormal state of a machine, crime prevention, detection of fire or smoke, and detection of gas leakage. However, it is not realistic to detect an abnormality in the posture of the workpiece by the techniques described in these patent documents. In addition, these techniques do not measure the time when the work is put in, and the time information remains in a description included in the image information. Therefore, it is also impossible to monitor the workpiece input interval using the techniques described in these patent documents. The technical problems described in Patent Documents 3 to 7 will be specifically described below.

  Patent Document 3 discloses a crime prevention monitoring method and apparatus that monitors a line surrounding a region of interest while detecting images at regular intervals, detects a luminance change due to crossing of a person or an object as an abnormality, and reports the abnormality. . In the technique described in Patent Document 3, an image change due to the intrusion of a person or an object is detected. It is not practical to detect whether or not a predetermined object has passed a certain position by this method. This is because, when the passage of an object is detected by this method, it is necessary to analyze an enormous image and extract a predetermined position passing image every time the object passes.

  Patent Document 4 is a method of taking an image of a product flowing on a belt conveyor in a factory production line and inspecting with the image. In the technique described in Patent Document 4, a signal from a sensor that detects that a product has passed an arbitrary point is received, and time-series image data is captured by a CCD camera at certain intervals. Then, the captured image is sent to the image analysis apparatus, and the quality is determined. When an NG (abnormality) is detected in the pass / fail judgment, an NG signal is output, and 10 frames before the occurrence of the event and 40 frames after that are stored among the image frames recorded in time series in the cache. It is recorded in the memory of the medium. As a result, an image of 50 frames including an abnormal image can be recorded, and the administrator can analyze the cause investigation image later.

  However, Patent Document 4 does not disclose an image processing method and can determine whether a product is good or bad, but does not indicate whether or not the product itself is present. In addition, sensor information (that is, NG signal) indicating that a specific position has been passed can be acquired, but since the sensor information is output only when there is an abnormality, there is no normal product passing record.

  When trying to record the passing records of normal products using this technology, a large amount of images of at least a plurality of products, 50 frames in the embodiment, are recorded for each product, which requires a huge memory capacity. turn into. Therefore, it is not realistic to detect whether or not the predetermined object has passed a certain position with the technique described in Patent Document 4.

  The technique described in Patent Document 5 relates to an apparatus for automatically monitoring an abnormality such as a fire in a factory by an image. A fluctuation (change) is extracted from (input image-reference image), binarized, and the image Find the center of gravity. If the number of centroids is plural, it is determined as smoke and the generation of smoke is automatically detected. Smoke is a technique that uses the fact that the density of the smoke occurs and the region is divided into a plurality of lumps, so that there are a plurality of centers of gravity.

  The technique described in Patent Document 6 is a technique for automatically detecting an abnormality such as a fire, steam ejection, and black smoke generation in a plant facility and the like to reduce the burden on a supervisor. Specifically, in the technique, the following processing is performed. That is, video is taken at a certain time interval, and abnormality of the object is detected from the accumulated image of the time difference of the monochrome image. The fluctuation area of the color image is extracted from the fluctuation area extracted from the binary image of the difference image, and fire, steam, and black smoke are discriminated. The logical product of the logical product of the bright part of the monochrome image and the difference binary image and the color image, and the logical product of the logical product of the dark part image of the previous monochrome image and the difference binary image and the previous color image. Find the product and detect light and dark fluctuations.

  These techniques described in Patent Documents 5 and 6 have a drawback in that the amount of calculation for a captured image is large. Also, with these techniques, the number of captured images themselves is enormous, so the number of file reference operations increases, and the processing is complicated, so processing time is required. In addition, since image processing is always performed, there is a problem in that the load on the apparatus that performs image processing is high. Furthermore, with these techniques, it is difficult to extract an image of each product when it passes through the production line.

  Finally, Patent Document 7 is an apparatus that images solid objects conveyed at intervals on a conveyance path, determines the quality of appearance based on the image processing, and distributes the objects according to the result. Therefore, an abnormality treatment is appropriately performed when the next object passes before the pass / fail determination process due to some abnormality. In other words, in an apparatus that determines the product that passes through a certain interval by inspecting the image, if there is an abnormality that the subsequent product enters the inspection area at the time when the determination does not end, the subsequent product is defective, It is characterized in that the determination is performed normally and the distribution is performed, and the pass / fail determination is a method of binarizing the acquired image and comparing it with reference data. Even in this method, it is difficult to identify the image to be solved.

  In summary, with the techniques described in Patent Documents 1 to 7, at least a work input loss cannot be practically detected. Here, the input loss refers to a state in which the input of the work from the accumulator that stores the work before being input to the production line is delayed and the work interval becomes larger than usual. The present disclosure has been made in view of the above-described problems, and an object thereof is to realize a control device or the like that can detect an abnormality of a workpiece caused by a workpiece insertion loss.

  In order to solve the above-described problem, the control device according to one aspect of the present invention provides a position at which the workpiece is expected to reach on the conveyance path when the workpiece is inserted into the conveyance path at a normal posture and time interval. An image acquisition unit that acquires a captured image obtained by capturing a specific position at a specific angle, an image of at least a part of a target area of the captured image, and a partial image including a characteristic contour shape portion of the workpiece By collating, an analysis unit that analyzes the similarity between the image of the target area and the partial image, and a work that determines whether the work is included in the target area according to the similarity A presence / absence determination unit.

  According to the above configuration, it is possible to determine whether or not the workpiece has been normally input into the conveyance path from the photographed image of the camera. Therefore, it is possible to detect an abnormality of the workpiece caused by, for example, a workpiece insertion loss in the conveyance path.

  In the control device, the photographed image is taken when the work is input to the transport path, and when the work is input to the transport path at a normal posture and time interval, a part of the input work is Further, it may be an image photographed so as to include a part or all of the work put in front of the work.

  According to the above configuration, by including a plurality of workpieces in the captured image, the distance between the workpieces can be recorded as an image.

  In the control device, the photographed image is taken when the work is input to the transport path, and when the work is input to the transport path at a normal posture and time interval, a part of the input work or It may be an image taken so as to include the whole and a part of the workpiece input after the workpiece.

  According to the above configuration, by including a plurality of workpieces in the captured image, the distance between the workpieces can be recorded as an image.

  When the workpiece presence / absence determination unit determines that the workpiece is included in the captured image, the control device captures the captured time or acquisition time of the captured image, and the image acquisition unit immediately before the captured image. Depending on whether or not the time interval calculation unit that calculates the time interval between the shooting time or the acquisition time of the acquired captured image and the time interval calculated by the time interval calculation unit is within a preset range And an interval abnormality determination unit that determines normality or abnormality of the workpiece input interval.

  According to the above-described configuration, a photographed image taken with the input of a certain workpiece and a previous photographed image, that is, a photographed image taken with the introduction of the workpiece before the certain workpiece, taken. The time interval can be calculated. Then, based on the time interval, it can be determined whether the workpiece input interval is normal. As a result, it is possible to accurately specify the delay time of input when a work input loss occurs.

  In the control device, the file name of the photographed image includes information indicating a photographing time or an acquisition time of the photographed image, and the file name of the photographed image includes the work in the photographed image. A file name changing unit that adds a work presence / absence identifier indicating whether or not a workpiece is present, and the time interval calculating unit calculates the time interval based on the file name of the captured image and the file name of the captured image acquired immediately before May be.

  According to the above-described configuration, it is possible to determine the presence / absence of a work and calculate the time interval simply by referring to the file name without referring to the contents of the captured image file. Therefore, it is possible to reduce the processing load related to the determination of the presence or absence of a workpiece and the calculation of the time interval.

  In the control device, the analysis unit may collate the partial image with the image of the target area on which binarization, color inversion, and edge detection are performed.

  According to the above configuration, it is possible to perform the analysis of the analysis unit by eliminating the influence of the lighting environment where the conveyance path is placed as much as possible. Therefore, the analysis unit can perform image analysis more accurately.

  In the control apparatus, the analysis unit includes a workpiece abnormality determination unit that calculates a score according to the similarity and determines whether or not the posture of the workpiece in the captured image is normal according to the score. May be.

  According to the above configuration, it is possible to determine whether or not the posture (at least one of the position and the shape) of the loaded workpiece is normal as well as the detection of the loading loss. Therefore, it is possible to detect the abnormality of the work in the conveyance path in more detail.

  The control apparatus may include a file name changing unit that attaches a posture identifier indicating whether or not the posture of the workpiece in the photographed image is normal to the file name of the photographed image.

  According to the above configuration, when the captured image is collected and analyzed later, it is possible to determine whether or not the posture of the workpiece is normal without referring to the contents of the captured image file. Therefore, it is possible to reduce the processing load related to the analysis of the captured image.

  In order to solve the above-described problem, the abnormality detection system according to one aspect of the present invention includes the control device, a detection sensor that detects insertion of the workpiece into the conveyance path, and the detection sensor that inputs the workpiece. And a camera that captures the captured image in response to detection of

  In order to solve the above-described problem, a control method executed by the control device according to one aspect of the present invention is such that, when a work is inserted into the transport path at a normal posture and time interval, the work reaches the transport path. An image acquisition step of acquiring a captured image obtained by capturing a specific position, which is a position to be performed, at a specific angle; an image of at least a part of a target region of the captured image; and a characteristic contour shape portion of the workpiece. An analysis step for analyzing the similarity between the image of the target area and the partial image by collating with the partial image to include, and whether or not the work is included in the captured image according to the similarity And a work presence / absence determination step for determining whether or not.

  In order to solve the above-described problem, a control program for a control device according to an aspect of the present invention is a control program for causing a computer to function as the control device, the image acquisition unit, the analysis unit, and the The computer functions as a work presence / absence determination unit.

  According to one aspect of the present invention, it is possible to detect a workpiece abnormality caused by a workpiece insertion loss.

It is a block diagram which shows the principal part structure of the personal computer contained in the abnormality detection system which concerns on Embodiment 1 of this invention. It is a figure which shows an example of a structure of a production line. It is a figure which shows the structure of a workpiece | work. It is a figure which shows the outline | summary of an abnormality detection system. It is a figure which shows the relationship between a workpiece | work and the imaging | photography area | region of a camera. It is a figure which shows an example of the flow of an image analysis process. It is a figure which shows the specific example of the image obtained by an image analysis process, and various scores. It is a figure which shows an image and various filters. It is a flowchart which shows the flow of a process of the said abnormality detection system. The average value, standard deviation, and skewness score distribution of each captured image of the camera when image analysis processing is executed is shown.

  The abnormality detection system according to the present invention will be described in detail. When the abnormality detection system according to the present invention detects that a workpiece has passed a specific position on the production line (a certain position on the conveyance path on which the workpiece is conveyed on the conveyor), the workpiece is in a normal posture and time interval. A specific position, which is a position where the workpiece is scheduled to arrive, is photographed when it is put in. Then, the abnormality detection system determines the presence or absence of the workpiece at the specific position at the time of shooting, and the normality or abnormality of the posture (direction, shape, and position) of the workpiece when there is a workpiece from the captured image. In addition, the abnormality detection system captures sequentially loaded workpieces as described above, and captures a captured image (or a captured image acquisition time on a personal computer (PC, described later)) and a previously captured image. The time interval at which the workpiece is inserted is calculated from the shooting time or the acquisition time. Then, it is determined whether the time interval is normal or abnormal.

  In the following description, an example in which the abnormality detection system is realized in a production line that performs a brazing process of aluminum parts using a continuous brazing furnace will be described. However, the application example of the abnormality detection system is not limited to this. The abnormality detection system can be applied to a production line in which workpieces are automatically input by a belt conveyor or the like.

[Embodiment 1]
≪Outline of production line≫
Hereinafter, an example of a specific embodiment of the abnormality detection system will be described with reference to FIGS. First, an outline of a production line that is an application target of the abnormality detection system 500 according to the present embodiment will be described. FIG. 2 is a diagram illustrating an example of the configuration of the production line 100. The production line 100 includes an automatic workpiece input device 5, a brazing processing unit 6, and a delivery side accumulator 7. The workpiece 4 is sequentially conveyed from the left direction to the right direction in the drawing.

  In the following description, the right end of the workpiece 4 in the drawing, that is, one end in the direction in which the production line 100 flows is referred to as the “tip portion of the workpiece 4”. In addition, the left side of the workpiece 4 in the drawing, that is, once in the direction opposite to the flow of the production line 100 is referred to as “tail end portion of the workpiece 4”.

  The automatic workpiece loading device 5 is a device group and a conveyance path for adjusting the loading interval of the workpiece 4 and sending the workpiece 4 to the brazing processing unit 6. The automatic workpiece loading device 5 includes an entry side accumulator 11, a stopper gate 12, and a conveyor belt 13. The workpiece 4 put into the production line 100 is transported on the entry side accumulator 11. A stopper gate 12 is provided at the end of the entry-side accumulator 11. The workpiece 4 is dammed up by the stopper gate 12 and accumulated. When the stopper gate 12 is lowered and the gate is opened, the workpiece 4 is conveyed from the entrance-side accumulator 11 to the conveyor belt 13. The conveyor belt 13 adjusts the loading interval of the workpiece 4 and sends it to the entry side table unit 14 of the brazing processing unit 6. The “work 4 input interval” indicates an interval of time during which the work 4 is input to the brazing furnace 20. The adjustment of the input interval of the workpiece 4 will be described in detail later.

  The brazing processing unit 6 is a group of devices and a conveyance path for performing a brazing process on the workpiece 4. The brazing processing unit 6 includes an entry side table unit 14, a brazing furnace 20, and an exit side table 17. As will be described in detail later, the entry table section 14 is provided with a detection sensor 61 that detects the presence of the workpiece 4 at the entry table end 16. The workpiece 4 sent from the conveyor belt 13 is placed on the mesh-like conveyor metal belt 15 at a constant interval in the entry side table section 14 of the brazing processing section 6 and is conveyed to the brazing furnace 20.

  The brazing furnace 20 includes a preheating zone 21, a heating zone 22, and a cooling zone 23. The workpiece 4 conveyed to the brazing furnace 20 is heated in the preheating zone 21 and reaches a predetermined temperature (for example, 600 degrees or more) in the heating zone 22. Thereby, the parts to be brazed included in the workpiece 4 are brazed. Thereafter, the workpiece 4 is conveyed to the cooling zone 23 and cooled to near room temperature. The cooled workpiece 4 is carried out to the delivery side table 17.

  The workpiece 4 that has been subjected to the brazing process is conveyed from the delivery side table 17 to the delivery side accumulator 7 and accumulated in the delivery side accumulator 7. The workpiece 4 accumulated in the delivery-side accumulator 7 is taken out by an operator and processed at another place. For example, the workpiece 4 is moved to a production line different from the production line 100 and subjected to processing such as removal of jigs, post-processing, and inspection.

(Adjustment of workpiece loading interval)
The adjustment of the input interval of the workpiece 4 in the production line 100 shown in FIG. 2 is performed as follows, for example. The workpiece 4 accumulated in the entry-side accumulator 11 is blocked by the stopper gate 12 as described above. In this state, when the tail end portion of the workpiece 4 (referred to as the nth workpiece 4) that has already passed the stopper gate 12 (and the conveyor belt 13) and placed on the conveyor metal belt 15 has passed through the entry side table end 16. The detection output of the detection sensor 61 is turned off. When the detection output of the detection sensor 61 is turned off, the stopper gate 12 is lowered and the gate is opened. As a result, the (n + 1) th work 4 is sent onto the transport belt 13.

  The (n + 1) th workpiece 4 sent to the conveyor belt 13 is conveyed to a position where the distance from the nth workpiece 4 is a predetermined distance. When the (n + 1) th workpiece 4 is transported to this position, the speed of the transport belt 13 is switched to the same speed as that of the transport metal belt 15 of the brazing furnace 20. The (n + 1) th work 4 is conveyed to the entry side table unit 14 while maintaining a predetermined distance from the previous nth work 4 at the speed.

  The detection sensor 61 detects the workpiece 4 when the tip of the (n + 1) th workpiece 4 reaches the entry table end 16. Thereafter, the detection output of the detection sensor 61 is turned on until the tail end of the (n + 1) th work 4 completes passing through the entry table end 16.

  When the tail end of the (n + 1) th work 4 completes passing through the entry side table end 16, the detection output of the detection sensor 61 is turned off again, and the stopper gate 12 is opened. Then, the (n + 2) th workpiece 4 passes through the stopper gate 12 and the conveyor belt 13 and is conveyed to the entry side table unit 14. Thus, the production line 100 repeats opening and closing (up and down) of the stopper gate 12 in accordance with the detection output of the detection sensor 61, thereby realizing the loading of the workpiece 4 at a constant distance interval.

  In addition, the adjustment method of the insertion interval of the workpiece | work 4 is not limited to the above-mentioned example. For example, the workpiece 4 may be taken from a stock table by a 6-axis robot or the like and placed on the transport metal belt 15 of the entry side table unit 14. Alternatively, the workpiece 4 may be taken from the stock table and placed on the transport metal belt 15 of the entry table section 14 by a human hand.

  Further, the work 4 may be transported on a predetermined base plate. In this case, the base plate is also handled as a part of the workpiece 4. In particular, the rectangular base plate is suitable and suitable for adjusting the orientation, shape, and position of the workpiece 4 and the distance between the workpieces 4. Further, the detection sensor 61 may detect the tail end of the base plate as the tail end portion of the workpiece 4. By using a base plate having a shape that the detection sensor 61 can easily detect, the sensitivity and detection accuracy of the detection sensor 61 can be increased.

≪Work structure≫
Next, a typical structure of the workpiece 4 will be described. FIG. 3 is a diagram illustrating the structure of the workpiece 4. FIG. 3A shows the three-dimensional structure and arrangement of each part of the workpiece 4. On the other hand, FIG. 3B shows an appearance of the workpiece 4 with the jig tightened (that is, the workpiece 4 flowing on the production line 100). In addition, the jig | tool shown to (a) and (b) of FIG. 3 is an example to the last. The material, mass, spring strength, and structure of the jig are appropriately selected according to the type of the part to be brazed. A jig having a small heat capacity, good heat conduction, and capable of heating the brazed component uniformly is desirable.

  As shown in FIGS. 3A and 3B, the workpiece 4 is sandwiched between two parts to be brazed by a jig base upper plate and a jig base lower plate, and is arranged at four corners around the brazed part. It is fixed by tightening the jig fixing bolt and the jig fixing nut. The size of the jig is not particularly limited, but may be, for example, about 250 mm square × 100 mm height. In order to prevent distortion of the jig fixing bolt, the jig fixing bolt may be fixed using a bar or a spring (not shown). The brazed part fastened with a jig as shown in FIG. 3B is heated and cooled in the brazing furnace 20 together with the jig.

≪Overview of monitoring system≫
Next, an abnormality detection system 500 for imaging and analyzing the workpiece 4 will be described with reference to FIG. FIG. 4 is a diagram showing an outline of the abnormality detection system 500. The abnormality detection system 500 includes a detection sensor 61, a contact input detector 62, a camera 63, a PC (control device) 64, a display device 65, and an NTP (Network Time Protocol) server 66. The detection sensor 61 and the contact input detector 62 are connected by a contact signal line. Each device other than the detection sensor 61 is interconnected via a LAN cable and a switching hub as shown in the figure.

  The display device 65 is a device for displaying the result of image analysis processing (described later) on the PC 64. The display device 65 may be configured integrally with the PC 64. The display device 65 is also a warning device that posts a warning of the abnormality on the display screen when an abnormality occurs in the production line 100. Further, the display device 65 may include a speaker. In the case where a speaker is provided, the display device 65 may output a warning of the abnormality by sound when an abnormality occurs in the production line 100.

  The NTP server 66 is a device that adjusts the time of the PC 64 by transmitting time information to the PC 64. When the NTP server 66 receives a time adjustment request from the PC 64, the NTP server 66 adjusts the time of the PC 64. This request is issued periodically. Note that the time adjustment interval (that is, the time motive interval between the NTP server 66 and the PC 64) is preferably several hours or half a day, but may be, for example, one week interval.

  As described above, the NTP server 66 sets the time of the PC 64 to the correct time, so that the time (that is, the timing of various processes) with other production management information can be collated accurately. Further, in order to statistically analyze the abnormality of the workpiece 4, it is required that the time of the PC 64 is an accurate time.

  As shown in the description of FIG. 2, the detection sensor 61 is a sensor that detects the presence or absence of the workpiece 4. The detection sensor 61 transmits a detection output (ON signal or OFF signal) to the contact input detector 62. The contact input detector 62 is a device that monitors changes in the detection output of the detection sensor 61. As described above, when the tail end portion (downstream end in the transport direction) of the work 4 has passed through the entry table end 16, the detection output of the detection sensor 61 changes from on to off. When the contact input detector 62 detects this output change from on to off, it sends a signal to the camera 63 by socket communication.

  The camera 63 is a device that captures an image of the workpiece 4 passing through the entry side table unit 14. For example, a CMOS (complementary metal-oxide semiconductor) camera or the like can be suitably used as the camera 63. When receiving a signal from the contact input detector 62 through socket communication, the camera 63 captures a specific position of the entry side table unit 14 at a specific angle. Note that the camera 63 may photograph not only the entry side table unit 14 but also the exit side table 17. That is, you may image | photograph the workpiece | work 4 after a process instead of the workpiece | work 4 before the process in the production line 100. FIG.

  Note that the camera 63 may be provided with a certain delay time from when a signal is received until shooting is performed according to the arrangement position of the sensor 61 and the arrangement position and angle of the camera 63. This can be realized by providing the camera 63 with a timer relay or the like. Or you may implement | achieve by delaying the signal output from the contact input detector 62. FIG.

  The camera 63 transmits a captured image file to the PC 64. At this time, the camera 63 measures the shooting time based on the internal clock of its own device, and transmits the file to the PC 64 with the shooting time attached to the file name of the shot image. For example, the camera 63 adds numbers indicating “year”, “month”, “day”, “hour”, “minute”, and “second” in order from the top of the file name, and sends them to the PC 64.

  Note that the shooting time measurement process and the process of assigning the shooting time to the file name of the shot image may be performed by the PC 64 instead of the camera 63. In this case, the camera 63 appropriately assigns a distinguishable file name to the captured image and transmits the file to the PC 64. When receiving the file, the PC may attach the shooting time to the file name of the captured image as described above based on the internal clock of the PC 64.

(Camera shooting angle)
Here, the shooting angle of the camera 63 will be described in detail with reference to FIG. FIG. 5A is a diagram illustrating the relationship between the workpiece 4 passing through the entry side table unit 14 and the imaging region of the camera 63. FIG. 5B shows an image when the camera 63 captures the area A1 shown in FIG. FIG. 5C shows an image when the camera 63 captures the area A2 shown in FIG. Broken line arrows in FIGS. 5A to 5C indicate the traveling direction of the workpiece 4. In the following description, it is assumed that the workpiece 4 that has caused a change in the detection output of the detection sensor 61 (that is, the workpiece 4 that has just passed through the entry-side table end 16) is the n-th workpiece, and the workpiece 4 immediately before that. Is the (n-1) th workpiece 4, and the immediately following workpiece 4 is the (n + 1) th workpiece 4.

  The camera 63 enters the camera at a shooting angle that includes the nth workpiece 4 and either the (n−1) th workpiece 4 or the (n + 1) th workpiece 4 when the workpiece 4 is normally conveyed. A specific position of the side table unit 14 is photographed. Thus, by photographing so that a plurality of workpieces 4 are included, not only the workpiece 4 itself but also the distance interval between the workpiece 4 and the immediately preceding or immediately following workpiece 4 can be recorded. Furthermore, since each workpiece 4 is transported through the entry side table unit 14 at a constant speed, the distance interval between the workpieces 4 can be regarded as the interval of the passage time of the workpiece 4 at a specific position of the entry side table unit 14. it can.

(1: When shooting area A1)
For example, the camera 63 captures the area A1 in FIG. That is, as shown in FIG. 5B, the camera 63 sets the (n-1) th workpiece 4 as a main photographing target (preferably, the entire (n-1) th workpiece 4 is shown), and the nth A specific position of the entry side table unit 14 is photographed at an angle at which a part (tip portion) of the work 4 is reflected. Thus, at the specific position of the nth workpiece 4 and the (n−1) th workpiece 4 together with a reliable record that the nth workpiece 4 has been conveyed on the conveying metal belt 15 of the entry side table unit 14. Distance interval (that is, the interval between the n-1th workpiece 4 and the nth workpiece 4).

  By the way, when an abnormality occurs in the input interval of the workpiece 4, a change occurs in an image captured by the camera 63 at a fixed point. For example, it is assumed that the workpiece 4 is temporarily depleted in the entry-side accumulator 11 and that the entry of the nth workpiece 4 into the entry-side table unit 14 is delayed more than usual. In this case, the time interval from when the (n-1) th workpiece 4 is input to when the nth workpiece 4 is input is longer than usual. Accordingly, the distance between the two workpieces 4 included in the captured image of the camera 63 is widened. Alternatively, since the (n-1) th workpiece 4 has already passed through the shooting area A1, a shot image in which only a part of the nth workpiece 4 is shown is obtained.

  In this case, a captured image mainly showing the (n-1) th workpiece 4 cannot be obtained, but an image captured by the camera 63 immediately before that, that is, a captured image mainly including the (n-2) th workpiece 4 is The (n-1) th work 4 is shown. Therefore, it can be confirmed that the (n-1) th work 4 has passed the specific position of the entry side table unit 14. In the actual production line 100, it is assumed that such an abnormality does not occur frequently. Therefore, as long as it can be confirmed that at least the (n-1) th workpiece 4 has passed and that there is an abnormality in the subsequent nth workpiece 4, it is sufficiently effective as information used for improving the production line 100. is there.

(2: When shooting area A2)
Further, for example, the camera 63 may photograph the area A2 in FIG. That is, as shown in FIG. 5C, the camera 63 sets the n-th workpiece 4 as a main photographing target (preferably, the entire n-th workpiece 4 is shown), and one of the n + 1-th workpiece 4. The specific position of the entry table portion 14 may be photographed at an angle such that the portion (tip portion) is reflected. Thereby, the distance between the nth workpiece 4 and the (n + 1) th workpiece 4 at the specific position as well as the reliable recording that the nth workpiece 4 has been conveyed on the conveying metal belt 15 of the entry side table unit 14. The interval (that is, the interval between the nth workpiece 4 and the (n + 1) th workpiece 4) can be recorded. In this case, after detecting the tail end of the nth workpiece 4, it is necessary to delay the imaging until a certain time interval (a time interval at which the (n + 1) th workpiece 4 is in frame). Therefore, this photographing angle is effective for the production line 100 in which the change in the conveyance speed of the conveyance metal belt 15 and the work interval setting is small, for example. In this case, the nth workpiece 4 is photographed at a position where the entry side table end 16 has been moved by the delay time. Therefore, if there is a work input delay, a part of the next (n + 1) th work is not shown in the lower frame position (left in (a) of FIG. 5, upward in (c)). Is possible. Therefore, it is possible to detect an abnormality in real time and output an alarm, and if an operator enters in the vicinity, there is an advantage that the abnormality can be corrected immediately depending on the case.

  In the case of photographing either of the areas A1 and A2, when the abnormality detection system 500 includes an alarm device, when the detection output of the detection sensor 61 does not turn on for a predetermined time or longer, the contact input detector 62 May be activated to activate the alarm device. Alternatively, when there is no signal output from the contact input detector 62 to the camera 63 within a predetermined time, the camera 63 may activate the alarm device. Alternatively, when a captured image from the camera 63 cannot be obtained for a predetermined time or more, the PC 64 may activate the alarm device. Thus, by outputting an alarm when an abnormality occurs in the detection timing of the workpiece 4, an abnormality in the production line 100 can be notified to the manager of the line in real time.

  Furthermore, you may make the camera 63 perform an image | photographing at the timing which operates an alarm device. Thereby, for example, when the nth workpiece 4 is not input, it is possible to capture a captured image mainly of the n−1th workpiece input first. Of course, the image taken at this time does not include the n-th workpiece 4 and includes only the (n-1) -th workpiece 4.

≪Main components of PC≫
FIG. 1 is a block diagram showing a main configuration of the PC 64. The PC 64 includes a communication unit 641, a time acquisition unit 642, a control unit (image acquisition unit) 643, and a storage unit 644.

  The communication unit 641 performs communication between the PC 64 and other devices. The storage unit 644 stores a captured image file acquired by the PC 64. In addition, the storage unit 644 stores various formulas, setting values, and threshold values used for image analysis processing or determination processing executed by each unit of the control unit 643.

  The time acquisition unit 642 reads the acquisition time from the file name of the captured image. The time acquisition unit 642 may be included in the control unit 643. It is also conceivable that the camera 63 counts the shooting time and attaches the shooting time to the file name of the shot image instead of attaching the shooting time to the file name of the shot image. In this case, a time based on an internal clock (not shown) of the PC 64 is given to the captured image file. The time acquisition unit 642 reads the shooting time after the shooting time is given to the file name of the shot image in the PC 64.

  In addition, it is desirable that the internal clock of the PC 64 periodically acquires time information from the NTP server 66 via the communication unit 641 to adjust the time. An error in the input interval of the workpiece 4 in the production line 100 may be a problem even within a few seconds. By adjusting the time, the time acquisition unit 642 can accurately measure the acquisition time of the captured image. Therefore, it is possible to accurately record the time when the workpiece abnormality occurs.

  The control unit 643 controls the PC 64 in an integrated manner. The control unit 643 includes an analysis unit 643A, a workpiece presence / absence determination unit 643B, a workpiece abnormality determination unit 643C, a time interval calculation unit 643D, and an interval abnormality determination unit 643E.

  The control unit 643 also determines that there is no workpiece by the workpiece presence / absence determination unit 643B, or determines that there is an abnormality by either the workpiece abnormality determination unit 643C or the interval abnormality determination unit 643E via the communication unit 641. A warning may be sent to the display device 65. As a result, the PC 64 can warn in real time of the workpiece insertion loss (no workpiece or abnormal workpiece insertion interval) and the workpiece posture abnormality via the display device 65.

≪Image analysis algorithm≫
The analysis unit 643A performs an image analysis process on the image acquired by the PC 64 from the camera 63. FIG. 6 shows an example of the flow of image analysis processing.

  First, the analysis unit 643A determines a region (target region) to be subjected to image analysis in the captured image. It is desirable that the target area is an area that does not include extra persons and objects other than the production line 100 and the workpiece 4 (for example, intruders other than the workpiece and the in-furnace adjusting agent) in the captured image.

  The analysis unit 643A performs binarization processing (S201), inversion processing (S202), and edge detection processing (S203) in this order on the determined target region (image thereof). As a result, the image of the target area is binarized and then the color is inverted (white “1” = 255 bit is converted to black “0” = 0 bit), and the image is further enhanced with a simple edge. . That is, the analysis unit 643A performs image processing so that the detected edge becomes “1”.

  At the site of the production line 100, the lighting environment changes depending on the time and season, the arrangement position of the production line 100 (whether or not it is affected by outdoor sunlight), and the like. There is a problem that complex equipment is required to take into account the change due to the lighting environment. The analysis unit 643A according to the present embodiment can eliminate the influence of the factory lighting environment on the image of the target region as much as possible by performing binarization, inversion, and edge detection processing on the target region.

  The analysis unit 643A further performs a mask process (S204). Mask processing refers to the degree of similarity between the image of the target region and the partial image by comparing the image of the target region with a partial image (reference mask image) including a characteristic contour shape portion of the workpiece 4. Is a process of analyzing.

  In the case of the workpiece 4 assembled from a material including many straight portions such as a square plate and a square member like the workpiece 4 according to the present embodiment, if the workpiece 4 is inserted at a normal interval and posture, the target of the photographed image In the region, many straight portions are seen. Therefore, as a reference mask image, for example, an image of a region (mask region) including a portion where a straight edge is most clearly detected in the work 4 and having an arbitrary width is prepared in advance and stored in the storage unit 644. It is desirable to keep it. The “arbitrary width” referred to here is a buffer for allowing a small amount of positional deviation (pixel level error) of the work 4 due to a fluctuation during conveyance or a slight deviation in imaging timing.

  The reference mask image is a black and white binary image having the same size or larger than the image of the target area. The mask area of the reference mask image is defined as white (that is, “1” = 255 bits), and the area other than the mask area is defined as black (= “0” = 0 bits). The analysis unit 643A multiplies the image of the target area whose edge has been detected (that is, an image indicating whether each pixel is a monochrome binary) and the reference mask image. As a result, only the pixels that are white in both the image of the target area where the edge is detected and the reference mask image are extracted, and the other pixels are black.

The analysis unit 643A calculates various scores using the image of the target area subjected to the image analysis processing of S201 to S204 (S205). For example, the analysis unit 643A calculates an average luminance (hereinafter referred to as an average value), a luminance standard deviation (hereinafter referred to as a standard deviation), and a skewness of the image of the target region. Here, the skewness is a value of a scale representing the asymmetry of data around the average value. When the average value is μ, the standard deviation is σ, and the expected value is E (), the skewness value Skew can be defined by the following equation (1). In addition, when the skewness is a negative value, it indicates that the data is spread to a lower side than the average value.
S = E (X-μ) ³ / σ ³ ··· Equation (1)
≪Example of analysis results≫
FIG. 7 is a diagram illustrating specific examples of images and various scores obtained by the image analysis processing by the analysis unit 643A. The “binarization / inversion” column in the figure indicates a captured image (target area) after the inversion processing shown in S202 of FIG. 6 is executed. The “edge detection / mask” column in the drawing shows a captured image (target region) after the mask processing shown in S204 of FIG. 6 is executed. The “Mean” column indicates an average value, the “Stdev” column indicates a standard deviation, and the “Skew” column indicates a skewness value. In addition, the values in the “binarization” row of the “Mean” column, the “Stdev” column, and the “Skew” column are compared with the conventional image analysis processing (binarization) on the same captured image (target region). The values when an image processing method using an image is performed are shown.

  When the posture (direction and shape) of the workpiece 4 is abnormal, the jig of the workpiece 4 appears to be tilted as compared with a normal case. For example, only a part of the handle part of the jig fits in the target area, and a part of the jig protrudes outside the target area. In this case, as shown in the figure, the number of contours that can be detected when an edge is detected is smaller than that in a normal case, and thus the average score is smaller than that in a normal case.

  Further, when the photographed image does not include the workpiece 4 (mainly the workpiece 4 to be photographed), when the edge of the target region is detected, the contour of the mesh of the transport metal belt 15 that should be hidden under the workpiece originally. Is extracted. Therefore, the number of contours that can be detected when an edge is detected is larger than that in a normal case, and thus the average score is larger than that in a normal case.

  In addition, when the image is taken with the adjusting agent instead of the workpiece 4, the adjusting agent has a large amount of powder, and thus the contour cannot be clearly detected even by edge detection as shown in the figure. In this case, therefore, the average score is also smaller than in the normal case.

(Judgment of presence / absence of workpiece)
The workpiece presence / absence determination unit 643B determines whether or not a workpiece is included in the target area according to the analysis result of the analysis unit 643A. Here, “a workpiece is included (with a workpiece)” means a workpiece to be photographed mainly (for example, the n−1th workpiece 4 in the case of photographing the area A1 in FIG. 5A). It shows that it is included in the target area at a predetermined ratio or more. On the other hand, “a workpiece is not included (no workpiece)” indicates a case where a workpiece to be photographed is not included in the target area at all or less than the predetermined ratio.

  In addition, in the production line 100 according to the present embodiment, an in-furnace atmosphere adjusting agent for adjusting the atmosphere in the brazing furnace in accordance with the input interval of the work 4 (that is, at a timing when a certain work 4 is input). (Hereinafter simply referred to as a regulator). In the case of the production line 100 in which the adjustment agent is charged in this way, the workpiece presence / absence determination unit 643B includes whether or not the workpiece 4 is included in the target area (with a workpiece) according to the analysis result. It may be determined whether the workpiece 4 and the adjusting agent are not included (no workpiece).

  Further, in the production line 100 according to the present embodiment, among the workpieces 4 carried into the production line, the workpiece 4 (first workpiece) to be carried in first is fixed with a dummy jig and is not shipped as a product. (Dummy work) may be carried in. When the workpiece presence / absence determination unit 643B determines that the workpiece is present in the production line 100 in which the dummy workpiece is carried in as the first workpiece, whether the workpiece 4 included in the target area is a dummy workpiece or not. It may be determined whether or not. The determination result of the workpiece presence / absence determination unit 643B is sent to the workpiece abnormality determination unit 643C and the time interval calculation unit 643D.

  Specifically, the workpiece presence / absence determination unit 643B previously determines, for various scores, areas with workpieces (and whether or not they are dummy workpieces), regulators, and workpieces, and the respective determination results. Then, the determination result is given according to which determination result area the various score values calculated by the analysis unit 643A are included in. The area may be appropriately set according to the size, shape, carry-in speed, etc. of the workpiece 4 to be determined. For example, when a score as shown in FIG. 7 is obtained, if Mean> 5 and Skew <10, or Mean> 5 and StdDev> 25, it can be said that the influence of the metal network is strong, Judged as nothing. In addition, when Mean <1 and Skew> 24, it is determined that the adjusting agent is present. In addition, since the jig is different between the normal workpiece 4 and the dummy workpiece, the values of various scores are different. Based on the difference in the scores, the workpiece presence / absence determination unit 643B determines whether the workpiece is a dummy workpiece when the workpiece is present.

  The workpiece presence / absence determination unit 643B may perform determination by setting the above-described region for any one of the average value, standard deviation, and skewness, or may perform AND determination of these scores. By performing region determination for many types of scores, the workpiece presence / absence determination unit 643B can improve the accuracy of determination of the presence / absence of workpieces and adjusting agents.

  Note that the control unit (file name changing unit) 643 identifies the presence / absence of a work indicating the presence / absence of a work in the file name of the captured image 644A stored in the storage unit 644 when the work presence / absence determination unit 643B finishes the determination. A tag (work presence / absence identifier) is attached. For example, the control unit 643 adds the character string of the work presence / absence identification tag to the end of the file name of the captured image 644A. Specifically, character strings such as “with workpiece”, “with workpiece (dummy)”, “without workpiece”, and “adjusting agent” may be added as identification tags.

(Determination of workpiece posture abnormality)
The workpiece abnormality determination unit 643C determines abnormality of the posture (direction, shape, and position) of the workpiece according to the analysis result of the analysis unit 643A. Specifically, the work abnormality determination unit 643C previously determines areas for determining the presence or absence of the work posture abnormality and the respective determination results for various scores. Then, the determination result is given according to which determination result area the various score values calculated by the analysis unit 643A are included in.

  Note that the control unit (file name changing unit) 643 has a normal posture of the workpiece in the captured image based on the file name of the captured image 644A stored in the storage unit 644 when the workpiece abnormality determination unit 643C finishes the determination. A posture identification tag (posture identifier) indicating whether or not is attached. For example, the control unit 643 adds the character string of the posture identification tag to the end of the file name of the captured image 644A. Specifically, a character string of “work posture normal” or “work posture abnormal” may be attached as the posture identification tag.

  Note that the workpiece abnormality determination unit 643C may perform determination by setting the above-described region for any one of the average value, standard deviation, and skewness, similar to the workpiece presence / absence determination unit 643B, or AND determination of these scores. May be performed. By performing region determination for many types of scores, it is possible to improve the determination accuracy of workpiece posture abnormality. For example, when the score as shown in FIG. 7 is obtained, it is determined that the work posture is abnormal because the adjusting agent or the work is inclined more than usual (meaning an abnormal posture) when Mean <1.

  Furthermore, when the types of scores used for the determination of the workpiece presence / absence determination unit 643B and the determination of the workpiece presence / absence determination unit 643B are the same, the control unit 643 determines the region determination of the workpiece presence / absence determination unit 643B and the region determination of the workpiece abnormality determination unit 643C. May be performed simultaneously.

(Time interval calculation and workpiece input interval determination)
The time interval calculation unit 643D acquires the shooting time (or acquisition time) attached to the file name of the captured image 644A and the acquired image immediately before the captured image 644A (that is, attached to the file name). The time interval (time interval) with the shooting time (or acquisition time) of the shot image 644A (which has the latest shooting time or acquisition time) is calculated.

  For example, the time interval calculation unit 643D acquires the file name having the latest shooting time for the file whose identification tag of the captured image 644A is “working” or “adjusting agent”, and displays the “year” at the beginning of the file name. The information of “month” “day” “hour” “minute” “second” is converted into time, and a time difference, that is, an interval (time interval) value is calculated. As a result, the shooting time between the shot image shot with the input of a certain work 4 and the previous shot image, that is, the shot image shot with the input of the workpiece 4 before the given work 4 is set. The interval can be calculated. In other words, it can be said that this is a time interval between the workpieces. The time interval calculation unit 643D sends the calculation result to the interval abnormality determination unit 643E.

  The interval abnormality determination unit 643E determines whether or not the workpiece input interval is normal. The interval abnormality determination unit 643E determines that the workpiece input interval is normal when the time interval value calculated by the time interval calculation unit 643D is within a preset range. On the other hand, when the value of the interval is out of the range, it is determined that an abnormality has occurred in the workpiece input interval. For example, if the absolute value of the difference between the time interval value and the preset standard time of the time interval exceeds a predetermined threshold, the interval abnormality determination unit 643E abnormally detects the workpiece input interval. Is determined to have occurred.

≪Details of image analysis process≫
(S201 to S202: binarization and inversion)
Hereinafter, an algorithm of image analysis processing executed by the analysis unit 643A will be described in more detail. Note that the algorithm described below is an example, and specific processing contents of binarization, inversion, edge detection, and mask processing are not limited to the following example. For binarization, inversion, and edge detection, a well-known technique such as a commercially available image processing program may be used.

The analysis unit 643A specifies a target region and obtains a binarization threshold value by the following equation (R1-1).
(Threshold) = (Average luminance value of the target area + Average luminance value of the background area) / 2 (R1-1)
In addition, each pixel brightness | luminance is calculated | required by calculating Gray = (R + G + B) / 3 in RGB. The analysis unit 643A inverts the color of the binarized target region image (that is, white and black), and then executes the next edge detection process.

(S203: Edge detection)
Next, the analysis unit 643A performs edge detection processing on the binarized and inverted image of the target region. This process is a process for emphasizing a sudden luminance change in an active image (or selection range) using a contour detector. Specifically, vertical and horizontal spatial filtering (luminance change rate generation) is performed using two 3 × 3 convolution kernels (convolution filters), and the square root of the sum of squares is obtained. Thereby, an image subjected to edge detection (hereinafter referred to as an edge detection image) is generated. FIG. 8A shows 3 × 3 pixels, and FIG. 8B shows a spatial filter. X takes a value of 0 or 1, and the other a to d take an integer value.

The pixel Pi, j after processing by spatial filtering is expressed by the following equation (R2-1).
P _{i, j} = a (P _{i-1, j-1} −P _{i + 1, j} ) + b (P _{i-1, j} −P _{i + 1, j} ) + c (P _{i-1, j + 1} −P _{i + 1, j-1} ) +
d (P _{i, j + 1} -P _{i, j-1} ) + xP _{i, j} (R2-1)
For example, two types of filters, S / Edge (FIG. 8C) and E / Edge (FIG. 8D) of the Sobel kernels filter, can be used for this spatial filter. Assuming that Pi, j after the former process and Pi, j after the latter process are SP _{i, j} and EP _{i, j} , these are expressed by the following equations (R2-2) and (R2-3).
SP _{i, j} = (P _{i-1, j-1} −P _{i + 1, j} ) − (P _{i−1, j + 1} −P _{i + 1, j−1} ) −2 (P _{i, j + 1} -P _{i, j-1} ) ... (R2-2)
EP _{i, j} = (P _{i-1, j-1} −P _{i + 1, j} ) +2 (P _{i-1, j} −P _{i + 1, j} ) + (P _{i-1, j + 1} − P _{i + 1, j-1} ) (R2-3)
Expression R2-2 calculates the luminance difference between the pixels in the vertical direction, and Expression (R2-3) calculates the luminance difference in the horizontal direction. Therefore, the pixel Pi, j after the edge detection process generated by these two 3 × 3 convolution kernels is expressed by the following equation (R2-4).
P _{i, j} = (SP _{i, j} ² + EP _{i, j} ² ) ^1/2 (R2-4)
(S204: Score processing after mask processing)
Next, the analysis unit 643A performs a mask process on the image on which the edge detection process has been performed. The contents of the mask processing are as described above. Finally, the analysis unit 643A calculates an average value, standard deviation, and skewness.

  The standard deviation Stdv is a standard deviation value of the target image. When the average gray value of all pixels is Xave, the total number of pixels is m × n, and the gray value (luminance value) of each pixel is Xi, j, Stdv (σ) is expressed by the following equation (R2-5) or (R2-6) ).

σ = {[(X ₁₁ −X _ave ) ² + •• + (X _1n −X _ave ) ² + •• + (X _m1 −X _ave ) ² +
(X _m2 −X _ave ) ² + ·· + (X _mn −X _ave ) ² ] / (m × n)} ^0.5 (R2-6)
Skewness is a measure of the asymmetry of the data around the sample mean. When the skewness is negative, it can be said that the data spreads to the left of the average value. On the other hand, when the skewness is positive, the data spreads further to the right. The skewness of the normal distribution (or all distributions that are completely symmetric) is zero. Distribution skewness (third moment regarding the mean) is defined by the following equation (R2-7).

  Here, μ is an average value of x, σ is a standard deviation of x, and E (t) represents an expected value of a statistic. skewness computes a sample version of this population value. If each element is Xi, the following equation (R2-8) is applied.

≪Process flow≫
Finally, the processing flow of the entire system will be described with reference to FIG. FIG. 9 is a flowchart showing a process flow of the abnormality detection system 500 according to the present embodiment. The detection sensor 61 detects that the tail end of the workpiece 4 has passed through the entry table end 16 and is turned off (S101). When the contact input detector 62 detects an output change from ON to OFF of the detection sensor, it transmits a signal to the camera 63 by socket communication (S102). The camera 3 executes shooting and acquires the shooting time (S103). After attaching the shooting time to the file name of the shot image, the camera 3 sends the shot image data to the PC 64 (S104).

  The control unit 643 of the PC 64 acquires a captured image file (S105, image acquisition step), and executes image analysis processing (S106, analysis step). This image analysis process has been described in each step of FIG. The analysis unit 643A sends the analysis result to the workpiece presence / absence determination unit 643B, the workpiece abnormality determination unit 643C, the time interval calculation unit 643D, and the interval abnormality determination unit 643E. The workpiece presence / absence determination unit 643B performs the workpiece presence / absence determination (work presence / absence determination step), and the workpiece abnormality determination unit 643C determines whether or not the workpiece posture is normal. The control unit 643 adds the workpiece presence / absence identification tag and the posture identification tag to the file name of the captured image (S107). The time interval calculation unit 643D calculates a time interval (Ti) (S108). The interval abnormality determination unit 643E determines whether or not the absolute value of the difference between the value of Ti and the standard interval time (standard time of Ti) Tis exceeds a set threshold value (for example, 10 s) (S109). When it exceeds (YES in S109), the control unit 643 notifies the warning to the display device 65 and the like, and the display device 65 displays the warning (S110).

  On the other hand, if the value is equal to or less than the threshold value (NO in S109), the control unit 643 ends the process without notifying the warning. Thereafter, the abnormality detection system 500 repeats the processing from S101.

  According to the above processing, it can be determined from the captured image of the camera 63 whether or not the workpiece 4 has been normally input to the production line 100. Therefore, in the production line 100, it is possible to detect an abnormality of the workpiece 4 due to, for example, the insertion loss of the workpiece 4 or the like.

  Further, according to the above processing, a photographed image taken when a certain work 4 is input and a previous photographed image, that is, a photographed image photographed when the work 4 before the certain work 4 is thrown. The photographing time interval can be calculated. Then, based on the time interval, it can be determined whether the workpiece input interval is normal. As a result, it is possible to accurately specify the delay time of input when a work input loss occurs.

  Further, according to the above processing, it is possible not only to detect the insertion loss but also to determine whether or not the posture (direction, shape and position) of the input workpiece is normal. Therefore, it is possible to detect a workpiece abnormality in the production line in more detail.

  Furthermore, when performing high-temperature heat treatment such as brazing, it is difficult to attach a tag such as an RFID tag to the workpiece 4 and to monitor the loading interval and loading posture of the workpiece 4. The abnormality detection system 500 according to the present embodiment can monitor the input interval and the input posture of the workpiece 4 even in an environment where it is difficult to monitor the workpiece 4 with a tag or the like.

[Example of software implementation]
The control block of the PC 64 may be realized by a logic circuit (hardware) formed on an integrated circuit (IC chip) or the like, or may be realized by software.

  In the latter case, the PC 64 includes a computer that executes instructions of a program that is software for realizing each function. The computer includes, for example, one or more processors and a computer-readable recording medium storing the program. In the computer, the processor reads the program from the recording medium and executes the program, thereby achieving the object of the present invention. As the processor, for example, a CPU (Central Processing Unit) can be used. As the recording medium, a “non-temporary tangible medium” such as a ROM (Read Only Memory), a disk, a card, a semiconductor memory, a programmable logic circuit, or the like can be used. Further, a RAM (Random Access Memory) for expanding the program may be further provided. The program may be supplied to the computer via an arbitrary transmission medium (such as a communication network or a broadcast wave) that can transmit the program. Note that one aspect of the present invention can also be realized in the form of a data signal embedded in a carrier wave, in which the program is embodied by electronic transmission.

  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.

  One embodiment of the present invention will be described below. FIG. 10 shows a score distribution of the average value, standard deviation, and skewness of each captured image of the camera 63 when the image analysis process is executed. FIG. 10A shows the score distribution when the image analysis process (a series of processes shown in FIG. 7) according to the present invention is executed. FIG. 10B shows a score distribution when an image analysis process (an image processing method using a binarized image) that has been often used conventionally is executed.

  In FIG. 10, Mean represents an average value, Stdev represents a standard deviation, and Skew represents a skewness. These are defined by the formula (1) described in the embodiment.

  When the image analysis processing according to the present invention was executed, the average value, the standard deviation, and the skewness varied greatly depending on the presence or absence of a workpiece, as shown in FIG. In addition, the average value, standard deviation, and skewness were significantly different between the work piece and the conditioner. Furthermore, the average value, standard deviation, and skewness also differed significantly when a workpiece posture abnormality occurred. As a result, as shown in the figure, the plots of the respective photographed images were dispersed, and it was possible to accurately determine whether the workpiece was present, whether the adjusting agent was present, and whether the workpiece was present, and whether the workpiece posture was abnormal or normal when the workpiece was present.

  On the other hand, when the conventional image analysis processing is executed, as shown in FIG. 10B, there is a portion where the distribution areas of the plot without work and the plot with work and normal posture overlap. As described above, in the conventional image analysis processing, depending on the average value, standard deviation, and skewness score, it may be difficult to discriminate between the absence of a workpiece and the presence of a workpiece and normal posture. Further, as shown in the figure, it is also difficult to discriminate between the absence of a workpiece and the adjusting agent and the discrimination between the absence of a workpiece and a workpiece posture abnormality.

  The image analysis processing according to the present invention designates a region that should show the characteristic contour shape of the workpiece 4 from the captured image, and extracts and discriminates the feature amount of the region, whereby (a) in FIG. As shown in Fig. 5, it is possible to determine the presence / absence of the workpiece and the posture of the workpiece with high accuracy.

61 Detection sensor 63 Camera 64 PC (control device)
643 Control unit 643A Analysis unit 643B Work presence / absence determination unit 643C Work abnormality determination unit 643D Time interval calculation unit 643E Interval abnormality determination unit 500 Abnormality detection system

Claims (11)

An image acquisition unit that acquires a photographed image obtained by photographing a specific position, which is a position where the work is expected to reach on the conveyance path, at a specific angle when the workpiece is inserted into the conveyance path at a normal posture and time interval; ,
Analyzing the similarity between the image of the target area and the partial image by collating the image of the target area of at least a part of the captured image with a partial image including a characteristic contour shape portion of the workpiece An analysis unit to
A control device comprising: a work presence / absence determination unit that determines whether or not the work is included in the target area according to the similarity.   The photographed image is taken in response to the loading of the workpiece into the conveyance path. When the workpiece is loaded into the conveyance path at a normal posture and time interval, a part of the loaded workpiece and one of the workpieces are captured. The control device according to claim 1, wherein the control device is an image photographed so as to include a part or all of a workpiece that was input immediately before.   The photographed image is taken when the work is input to the transport path, and when the work is input to the transport path at a normal posture and time interval, a part or all of the input work and the work 2. The control device according to claim 1, wherein the control device is an image photographed so as to include a part of a workpiece that has been input one after the next. When the workpiece presence / absence determination unit determines that the workpiece is included in the captured image, the captured time or the acquisition time of the captured image and the captured image acquired immediately before the captured image by the image acquisition unit A time interval calculation unit that calculates a time interval from the shooting time or the acquisition time of
An interval abnormality determination unit that determines whether the workpiece input interval is normal or abnormal according to whether the time interval calculated by the time interval calculation unit is within a preset range. The control device according to any one of claims 1 to 3, wherein the control device is characterized. The file name of the photographed image includes information indicating the photographing time or the acquisition time of the photographed image,
A file name changing unit for attaching a work presence / absence identifier indicating whether or not the photographed image includes the workpiece to the file name of the photographed image;
The control device according to claim 4, wherein the time interval calculation unit calculates the time interval based on a file name of the captured image and a file name of the captured image acquired immediately before.   The said analysis part collates the image of the said target area | region which performed binarization, color inversion, and edge detection, and the said partial image, The any one of Claims 1-5 characterized by the above-mentioned. The control device described. The analysis unit calculates a score according to the similarity,
The control apparatus according to claim 1, further comprising a work abnormality determination unit that determines whether or not a posture of the work in the captured image is normal according to the score.   The control apparatus according to claim 7, further comprising: a file name changing unit that attaches a posture identifier indicating whether or not the posture of the workpiece in the photographed image is normal to the file name of the photographed image. The control device according to any one of claims 1 to 8,
A detection sensor for detecting the loading of the workpiece into the conveyance path;
An abnormality detection system comprising: a camera that captures the captured image when the detection sensor detects the loading of the workpiece. An image acquisition step of acquiring a photographed image obtained by photographing a specific position, which is a position where the work is expected to reach on the transport path, at a specific angle when the work is put into the transport path at a normal posture and time interval; ,
Analyzing the similarity between the image of the target area and the partial image by collating the image of the target area of at least a part of the captured image with a partial image including a characteristic contour shape portion of the workpiece An analysis step to perform,
And a work presence / absence determination step of determining whether or not the photographed image includes the work in accordance with the similarity.   A control program for causing a computer to function as the control device according to claim 1, wherein the control function causes the computer to function as the image acquisition unit, the analysis unit, and the work presence / absence determination unit.

Images