JPH08118273A - Judgment of placement direction of objective substance and device therefor - Google Patents

Abstract

PURPOSE: To improve the reliability in the judgment of the placement direction of an objective substance, by selecting the standard data set in the region of the standard position on the objective substance and carrying out comparison and collation, in the collation of the standard data and the density data in the specific region of the objective substance which is image picked-up. CONSTITUTION: The image pick-up viewfield 30 is divided to three areas 30a-30c. A material is image-picked-up, and the position of a valve lifter hole 10 is extracted from the obtained image, and the type of the material is recognized through collation and specification. Then, the placement direction is judged. At first, the position 11 of the center of gravity of the valve lifter hole 10 is calculated, and then the tilt angle of the material on a pallet is calculated. Then, each area 30a-30c where the position 11 of the center of gravity exists is inspected, and the correlation model corresponding to the area is loaded. Then, the load of the density data at the first point 12, comparison and collation with the thresh level are carried out. At the second point, comparison and collation are carried out similarly.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an object placement direction determining method and apparatus for accurately determining the orientation of an object placed movably on a pallet.

[0002]

2. Description of the Related Art In order to manufacture an engine which is a part of an automobile part, a plurality of rough materials, which are the head parts of the engine prepared in the preceding step of the processing line, are placed on a pallet and placed at the entrance of the processing line. There is also a process for loading the material to the traverser and sequentially loading it into the processing line. Since multiple types of engine heads are loaded into the traverser, it is necessary to determine the type of engine in which each rough material is used before it is put into the processing line. Further, since the rough materials are placed on the pallet in a mixed state in the vertical and horizontal directions, all the rough materials of the types of the respective engines are arranged in the same direction and then put into the processing line. The determination of the type of the rough material on the pallet and the determination of the mounting direction on the pallet also play an important role in automating the engine manufacturing.

As a method for discriminating the type and placement direction of the rough material placed on the pallet, a method based on the light and shade pattern matching is used. This is a method of detecting a grayscale pattern of image data obtained by capturing an image of a rough material as an object and comparing the grayscale pattern of the image data with a grayscale pattern of the image data serving as a model to determine the model and direction. .

FIG. 6 is a diagram showing an orthogonal robot which sequentially determines the type and placement direction of the rough material placed on the pallet by using the light and shade pattern matching, and sequentially inputs the rough material to the processing line. This orthogonal robot is a four-axis robot in the X, Y, Z, and θ directions, and takes out the rough material on the pallet carried into the traverser by the forklift, aligns it in a predetermined direction, and then sequentially inserts it into the processing line. . Figure 6
In the above, the pallets are stacked in four stages, and ten rough materials are placed on each pallet, but it is also possible to arrange ten rough materials for each layer in one pallet and stack them in four layers. . The pallets that have been emptied of the rough material and are emptied are discharged. At the tip of the orthogonal robot, a rough material take-out section 2 for picking up the rough material by hooking it from the top of the pallet and a ring illumination 4 for irradiating light to the rough material are arranged, and further the rough material is imaged. A CCD camera 6 and a light source unit 8 are provided. An image will be input using these configurations.

FIG. 7 is a schematic view of a head portion of an engine, which is an object handled by an orthogonal robot, as viewed from above. As is apparent from FIG. 7, the valve lifter holes 10 are provided in a specific position or number depending on the type of engine. The valve lifter hole 10 is rarely changed in design. Moreover, since it has a conical shape, a light and shade pattern is likely to appear. Therefore, when the object is the head portion of the engine, it is possible to determine the type of rough material by recognizing the positional relationship of the valve lifter hole 10.

FIG. 8 is a flow chart showing a process for discriminating the type and placing direction of the rough material placed on the pallet. The method for discriminating the placing direction of the rough material will be described below based on this figure. Will be described.

The light from the ring illuminator 4 is illuminated on the rough material to obtain C
An image of the rough material is input by being imaged by the CD camera 6 and sent to an image processing device (not shown) (step 101). The picked-up image becomes black-and-white pattern data, and the contour line of the black-and-white pattern is extracted, and the site of the valve lifter hole 10 which is the characteristic of the model of each engine is extracted from the contour line (step 102). This can be achieved by extracting from the contour line only the line forming the closed figure having the circumference length and area of a specific size. By performing so-called model matching, in which the extracted data of the portion of the valve lifter hole 10 is compared with the data that is the model of each engine model, the rough material as the object corresponds to any engine model. Recognize whether (step 10
3).

As described above, by comparing the closed figure serving as the model with the specific contour line (closed figure) extracted from the contour line obtained from the picked-up image, an extra contour line is extracted from the picked-up image. Even if it is, the type of the rough material placed on the pallet can be discriminated. This can be determined even if the direction in which the rough material is placed on the pallet is not constant. Unlike the conventional model matching process using the binarization process, the method of extracting and collating a contour line is strong against ambient light and strong against illuminance change because the contour line of the object is emphasized and extracted. Therefore, it can be said that it is suitable when the object has irregularities such as an engine head.

When the model is determined, the placement direction of the rough material is determined (step 104). The arrangement and number of valve lifter holes differ depending on each model, so it is effective to use as a model identification, but since each valve lifter hole in each rough material does not have a feature that distinguishes it from other valve lifter holes, the valve lifter hole is not available. It is difficult to identify the mounting direction only by the shape and arrangement of the holes. Therefore, the mounting direction is determined by performing the light and shade pattern matching at another position on the rough material instead of the valve lifter hole. Conventionally, as shown in FIG. 7, a position (specific region) where a specific offset is provided from the center of gravity position 11 of the valve lifter hole 10.
In, the gradation pattern matching process is performed. Since this specific region should have a unique shape in each model, it is possible to determine the direction by comparing the grayscale data that is the model at this position with the grayscale data obtained from the captured image. it can. The first point 12 and the second point 14 shown in FIGS. 7A and 7B are the specific areas. The second point 14 is a point-symmetrical position with respect to the first point 12 with the center of gravity 11 as the center. Note that the grayscale data that serves as a model is called a correlation model.

FIG. 9 is a flow chart showing the details of the direction discriminating process shown in FIG. 8, and the direction discriminating process will be described with reference to this figure.

First, the correlation model previously determined for each model is loaded into the image processing apparatus. Since it has already been determined which type is the type, only the correlation model of the corresponding type is loaded (step 201). next,
The grayscale data obtained from the captured image of the first point 12 is loaded (step 202), and the correlation value (degree of coincidence) at the first point 12 is calculated by comparing and collating with the correlation model (step 203). . The correlation value is compared with a predetermined threshold level (step 204), and if the correlation value is larger than the threshold level, it is judged to be in the normal direction (step 205). FIG. 10 is a diagram showing the correlation value for each type for each rough material, and normally, the correlation value in the normal case stably exceeds the threshold level as shown in FIG. FIG. 11 is a diagram showing the correlation value for each rough material, and shows the relationship with the threshold level 0.72.
If the correlation value is equal to or lower than the predetermined threshold level as in the case of the rough material A in FIG. 11, the grayscale data obtained from the imaged image of the second point 14 is loaded (step 20).
6) Then, the correlation value (degree of coincidence) at the second point 14 is calculated by comparing and collating with the correlation model (step 207). Note that if all the rough materials are placed in the opposite direction to the normal direction, the result is as shown in FIG. The correlation value is compared with a predetermined threshold level (step 208), and if the correlation value is larger than the predetermined threshold level, it is determined that the rough material is placed in the inverted direction (step 20).
9). If the correlation value is below a predetermined threshold level,
It is determined that the direction cannot be determined (step 210).

As described above, first, the correlation value of the first point 12 is calculated and compared with the threshold level to determine whether it is in the normal direction. If it is not the normal direction, the second point at the point symmetry position is located. The correlation value of 14 is calculated, and if the correlation value is larger than the threshold level, it is determined that the rough material is inverted. In this way, when the placement direction of the rough material is determined, the image processing device calculates the position of the rough material (step 105), and when the position is transmitted to the robot control device (not shown), the robot control device The rough material take-out section of the orthogonal robot is operated to take out the rough materials on the pallet, align them in a predetermined direction, and sequentially feed them into the processing line.

[0013]

By the way, even if the same position is imaged by the CCD camera, the light and shade data is different because the appearance is slightly different depending on the position in the imaging visual field. For example, FIG. 13A is a diagram in which the characteristic part of the first point 12 in FIG. 7B is extracted.
Assuming that the cross-sectional view taken along the line AA ′ of FIG.
When the D camera is located at the positions p1, p2, and p3, the length d of the same portion looks different. In addition, the grayscale data also varies depending on the placement of the rough material itself, the influence of shadows, and the like. Thus, even if the parts have the same shape, C
It is obvious that the grayscale data of the image captured differs depending on the positional relationship between the CD camera and the rough material.

However, in the conventional method, since only one correlation model is prepared for each model,
The correlation value calculated by the above-mentioned method differs depending on the positions of the points 12 and 14 within the imaging visual field. Therefore, the correlation value may be equal to or lower than the threshold level depending on the positions of the points 12 and 14 in the imaging visual field even if they are the same part. That is, although the rough material is placed in the normal or reverse direction, the case where the direction cannot be normally determined occurs. If it cannot be discriminated normally, it is treated as a rough material that cannot be discriminated. Therefore, it is necessary to stop the Cartesian robot once and remove the rough material.

If the threshold level is set low (for example, to 0.55 as shown in FIG. 11), it is possible to surely recognize the normal or reverse direction by giving a large allowable range. Therefore, there is a possibility that the direction is erroneously determined to be correct even if the product is placed in the opposite direction or is a defective product. Therefore, the threshold level cannot be set too low.

The present invention has been made to solve the above problems, and an object thereof is to perform a discrimination process of a mounting direction of an object according to a position within a visual field of the object. An object of the present invention is to provide a method and apparatus for discriminating a placement direction of an object for improving its reliability.

[0017]

In order to achieve the above-mentioned object, the method for discriminating the object placement direction according to the present invention is
An object placement direction determining method for determining the orientation of an object by comparing the grayscale data of a specific area of an image obtained by capturing the object with the reference data. Of detecting the reference position, and selecting the reference data corresponding to the reference position from the reference data set for each of the areas obtained by dividing the imaging visual field into a plurality of areas.

Further, the object placement direction discriminating apparatus according to the present invention determines the direction in which the object is placed by collating the grayscale data of the specific area of the image obtained by imaging the object with the reference data. In an object placement direction discriminating apparatus for discriminating, an image pickup means for picking up an object, a reference position detection means for detecting a reference position on the object within the image pickup field of view, and setting for each area obtained by dividing the image pickup viewfield Reference data storing means for storing the selected reference data, and reference data selecting means for selecting the reference data corresponding to the reference position detected by the reference position detecting means from the reference data stored in the reference data storing means. And comparing the grayscale data of the specific area with the reference data selected by the reference data selecting means.
And a grayscale data collating means for judging.

[0019]

Therefore, in the object placement direction determining method according to the present invention, when the grayscale data of the specific region of the object obtained by imaging is compared with the reference data, the region having the reference position on the object is detected. Since the reference data set in the above is selected and compared and compared with the grayscale data, even if the appearance of the same part may change depending on the positional relationship between the imaging means and the target, the placement of the target It is possible to reliably discriminate the performed direction.

Further, according to the object placement direction discriminating apparatus of the present invention, the reference position detecting means detects the reference position on the object within the imaging visual field of the imaging means. The reference data selection means selects the reference data set in the area having the detected reference position from the reference data stored in the reference data storage means. Since the grayscale data collating means collates and determines the grayscale data of the specific region of the imaged image of the object and the selected reference data,
Even if the appearance of the same part may change depending on the positional relationship between the imaging unit and the object, the direction in which the object is placed can be reliably determined.

[0021]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT A preferred embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of an object placement direction discriminating apparatus according to the present invention. The present apparatus includes an image pickup unit 20 as an image pickup unit that picks up an image of a rough material that is an object, a reference position detection unit 22 that serves as a reference position detection unit that detects a reference position on the rough material within the imaging field of view,
Memory 24 as reference data storage means for storing the correlation model set for each of the regions obtained by dividing the imaging field of view.
And a correlation model selection unit 26 as reference data selection means for selecting a correlation model corresponding to the reference position detected by the reference position detection unit 22 from the correlation models stored in the memory 24, and details will be described later. A grayscale data collating unit 28 as a grayscale data collating unit that collates / determines the grayscale data of a specific area on the captured image with the correlation model selected by the correlation model selecting unit 26. Imaging unit 20
Corresponds to the CCD camera 6 in FIG. In addition, the reference position detection unit 22, the memory 24, the correlation model selection unit 26
The grayscale data collating unit 28 is configured in a normal image processing apparatus.

When determining the direction of the rough material placed on the pallet in this apparatus, the grayscale data of the specific area on the captured image and the reference data are collated as in the conventional case. What is characteristic of is that the correlation model set in the area (reference area) where the reference position exists on the rough material is selected, and the selected correlation model and the grayscale data are compared and collated. Even if the appearance of the same portion, that is, the specific region may change due to the influence of the shadow or the like depending on the positional relationship between the portion 20 and the object, the direction in which the rough material is placed can be reliably determined.

First, the correlation model in this embodiment will be described. Conventionally, one correlation model was prepared for each model. Then, in the grayscale data obtained by extracting the contour line from the captured image, by comparing the grayscale data of the specific region, that is, the first and second points 12 and 14 in FIG. The placement direction was determined. In the present embodiment, attention is paid to the fact that the length of a specific part appears to change depending on the position of the rough material in the imaging field of view, and that the grayscale data of the rough material changes even if the same rough material is imaged due to the influence of shadows and the like. , The field of view is divided into multiple areas,
The feature is that a correlation model is set for each area. That is, a plurality of correlation models corresponding to the area within the imaging visual field are set for one model. Figure 2
It is the figure which showed the example at the time of dividing an imaging visual field into three. In FIG. 2, the imaging visual field 30 is divided into areas 30a, 30b,
Divide into 30c. In addition, this FIG. 2 (a), (b),
(C) shows the valve lifter hole 1 on the rough material shown in FIG.
0, center of gravity position 11, first point 12 and FIG.
Although part of the outline of the rough material shown in (a) is also shown at the same time, it is clear that the appearance changes depending on the position of the first point 12 in the imaging visual field. Therefore, in the present embodiment, one of the correlation models set for each region is selected, and the first or second points 12, 1 are selected.
It was made to compare and collate with the grayscale data of 4.

FIG. 3 is a flow chart showing the processing of the object placement direction determining method in this embodiment.
The processing will be described based on this figure. Since the outline of the discrimination processing in the present embodiment is the same as the conventional one, FIG. 8 is used as it is, and the same reference numerals are given to the same processing as the conventional one shown in FIG.

The type of the rough material is recognized by extracting the part of the valve lifter hole 10 from the image obtained by picking up the image of the rough material, collating and specifying it (steps 101 to 103). When the model is determined, the placement direction of the rough material is determined next (step 104), but in the present embodiment, the center of gravity position 11 of the valve lifter hole 10 is first calculated as the reference position (step). 301). Since the position of each valve lifter hole 10 is detected when the type of the rough material is determined,
The center of gravity position 11 can be easily obtained.

Next, when the gravity center position 11 is obtained, the inclination angle of the coarse material on the pallet is calculated (step 30).
2). That is, the forklift may be displaced from the position when the rough material is placed on the pallet due to a vibration or the like when the pallet is transported. In this case, when the rough material is displaced from the original position and largely deviates from the imaging field of view of the CCD camera 6, the CCD camera 6 can be moved in the X and Y directions to find it. However, when imaging the rough material,
It is ideal if the rough material is in a reference position on the pallet, such as the arrangement of the valve lifter holes 10 and the normal direction of the rough material as shown in (a) and (b). , Fig. 4
In many cases, it is deviated from the direction which is the reference for the direction determination as in (c) and (d). In order to get each point 12, 14 correctly, its tilt angle must be obtained. This inclination angle can be easily obtained by calculating the deviation amount between the alignment of the valve lifter holes 10 and the normal direction. Once the tilt angle is obtained, the coordinate direction in the image processing apparatus can be adjusted by that amount to perform adjustment in the normal direction or the opposite direction. Therefore, in the subsequent processing, the first and second points can be adjusted. The grayscale data of 12 and 14 can be obtained correctly.

Next, it is checked in which area in the visual field of view the barycentric position 11 of the rough material exists (step 30).
3). This is easily understood because the position of the center of gravity 11 in the imaging visual field is known. When it is known which area the center of gravity 11 is included in, the correlation model of the corresponding area is loaded. The correlation model has already been described. After that, the grayscale data of the first point 12 is loaded and compared with the threshold level (steps 202 to 20).
5), loading the grayscale data of the second point 14 and comparing / verifying with the threshold level (steps 206 to 21)
0) and the like are the same as conventional ones, and therefore omitted. In the present embodiment, as described above, the correlation model having the same appearance is selected, so that a more accurate correlation value can be obtained, and as a result, highly reliable direction determination can be performed.

FIG. 5 is a diagram showing discrimination results in the conventional example and the present example when the rough material is placed in the normal direction and the center of gravity position 11 exists in each of the areas 30a, 30b, 30c. As is clear from this content, it can be understood that according to the present embodiment, it is possible to reliably determine that the rough material is placed in the normal direction in any area.

As described above, according to the present embodiment, a correlation model is set for each of the three divided areas 30a, 30b, 30c, and the center of gravity position 11 of the valve lifter hole 10 in the captured image is present in the correlation model. Since the correlation model corresponding to the area to be loaded is calculated to calculate the correlation value, the placement direction of the rough material on the pallet can be reliably performed without being affected by the positional relationship between the CCD camera 6 and the rough material, shadow, and the like. Can be determined. Therefore, it is not necessary to set the threshold level low, and the reliability of the determination process can be improved.

Since the present invention is characterized in that the imaging field of view is divided into a plurality of areas and the correlation model is set for each area, the imaging field of view is along the normal direction as in the above embodiment. However, when the object is the head portion of the engine on the pallet and has a shape close to a rectangle as in the present embodiment, normal division by three divisions along the normal direction is performed normally. The mounting direction can be determined. The division method will be determined by the shape of the object to be treated, the characteristic portion, and the like.

Further, in the above embodiment, the reference position is the center of gravity of the valve lifter hole and the specific region is the point-symmetrical position from the center of gravity. However, these may be arbitrarily set depending on the shape of the target object, the characteristic portion, etc. . Therefore, in the above-mentioned embodiment, the center of gravity position and the specific region may be divided into different regions, but this can be done by changing the setting of the reference position or the specific region. In this embodiment, it is within the threshold level range.

[0033]

According to the present invention, the imaging field of view is divided into a plurality of areas, reference data is set for each object even for one object, and the grayscale data and the reference data of a specific area of the object are set. Since the reference data set in the area having the reference position on the object is selected and compared and compared with the grayscale data, the positional relationship between the image pickup means and the object, the shadow, etc. Even if the appearance of the same specific area may change due to the influence, it is possible to reliably determine the direction in which the target object is placed without being affected by them.

Therefore, it is not necessary to set the threshold level used for discriminating the orientation low, and the reliability of the discrimination processing can be improved.

[Brief description of drawings]

FIG. 1 is a block configuration diagram showing an embodiment of an object placement direction determination device according to the present invention.

FIG. 2 is a diagram showing an example when a rough material is imaged in each area by dividing the imaging visual field into three in the present embodiment.

FIG. 3 is a flowchart showing a process of a method for determining an object placement direction in the present embodiment.

FIG. 4 is a diagram showing a mounting direction of a rough material in an imaging visual field in the present embodiment.

FIG. 5 is a diagram showing a discrimination result in a conventional example and a present example when a rough material is placed in a normal direction and a barycentric position exists in each area in the present example.

FIG. 6 is a diagram showing an orthogonal robot.

FIG. 7 is a schematic view of a head portion of each engine, which is an object, as viewed from above.

FIG. 8 is a flowchart showing a process of determining the type and placement direction of a rough material placed on a pallet.

FIG. 9 is a flowchart showing a process of a conventional object placement direction determination method.

FIG. 10 is a diagram showing a correlation value for each model for each rough material.

FIG. 11 is a diagram showing a relationship between a correlation value of each rough material and a threshold level.

FIG. 12 is a diagram showing a relationship between a correlation value of each rough material and a threshold level.

13A is a schematic view of a characteristic portion of the first point shown in FIG. 7B, and FIG. 13B is a sectional view taken along the line AA ′ of FIG.

[Explanation of symbols]

 6 CCD Camera 10 Valve Lifter Hole 11 Center of Gravity Position 12 First Point 14 Second Point 20 Imaging Section 22 Reference Position Detection Section 24 Memory 26 Correlation Model Selection Section 28 Grayscale Data Collating Section 28 30 Imaging Field of View 30a, 30b, 30c Area

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI Technical display location G06T 1/00 7/60 9061-5H G06F 15/70 370

Claims (2)

[Claims] 1. An object placement direction determining method for determining a direction in which an object is placed by comparing grayscale data of a specific region of an image obtained by picking up an image of an object and reference data, Detecting a reference position on the object in the image, and selecting reference data corresponding to the reference position from reference data set for each of the areas obtained by dividing the imaging visual field into a plurality of areas. A method for distinguishing a target object placement direction. 2. An object placement direction discriminating apparatus for discriminating a direction in which an object is placed by comparing grayscale data of a specific region of an image obtained by picking up an image of the object with reference data. An image pickup means for picking up an image, a reference position detection means for detecting a reference position on an object in the image pickup field of view, and a reference data storage means for storing reference data set for each of the regions obtained by dividing the image pickup field of view, Reference data selection means for selecting reference data corresponding to the reference position detected by the reference position detection means from the reference data stored in the reference data storage means, grayscale data of the specific area, and the reference data selection An object placement direction discriminating apparatus comprising: a grayscale data collating unit that collates and determines the reference data selected by the unit.