JP2014092904A - Method and apparatus for identifying workpiece - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a workpiece identifying method for identifying the kind of a workpiece with low rigidity.SOLUTION: A workpiece identifying method of identifying the kind of a workpiece to be identified with low rigidity includes: a step S100 of acquiring at least one local image of the workpiece to be identified; a step S200 of generating a feature image from the local image; a step S300 of collating the feature image with a plurality of reference images stored for each kind of the workpiece, to calculate correlation ratios between the reference images and the feature image, respectively; a step S800 of additionally storing the feature image as a new reference image when the maximum value of the calculated correlation ratios is lower than a set value; and a step S500 of identifying the kind of the workpiece of the reference image having the maximum correlation ratio value, as the kind of the workpiece to be identified, when the maximum value of the calculated correlation ratios is equal to or larger than the set value.

Description

  The present invention relates to a workpiece recognition method and a workpiece recognition apparatus.

  Conventionally, bumpers molded by a resin molding machine are subjected to processing such as gate cutting, and are then classified by vehicle type and type and transported to a production line. A work recognition device that recognizes the shape of the bumper from the captured image is used for classification of the bumper by vehicle type and type.

  In a conventional workpiece recognition apparatus, a real image of a workpiece to be recognized is acquired, a comparison model is generated from the real image, and the type of the workpiece is specified by collating with a stored reference model. Therefore, in the conventional workpiece recognition apparatus, the superiority or inferiority of the workpiece recognition rate and the recognition processing time are determined by the type and quantity of the reference model.

  In Patent Document 1 below, a reference point that can identify the type of workpiece is extracted from video data of all postures of the workpiece, and a standard model is determined based on this reference point. Further, in Patent Document 2, a plurality of feature points that are characteristic in shape are determined from a workpiece image, the geometric feature amounts and positional relationships of the feature points are calculated, and the feature amounts and the feature points are calculated. The positional relationship is used as a reference for collation. In Patent Document 3, a plurality of reference models are created by processing one reference model based on a predetermined condition.

Japanese Examined Sho 56-40981 JP-A-2-167410 JP-A-10-213420

  The above-described patent document recognition methods are based on the premise that all reference models to be compared with the comparison model are stored in advance. However, a work with low rigidity such as a resin bumper has a different shape even if it is placed differently. For this reason, in order to reliably specify the vehicle type of the bumper, it is necessary to store all deformation modes as a reference model. The verification that can cope with all the deformations requires not only a huge storage capacity that can store all the reference models, but also takes a lot of time to process the bumper model. For the above reasons, in the bumper manufacturing process adopting the tact method, the above-described recognition method of the patent document cannot be adopted as it is.

  The present invention has been made in view of the circumstances as described above, and does not store all reference models in advance as a workpiece recognition method and recognition device for identifying a type of workpiece having low rigidity, but in a recognition process. It is an object of the present invention to provide a workpiece recognition apparatus and a recognition method capable of additionally storing a reference model and improving a recognition rate.

  In order to achieve the above object, a workpiece recognition method of the present invention stores at least one local image of a workpiece to be recognized, a feature image is generated from the local image, and stored for each type of workpiece. Comparing the plurality of reference images with the feature image, calculating the correlation rates of the plurality of reference images and the feature image, respectively, and the maximum value of the correlation rates among the plurality of calculated correlation rates from the set value If it is low, the step of additionally storing the feature image as a new reference image, and if the maximum value of the correlation rate is greater than or equal to the set value among the plurality of calculated correlation rates, Identifying the type of workpiece as the type of workpiece to be recognized.

  In order to achieve the above object, a workpiece recognition apparatus of the present invention includes a camera, a light source, a recognition unit, a storage unit, and an input / output unit.

  The camera acquires a local image of a work to be recognized with low rigidity. The light source irradiates the workpiece during imaging of the workpiece by the camera. The storage unit stores a plurality of reference images for each type of work. The recognizing unit generates a feature image from the local image, compares the feature image with a plurality of reference images to calculate a correlation rate, and among the plurality of calculated correlation rates, the maximum value of the correlation rate is greater than a set value. If it is low, the feature image is additionally stored as a new reference image. If the maximum correlation rate is greater than or equal to the set value, the workpiece type of the reference image that calculated the maximum correlation rate is specified as the type of workpiece to be recognized. To do. The input / output unit performs an interface function between the recognition unit and the outside.

  According to the workpiece recognition method and workpiece recognition device of the present invention, a workpiece having low rigidity and easily deformed, and even if the workpiece position cannot be made constant, the workpiece recognition rate can be improved and the time required for recognition processing can be shortened. .

It is a figure which shows the structure of the workpiece | work recognition apparatus which concerns on Embodiment 1. FIG. It is a top view which shows the camera and light source which concern on Embodiment 1. FIG. It is a figure which shows the structure of a common parts stand. It is a figure which illustrates the position and orientation of the bumper mounted on the common parts table. It is a figure which illustrates the imaging part of a bumper. It is a figure where it uses for description of the light irradiation method in a light source. It is a figure which shows the image of a reference | standard image table. 3 is an operation flowchart of the workpiece recognition apparatus according to the first embodiment. It is a figure provided in order to demonstrate an example which recognizes the vehicle type of a bumper with the workpiece | work recognition apparatus which concerns on Embodiment 1. FIG. It is a figure which shows the image of the reference | standard image table in the example of FIG. It is a figure provided in order to demonstrate another example which recognizes the vehicle type of a bumper with the workpiece | work recognition apparatus which concerns on Embodiment 1. FIG. It is a figure which shows the image of the reference | standard image table in the example of FIG. It is a figure provided in order to demonstrate another example which recognizes the vehicle type of a bumper with the workpiece | work recognition apparatus which concerns on Embodiment 1. FIG. It is a figure which shows the image of the reference | standard image table in the example of FIG. 9 is a subroutine flowchart of an image acquisition step S100 of the first embodiment in the operation flowchart of FIG. It is a subroutine flowchart of step S300 of Embodiment 1 in the operation flowchart of FIG. It is a figure which shows the structure of the workpiece | work recognition apparatus which concerns on Embodiment 2. FIG. It is a top view which shows the camera and light source which concern on Embodiment 2. It is a figure which shows an example of the image which images a bumper with the workpiece | work recognition apparatus which concerns on Embodiment 2. FIG. It is a figure which shows the image which images the position and orientation when a bumper is distorted from the example of FIG. 19 with the workpiece | work recognition apparatus which concerns on Embodiment 2. FIG. It is a figure which shows the image which images the position and orientation when a bumper twists from the example of FIG. 19 with the workpiece | work recognition apparatus which concerns on Embodiment 2. FIG. 9 is a subroutine flowchart of an image acquisition step S100 of the second embodiment in the operation flowchart of FIG. It is a subroutine flowchart of step S300 of Embodiment 2 in the operation flowchart of FIG. It is a figure provided in order to demonstrate an example which recognizes the vehicle type of a bumper with the workpiece | work recognition apparatus which concerns on Embodiment 2. FIG. It is a figure which shows the image of the reference | standard image table in the example of FIG.

  Hereinafter, a workpiece recognition method and a workpiece recognition apparatus for recognizing a vehicle type of a resin bumper in a bumper manufacturing process will be described in detail with reference to the accompanying drawings, divided into (Embodiment 1) and (Embodiment 2). (Embodiment 1) recognizes a vehicle type from a local image acquired by imaging a bumper from one direction, and (Embodiment 2) recognizes a vehicle type from a local image acquired by imaging a bumper from two directions.

(Embodiment 1)
<Configuration of workpiece recognition device>
FIG. 1 is a diagram illustrating a configuration of a workpiece recognition apparatus according to the first embodiment. FIG. 2 is a plan view showing the camera and the light source according to the first embodiment. FIG. 3 is a diagram showing the structure of the common parts table. FIG. 4 is a diagram illustrating the position and orientation of the bumper mounted on the common parts table.

  The workpiece recognition apparatus 100 according to the first embodiment can improve the recognition rate of a bumper vehicle type even if the bumper is made of a resin bumper having low rigidity and easily deformed, and the bumper's position and orientation cannot be made constant.

  The configuration of the workpiece recognition apparatus 100 will be described in detail. As illustrated in FIG. 1, the workpiece recognition apparatus 100 includes a camera 110, a light source 120, a recognition unit 130, a storage unit 140, and an input / output unit 150.

  The camera 110 images the bumper from the front side of the bumper. As the camera 110, for example, a monochrome camera using an element such as a CCD or a CMOS is used. The camera 110 has a lens.

  The light source 120 irradiates the bumper from the back side of the bumper. The light source 120 uses a surface light source with uniform illuminance. As the light source 120, for example, a surface light source in which a diffusion plate is attached to a fluorescent lamp, an LED, or the like may be used.

  The recognition unit 130 communicates with the camera 110, the light source 120, the storage unit 140, and the input / output unit 150. The recognizing unit 130 generates a feature image for recognizing the vehicle type from the local image acquired by the camera 110, and compares the feature image with a plurality of reference images stored in the storage unit 140 to calculate a correlation rate. Recognize bumper models. When the maximum value of the correlation rate is lower than the set value, the recognizing unit 130 additionally stores the feature image in the storage unit 140 as a new reference image of the vehicle type specified by the input / output unit 150. On the other hand, when the maximum value of the correlation rate is equal to or greater than the set value, the vehicle type corresponding to the reference image for which the maximum value of the correlation rate is calculated is automatically specified as the vehicle type of the bumper to be recognized. The recognition unit 130 may be configured by a computer including a CPU, for example.

  The storage unit 140 stores a reference image for automatically specifying the vehicle type of the bumper for each vehicle type. The storage unit 140 has a reference image table. The storage unit 140 may be a storage device such as an HDD.

  The input / output unit 150 displays the processing contents of the recognition unit 130 and the result of recognizing the vehicle type so as to be visually confirmed in real time, and outputs the recognized bumper vehicle type information to the external device on the downstream side. The input / output unit 150 inputs designation information from the outside to the recognition unit 130. The input / output unit 150 is an I / O device such as a touch panel, for example.

  As shown in FIG. 2, since the bumper has a three-dimensional shape having a concave portion, the work recognition device 100 normally recognizes the vehicle type while being supported by the common component base M. The outer surface of the bumper has a plurality of holes for taking in air and attaching a lamp.

  As shown in FIG. 3, the common parts table M only supports the bumper from below with a plurality of frames so that bumpers of different vehicle types can be mounted, and does not fix the position and orientation of the bumper with a jig. For this reason, when the bumper is mounted on the common parts table M by hand or the like, the position and orientation of the bumper cannot be made constant, so even the bumper of the same vehicle type is common as illustrated in FIGS. 4 (a) to 4 (d). On the parts table M, they have different positions and orientations.

  As described above, in order to accurately grasp the vehicle type of the bumper whose position and orientation are not determined, the following ingenuity has been devised for imaging the bumper.

  FIG. 5 is a diagram illustrating an imaging region of the bumper. 5A is a diagram showing a part of the outer surface of the bumper of the vehicle type A, and FIG. 5B is a diagram showing a part of the outer surface of the bumper of the vehicle type B. The bumpers of the vehicle type A and the vehicle type B are very similar vehicle types that differ only in the opening part surrounded by the dotted line. As described above, the shape of the opening portion is different between bumpers of different vehicle types.

  If a bumper image obtained by capturing a wide range with the camera 110 is used for vehicle type recognition, the shape changes at low rigidity even in the same vehicle type, making it impossible to distinguish it from a very similar vehicle type and recognition is not possible. It becomes stable.

  Therefore, only the opening portion that is relatively rigid and difficult to deform is imaged, and the image of the opening portion is used for vehicle type recognition. By collating images in a narrow range, a slight shape difference of a very similar vehicle type is determined.

  FIG. 6 is a diagram for explaining a light irradiation method in a light source.

  FIG. 6A is a diagram illustrating the difference Δd in the imaging distance from each of the hole, which is the opening of the bumper, and the notch to the camera 110. Suppose that the bumper is irradiated with the light source 120 from the front side and the outer surface of the bumper is imaged with the camera 110 from the same front side. When the imaging distance difference Δd exceeds the focal depth of the lens, the image acquired by the camera 110 is an image in which at least one of the hole and the notch is out of focus.

  FIG. 6B is a diagram showing a positional deviation that occurs when the bumper is mounted on the common parts table M. FIG. Similar to the case of FIG. 6A, when the bumper positional deviation amount Δd ′ in the imaging direction exceeds the focal depth of the lens, the image acquired by the camera 110 may be out of focus depending on the bumper position. In either case of FIGS. 6A and 6B, when the blur occurs, the edge shape of the opening cannot be accurately recognized.

  Therefore, as shown in FIG. 6, the bumper is irradiated with the surface light source from the other side of the light source 120 so as to face the camera 110, thereby creating a bumper silhouette. Since the light from the light source 120 passes through the opening portion of the bumper, the edge shape of the opening portion appears when viewed from the front side of the bumper. The camera 110 images the brightness of the end face of the bumper. Therefore, it becomes easy to recognize the edge shape of the opening regardless of the focal depth of the lens.

FIG. 7 is a diagram illustrating an image of the reference image table. Line numbers A, B,..., X indicate bumper vehicle types, and column numbers 1, 2,..., Y indicate the order of reference images. RM _XY located at the matrix coordinates (X, Y) indicates the Yth reference image of the vehicle type X. For example, RM _C2 located at the matrix coordinates (C, 2) indicates the second reference image of the vehicle type C. The reserve position, such as the matrix coordinates (A, 3), indicates a position where the reference image can be additionally stored. The storage unit 140 manages the reference image for each vehicle type according to the matrix coordinates of the reference image table. The reference image table can be accessed from the recognition unit 130 to add or delete the reference image.

<Operation of workpiece recognition device>
FIG. 8 is an operation flowchart of the workpiece recognition apparatus according to the first embodiment. The operation flowchart shown in FIG. 8 also shows the procedure of the workpiece recognition method according to the first embodiment.

  In order to make the workpiece recognition method according to the first embodiment easier to understand, an example in which three bumpers that are the same vehicle type B are sequentially recognized by the workpiece recognition apparatus 100 will be described. 9 and 10 are diagrams for explaining recognition of the first bumper. 11 and 12 are diagrams for explaining recognition of the second bumper. FIGS. 13 and 14 are diagrams for explaining recognition of the third bumper.

Here, as the position and orientation of the bumper on the common parts table M, a relatively large positional deviation occurs in the first to second cases, and a relatively small positional deviation occurs in the second to third cases. . That is, in the images acquired by the camera 110, the second and third bumper local image shown in FIG. 9 (a) 1-th and the local image _{I A} bumper shown in FIGS. 11 (a) and 13 (a) different relatively large and I _B and _{I C,} but the second local image _{I B} and third local image _{I C} different relatively small.

  In the workpiece recognition method according to the first embodiment, the reference image is additionally stored in the storage unit 140 in the operation of the workpiece recognition apparatus 100 in the recognition of the second bumper. Therefore, the recognition rate of the second and subsequent vehicle type B bumpers is improved. I am letting.

  Hereinafter, the operation of the workpiece recognition apparatus 100 according to the first embodiment will be described in detail with reference to FIGS. 9 to 14 based on the operation flowchart of FIG.

<Step S100>
The camera 110 acquires a local image of the bumper. Specifically, the camera 110 acquires a silhouette image of the opening portion of the bumper from the front side of the bumper. For example, in the recognition of the first bumper, it obtains a local image I _A shown in Figure 9 (a). Similarly, in the recognition of the second and third bumper, it obtains a local image _{I B} and _{I C} are shown in FIGS. 11 (a) and 13 (a). Details of the local image acquisition processing will be described later with reference to a subroutine flowchart of FIG.

<Step S200>
The recognition unit 130 generates a feature image from the local image acquired in step S100. Specifically, the recognition unit 130 performs edge processing on the silhouette image of the opening to generate an edge line image of the opening. An edge line image of the opening is used as a feature image. For example, in the recognition of the first bumper, it generates a feature image SM _A shown in Figure 9 (b). Similarly, the feature images SM _B and SM _C shown in FIGS. 11B and 13B are generated in the recognition of the second and third bumpers.

<Step S300>
The recognizing unit 130 accesses the reference image table, sequentially calls a plurality of reference images stored in the storage unit 140, and calculates the maximum value R _MAX of the correlation rate with the feature image generated in step S200. For example, calculated in the recognition of the first bumper accesses the reference image table shown in Figure 10, the maximum value R _MAX of the correlation coefficient between the feature images SM _A already with all the reference image stored. Similarly, in the recognition of the second and third bumpers, the reference image tables shown in FIGS. 12 and 14 are accessed to calculate the maximum correlation value R _MAX . FIG. 9C, FIG. 11C, and FIG. 13C show reference images RM _B1 , RM _B1 , and RM _{B2 in} which the maximum value R _MAX of the correlation rate is calculated in each recognition. FIG. 9D, FIG. 11D, and FIG. 13D show a part of the result of recognizing the bumper vehicle type by the recognition unit 130 in each recognition. Details of the processing for calculating the maximum value R _MAX of the correlation rate will be described later using the subroutine flowchart of FIG.

<Step S400>
The recognizing unit 130 compares the maximum correlation value R _MAX calculated in step S300 with the set value R _Th . If R _MAX is equal to or greater than R _Th , the process proceeds to step S500, and otherwise, the process proceeds to step S600.

<Step S500>
Since R _MAX calculated in step S300 is equal to or greater than R _Th , the recognition unit 130 automatically identifies the bumper vehicle type corresponding to the reference image for which the R _MAX is calculated as the bumper vehicle type to be recognized. For example, in the recognition of the first bumper, since the maximum value 75% of the correlation rate is equal to or more than the set value 60% as shown in FIG. 9D, FIG. 10 of the reference image RM _B1 shown in FIG. The line coordinates B in the reference image table shown in FIG. In the recognition of the third bumper, since the maximum value 88% of the correlation rate is 60% or more as shown in FIG. _13D , FIG. 14 of the reference image RM _B2 shown in FIG. The line coordinates B in the reference image table shown in FIG.

In the above example, the reference image table shown in FIG. 10 is the same as the reference image table shown in FIG. 7, and the number of reference images is not changed. However, the reference image table shown in FIG. The number of reference images is increased by one. The increased reference image is a reference image RM _B2 additionally stored in step S800 described later.

<Step S600>
Since R _MAX calculated in step S300 is lower than R _Th , the recognition unit 130 interrupts the recognition process, and causes the input / output unit 150 to display the result of recognizing the vehicle type with the reference image for which the R _MAX is calculated. The recognition result includes information such as a recognition result (OK / NG), a reference image, a correlation rate R _MAX , and a set value R _Th . For example, in the recognition of the second bumper, as shown in FIG. 11 (d), the maximum value 44% of the correlation rate is lower than the set value 60%. Therefore, the vehicle type in the reference image RM _B1 shown in FIG. The result of recognizing is displayed. Here, the reference image table shown in FIG. 12 is the same as the reference image table of FIG. 7, and the number of reference images does not change.

<Step S700>
The operator visually confirms the bumper, manually designates the vehicle type of the bumper, and inputs the recognition unit 130 through the input / output unit 150. For example, in recognition of the second bumper, the vehicle type B is manually designated and input to the recognition unit 130.

<Step S800>
The recognition unit 130 operates the reference image table to store the feature image in the storage unit 140 as a new reference image of the vehicle type specified by the input / output unit 150, and restarts the recognition process. For example, in the recognition of the second bumpers, looking for a matrix coordinate of the reserve position of the vehicle type B from the reference image table shown in FIG. 12 (B, 2), stores the feature image SM _B as the new reference image RM _B2 models B It is additionally stored in the unit 140.

Therefore, in the reference image table, the reference image of the vehicle type B is increased by one, so that the recognition rate of the second and subsequent vehicle type B bumpers can be improved. For example, consider a case where the reference image RM _B2 is not additionally stored in the recognition of the third bumper. As described above, since the position / posture of the third bumper on the common parts table M is not so different from the position / posture of the second bumper, the vehicle type cannot be automatically specified as in the result of recognizing the vehicle type of the second bumper. It should be. However, in the recognition of the second bumper, the reference image RM _B2 of the vehicle type B is additionally stored in the reference image table in step S800, so that the third bumper can be automatically specified as the vehicle type B in step S500 described above. .

<Step S900>
The input / output unit 150 outputs bumper vehicle type information to an external device on the downstream side.

  As described above, since the workpiece recognition apparatus 100 additionally stores the reference image in the storage unit 140 only when performing the operation of step S800, it is not necessary to store the reference image more than necessary. For example, when the vehicle type can be automatically specified only by comparing the reference image already stored in the reference image table with the feature image generated in step S200, no further reference image is stored for the vehicle type. . By storing the reference images reasonably and flexibly, the number of reference images can be minimized while improving the recognition rate of the bumper vehicle type. Therefore, the processing time for calculating the correlation rate with the feature image can be shortened.

  Further, even when the bumper vehicle type is changed or the reference image of the bumper vehicle type is changed due to a change in production conditions, the reference image can be deleted or additionally stored only by operating the reference image table each time. Therefore, it is possible to avoid stopping the production line for re-storing the reference image.

  Next, details of the local image acquisition process step S100 will be described. FIG. 15 is a subroutine flowchart of the image acquisition step S100 of the first embodiment in the operation flowchart of FIG.

<Step S1101>
The light source 120 irradiates the bumper from the back side of the bumper to create a bumper silhouette. The light source 120 may irradiate the bumper while the camera 110 is imaging.

<Step S1102>
The camera 110 acquires a silhouette image of the opening portion of the bumper from the front side of the bumper. After imaging by the camera 110, the light source 120 may be turned off.

Next, detailed processing of the maximum value R _MAX calculation process step S300 of the correlation rate between the feature image and the reference image will be described in detail. FIG. 16 is a subroutine flowchart of step S300 of the first embodiment in the operation flowchart of FIG. The recognition unit 130 performs the following operations in steps S1301 to S1307.

<Step S1301>
Initialize variables used in the calculation process. For example, the maximum correlation value R _MAX is initialized to zero.

<Step S1302>
The number N of reference images currently stored is acquired from the reference image table in the storage unit 140.

<Step S1303>
One reference image is selected from the N reference images. The method for selecting the reference image may be arbitrary. However, the same reference image is not selected repeatedly in a single subroutine process.

<Step S1304>
A correlation rate between the feature image and the selected reference image is calculated. The correlation rate is calculated using a normal pattern matching method. For example, paying attention only to the shape of the edge line, pattern matching for correcting the positional deviation and the rotation amount is performed, and the similarity of the shape feature called the edge line is calculated.

<Step S1305>
The correlation rate calculated in step S1304 is compared with R _MAX . If the correlation rate is greater than R _{MAX, the process} proceeds to step S1306, and otherwise, the process proceeds to step S1307.

<Step S1306>
Substituting the correlation rate into R _MAX . Therefore, R _MAX always has a maximum value every time the correlation rate between the feature image and the reference image is calculated.

<Step S1307>
It is determined whether or not calculation of correlation rates with all reference images has been completed. If all the processes are completed, the process proceeds to step S400 in FIG. 8, and otherwise, the process proceeds to step S1303.

  In the first embodiment, the case where the local image is acquired from the front side of the bumper has been described. However, the present invention is not limited thereto, and the local image may be acquired from another direction, for example, the side surface side of the bumper.

  Although the first embodiment has been described based on the assumption that some reference image is stored in the reference image table, the operation may be started from a blank state in which no reference image is stored. In this case, since there is no reference image, the first feature image is stored as it is as the reference image.

  The workpiece recognition method and workpiece recognition apparatus according to the first embodiment recognize the bumper vehicle type while additionally storing a reference image as necessary. Even if the bumper of the same vehicle type changes in shape or displacement due to deformation or the like, it is additionally stored in the storage unit 140 as the same vehicle type whenever a difference in recognized shape occurs. Therefore, the recognition rate of the bumper can be improved.

  The workpiece recognition method and workpiece recognition apparatus according to the first embodiment create a bumper silhouette with a surface light source, and capture a silhouette image of an opening with a camera. Therefore, the shape feature of the opening that can recognize the vehicle type is stable, and the recognition rate of the bumper can be improved.

(Embodiment 2)
<Configuration of workpiece recognition device>
FIG. 17 is a diagram illustrating a configuration of a workpiece recognition apparatus according to the second embodiment. FIG. 18 is a plan view showing a camera and a light source according to the second embodiment.

  The workpiece recognition apparatus 200 includes cameras 210 and 211, light sources 220 and 221, a recognition unit 230, a storage unit 240, and an input / output unit 250. The workpiece recognition apparatus 200 may be the same as the workpiece recognition apparatus 100 according to the first embodiment.

The difference in configuration between the workpiece recognition apparatuses according to the second embodiment and the first embodiment is that the workpiece recognition apparatus 200 according to the second embodiment recognizes the vehicle type by imaging the bumper from two directions. A feature image is generated from a local image of a bumper acquired from two directions, a correlation rate between each feature image and a reference image stored in the storage unit 240 is calculated, a maximum value R _MAX of the correlation rate is calculated, Recognize bumper models.

  The image which image | photographs the bumper of the same vehicle type B with the workpiece | work recognition apparatus 200 which concerns on Embodiment 2 is shown using FIGS. 19A to 21B, the positional relationship between the cameras 210 and 211 and the bumper is shown in FIGS. 19A and 21B, and the edge line shape of the feature image generated from the image acquired by the camera 210 is shown in FIG. (D) shows the edge line shape of the feature image generated from the image acquired by the camera 211.

  FIG. 19 is a diagram illustrating an example of an image in which a bumper is imaged by the workpiece recognition apparatus according to the second embodiment.

  FIG. 20 is a diagram showing an image of the position and orientation when the bumper is distorted from the example of FIG. 19 in the work recognition apparatus according to the second embodiment. Comparing FIG. 20C and FIG. 19C, when the bumper is distorted, the edge line shape of the opening portion on the front side does not change. On the other hand, comparing FIG. 20D and FIG. 19D, when the bumper is distorted, the edge line shape of the opening portion on the side surface side changes.

  FIG. 21 is a diagram showing an image of the position and orientation when the bumper is twisted from the example of FIG. 19 in the workpiece recognition apparatus according to the second embodiment. Comparing FIG. 21C and FIG. 19C, when the bumper is twisted, the edge line shape of the opening portion on the front side changes. On the other hand, comparing FIG. 21D and FIG. 19D, when the bumper is twisted, the edge line shape of the opening portion on the side surface side does not change.

  When the bumper is distorted, it is easier to recognize the vehicle type of the bumper from the image acquired by the camera 210 that images the bumper from the front side than the camera 211. On the other hand, when the bumper is twisted, it is easier to recognize the vehicle type of the bumper from the image acquired by the camera 211 that images the bumper from the side than the camera 210.

  Assume that a feature image having the edge line shape of FIG. 19C is stored as a reference image in the reference image table. When the bumper is twisted as shown in FIG. 21, the edge line shape of FIG. 21 (c) and the edge line shape of FIG. 19 (c) are relatively different depending on the twist angle, and the correlation between the two can be low. . If the maximum correlation rate is lower than the set value, the bumper model cannot be automatically identified. Therefore, even in the case of the bumper of the same vehicle type, in order to improve the recognition rate, a large number of reference images must be additionally stored in the twisted direction.

  On the other hand, if the feature image having the edge line shape of FIG. 19D is stored as the reference image in the reference image table, the edge line shape of FIG. 21D and the edge line shape of FIG. Since there is no change, the correlation rate between the two is high, and the vehicle type can be sufficiently automatically specified without additionally storing a large number of reference images as in the front side.

<Operation of workpiece recognition device>
The operation of the workpiece recognition apparatus according to the second embodiment can also be shown in the operation flowchart of FIG.

The difference in operation between the workpiece recognition apparatus according to the second embodiment and the first embodiment is that the subroutine of the image acquisition step S100 and the maximum correlation rate R _MAX calculation step S400 in FIG. 8 is different.

  FIG. 22 is a subroutine flowchart of the image acquisition step S100 of the second embodiment in the operation flowchart of FIG.

  Steps S2101 and S2103 perform the same operation as step S1101 of the first embodiment, and create a bumper silhouette with the light sources 220 and 221.

  Steps S2102 and S2104 perform the same operation as in step S1102 of the first embodiment, and the silhouette images of the opening portions of the bumper are acquired by the cameras 210 and 211.

  The operations of step S2103 and step S2104 may be performed first, and then the operations of step S2101 and step S2102 may be performed. In order to obtain a clearer silhouette image, it is preferable to irradiate the light sources 220 and 221 at the same time and not shoot with the cameras 210 and 211 but irradiate them in order.

  FIG. 23 is a subroutine flowchart of step S300 of the second embodiment in the operation flowchart of FIG.

  Step S2301 of the second embodiment performs the same operation as step S1301 of the first embodiment, and steps S2304 to S2308 of the second embodiment perform the same operations as steps S1302 to S1306 of the first embodiment.

  Hereinafter, only operations of Step S2302, Step S2303, Step S2309, and Step S2310 that perform operations different from those of the first embodiment will be described in the second embodiment.

<Step S2302>
The number M of feature images generated from the local images is acquired. For example, when a feature image is generated from two local images acquired using the two cameras 210 and 211, the number M of feature images is set to two.

<Step S2303>
One feature image is selected from the M feature images. The method for selecting the feature image may be arbitrary. According to the order in which the bumpers are imaged in step S100, for example, a feature image generated from the image acquired by the camera 210 previously captured may be selected. However, the same feature image is not selected repeatedly in a single subroutine process.

<Step S2309>
It is determined whether or not calculation of correlation rates with all reference images has been completed. If all are completed, the process proceeds to step S2310. Otherwise, the process proceeds to step S2305.

<Step S2310>
It is determined whether or not calculation of correlation rates between all feature images and N reference images has been completed. If all the processes are completed, the process proceeds to step S400 in FIG. 8, and otherwise, the process proceeds to step S2303.

  In order to make the understanding of the workpiece recognition method according to the second embodiment easier to understand, a step in which a bumper that is a vehicle type B is recognized by the workpiece recognition device 200 according to the second embodiment will be described.

FIG. 24 is a diagram for explaining an example of recognizing a bumper vehicle type by the workpiece recognition apparatus according to the second embodiment. Shows the local image _{I D} obtained by the camera 211 in FIG. (A), shows the feature image SM _D generated by the edge processing from the local image _{I D} in FIG. (B). The reference image RM _B2 ′ from which the maximum correlation rate value R _MAX is calculated is shown in FIG. (C), and part of the result of recognizing the vehicle type is shown in FIG. As a comparison, a local image _{I B,} and wherein the image SM _B obtained in the front side of the camera 210 of the workpiece recognition device 200 according to the second embodiment is assumed to be those shown in FIG. 11 (a), FIG. 11 (b). FIG. 25 is a diagram showing an image of the reference image table in the example of FIG.

Camera 210 and 211 obtains a local image _{I B} and _{I D} of the bumper, respectively (S100).

The recognition unit 230 generates feature images SM _B and SM _D from the local images I _B and I _D (S200).

Recognition unit 230, a plurality of reference images stored in the storage unit 240 accesses the reference image table shown in Figure 25 call in turn, the maximum value R _MAX of the correlation coefficient between the feature images SM _B and SM _D Calculate (S300).

As shown in FIGS. 11D and _24D , since the correlation rate 80% of the reference model RM _B2 ′ is larger than the correlation rate 44% of the reference model RM _B1 , the correlation rate calculated in step S300 is The maximum value R _MAX is 80%.

  The recognizing unit 130 compares the maximum correlation rate value of 80% with the set value of 60% (S400).

Since R _MAX is equal to or greater than R _Th , the recognition unit 130 can automatically specify the bumper vehicle type B corresponding to the reference image RM _B2 ′ for which the R _MAX is calculated as the bumper vehicle type to be recognized (S500).

  The input / output unit 250 outputs bumper vehicle type information to an external device on the downstream side (S900).

  As described above, the workpiece recognition apparatus 200 according to the second embodiment captures the bumper from two directions, compares the bumper with the reference image in the two directions in the reference image table, and compares the feature image and the reference image in each direction. By calculating the maximum correlation rate and recognizing the bumper model, the bumper recognition rate is improved.

Consider the case where the bumper is twisted as shown in FIG. As in the case of the first embodiment, when the local image is acquired by the camera 210 only on the front side, the characteristic image SM _B shown in FIG. The new reference image RM _B2 for _B must be additionally stored in the storage unit 240. Depending on the amount of positional deviation on the front side, the calculated correlation rate may vary greatly, so that it becomes necessary to additionally store a large number only with the reference image on the front side. On the other hand, as in the second embodiment, when the local images are acquired by the cameras 210 and 211 on both the front side and the side side, only one reference image RM _B2 ′ shown in FIG. Even if a misalignment occurs, the bumper model can be automatically identified.

  Therefore, the number of reference images stored in the storage unit 240 can be minimized, and the correlation time calculation time can be shortened.

  In the workpiece recognition apparatus 200 according to the second embodiment, the case where the local part of the bumper is imaged from two directions using the two cameras 210 and 211 and the light sources 220 and 221 has been described. The workpiece recognition method according to the present embodiment can be applied even when using a direction, two or more cameras, and a light source. From the viewpoint of cost performance, it is better to use two cameras (center and end) on the front side and one camera on the side side from the shape characteristics of the opening of the bumper vehicle type.

  The workpiece recognition apparatus and the recognition method according to the second embodiment acquire a plurality of situation images using a plurality of cameras. Therefore, even when the bumper is misaligned in the three-dimensional direction, the number of reference images to be stored can be minimized, the correlation rate calculation time can be shortened, and the bumper recognition rate can be improved.

100, 200 Work recognition device,
110, 210, 211 cameras,
120, 220, 221 light source,
130, 230 recognition unit,
140, 240 storage unit,
150, 250 Input / output unit.

Claims (9)

A workpiece recognition method for identifying the type of workpiece to be recognized with low rigidity,
Obtaining at least one local image of the workpiece to be recognized;
Generating a feature image from the local image;
Collating a plurality of reference images stored for each type of workpiece and the feature image, and calculating a correlation rate between the plurality of reference images and the feature image, respectively;
A step of additionally storing the feature image as a new reference image when a maximum value of the correlation rate among the plurality of calculated correlation rates is lower than a set value;
Of the plurality of calculated correlation rates, when the maximum value of the correlation rate is equal to or greater than the set value, the type of workpiece of the reference image for which the maximum value of the correlation rate has been calculated is A step to identify as a type,
A workpiece recognition method characterized by comprising:   2. The work recognition method according to claim 1, wherein a work type of the feature image is designated, and the feature image is additionally stored as the new reference image of the designated work type.   The work recognition method according to claim 1, wherein the correlation rate is a correlation rate between shape characteristics of the reference image and the feature image.   The work recognition method according to claim 1, wherein the work is a resin molded product.   The work recognition method according to claim 1, wherein the reference image and the local image are silhouette images of the work.   The said reference | standard image and the said local image include the front image which imaged the said workpiece | work from the front side, and the side image which imaged the said workpiece | work from the side surface side. Work recognition method. A camera that acquires a local image of a work to be recognized with low rigidity;
A light source for illuminating the workpiece during imaging of the workpiece by the camera;
A storage unit that stores a plurality of reference images for each type of the workpiece;
A feature image is generated from the local image, a correlation rate is calculated by collating the feature image with a plurality of the reference images, and a maximum value of the correlation rate among a plurality of the calculated correlation rates is greater than a set value If it is low, the feature image is additionally stored as a new reference image. If the maximum value of the correlation rate is greater than or equal to the set value, the type of workpiece of the reference image for which the maximum value of the correlation rate has been calculated is recognized. A recognition unit that identifies the type of workpiece to be
An input / output unit that performs an interface function between the recognition unit and the outside;
A work recognition apparatus comprising:   The workpiece recognition apparatus according to claim 7, wherein the light source is a surface light source.   9. The workpiece recognition apparatus according to claim 7, wherein the workpiece is a resin molded product.