JP2016114433A - Image processing device, image processing method, and component supply device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image processing device for determining the posture of each of components individually.SOLUTION: Provided is an image processing device for recognizing the captured image 11 of a component 100 transported in a prescribed transfer direction and thereby determining the posture of the component 100, the image processing device having: a procedure for importing the detection area image of a detection area 12 divided at a detection position in the transfer path 2a of the component 100 in the captured image 11; a procedure for importing the recognition area image of a recognition area 13 divided at a position adjacent to the detection area 12 on the upstream side of the transfer direction in the captured image 11; a procedure for detecting whether the component 100 exists at the detection position on the basis of the detection area image; and a procedure for determining the posture of the component 100 on the basis of the recognition area image when the presence of the component 100 at the detection position is detected.SELECTED DRAWING: Figure 2

Description

  Embodiments disclosed herein relate to an image processing apparatus, an image processing method, and a component supply apparatus.

  Patent Document 1 describes a configuration in which a product to be transported that is transported by a vibration feeder is photographed by a camera and the shape direction of the product to be transported is correctly recognized by recognizing the photographed image.

JP-A-5-164529

  However, since the vibratory feeder cannot control the transfer timing and transfer speed of each transferred product, there are cases in which the two transferred products are transferred in contact with each other in the transfer direction. The direction (posture) could not be recognized individually.

  The present invention has been made in view of such problems, and an object thereof is to provide an image processing device, an image processing method, and a component supply device that can individually determine the posture of each transported body. To do.

  In order to solve the above-described problem, according to one aspect of the present invention, an image processing apparatus that recognizes an image of a captured image of a transferred object that is transferred in a predetermined transfer direction to determine the posture of the transferred object. A detection image capturing means for capturing a detection region image of a detection region partitioned at a predetermined position on a transport path of the transported object in the captured image, and the transport direction of the detection region in the captured image A recognition image capturing unit that captures a recognition area image of a recognition area partitioned at a position adjacent to the upstream side of the detection object, and the transferred object at the predetermined position based on the detection region image captured by the detection image capturing unit. Detecting means for detecting whether or not there is present, and when the detecting means detects the presence of the transported object at the predetermined position, the detection means is based on the recognition area image captured by the recognized image capturing means. And discriminating means for discriminating the orientation of the conveying member, an image processing apparatus having applied.

  According to another aspect of the present invention, there is provided an image processing method for discriminating a posture of the transported body by recognizing an image of a captured image of the transported body that is transported in a predetermined transport direction. Capturing a detection region image of a detection region partitioned at a predetermined position on a transport path of the transported object in the image, and partitioning at a position adjacent to the upstream side of the detection region in the transport direction in the captured image Capturing a recognition area image of the recognized recognition area, detecting whether the transported object is present at the predetermined position based on the captured detection area image, and existence of the transported object at the predetermined position An image processing method is executed to detect the posture of the transferred object based on the captured recognition area image when the image is detected.

According to another aspect of the present invention, the image processing apparatus includes an image processing device that determines an attitude of the transported body by recognizing a captured image of the transported body that is transported in a predetermined transport direction. An image processing program for capturing a detection area image of a detection area partitioned at a predetermined position on a transfer path of the transferred object in the captured image, and transferring the detection area in the captured image Capturing a recognition area image of a recognition area partitioned at a position adjacent to the upstream side of the direction, detecting whether the transported object is present at the predetermined position based on the captured detection area image, When detecting the presence of the transported body at a predetermined position, determining the posture of the transported body based on the captured recognition area image;
An image processing program for executing is applied.

  Further, according to another aspect of the present invention, a transfer unit that transfers a plurality of components stored therein in a line on a transfer path, an image pickup unit that images the transfer path, and the image picked up by the image pickup unit An image processing device that determines the posture of the component by recognizing a captured image of the component, and an exclusion unit that excludes the component determined to be not in a predetermined posture by the image processing device from the transfer path. A component supply device, wherein the image processing device includes a detection image capturing unit that captures a detection region image of a detection region partitioned at a predetermined position on the transfer path in the captured image, and A recognition image capturing means for capturing a recognition area image of a recognition area partitioned at a position adjacent to the upstream side of the detection area in the transfer direction of the component, and before the detection image capturing means captures Detection means for detecting whether the part is present at the predetermined position based on a detection area image, and when the detection means detects the presence of the part at the predetermined position, the recognition image capturing means captures the part. A component supply apparatus having a determining unit that determines the posture of the component based on the recognition area image is applied.

  According to the present invention, the posture can be individually determined for each of the transferred objects.

It is a figure showing the external appearance plane and control structure of the parts feeder of embodiment. It is a figure which shows an example of the captured image of the transfer path | route which the camera imaged. It is a figure which shows an example of the captured image which drops components from a conveyance path | route and excludes them. It is a figure which shows the normal attitude | position of the component of the object which carries out attitude | position discrimination | determination by embodiment. It is a figure which shows the upside down attitude | position of the component of the object which carries out attitude | position discrimination | determination by embodiment. It is a figure which shows the state immediately after components arrive at a recognition area. It is a figure which shows the state immediately before components reach | attain a detection area. It is a figure which shows the state from which components can reach | attain a detection area and attitude | position discrimination | determination is possible. It is a figure which shows the state in which the downstream part in a connection transfer state exists in a detection position. It is a figure which shows the state in which the upstream part in a connection transfer state exists in a detection position. It is the figure which showed typically the periodic process which an image processing apparatus performs. It is a figure which shows the state just before the components with a short transfer direction length arrive at a detection area at the time of the n-th imaging. It is a figure which shows the state which the components with a short transfer direction length arrived at the detection area at the time of the (n + 1) th imaging. It is a figure which shows the state which divides and sets the re-recognition area | region according to the position of the upstream edge part of downstream components. It is a figure which shows the state which returned the recognition area to the original position after carrying out attitude | position discrimination | determination in the re-recognition area. It is a figure which shows the state in which two components of the articulated transfer state exist in the recognition area. It is a figure which shows the state which divided and set the re-recognition area | region in order to carry out attitude | position discrimination | determination of the downstream components in the articulated transfer state. It is a flowchart which shows the control procedure which CPU of an image processing apparatus performs. FIG. 4 is a front view showing a normal posture in both the front and back directions and the vertical direction, with components having a posture distinction between the vertical direction and the front and back directions. It is a figure which shows the arrow XX-XX cross section in FIG. FIG. 6 is a front view showing a posture that is distinguished in at least one of a front / back direction and an up / down direction, with parts having a posture distinction between the vertical direction and the front / back direction. It is a figure explaining the attitude | position discrimination | determination method of the image recognition with respect to components with the difference in attitude | position in an up-down direction and a front-back direction. It is a figure explaining the attitude | position discrimination | determination method of the image recognition with respect to components with the difference in attitude | position in an up-down direction and a front-back direction. FIG. 10 is a front view showing a normal posture and a reversed posture, with parts having a posture distinction in a transfer direction. It is a figure explaining the attitude | position discrimination | determination method of the image recognition using the comparison pattern of a shape feature part with respect to components with the attitude | position distinction in a transfer direction. It is a figure which shows the state which divided and set the re-recognition area | region with the connection transfer state of a normal attitude | position with respect to components with the distinction of an attitude | position in the transfer direction. It is a figure which shows the state which divided and set the re-recognition area | region in the articulated transfer state of the attitude | position reversed together with respect to the components with the attitude | position distinction in a transfer direction. FIG. 5 is a front view showing a normal posture and a reversed posture, with parts having a posture distinction in the vertical direction. It is a figure explaining the attitude | position discrimination | determination method of the image recognition using the comparison pattern of a shape feature part with respect to components with the attitude | position distinction in an up-down direction. It is a figure which shows the state which divided and set the re-recognition area | region with the connection transfer state of a normal attitude | position with respect to the components with the attitude | position distinction in an up-down direction. It is a figure which shows the state which divided and set the re-recognition area | region with the connection transfer state of the attitude | position which the upstream determined with respect to the components with the attitude | position distinction in the up-down direction in the downstream.

  Hereinafter, an embodiment will be described with reference to the drawings. In the following description, directions such as up, down, left, and right are used as appropriate for convenience of description of the configuration of the component supply device and the like, but the positional relationship between the components of the component supply device and the like is not limited.

<Schematic configuration of parts feeder>
FIG. 1 shows an external appearance plane and a control configuration of a parts feeder which is a component supply apparatus of the present embodiment. In FIG. 1, a parts feeder 1 includes a vibration feeder 2, an illumination 3, a camera 4, an air ejection nozzle 5, an electromagnetic valve 6, an image processing device 7, an image display device 8, and an operation interface 9. And have.

  The vibration feeder 2 (corresponding to the transfer means) is a so-called mortar-shaped (concave-shaped) container body that has a bottom portion on the back side in plan view, and is arranged in a spiral manner on the inner wall on the outer peripheral side thereof A path 2a is provided. The inner peripheral end portion of the transfer path 2a communicates with a storage portion 2b located at the center bottom of the vibration feeder 2, and the outer peripheral end portion of the transfer path 2a extends linearly to form a discharge port 2c. doing. The storage unit 2b stores a large number of parts 100 all having the same shape to be supplied, and by applying vibration such as centrifugal vibration to the entire vibration feeder 2 by a vibration mechanism (not shown) in particular, the storage unit 2b. The inner part 100 enters the transfer path 2a. Since the transfer path 2a is formed in a rail shape whose width dimension is substantially equal to the width dimension of the part 100 in a predetermined direction, a large number of parts 100 are aligned in a line along the longitudinal direction on the transfer path 2a. Get stuck in. Furthermore, fine irregularities of a predetermined pattern are formed on the road surface of the transfer path 2a, and each component 100 on the transfer path 2a is given a propulsive force from the irregularities by the vibration, and the direction along the transfer path 2a ( It is transferred in the path direction toward the outer peripheral side of the spiral;

  The illumination 3 is an illumination fixture that projects light toward a predetermined position of the transfer path 2a.

  The camera 4 (corresponding to the image pickup means) has a function of picking up the projected portion of the illumination 3 and transmitting the data of the picked-up image to an image processing device 7 described later. Note that the camera 4 captures an image in the imaging direction that is the same height as the transfer path 2a of the light projecting location and is orthogonal to the transfer direction of the component 100.

  The air ejection nozzle 5 (corresponding to the exclusion means) is disposed slightly downstream in the transfer direction from the light projecting portion, and has a function of ejecting compressed air supplied from an air pump (not shown).

  The electromagnetic valve 6 has a function of controlling the supply and suppression of compressed air from an air pump (not shown) to the air ejection nozzle 5 based on a command from the image processing device 7 described later.

  The image processing device 7 includes a computer having a CPU and a memory (not shown), and has a function of discriminating the posture of each component 100 transferred on the transfer path 2a by recognizing a captured image acquired from the camera 4. Have

  The image display device 8 has a function of displaying captured images of the camera 4 and various command and parameter settings via the image processing device 7.

  The operation interface 9 includes a pointing device such as a keyboard and a mouse, and has a function of accepting various commands and setting inputs from the operator.

  The parts feeder 1 having the above-described configuration is configured such that a large number of parts 100 of the same shape stored in a random posture in the central storage portion 2b of the vibration feeder 2 are all aligned in a row with the same predetermined posture. It is an apparatus which supplies from the exit 2c. However, as described above, the transfer path 2a of the vibration feeder 2 has only a simple rail structure having the same width as the width of each part 100 in a predetermined direction, and the fitting of the part 100 into the transfer path 2a is only the momentum of vibration. Therefore, each component 100 often gets stuck in the transfer path 2a in various postures different from the predetermined posture and is transferred.

  On the other hand, the image processing device 7 recognizes the image captured by the camera 4 to determine the posture of each component 100 on the transfer path 2a (see FIG. 2 described later), and for the component 100 that is not in an appropriate posture. Controls the solenoid valve 6 so as to apply the compressed air through the air ejection nozzle 5 (see FIG. 3 described later). The component 100 to which the compressed air is applied is removed from the transfer path 2a and returned to the storage portion 2b in the center of the vibration feeder 2. By repeating this, the components 100 finally supplied from the discharge port 2c are all supplied in a line with the same predetermined posture.

<Features of this embodiment>
However, in the vibration feeder 2 having the above-described configuration, the timing at which the parts 100 stored in the storage unit 2b are sequentially placed on the transfer path 2a and the transfer speed of the parts 100 on the transfer path 2a cannot be accurately controlled. On the transfer path 2a, each component 100 is transferred at an uncertain speed at random transfer intervals. For this reason, there is a case where the two parts 100 arranged in the front-rear direction in the transfer direction are in a connected transfer state in which the two parts 100 are in contact with each other. In this case, it is difficult to reliably determine the posture of each. Further, even when the length of each component 100 in the transfer direction is short, it is difficult to reliably determine the posture of each component 100. In the parts feeder 1 of the present embodiment, the image processing device 7 recognizes the captured image so as to deal with these situations, so that the posture of each component 100 can be individually determined. Hereinafter, image recognition techniques for that purpose will be described in order.

<Basic method of image recognition>
FIG. 2 shows an example of a captured image of the transfer path 2 a captured by the camera 4. This captured image 11 is an image captured from the side of the transfer path 2a in which the direction from the right side to the left side in the figure is the transfer direction. The X coordinate is set in the horizontal direction and the Y coordinate is set in the vertical direction. -Y coordinate is set.

  The image processing apparatus 7 divides the detection area 12 at a predetermined detection position (corresponding to the predetermined position) in the transfer path 2a in the captured image 11, and captures the partial image as a detection area image. This detection area image is for detecting whether or not the part 100 exists in the detection area 12, that is, whether or not the part 100 has reached the detection position. Therefore, the detection region 12 is set to have a sufficiently short length Lk (length in the X direction) in the transfer direction so that only the presence / absence of the component 100 can be determined. Conversely, in order to reduce the possibility that the downstream end portion of the component 100 is positioned in the detection area image and the determination process of the presence / absence thereof becomes complicated, the length of the detection area 12 in the transfer direction is not set long. .

  Further, the image processing apparatus 7 divides the recognition area 13 at a position adjacent to the upstream side of the detection area 12 in the transfer direction in the captured image 11 and captures the partial image as a recognition area image. This recognition area image is for determining the posture of the component 100 existing in the recognition area 13. Therefore, the recognition region 13 sets the length Ln in the transfer direction to the extent that the posture of the component 100 can be determined. This point will be described in detail later.

  The captured image 11 captured by the camera 4 is image data that can be handled in units of pixels, and the detection area image and the recognition area image are captured by software processing of the image processing device 7. Also, if both the detection area image and the recognition area image are set to a height Hy (length and position in the Y direction) that covers the entire height of the component 100 transferred on the transfer path 2a. Good. When the image processing apparatus 7 detects the arrival of the component 100 at the detection position based on the detection region image, the image processing device 7 determines the posture of the component 100 based on the recognition region image. As described above, since the image processing apparatus 7 included in the present embodiment performs image recognition on a partial image captured from a part of the captured image 11 captured by the camera 4, the posture determination of the component 100 is short. It can be done in processing time. If the determined posture of the component 100 is not in a predetermined correct posture, the solenoid valve 6 is opened to release compressed air when the component 100 is positioned in front of the air ejection nozzle 5 as shown in FIG. The part 100 is dropped from the transfer path 2a and removed.

  Here, FIGS. 4 and 5 show examples of the shape of the component 100 to be supplied in the present embodiment. The component 100 is a rectangular parallelepiped-shaped main body, and the nine round holes 101 pass through in the thickness direction of the component 100 (the direction orthogonal to the paper surface in the drawing) with an even arrangement throughout the longitudinal direction (transfer direction). It is provided as follows. All the round holes 101 are uniformly arranged at the same height, and the height position is offset in the height direction (Y direction) with respect to the center of gravity position 102 of the entire component 100. Thus, since the shape of the component 100 in this embodiment is symmetric in the thickness direction of the component 100 (the direction orthogonal to the paper surface in the drawing), there is no distinction between the front and back with respect to the imaging direction of the camera 4, and each round hole 101. Only the difference between the normal posture and the inverted posture is caused by the difference in the arrangement height of the. In the example of the present embodiment, the posture in which each round hole 101 is located higher than the center of gravity position 102 as shown in FIG. 4 is a normal posture, and each round hole 101 is placed lower than the center of gravity position 102 as shown in FIG. A certain posture is set as a reversed posture.

  In the posture determination based on the recognition area image, only two types of posture determinations of a normal posture and a reversed posture are performed by comparing the average height position of each round hole 101 and the height position of the center of gravity position 102 of the entire component 100. Shall be performed. For this posture discrimination, it is basically sufficient to recognize one round hole 101. However, in order to ensure the discrimination accuracy, it is preferable to recognize at least four round holes 101. At this time, the length Ln in the transfer direction of the recognition area 13 is set to be equal to or longer than the length L4 corresponding to the four pitches of the round holes 101, and the length L4 corresponding to the four pitches of the round holes 101 is included in the recognition area image. It is preferable that the component 100 exists. In addition, said component 100 is corresponded to the to-be-transferred body as described in each claim, and the round hole 101 is corresponded in the shape characteristic part as described in each claim.

  In the present embodiment having the above-described region section and component configuration, the arrival detection and posture determination of the component 100 are performed in the processes shown in FIGS. First, as shown in FIG. 6, immediately after the downstream end of the component 100 reaches the recognition region 13 by the transfer by the vibration feeder 2, the component 100 is still detected to reach the detection position based on the detection region image. Because it is not, posture discrimination is not performed. The posture determination is not performed in the same manner even immediately before the part 100 is transferred and the downstream end reaches the detection region 12 as shown in FIG. When the component 100 reaches the detection area 12 as shown in FIG. 8, the arrival of the component 100 at the detection position is detected by the image recognition device, and the posture is determined based on the detection. At this time, since at least a part of the component 100 is almost certainly present in the recognition area 13 adjacent to the upstream side of the detection area 12 in the transfer direction, the recognition area image captured from the same captured image 11 is displayed. Based on this, the posture of the component 100 can be determined.

<Periodic processing for articulated transfer status>
However, for the reasons described above, there may be a connected transfer state in which the two parts 100 arranged in the front-rear direction in the transfer direction are in contact with each other as shown in FIGS. In this case, in the detection of the component 100 in the detection region image described above, it is difficult to detect the upstream component 100 separately from the previously detected downstream component 100, and it is possible to individually determine the posture. It becomes difficult. That is, it is difficult to detect the timing at which the upstream component 100 reaches the detection position as shown in FIG. 10 from the state where only the downstream component 100 exists at the detection position as shown in FIG. On the other hand, in the present embodiment, the image processing device 7 causes the camera 4 to sequentially capture images with the shortest cycle so that the posture can be individually determined for each component 100 in the articulated transfer state. The posture is determined.

  Such periodic processing performed by the image processing apparatus 7 of the present embodiment is schematically shown in FIG. As shown in the figure, the image processing apparatus 7 basically repeats the detection process, and performs the determination process when the detection (detection) detects the presence (arrival) of the component 100 at the detection position. As main steps in the detection process, first, an imaging step of acquiring the captured image 11 from the camera 4 is performed, then a step of capturing a detection region image from the captured image 11 is performed, and then the detection region image is detected. The step of detecting the presence of the component 100 (arrival at the detection position) is performed based on the above. In this way, the detection process is repeated with the shortest processing cycle T in which only the minimum steps are performed. Further, as a main process in the discrimination process, a process of first capturing a recognition area image is performed, and then a process of determining the posture of the component 100 together with the recognition area image is performed. Note that the recognition area image is captured from the captured image 11 captured in the detection process immediately before the determination process. As a result, the time required for the discrimination process can be shortened by omitting the imaging process, and more accurate posture discrimination with a small time lag from the time of the previous detection is enabled.

  As described above, by repeatedly performing the detection process with the shortest processing cycle T, it is possible to determine the posture of each of the parts 100 even when the plurality of parts 100 are in the connected transfer state. Although there is a possibility that the posture determination is performed a plurality of times for the same component 100, only the component 100 in the normal posture passes through and the component 100 in the reversed posture is simply eliminated immediately. no problem.

<Re-recognition processing for asynchronous transport>
However, even when the detection process is repeated in the shortest cycle as described above, the image capturing process in the camera 4 and the image recognition process in the image processing apparatus 7 each take time, and thus the processing period T becomes somewhat long. Therefore, as described above, each component 100 is transferred asynchronously (in phase mismatch) with this processing cycle T and at indefinite intervals, and particularly when the length of each component 100 in the transfer direction is relatively short, It may be difficult to reliably determine the posture.

  For example, as shown in FIGS. 12 and 13, consider the case of a component 100 ′ whose overall transfer direction length is only 5 pitches of the round hole 101. In the n-th detection process shown in FIG. 12, the downstream end of the component 100 ′ is close to the upstream side of the detection region 12 when the camera 4 captures an image. In this case, at the time of the next n + 1 detection processing after the processing cycle T of the detection processing has elapsed, as shown in FIG. 13, the length in the transfer direction of the component 100 ′ in the recognition area image indicates the round hole required for posture determination. There is a high possibility that the length is shorter than the length of 4 pitches of 101. Thus, even if the image of the component 100 ′ is recognized in the recognition region 13 with a length in the transfer direction that is less than the four pitches of the round holes 101, the posture determination accuracy cannot be obtained as described above.

  On the other hand, in the present embodiment, in the recognition process for recognizing the recognition area image, first, the image recognition of the end portion on the upstream side in the transfer direction of the component 100 ′ located downstream in the transfer direction in the recognition area image is prioritized. To do. Then, when it is determined that the length in the transfer direction of the part 100 ′ located downstream in the feed direction in the recognition area image is shorter than the discrimination reference length L 4 for the four pitches of the round holes 101, The recognition area image is newly captured in the re-recognition area 14 that is partitioned at a position shifted to the downstream side in the transport direction from the area 13.

  For example, as shown in FIG. 14, the position of the upstream end portion in the transfer direction of the component 100 ′ on the downstream side in the transfer direction in the recognition region image at the initial position is detected. The re-recognition area 14 is newly set by shifting so that the upstream end in the transfer direction of the recognition area 13 is positioned in the vicinity of the upstream end position, and the partial image inside is newly set as a new recognition area image. take in. If the length Ln in the transfer direction of the recognition area 13 is set to be greater than or equal to the discrimination reference length L4 that is a reference for attitude discrimination (4 pitches or more of the round holes 101), the component 100 ′ is always included in the re-recognition area 14. It exists with a length equal to or longer than the discrimination reference length L4. Thereby, posture discrimination can be performed with high accuracy.

  The section setting of the re-recognition area 14 and the capture of the recognition area image in the re-recognition area 14 may be captured from the captured image 11 captured in the immediately preceding detection process, and in this case, the determination process can be performed quickly. . Further, as shown in FIG. 15, even when the two parts 100 ′ are transferred in the proximity of each other in the transfer direction, the previous part 100 ′ is always present when the posture is determined by the recognition area image of the re-recognition area 14. Since the detection area 12 is missing, returning the recognition area 13 to the original position does not affect the detection process and the discrimination process for the rear part 100 ′. Also, as shown in FIGS. 16 and 17, even when two parts 100 ′ aligned in the transfer direction are connected and transferred, an appropriate re-recognition area 14 can be set for each part 100 ′. 100 'detection processing and discrimination processing can be performed normally.

<Control flow>
A control procedure executed by a CPU (corresponding to a computing device; not shown in particular) included in the image processing device 7 in order to realize the functions as described above will be described step by step with reference to FIG. In FIG. 18, the processing shown in this flow is started when the operation of the parts feeder 1 is started.

  First, in step S <b> 105, the CPU of the image processing device 7 acquires a captured image 11 captured by the camera 4.

  Next, the process proceeds to step S110, and the CPU of the image processing apparatus 7 captures a detection area image from the captured image 11 acquired in step S105.

  Next, the process proceeds to step S115, and the CPU of the image processing apparatus 7 recognizes the presence / absence of the component 100 in the detection area image captured in step S110.

  Next, the process proceeds to step S120, and the CPU of the image processing apparatus 7 determines whether or not it is recognized that the component 100 is present in the detection area image by the image recognition in step S115. If the component 100 is not present in the detection area image, the determination is not satisfied and the process returns to step S105 and the same procedure is repeated.

  On the other hand, if the component 100 is present in the detection area image, the determination is satisfied, and the routine goes to Step S125.

  In step S125, the CPU of the image processing device 7 captures a recognition area image from the captured image 11 acquired in step S105.

  Next, the process proceeds to step S130, and the CPU of the image processing apparatus 7 recognizes the upstream end of the component 100 in the recognition area image captured in step S125.

  Next, proceeding to step S135, the CPU of the image processing apparatus 7 calculates the component length in the recognition area image.

  Next, the process proceeds to step S140, and the CPU of the image processing apparatus 7 determines whether or not the component length calculated in step S135 is equal to or greater than the discrimination reference length (four pitches of the round hole 101 in this example). If the component length in the recognition area image is less than the determination reference length, the determination is not satisfied and the routine goes to Step S145.

  In step S145, the CPU of the image processing apparatus 7 divides and sets the re-recognition area 14 shifted from the current recognition area 13 to the downstream side in the transfer direction. The shift amount may be shifted so that, for example, the upstream end of the re-recognition region 14 is positioned in the vicinity of the position of the upstream end recognized in step S130 (see FIGS. 14 and 17).

  Next, the process proceeds to step S150, and the CPU of the image processing apparatus 7 captures the partial image of the re-recognition area 14 set in step S145 as a recognition area image. Then, the process proceeds to step S155.

  On the other hand, if it is determined in step S140 that the component length in the recognition area image is equal to or greater than the determination reference length, the determination is satisfied, and the process proceeds to step S155.

  In step S155, the CPU of the image processing device 7 determines the posture of the component 100 in the recognition area image by image recognition (see FIGS. 4 and 5 above).

  Next, the process moves to step S160, and the CPU of the image processing apparatus 7 determines whether or not the posture of the component 100 is normal in the posture determination in step S155. If it is determined in the posture determination that the posture of the component 100 is not normal (inverted posture), the determination is not satisfied and the routine goes to Step S165.

  In step S165, the CPU of the image processing device 7 opens the electromagnetic valve 6 at an appropriate timing, and ejects the compressed air from the air ejection nozzle 5 to remove the component 100 whose posture is determined from the transfer path 2a (FIG. 3 above). reference). Then, the process proceeds to step S170.

  On the other hand, if it is determined in the determination in step S160 that the posture of the component 100 is normal in the posture determination, the determination is satisfied, and the process proceeds to step S170.

  In step S <b> 170, the CPU of the image processing apparatus 7 determines whether an operation for ending the operation of the parts feeder 1 is input from the operator via the operation interface 9. If the end operation has not been input, the determination is not satisfied and the process returns to step S105 and the same procedure is repeated. On the other hand, when the end operation is input, the determination is satisfied, and this flow ends.

  In the above, the procedure of step S110 corresponds to the detected image capturing unit described in each claim, the procedures of steps S125, S145, and S150 correspond to the recognized image capturing unit described in each claim, and the procedure of step S115 is performed. It corresponds to the detection means described in each claim, and the procedures of steps S130, S135, and S155 correspond to the determination means described in each claim.

<Effect of this embodiment>
As described above, according to the parts feeder 1 of the present embodiment, the image processing device 7 does not recognize the entire captured image 11 in the procedure of step S115, but based on a detected image captured from a part thereof. The arrival of the component 100 at the detection position can be detected in a short processing time. For this reason, it is possible to sufficiently shorten the processing cycle T of the detection process in which the imaging process in step S105, the detection area image capturing process in step S110, and the detection process in step S115 are repeated (see FIG. 11 above). As a result, even if the length in the transfer direction of each component 100 transferred asynchronously (in phase mismatch) and at indefinite intervals with this processing cycle T is relatively short, the arrival of each component 100 at the detection position is almost assured. Can be detected. In this case, the arrival of the component 100 at the detection position can be detected with a simple configuration using only the imaging means such as the camera 4 without providing another sensor such as a photosensor.

  Since the recognition area 13 is adjacent to the upstream side of the detection area 12 in the transfer direction, at least one of the parts 100 is detected when the detection process (with a short processing cycle T) detects the arrival of the part 100 at the detection position. Is also present in the recognition area image. Accordingly, it is possible to individually determine the posture of each component 100 based on the recognition area image in the steps S130, S135, and S155. Further, this determination process is performed only when the arrival of the component 100 at the detection position is detected by the detection process, so that the processing cycle T of the detection process described above can be kept sufficiently short, and each component at the detection position can be maintained. The arrival of 100 can be detected almost certainly.

  As a result, according to the image processing apparatus 7 of the present embodiment, even when a plurality of parts 100 having a short length in the transfer direction are transferred asynchronously and at irregular intervals, the posture of each part 100 is individually determined. it can.

  In this embodiment, in particular, the recognition process can capture the recognition area image from the same captured image 11 as that used in the immediately preceding detection process, thereby shortening the processing cycle T by omitting the time for re-imaging. In addition, it is possible to acquire a recognition area image in which at least a part of the component 100 exists reliably and determine its posture.

  In the present embodiment, in particular, in the discrimination process, the end of the part 100 located on the downstream side in the transport direction in the recognition area image is recognized in the image (step S130). Thereby, even when the two parts 100 are connected and transferred in the transfer direction, the boundary (the upstream end of the downstream part 100) is recognized, and only the posture of the downstream part 100 is recognized. Can be determined individually.

  In the present embodiment, in particular, in the determination process, when it is determined that the transfer direction length of the component 100 located on the downstream side in the transfer direction in the recognition area image is shorter than the predetermined length, the transfer direction is higher than the recognition area 13. The recognition area image of the re-recognition area 14 partitioned at a position shifted to the downstream side of the image is newly captured. As a result, the length in the transfer direction of the component 100 in the recognition area image can be made longer, and the posture determination of the component 100 in the determination process can be performed reliably.

  In the present embodiment, in particular, the predetermined length is a determination reference length that is a reference for determining the posture of the component 100 in the determination process. In the determination process, the component located downstream in the transfer direction in the recognition area image. The re-recognition area 14 is defined at a position where the length of 100 in the transfer direction is equal to or greater than the discrimination reference length. As a result, the length in the transfer direction of the component 100 in the recognition area image can be greater than or equal to the determination reference length, and the posture determination of the component 100 in the determination processing can be performed reliably.

  In the present embodiment, in particular, in the discrimination process, the circular hole 101 included in the component 100 is image-recognized as a shape feature portion, and the posture of the component 100 in the transport direction and the vertical direction based on the presence and position of the shape feature portion. And the above-mentioned discrimination reference length is an area length in which a predetermined number (four in this example) of round holes 101 can be provided in the component 100. Thereby, the attitude | position discrimination | determination of the components 100 in a discrimination | determination process can be performed more reliably.

<Application examples to other parts shapes>
The parts feeder 1 of the above embodiment can be applied to other various shaped parts.

<1: If there is a distinction between the vertical and front / back orientations>
For example, the parts feeder 1 of the above embodiment can also be applied to a part 200 having a shape as shown in FIG. 19 which is a front view and FIG. 20 which is a cross-sectional view taken along arrow XX-XX. The component 200 is formed by forming an embossed notch 202 only on the upper edge on one side of the front and back with respect to the imaging direction of the camera on the component 100 (see FIGS. 4 and 5) used in the above embodiment. Yes. In the component 200 having such a shape, the normal posture in both the front and back directions and the vertical direction shown in FIG. 19 is reversed in the front and back directions as shown on the left side in FIG. 21 (the notch 202 is exposed to the front side) and the vertical direction. And the normal posture in the front and back direction as shown in the center in FIG. 21, and the posture reversed in the vertical direction as shown in the right side in FIG. I can take it.

  For the component 200 having such a shape, the distance between each round hole 201 and the edge portion closest to them may be measured by image recognition in the discrimination process. As shown in FIG. 22, if the separation distances are uniformly the same, it can be determined that the component 200 is normal in the front and back direction. Further, as shown in FIG. 23, when the separation distance is measured by the round hole 201 close to the notch 202 (that is, when there is a difference or variation in the separation distance), the part 200 is reversed in the front and back direction. Can be determined. Note that the posture in the vertical direction may be compared by measuring the separation distance from the round hole 201 at each of the upper edge and the lower edge, or the center of gravity of the entire component 200 as in the above embodiment. You may discriminate | determine by comparing the position (illustration omitted) and the height position of each round hole 201. FIG.

<2: When there is a difference in posture in the transfer direction>
Moreover, as shown in FIG. 24, the parts feeder 1 of the said embodiment is applicable also to the components 300 of a shape with the distinction of an attitude | position in a transfer direction. The component 300 is formed in a shape in which the central axes of the large-diameter long portion 301 and the small-diameter short portion 302 are aligned with each other and aligned in the transfer direction. In the component 300 having such a shape, a normal posture in the transfer direction in which the small-diameter short portion 302 is positioned on the downstream side as shown on the left side in FIG. 24, or the small-diameter short portion 302 on the upstream side as shown on the right side in FIG. Two postures can be taken: a posture reversed in the transfer direction located on the side.

  For a part 300 having such a shape, a position that matches the shape feature is first detected by image recognition in the discrimination process. In the example shown in FIG. 25, the upper corner of the downstream end of the large-diameter long portion 301 is a shape feature, and the comparison pattern 303 corresponding thereto has only the upstream and lower sides of the component 300 in a substantially rectangular small region. Use a pattern that is partially present. In the recognition area image, the position in the transfer direction at a location that matches the shape feature of the comparison pattern 303 corresponds to the position of the downstream end of the large-diameter long portion 301. Note that the comparison pattern 303 is scanned to limit the height of the comparison range so that the downstream end of the small diameter short portion 302 is not detected. Then, the Y-direction existence length of the component 300 is detected at two positions before and after the detected downstream end of the large-diameter long portion 301 and compared, whereby the posture in the transfer direction of the component 300 can be determined. . In particular, the posture in the transfer direction can be determined based on whether or not the small-diameter short portion 302 is present downstream of the downstream end of the large-diameter long portion 301 in the transfer direction.

  When the posture of the component 300 is normal, the position where the shape feature comparison pattern 303 matches is limited to the downstream side in the recognition region 13 as shown in FIG. Therefore, by limiting the transfer direction range to be compared by scanning the comparison pattern 303 to the downstream side in the recognition area 13, the shape feature of the component 300 can be correctly corrected even after passing the recognition area 13, as shown in FIG. The posture can be determined by setting the re-recognition area 14. In this case, even if the two parts 300 are in the connected transfer state as shown in the drawing, it is possible to individually determine the posture of only the downstream part 300 without detecting the shape feature of the upstream part 300. In addition, as shown in FIG. 27, even when the two parts 300 in the articulated transfer state are both in the transfer direction reversal posture, it is possible to reliably determine the transfer direction posture of each part 300 in order from the downstream side.

<3: When there is a posture distinction in the vertical direction>
Moreover, as shown in FIG. 28, the parts feeder 1 of the said embodiment is applicable also to the components 400 of a shape with the distinction of an attitude | position in an up-down direction. The component 400 is formed in a shape in which a small-diameter portion 403 is sandwiched between a large-diameter long portion 401 and a large-diameter short portion 402, and the respective central axes are aligned so as to be aligned vertically and integrally joined. . In the component 400 having such a shape, a normal posture in the vertical direction in which the large-diameter long portion 401 is positioned on the lower side as shown on the left side in FIG. 28, or a large-diameter short portion 401 as shown on the right side in FIG. Two postures, that is, a posture reversed in the up-down direction, positioned on the lower side, can be taken.

  Even for the component 400 having such a shape, a portion that matches the shape feature portion is detected by image recognition in the discrimination processing. In the example shown in FIG. 29, a portion including the entire upper and lower direction of the large-diameter short portion 402 and the upper corner of the downstream end of the small-diameter portion 403 is used as a shape feature portion. A part of the part 400 (part of the large-diameter short part 402) is present in the center height position in the small region in the transfer direction, and a part of the part 400 (part of the small-diameter part 403 is also provided on the upstream and lower side. ) Is used. This comparison pattern 404 cannot be detected unless the entire component 400 is in a normal posture in the vertical direction. Therefore, if a shape feature that matches the comparison pattern 404 can be detected, it can be determined that the component 400 is in a normal posture in the vertical direction, and if it cannot be detected, it can be determined that the component 400 is in a posture inverted in the vertical direction.

  When the vertical posture of the component 400 is normal, the position where the comparison pattern 404 of the shape feature portion matches is limited to the downstream side in the recognition area 13 as shown in FIG. Therefore, by limiting the transfer direction range to be compared by scanning the comparison pattern 404 to the downstream side in the recognition region 13, the shape feature of the component 400 can be correctly corrected even after passing the recognition region 13, as shown in FIG. The posture can be determined by setting the re-recognition area 14. In this case, even when the two parts 400 are in the connected transfer state as shown in the drawing, the posture of only the downstream part 400 can be individually determined without detecting the shape feature of the upstream part 400. Further, as shown in FIG. 31, even when any one of the two components 400 in the articulated transfer state is in the vertically inverted posture, the vertical posture determination of each component 400 can be reliably performed in order from the downstream side.

<User interface and setting items>
The user interface and setting items that should be prepared in order to realize the functions of the above embodiment are listed below.

  First, there is a setting item for the component transfer direction. This is an item for setting the transfer direction of the components 100 to 400 in the camera field of view, and this determines the arrangement of the detection area 12 and the recognition area 13 in the captured image 11 and the shift direction of the re-recognition area 14.

  Second, there is a setting item for the re-recognition area shift amount. This is the amount of deviation from the first recognition area 13 to the downstream side in the transport direction when the posture of the parts 100 to 400 cannot be determined from the recognition area image while the arrival of the parts 100 to 400 is detected. Is a setting item as to whether or not the re-recognition area 14 is partitioned. In this setting item, an appropriate value is set according to the shape of the discriminating symmetrical parts 100 to 400, the length in the transfer direction, the arrangement of the shape feature portion, and the like. The numerical unit of the set value is basically a pixel unit or an actual size, but may be set in a unit specific to a part (for example, the number of shape feature portions such as a round hole).

  Third, there is a user interface for a dimension measurement menu. This is an operation menu for grasping the dimensions of the parts 100 to 400 in each pixel unit, and is operated when setting the re-recognition area deviation amount in pixel units or component-specific units.

  Fourth, there is a user interface for a pixel size calibration menu. This is an operation menu for obtaining the size relationship between the pixels and the actual size, and is operated when setting the re-recognition area deviation amount at the actual size.

  The fifth is a window border drawing user interface. This is an operation menu for drawing window frame lines in the partition ranges corresponding to the detection area 12 and the recognition area 13 when the captured image 11 of the camera 4 is displayed on the image display device 8. In order for the operator to visually confirm whether the detection process and the discrimination process are being performed at an appropriate timing while the parts feeder 1 is operating, a window frame line is drawn in each area.

  The operator performs the above-described user interface operation and setting item setting input via the operation interface 9 while confirming the transfer of the components 100 to 400 and the process of the detection process and the discrimination process on the image display device 8. As a result, it is possible to perform appropriate posture discrimination for the variously shaped parts 100 to 400.

  In the above-described embodiment, the posture determination method by image recognition in the parts feeder 1 using the vibration feeder 2 in which the transfer speed and transfer interval of the components 100 to 400 are uncertain has been described, but the present invention is not limited to this. In addition, for a component supply apparatus that uses a belt conveyor or the like and has a stable transfer speed or transfer interval. The same effect can be obtained by applying the above-described posture recognition method based on image recognition.

  Note that “equal” in the above description does not have a strict meaning. In other words, “equal” means that “tolerance and error in manufacturing are allowed by design and is“ substantially equal ”.

  In the above description, “vertical, orthogonal” does not mean exactly. In other words, “vertical and orthogonal” means that “tolerance and error in manufacturing are allowed in design and“ substantially vertical and orthogonal ”.

  Note that “parallel” in the above description does not have a strict meaning. That is, “parallel” means that “tolerance and error in manufacturing are allowed in design and“ substantially parallel ”.

  In addition to those already described above, the methods according to the above-described embodiments and modifications may be used in appropriate combination.

  In addition, although not illustrated one by one, the above-mentioned embodiment and each modification are implemented with various modifications within a range not departing from the gist thereof.

1 Parts feeder (part supply device)
2 Vibration feeder (transfer means)
2a Transfer route 2b Reservoir 2c Discharge port 3 Illumination 4 Camera (imaging means)
5 Air ejection nozzle (exclusion means)
6 Solenoid valve 7 Image processing device 8 Image display device 9 Operation interface 11 Captured image 12 Detection region 13 Recognition region 14 Re-recognition region 100 Parts (Transported object)
100 'Parts (Transported object)
101 Round hole (shape feature)
102 Center of gravity position 200 Parts (Transported object)
201 Round hole (shape feature)
202 Notch (shape feature)
300 Parts (Transported object)
301 Large-diameter long portion 302 Small-diameter short portion 303 Comparison pattern corresponding to shape feature 400 Component (Transported object)
401 Large diameter long part 402 Large diameter short part 403 Small diameter part 404 Comparison pattern corresponding to shape feature

Claims (9)

An image processing apparatus for discriminating a posture of a transferred object by recognizing a captured image of the transferred object transferred in a predetermined transfer direction,
Detection image capturing means for capturing a detection region image of a detection region partitioned at a predetermined position on the transport path of the transported object in the captured image;
A recognition image capturing means for capturing a recognition region image of a recognition region partitioned at a position adjacent to the upstream side of the detection region in the transport direction in the captured image;
Detecting means for detecting whether or not the object to be transported exists at the predetermined position based on the detection area image captured by the detected image capturing means;
A discriminating means for discriminating the posture of the transported body based on the recognition area image captured by the recognized image capturing means when the detecting means detects the presence of the transported body at the predetermined position;
An image processing apparatus comprising: The recognition image capturing means includes
2. The image according to claim 1, wherein among the captured images captured at a predetermined cycle, the recognition unit captures the recognition area image from a captured image in which the detection unit detects the presence of the transported body at the predetermined position. Processing equipment. The discrimination means includes
3. The image processing apparatus according to claim 1, wherein, in the recognition area image, an image recognition is performed on an upstream end portion in the transfer direction of a transfer object located downstream in the transfer direction. The recognition image capturing means includes
When the determination unit determines that the transfer direction length of the transfer object located downstream in the transfer direction in the recognition area image is shorter than a predetermined length, the determination unit is further downstream in the transfer direction than the recognition area. The image processing apparatus according to claim 3, wherein a recognition area image of a re-recognition area divided at a position shifted to the side is newly fetched. The predetermined length is a determination reference length that is used as a reference for determining the posture of the transported body by the determination unit.
The recognition image capturing means includes
5. The re-recognition area is defined at a position where a transfer direction length of a transfer object located downstream in the transfer direction in the recognition area image is equal to or greater than the determination reference length. Image processing device. The discrimination means includes
Recognizing an image of the shape feature portion of the transported body, and determining the posture of the transported body in the transport direction, the vertical posture, and the front and back posture based on the presence and position of the shape feature portion,
The discrimination reference length is
The image processing apparatus according to claim 5, wherein the transfer object has a region length that can include a predetermined number of the shape feature portions. An image processing method for discriminating the posture of the transferred object by recognizing a captured image of the transferred object transferred in a predetermined transfer direction,
Capturing a detection area image of a detection area partitioned at a predetermined position on a transfer path of the transfer object in the captured image;
Capturing a recognition area image of a recognition area partitioned at a position adjacent to the upstream side of the detection area in the transfer direction in the captured image;
Detecting whether the transferred object is present at the predetermined position based on the captured detection area image;
When detecting the presence of the transported object at the predetermined position, determining the posture of the transported object based on the captured recognition area image;
The image processing method characterized by performing. An image processing program to be executed by an arithmetic device provided in an image processing apparatus for recognizing a posture of the transferred object by recognizing a captured image of the transferred object transferred in a predetermined transfer direction,
Capturing a detection area image of a detection area partitioned at a predetermined position on a transfer path of the transfer object in the captured image;
Capturing a recognition area image of a recognition area partitioned at a position adjacent to the upstream side of the detection area in the transfer direction in the captured image;
Detecting whether the transferred object is present at the predetermined position based on the captured detection area image;
When detecting the presence of the transported object at the predetermined position, determining the posture of the transported object based on the captured recognition area image;
An image processing program characterized by executing Transfer means for transferring a plurality of parts stored therein in a row on a transfer path;
Imaging means for imaging the transfer path;
An image processing device for recognizing the orientation of the component by recognizing the captured image of the component imaged by the imaging means;
Exclusion means for excluding parts determined by the image processing apparatus that are not in a predetermined posture from the transfer path;
A component supply device comprising:
The image processing apparatus includes:
Detection image capturing means for capturing a detection area image of a detection area partitioned at a predetermined position on the transfer path in the captured image;
Recognition image capturing means for capturing a recognition area image of a recognition area partitioned at a position adjacent to the upstream side in the transfer direction of the part of the detection area in the captured image;
Detecting means for detecting whether or not the component is present at the predetermined position based on the detection area image captured by the detected image capturing means;
A discriminating unit for discriminating the posture of the component based on the recognition area image captured by the recognition image capturing unit when the detection unit detects the presence of the component at the predetermined position;
A component supply device comprising:

Images