JPH11272845A - Image recognition device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To securely recognize a work in photographing and recognizing the work even when the brightness of an operation position where the work is set varies. SOLUTION: In order to recognize the shape and position of a work through a camera 1, a learning mode is entered, and while a master work is photographed, the aperture of an iris diaphragm 2 provided to the camera 1 is varied to obtain the optimum aperture corresponding to illuminance; while the aperture is set at the optimum value, a parameter value is varied to obtain and store the optimum parameter value when the feature quantity of the master work is closest to a target value in a memory 5. Then a recognition mode is entered, and while the iris diaphragm 2 is adjusted to the optimum aperture stored in the memory 5 corresponding to the illuminance, the work is recognized according to the optimum parameter value.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image recognition apparatus for inputting an image of a work set at a predetermined position by a photographing means.

[0002]

2. Description of the Related Art In a facility using a visual device, it is necessary to capture a stable image even when the brightness around the facility changes. The brightness change was suppressed. In addition, as a light source for recognition, illumination with a small change in illuminance and good stability has been used.

[0003]

However, in the case of work by a robot device using a visual device, particularly a mobile robot device, if a light-shielding cover is provided, many operations such as putting an arm through a work window when the mobile robot device operates are required. There is a problem that the operation efficiency is deteriorated due to the limitation of the operation, the cost of the light shielding cover is high, and a light source with good stability is expensive.

In other words, when a work is illuminated by a stable light source in a light-shielded state by a light-shielding cover, the brightness of the work and the background is stabilized as shown in FIG. Although the recognition accuracy and reliability of the position and shape of the workpiece by image processing were secured,
There was a problem that the shading cost and the workability of the robot were poor.

On the other hand, in general illumination without light shielding, as shown in FIG. 9, the brightness of the work and the background not only changes due to the fluctuation of the brightness of the illumination light or disturbance light, but also the illuminance becomes extremely high. When it becomes bright, the work and the background exceed the upper limit (maximum luminance) of the dynamic range of the camera, and the brightness becomes the same, which causes a problem that the image cannot be recognized.

[0006] The present invention has been made in view of the above circumstances, and its object is to reduce the cost at a low cost even when the brightness of the work position where the work is set varies in a configuration in which the work is photographed and recognized. An object of the present invention is to provide an image recognition device capable of reliably recognizing a work.

[0007]

According to the first aspect of the present invention, the brightness of the work position where the work is set is mainly due to the illumination, but the brightness of the illumination light fluctuates or the sunlight When the disturbance light is inserted, the brightness may fluctuate, and the recognition of the work photographed by the photographing unit may become uncertain.

Therefore, the learning mode is executed before the photographing means recognizes the photographed image of the work. In this learning mode, the aperture value learning means obtains the difference (contrast) between the area and the workpiece photographed by the photographing means while sequentially changing the aperture value of the diaphragm means provided in the photographing means.

Here, the condition under which the workpiece can be reliably recognized is that the difference between the area of the workpiece and the background is desirably large. Therefore, the aperture value of the aperture means and the illuminance measured by the illuminance measuring means satisfying such conditions are satisfied. Find the corresponding relationship with
Then, the aperture value storage means stores the correspondence between the illuminance obtained by the aperture value learning means and the optimal aperture value.

On the other hand, the aperture value adjusting means adjusts the aperture means at the time of recognizing the work so that the optimal aperture value stored in the aperture value storage means is obtained in accordance with the illuminance measured by the illuminance measuring means. Thus, the difference between the region between the work photographed by the photographing means and the background is maximized, so that the work can be reliably recognized.

According to the second and third aspects of the present invention, the optimal feature amount when the photographing means optimally photographs the work is stored in the optimal characteristic amount storing means.

Here, since the optimal parameter value for extracting the characteristic amount varies depending on the illuminance, when the aperture value learning means obtains the optimal aperture value by the aperture means by executing the learning mode as described above, the parameter The value learning means changes the parameter values in order, and determines the optimal parameter value and the illuminance measured by the illuminance measuring means when the feature amount of the work photographed by the photographing means is closest to the optimal feature amount stored in the feature amount storing means. Find the corresponding relationship with The parameter value storage means stores the relationship between the illuminance obtained by the parameter value learning means and the optimum parameter value.

On the other hand, the parameter setting means stores the optimum parameter value stored in the parameter value storage means corresponding to the illuminance measured by the illuminance measuring means in a state where the iris value adjusting means adjusts the iris means when the workpiece is recognized. Set as a parameter. Thereby, the feature amount of the work can be appropriately recognized.

According to the fourth aspect of the present invention, the aperture value learning means executes the learning mode for a predetermined period in which the illuminance change of the work position falls within a predetermined range. Can be reliably obtained.

According to the fifth aspect of the present invention, the parameter value learning means executes the learning mode for a predetermined period in which the illuminance variation of the work position falls within a predetermined range. Can be reliably obtained.

According to the sixth aspect of the invention, when the work to be photographed by the photographing means is damaged or the paint is uneven during the learning mode, the adjustment of the diaphragm value to the diaphragm means based on the illuminance is inappropriate. Or the parameter setting becomes inappropriate, and the recognition of the work may become uncertain.

Therefore, by using an ideal reference work as a work to be photographed by the photographing means in the learning mode, it is possible to appropriately adjust the aperture value with respect to the diaphragm means and to appropriately set parameters. Work can be reliably recognized.

[0018]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below with reference to FIGS. FIG. 1 is a functional diagram of the present invention. In FIG. 1, a camera 1 (corresponding to photographing means) is provided with an aperture adjustment device 2 (corresponding to aperture means), and a visual device 3 (aperture value learning means, aperture value adjustment means, parameter value learning means, (Corresponding to parameter setting means) can adjust the aperture value of the aperture adjustment device 2. The visual device 3 has a learning function. In a learning mode, an optimal aperture value and an optimal parameter value are obtained as described later based on the illuminance captured from the illuminometer 4 (corresponding to the illuminance measuring means). 5 (corresponding to the aperture value storage means, the optimum feature amount storage means, and the parameter value storage means), and the illuminometer 4 in the recognition mode.
The optimal parameter value and the optimal parameter value stored in the memory 5 are extracted corresponding to the illuminance captured from the camera 1, and the optimal parameter value is characterized in a state where the aperture control is performed on the camera 1 so that the optimal aperture value is obtained. It is set as a parameter for quantity extraction.

Next, a specific example in which the present invention is applied to a mobile robot device will be described. In FIG. 2 schematically showing the overall configuration, a hand unit 8 of a robot device 7 mounted on an automatic guided vehicle 6 is provided with a camera 1 for photographing a work 10 set on a work table 9. The position and orientation of the camera 1 can be arbitrarily adjusted by controlling the robot device 7. In this case, the camera 1 is provided with an aperture adjustment device 2 (see FIG. 1), and the aperture value is adjusted by the visual device 3. The automatic guided vehicle 6 is provided with an illuminometer 4 for illuminating illumination light (including disturbance light) on the work 10 set at the work position.
Is to be detected.

The visual device 3 adjusts the aperture value of the camera 1 based on the illuminance of the illuminometer 4 so that the optimum photographing condition is obtained, and recognizes the position and the shape of the work 10 based on the image processing.

FIG. 3 shows the operation of the learning mode of the visual device 3. In FIG. 3, first, a master work 11 (corresponding to a reference work, see FIG. 4) is set on the work table 9, a camera position is adjusted so that the master work 11 can be photographed, and lighting is set (S101). . The master work 11 is an ideal work processed into a shape, a surface processing state, and a coating state according to design dimensions.

Subsequently, a work area and a background area are designated by teaching (S102). That is, as shown in FIG. 4, one master work 11 is photographed so as to be displayed in an image, and a work area and a background area are indicated by the visual device 3 on the image.

Subsequently, after inputting an image signal from the camera 1 (S103), a target value of a parameter and a characteristic amount in the current image is specified (S104). In this case, the parameters are a binary threshold, selection of a template used for recognition, and the like. Further, the feature amount is a position, an area, a shape, and the like of the work 10 to be finally detected. Then, the target values of these parameters and feature values are taught with the current brightness, and based on them, the optimal aperture value and optimal parameter value when the subsequent brightness changes in the following learning mode are learned. Will be decided.

First, the optimum aperture value is determined by learning based on the current brightness. That is, the illuminance is measured by the illuminometer 4 (S105), and it is determined whether the illuminance is within the already measured illuminance range (S106). In other words, the illuminance range to be learned is set in advance by dividing it into a plurality for each predetermined illuminance range, and if the current brightness belongs to the already learned illuminance range, there is no need to learn again. Then, it is monitored that the illuminance changes from the current illuminance range by shifting to step S105.

If the current brightness varies and the illuminance deviates from the already learned illuminance (S106: NO),
Since learning is required, an initial aperture value is set (S107). The initial aperture value is, for example, an aperture value with the smallest amount of incident light (for the second and subsequent times, the previous aperture value of the illuminance value may be referred to).

Subsequently, the difference (contrast) between the area of the workpiece 10 and the background is measured in the state of the initial aperture value (S108), and the aperture value is changed so as to increase by one step (S118). By shifting to S108, the difference between the areas of the master work 11 and the background is measured.
When the necessary aperture value is obtained by sequentially changing the aperture value in this way (S109: YES),
An aperture value that maximizes the difference between the areas is searched for and determined (S1).
10).

FIG. 5 shows the relationship between the magnitude of the illuminance and the optimum aperture value at which the difference between the regions is maximized. In this case, when the aperture value is one step smaller than the optimal aperture value, the difference between the areas becomes smaller because the overall image luminance becomes smaller. Conversely, if the aperture value is one step larger than the optimal aperture value, the overall image brightness becomes too large, and the master work 11
Is larger than the dynamic range of the camera 1, and the difference between the regions becomes smaller. Therefore, it is understood that there is only one optimal aperture value corresponding to the predetermined illuminance.

When the master work 11 or the background is too bright and the image luminance exceeds the upper limit of the dynamic range of the camera 1 for both the master work 11 and the background, the contrast becomes small. The maximum value of the range. When the illumination is too dark and the image brightness of the master work 11 or the background is lower than the minimum level, the contrast becomes small, and that point becomes the minimum value of the illuminance range.

When the optimum aperture value is obtained as described above, the corresponding relationship between the illuminance and the aperture value at that time is stored in the memory 5 (S111), and the minimum aperture value and the corresponding illuminance value are stored. The value and the maximum value are updated (S11
2). As a result, the upper and lower limits of the illuminance corresponding to the aperture value shown in FIG. 5 can be finally determined by the change in the illuminance. With the above operation, the illuminance and the optimal aperture value when the difference between the regions is the maximum at the current brightness are stored in the memory 5 in association with each other.

When the optimum aperture value corresponding to the current illuminance is obtained as described above, the optimum parameter value is subsequently determined by learning. In this case, the aperture value of the aperture adjustment device 2 is set to the optimal aperture value for the current illuminance.

First, initial parameter values are set (S11).
3) Master work 1 based on the initial parameter values
The first feature amount is measured (S114). This parameter is used for the master work 11 by binarizing the image signal.
And the background. In addition, the feature amount refers to the position, area, shape, and the like of the master work 11, and varies depending on the illuminance of illumination and parameters.

In this case, when the parameter value is appropriate, the position or area of the master work 11 comes closest to the target value, and when the parameter value is inappropriate, the position or area of the master work 11 deviates from the target value. .

Then, the parameter values are sequentially updated (S1).
19), when data acquisition is completed based on all parameter values (S115: YES), master work 1
The optimal parameter value that minimizes the target value and the current value of the feature value corresponding to 1 is searched and determined (S116), and the illuminance and the optimal parameter value at that time are stored (S11).
7).

With the above operation, at the current illuminance,
An optimum parameter value when the characteristic amount of the master work 11 is closest to the target value can be obtained. When the optimal aperture value and the optimal parameter value at the current illuminance are obtained in this manner, the process proceeds to step 105 to repeat the learning mode described above.

Here, considering the variation in the illuminance of the environment in which the work 10 is set, in an environment where the outdoor light hardly enters, the variation of the illuminance is only the variation of the brightness of the illumination, and the variation width is very small. not big. Further, in an environment where outdoor light enters, the outdoor brightness varies over time and varies depending on whether the weather is clear or cloudy, so that the fluctuation range is large.

Therefore, in an environment where almost no outdoor light enters, the learning mode is executed in response to a change in illuminance in one day, so that an optimum aperture corresponding to a change in illuminance at the time of actual recognition of the work 10 is obtained. Values and optimal parameter values can be determined. Further, in an environment where outdoor light enters, the learning mode is executed, for example, for about one week, so that an optimum aperture value and an optimum parameter value corresponding to a change in illuminance at the time of actual recognition of the work 10 can be obtained. . In the state where the learning mode as above is completed,
As shown in FIG. 6, the memory 5 stores the optimum aperture value and the optimum parameter value corresponding to the illuminance.

Although the learning mode described above needs to be executed only once, it is desirable to execute the learning mode every season since the amount of solar radiation in the room differs between summer and winter.

FIG. 7 shows the operation of the visual device 3 in the recognition mode. In FIG. 7, the visual device 3 includes an illuminometer 4
When the current illuminance is taken in from (S201), the optimal aperture value and the optimal parameter value stored in the memory 5 corresponding to the illuminance are read (S202), and the read optimal aperture value is set. After adjusting the aperture adjustment device 2 (S203), the optimal parameter value is set as a parameter (S204).

When the setting of the parameters is completed as described above, the work 10 is photographed by the camera 1 (S20).
5) Since the aperture value of the camera 1 is set to the optimal aperture value, the difference between the area of the workpiece 10 and the background is maximum. Also, since the parameters for binarizing the image signal from the camera 1 are set to optimal parameter values,
The work 10 and the background can be reliably identified based on the image signal from the camera 1 (S206). And
By controlling the robot device 7 based on the recognition result, the workpiece 10 can be reliably chucked and transported to a predetermined position.

According to the present embodiment, by executing the learning mode, the optimal aperture value at which the difference between the area of the workpiece 10 and the background becomes maximum corresponding to the illuminance is determined for a predetermined period during which the illuminance variation falls within a predetermined range. And in the memory 5 in the recognition mode, corresponding to the current illuminance.
Since the aperture adjustment device 2 of the camera 1 is controlled so as to have the optimal aperture value stored in the camera 10, the difference between the area of the workpiece 10 and the background can be increased, and the image signal from the camera 1 The work 10 can be reliably recognized as compared with a configuration in which the work 10 is simply recognized based on the base. In this case, since it can be implemented by changing software, it can be implemented at low cost as compared with a configuration in which a light shielding cover is provided.

In such a learning mode, the optimum parameter value for binarizing the image signal from the camera 1 is determined over a predetermined period in which the illuminance variation falls within a predetermined range. Since the optimum parameter value stored in the memory 5 is set as a parameter in correspondence with the above, the work 10 and the background are identified by binarizing the image signal from the camera 1 based on the optimum parameter value. Thus, the work 10 can be more reliably recognized.

Further, since the work to be recognized in the learning mode is the master work 11 serving as a reference,
By recognizing the master work 11 in the learning mode, the optimal aperture value and the optimal parameter value can be appropriately set.

The present invention is not limited to the above embodiment, but can be modified or expanded as follows. As the aperture adjustment device 2, a device capable of finely adjusting the aperture value in an analog manner may be employed. In this case, the aperture value corresponding to the illuminance can be set appropriately, though the learning mode period becomes longer.

[Brief description of the drawings]

FIG. 1 is an overall functional diagram according to an embodiment of the present invention;

FIG. 2 is a schematic diagram of a robot device and an image recognition device.

FIG. 3 is a flowchart showing a learning mode operation of the visual device;

FIG. 4 is a diagram showing recognition points of a work and a background during teaching.

FIG. 5 is a diagram illustrating a relationship between illuminance and image luminance for each aperture value.

FIG. 6 is a diagram showing an optimum aperture value and an optimum parameter value corresponding to illuminance.

FIG. 7 is a flowchart showing a recognition mode operation of the visual device;

FIG. 8 is a diagram showing image luminance corresponding to a work and a background when the brightness of illumination light is constant in a conventional example.

FIG. 9 is a diagram illustrating image luminance corresponding to a workpiece and a background when the brightness of illumination light varies.

[Explanation of symbols]

1 is a camera (imaging means), 2 is an aperture adjusting device (aperture means), 3 is a visual device (aperture learning means, aperture value adjusting means, parameter value learning means, parameter setting means), 4
Is an illuminance meter (illuminance measurement means), 5 is a memory (aperture value storage means, optimal feature quantity storage means, parameter value storage means), 7
Denotes a robot device, 10 denotes a work, and 11 denotes a master work (reference work).

Claims (6)

[Claims] 1. An image recognition apparatus for inputting an image of a work set at a work position by a photographing means, an aperture means provided in the photographing means, an illuminance measuring means for measuring illuminance at the work position, and learning. In the mode, the aperture value of the aperture means is changed in order, and the correspondence between the optimal aperture value when the contrast between the workpiece photographed by the photographing means and the background is maximized and the illuminance measured by the illuminance measuring means. Aperture value learning means to be sought; aperture value storage means for storing a correspondence relationship between the illuminance obtained by the aperture value learning means and the optimal aperture value in the learning mode; and illuminance measured by the illuminance measurement means for workpiece recognition. An image recognition apparatus comprising: an aperture value adjusting unit that adjusts the aperture unit so that the aperture value corresponds to the optimal aperture value stored in the aperture value storage unit. 2. An optimum feature amount storing means for storing an optimum feature amount when the photographing means optimally shoots a workpiece; and in the learning mode, the optimum aperture value obtained by the aperture value learning means so as to be an optimum aperture value. The parameter values for extracting the feature amount are sequentially changed in a state where the aperture means is adjusted, and the optimal value when the feature amount of the work photographed by the photographing means is most similar to the optimal feature amount stored in the feature amount storage means. Parameter value learning means for obtaining a correspondence between a parameter value and illuminance measured by the illuminance measurement means; and parameter value storage means for storing a relationship between illuminance obtained by the parameter value learning means and an optimum parameter value in a learning mode. When the workpiece is recognized, the parameter value storage means stores the parameter value storage means corresponding to the illuminance measured by the illuminance measurement means in a state where the aperture value adjustment means adjusts the aperture means. 2. The image recognition apparatus according to claim 1, further comprising parameter setting means for setting the obtained optimum parameter value as a parameter. 3. The image recognition apparatus according to claim 2, wherein said parameter value is a threshold value for binarizing an image signal from said photographing means. 4. The image according to claim 1, wherein the aperture value learning unit executes the learning mode for a predetermined period in which a change in illuminance at the work position falls within a predetermined range. Recognition device. 5. The image according to claim 1, wherein the parameter value learning unit executes the learning mode for a predetermined period during which a change in the illuminance of the work position falls within a predetermined range. Recognition device. 6. The image recognition apparatus according to claim 1, wherein the work to be photographed by the photographing means in the learning mode is an ideal reference work.