JP2012137474A - Method and device for recognizing work position - Google Patents

Abstract

PROBLEM TO BE SOLVED: To recognize a position of a work, even if there is no feature portion in the shape or the like of the work when recognizing the position of the work by a camera, and to accurately recognize the work to be an object even if a plurality of works enter a field of view of the camera.SOLUTION: A method for recognizing a work position includes: storing a basic shape of a work model viewed from a predetermined direction; calculating a correlation between the basic shape and the work shape acquired by the camera as image information and moving the camera in the direction of acquiring the basic shape, finding the correlation between the work shape of the image information newly acquired after the movement and the basic shape; and repeating the moving of the camera and the image acquisition till the correlation reaches a predetermined threshold. When a plurality of works are recognized in the image information, if the correlation after the movement is lower than the correlation before the movement, the work is changed to a work having a next-highest correlation.

Description

  The present invention relates to a workpiece position recognition technique that is effective when, for example, a workpiece having an irregular position or posture is gripped by a transfer robot.

  Conventionally, for example, when a workpiece such as a cylinder block or a transmission case is gripped and transferred to a predetermined location, a shape when the workpiece is viewed from a predetermined direction is stored, and an image captured by a camera is stored. A technique is known that determines the position and posture of a workpiece by pattern matching. At this time, even if multiple workpieces are in the camera's field of view, holes formed in the workpiece and multiple locations are known. Technology that can determine the position and orientation of the workpiece to be transferred by judging the characteristic part such as the interval of the characteristic part with the characteristic and the inclination of the straight line connecting the centroids of each shape part (for example, patent document) 1) is known.

  Further, when the position of the workpiece is recognized by the camera, the positional deviation from the reference position is detected by the number of pixels in the X direction and the Y direction based on the image of the first camera position, and this is converted into a distance. Based on the image, the camera is moved, and the positional deviation from the reference position of the workpiece is measured based on the image at that position, and errors such as spherical aberration are corrected based on the sum of the first error and the second error. Such a technique (see, for example, Patent Document 2) is also known.

JP-A-7-112383 Japanese Patent No. 3610771

However, as in the technique of Patent Document 1, the shape when the workpiece is viewed from a predetermined direction is stored, and the position and orientation are recognized by pattern matching with the shape of the workpiece imaged by the camera. Since the shape of the workpiece imaged by the camera differs depending on the direction in which the workpiece is viewed, it may not match the stored workpiece shape. For example, when the shape of the work viewed from directly above is stored, if the shape of the work viewed from an oblique direction is imaged by the camera, the shape of the side surface of the work can also be seen, and the two do not match. In such a case, even if pattern matching is performed, the correlation is low, and the position / posture of the workpiece cannot be accurately known. In some cases, the position and orientation of a workpiece different from the desired workpiece may be erroneously recognized.
In addition, in the technique such as Patent Document 2, when a plurality of workpieces are included in the field of view of the camera, it has not been possible to determine which workpiece should be targeted.

  In view of the above problems, an object of the present invention is to make it possible to recognize the position and posture of a workpiece with high accuracy by pattern matching using a single camera. The purpose is to make it possible to accurately recognize the target work even when entering the field of view.

  In order to achieve the above object, the present invention recognizes a workpiece shape from image information acquired by a camera, and calculates a correlation between the workpiece shape and a basic shape of a workpiece model viewed from a predetermined direction stored in advance. In the method for recognizing a work position in which the work is identified by recognizing the position of the work, the position of the work is recognized in the first process and the first process based on the image information acquired by the camera. A second step of moving the camera in the direction in which the basic shape is obtained based on the position of the workpiece, and after the second step, the position of the workpiece is recognized again based on the image information obtained from the camera. A third step is provided, and the second step and the third step are repeated until a predetermined condition is reached.

  As the predetermined condition, for example, the correlation between the work shape obtained in the third step and the basic shape reaches a predetermined threshold, or the position of the work obtained in the third step and the camera It is mentioned that the difference from the position of reaches a predetermined threshold value.

Here, the basic shape of the work model viewed from a predetermined direction is stored in advance, the work is identified by calculating the correlation between the basic shape and the work shape of the image information acquired by the camera, and the position And by moving the camera in the direction in which the basic shape is obtained, the correlation when pattern matching is performed can be improved, and the position / posture of the workpiece can be recognized with higher accuracy. That is, if the camera position that is stored in advance when viewed from a predetermined direction is set as the reference position of the camera, the amount of movement of the camera around the reference position is the amount of movement of the work even when the position of the work is out of order. Since this corresponds to a position error, the error amount of the workpiece can be indirectly known through the amount of movement of the camera.
For this reason, it is possible to grip a more accurate position of the workpiece by detecting the amount of movement of the camera and reflecting it on the gripping mechanism of the transfer robot.
At this time, the position error of the workpiece can be further reduced by increasing the correlation between the workpiece shape based on the image information and the basic shape to a predetermined threshold.
A specific example of the correlation threshold between the workpiece shape acquired by the camera and the basic shape is, for example, an area of the workpiece shape, a circumference, or the like, and can be arbitrarily set.

  In addition, as an apparatus at this time, the work is identified by calculating the correlation between the shape of the work recognized by the camera and the basic shape of the work model viewed from a predetermined direction in advance, and its position is recognized. And a moving mechanism for moving the camera in a direction in which the basic shape of the work model is obtained based on image information recognized by the camera. The image processing mechanism includes a moving mechanism. It is preferable to have a determination function for determining whether or not a predetermined condition calculated based on image information acquired by a camera moved in a direction in which the basic shape of the work model is obtained reaches a predetermined threshold. .

  Examples of the predetermined condition include a correlation calculated based on a workpiece shape, or a difference between a workpiece position and a camera position.

In the present invention, when a plurality of workpieces are recognized in the image information acquired by the camera in the first step, the workpiece having the highest correlation is selected and the position is recognized.
As described above, when a plurality of workpieces are recognized in the image information, if the workpiece having the highest correlation is selected, the processing can be performed more quickly and the processing efficiency can be improved. .
Further, as the apparatus at this time, it is preferable that the image processing mechanism has a selection function capable of selecting a workpiece having the highest correlation when a plurality of workpieces are recognized by the camera.

  In the present invention, the correlation in the first step is compared with the correlation in the third step, and as a result, the correlation in the third step is higher than the correlation in the first step. In some cases, the second step and the third step are repeated. If the correlation in the third step is lower than the correlation in the first step, the workpiece is changed to the next highest correlation. did.

  In the present invention, the difference between the workpiece position and the camera position in the first step and the difference between the workpiece position and the camera position in the third step are compared. As a result, in the third step If the difference between the workpiece position and the camera position is smaller than the difference between the workpiece position and the camera position in the first step, the second step and the third step are repeated, and the workpiece in the third step is repeated. When the difference between the position and the camera position is larger than the difference between the work position and the camera position in the first step, the work is changed to the work having the next highest correlation.

  In other words, if the camera is moved in the direction in which the basic shape can be obtained, if it is a regular workpiece, the correlation should be increased or the difference between the position of the workpiece and the camera position should be reduced. A large difference between the position of the workpiece and the camera position means that the workpiece is different.

Therefore, at that time, the loss time can be reduced by changing to the work having the next highest correlation.
In addition, as an apparatus at this time, the image processing mechanism is provided with a comparison function for comparing the correlation between before and after movement as the camera moves. When the correlation is lower than the correlation before the movement, it is preferable to have a workpiece changing function for switching to a workpiece having the next highest correlation.

  When recognizing the position of the workpiece based on the image information acquired by the camera, the basic shape of the workpiece model viewed from a predetermined direction is stored in advance, and this basic shape and the workpiece shape based on the screen information acquired by the camera The position of the workpiece is identified by calculating the correlation between the workpiece and the position of the workpiece. The position of the workpiece is recognized based on the image information acquired by the camera in the first step. In the second step, the camera is moved in the direction in which the basic shape is obtained, and then in the third step, the position of the workpiece is recognized based on the image information obtained from the camera. Until the correlation between the workpiece shape and the basic shape obtained in the step reaches a predetermined threshold, or until the difference between the workpiece position and the camera position obtained in the third step reaches a predetermined threshold, The second step and By repeating the third step, it is possible to detect accurate position and orientation relative to the desired workpiece.

  At this time, when a plurality of workpieces are recognized in the image information, the workpiece having the highest correlation is selected, or the correlation in the first step and the correlation in the third step or the position of the workpiece and the camera are selected. By comparing the difference with the position and changing to a work having the next highest correlation in a predetermined case, it can be processed more quickly and efficiently.

It is explanatory drawing for demonstrating the basic flow of the recognition method of the workpiece | work position which concerns on 1st invention. It is a basic flow figure of the 1st invention. It is explanatory drawing which shows the outline | summary of this apparatus structure, and the positional relationship of a workpiece | work and a camera. It is explanatory drawing which shows the relationship between the movement of a camera, and the threshold value of correlation. It is explanatory drawing when a some workpiece | work is recognized by image information. It is a flowchart in the case of FIG. It is explanatory drawing for demonstrating the basic flow of the recognition method of the workpiece | work position which concerns on 2nd invention. It is a graph which shows the relationship between the difference between the center position of the workpiece | work calculated | required by calculation, and the center position of a true workpiece | work, and the difference of the camera position and the center position of the workpiece | work calculated | required by calculation. Basic flow diagram of the second invention Application flow diagram of the second invention

Embodiments of the first invention will be described with reference to the accompanying drawings.
The work position recognition method according to the first aspect of the present invention recognizes an irregular position or posture of a work with a single camera even if there is no characteristic part in the work shape or the like. In addition, even when multiple workpieces fall within the field of view of the camera, the workpiece to be processed can be accurately recognized. The workpiece shape is acquired by the camera, and the workpiece shape is viewed from a predetermined direction stored in advance. The workpiece is identified by determining the correlation with the basic shape of the workpiece model, and the correlation is increased to a predetermined threshold by moving the camera in the direction in which the basic shape is obtained. The position is calculated.

  That is, as shown in FIG. 3A, a camera C is disposed above the work W, and a moving mechanism 2 for making the camera C movable in the X and Y directions is connected to the camera C. The movement mechanism 2 and the camera C are instructed to move the camera C, specify the workpiece based on the image acquired by the camera C, and recognize the position of the workpiece. A possible image processing mechanism 1 is connected.

  The image processing mechanism 1 includes at least a work shape acquired by the camera C when the camera C is positioned directly above the center of the work W set at the reference position (reference camera position). In the case of the basic shape, a work function obtained from actual image information is compared with the basic shape, and a determination function that can determine whether or not the correlation between the two has reached a predetermined threshold, or the camera C When a plurality of workpieces are included in the image acquired in step S1, the selection function that can select the workpiece C having the highest correlation with the basic shape, and the correlation before the movement with the movement of the camera C Comparison function that can compare with the correlation after movement, and work change function that switches to the next highest correlation when the correlation after movement is lower than the correlation before movement as a result of this comparison But It has been given.

  That is, as shown in FIG. 3A, when the workpiece W exists at a position shifted by α from the position immediately below the camera C, the workpiece shape viewed from the camera C at this time is such that the camera C is directly above the center of the workpiece W. It is different from the basic shape when positioned. In general, the larger the position error α, the lower the correlation between the workpiece shape and the basic shape acquired by the camera C at that time, and as shown in FIG. Also goes down.

  Conversely, as the position error α decreases, the correlation between the workpiece shape and the basic shape increases, and the evaluation value related to the correlation also increases.

  Therefore, in the first invention, the shape of the workpiece model viewed from a predetermined direction (in the present embodiment, directly above the workpiece W) is stored as a basic shape, and when handling an actual workpiece, FIG. As shown in FIG. 2, the workpiece C is recognized by the camera C, the correlation between the workpiece shape and the basic shape is calculated, the workpiece is identified and its position is recognized, and the camera C can obtain the basic shape. The position error α is reduced by moving in the direction (in FIG. 2, “the camera is moved closer to the model center”), the work shape is recognized again by the camera C, and as shown in FIG. The movement of the camera C and the recognition of the workpiece shape by the camera are repeated until the correlation between the workpiece shape and the basic shape reaches a predetermined threshold value.

  When it is detected by the determination function of the image processing mechanism 1 that the correlation has reached a predetermined threshold, the movement amount of the camera C at that time (the displacement amount from the reference camera position) is changed from the reference position of the workpiece W. By notifying the gripping mechanism or the like of the transfer robot of the movement amount of the camera, the gripping mechanism or the like can accurately recognize the position of the workpiece W. The state until the evaluation value related to the correlation reaches a predetermined threshold is also shown in FIG.

  By the way, the above flow is a basic flow when only one workpiece W is acquired by the camera C. Next, a flow when there are a plurality of workpieces W in the image information acquired by the camera C is shown. Will be described with reference to FIGS.

  As shown in FIG. 6, when a plurality of workpieces are recognized at the time of image acquisition, the workpiece having the highest correlation is selected by the selection function of the image processing mechanism 1 as a result of pattern matching, and based on the image of the workpiece. The image is processed, and the correlation is evaluated for this work. That is, the camera is moved toward the workpiece, and the correlation before movement and the correlation after movement are evaluated by the comparison function of the image processing mechanism.

  And, in the comparison of the evaluation value before movement and the evaluation value after movement, when the correlation after movement is higher than the correlation before movement, the same procedure as the above example is performed, but before the movement, When the evaluation value after movement is lower than the evaluation value, the work change function of the image processing mechanism 1 switches to the work having the next highest correlation. Then, it is determined whether or not the evaluation value is higher than the threshold value for the changed work in the same procedure as in the above example. If the evaluation value is higher than the threshold value as shown in FIG.

  FIG. 7 is an explanatory diagram for explaining the basic flow of the work position recognition method according to the second aspect of the invention. The camera C is movable in the X and Y directions by the moving mechanism 2, and the image processing mechanism 1. Is connected.

  The image processing mechanism 1 is a pattern matching mechanism that calculates the basic shape of a workpiece model and the position and orientation of an actual workpiece (a workpiece shape acquired by the camera C), and when an image acquired by the camera C includes a plurality of workpieces. , A selection function that can select the workpiece C having the highest correlation with the basic shape, and with the movement of the camera C, the difference between the workpiece position before the movement and the camera position, and the difference between the workpiece position after the movement and the camera position. A comparison function that can compare the difference between the workpiece position and the camera position, a determination function that can determine whether or not the difference between the workpiece position and the camera position has reached a predetermined threshold, and the difference between the workpiece position and the camera position after movement When the difference between the previous work position and the camera position is larger, a work change function for switching to the work having the next highest correlation is provided.

  As shown in FIGS. 7 and 8, in the second invention, first, the position and orientation of the workpiece are calculated by pattern matching in the first step. Next, the camera is moved toward the work center position obtained by the calculation in the second step. Then, pattern matching is performed again in the third step. This pattern matching reduces the difference (α ') between the current camera position (center position) and the calculated work center position, and as a result, the position between the calculated work center position and the true work center position. The error is also reduced.

  If the second step and the third step can be repeated until the position error becomes smaller than the threshold value, the workpiece position recognition is completed as shown in FIG.

Further, when the acquired image includes a plurality of workpieces, as shown in FIG. 10, as a result of performing pattern matching again in the third step, the current position (center position) of the camera and the calculation are obtained. If the difference (α ') from the workpiece center position is greater than the difference (α') before movement, it is determined that the wrong workpiece has been selected, and the correlation between the workpiece (object) and the basic shape Change to the next higher workpiece.
If the difference before movement (α ′) is smaller than the difference before movement (α ′), the second step and the third step are repeated until the position error becomes smaller than the threshold value as described above.

  Even if multiple cameras are found by the method described above, the desired workpiece can be reached more quickly by bringing the camera closer to the workpiece with the highest correlation from the first stage. Even if the selection of the workpiece is incorrect, since it is possible to determine erroneous recognition while the camera is approaching, it is possible to finally reach the desired workpiece at an early stage.

  In addition, this invention is not limited to the above embodiments. What has substantially the same configuration as the matters described in the claims of the present invention and exhibits the same operational effects belongs to the technical scope of the present invention.

  Since the work position can be recognized more quickly and accurately using only one camera, it is effective when applied to, for example, work recognition for a gripping robot that grips a work.

DESCRIPTION OF SYMBOLS 1 ... Image processing mechanism, 2 ... Movement mechanism, C ... Camera, W ... Workpiece | work.

Claims (12)

The workpiece shape is recognized from the image information acquired by the camera, the workpiece is identified by calculating the correlation between the workpiece shape and the basic shape of the workpiece model viewed from a predetermined direction stored in advance, and its position is recognized. A method for recognizing a workpiece position,
A first step of recognizing a position of a work based on image information acquired by the camera;
A second step of moving the camera in a direction in which the basic shape is obtained based on the position of the workpiece recognized in the first step;
After the second step, a third step of recognizing the position of the workpiece based on the image information obtained from the camera again is provided.
A work position recognition method characterized by repeating the second step and the third step until a predetermined condition is reached.   The workpiece position recognition method according to claim 1, wherein the predetermined condition is that a correlation between the workpiece shape obtained in the third step and the basic shape reaches a predetermined threshold value. A workpiece position recognition method characterized by   2. The work position recognition method according to claim 1, wherein the predetermined condition is that a difference between the work position and the camera position obtained in the third step reaches a predetermined threshold value. A workpiece position recognition method characterized by The workpiece position recognition method according to any one of claims 1 to 3,
In the first step, when a plurality of workpieces are recognized in image information acquired by a camera, the workpiece position recognition method is characterized in that the workpiece having the highest correlation is selected and the position is recognized. A method for recognizing a workpiece position according to claim 2, wherein:
The correlation in the first step and the correlation in the third step are compared. As a result, if the correlation in the third step is higher than the correlation in the first step, Step 2 and the third step are repeated, and when the correlation in the third step is lower than the correlation in the first step, the workpiece is changed to the workpiece having the next highest correlation. Recognition method.   5. The workpiece position recognition method according to claim 3, wherein a difference between a workpiece position and a camera position in the first step and a workpiece position and a camera in the third step are determined. If the difference between the position of the workpiece and the camera position in the third step is smaller than the difference between the position of the workpiece and the camera position in the first step, the second difference is determined. If the difference between the work position and the camera position in the third process is larger than the difference between the work position and the camera position in the first process, the correlation is next high. A method for recognizing a workpiece position, wherein the workpiece position is changed to a workpiece. A workpiece position recognition apparatus capable of recognizing a workpiece position based on image information acquired by a camera,
A camera capable of acquiring image information;
An image processing mechanism capable of identifying the workpiece by calculating the correlation between the shape of the workpiece recognized by the camera and the basic shape of the workpiece model viewed in advance from a predetermined direction, and recognizing the position;
A moving mechanism for moving the camera in a direction in which the basic shape of the work model is obtained based on image information recognized by the camera;
The image processing mechanism determines whether or not a predetermined condition calculated based on image information acquired by a camera moved in a direction in which a basic shape of the work model is obtained by the moving mechanism has reached a predetermined threshold value. A workpiece position recognizing device characterized by having a determination function.   8. The workpiece position recognition apparatus according to claim 7, wherein the predetermined condition is a correlation calculated based on a workpiece shape.   8. The workpiece position recognition apparatus according to claim 7, wherein the predetermined condition is a difference between a workpiece position and a camera position.   The workpiece position recognition device according to any one of claims 7 to 9, wherein the image processing mechanism selects a workpiece having the highest correlation when a plurality of workpieces are recognized by the camera. A workpiece position recognizing device having a selection function capable of being selected.   11. The workpiece position recognition apparatus according to claim 8, wherein the image processing mechanism compares the correlation before and after the movement with the movement of the camera. If the correlation after movement is lower than the correlation before movement as a result of comparison by this comparison function, the workpiece has a workpiece change function for switching to the workpiece having the next highest correlation. Position recognition device.   11. The workpiece position recognition apparatus according to claim 9, wherein the image processing mechanism compares a difference between the position before and after the movement with the movement of the camera. If the difference in the position after movement is larger than the difference in position before movement as a result of comparison by this comparison function, it has a work change function to switch to the next highly correlated work Work position recognition device.

Images