JP2021139723A - Identification device, identification method, and program - Google Patents

Abstract

To highly accurately identify the kind of a target in an identification apparatus, an identification method, and a program.SOLUTION: An identification apparatus 10 is configured to identify the kind of a target T. The identification apparatus 10 comprises a measurement device 100, a transfer characteristic calculation unit 212a, an identifier 213a and a determination unit 213. The measurement device 100 includes a vibration application section 121 which applies vibration to the target T, and an acceleration measurement section 131 which measures an acceleration of the target T. The transfer characteristic calculation section 212a calculates transfer characteristics for the acceleration measured by the acceleration measurement section 131 from the vibration applied to the target T by the vibration application section 121. The identifier 213a defines the transfer characteristics which are calculated by the transfer characteristic calculation section 212a for the target T, the kind of which is known, as supervisor input data, defines the kind of the target T as a supervisor output data and has learnt a correlation between the supervisor input data and the supervisor output data. The determination section 213 determines the kind of the target T, the kind of which is unknown, utilizing the identifier 213a while defining transfer characteristic which are calculated by the transfer characteristic calculation section 212a for the target T, the kind of which is unknown, as determination input data.SELECTED DRAWING: Figure 6

Description

The present invention relates to an identification device, an identification method, and a program.

Patent Document 1 discloses a general waste or industrial waste sorting system and a sorting method thereof. In the sorting system and the sorting method, the target object is sorted based on the teacher data accumulated in advance by using the image data of the target object captured by the imaging device as input data. In addition, the sorting accuracy is improved by further accumulating teacher data while performing daily sorting operations.

JP-A-2017-109161

However, in the sorting system and sorting method as in Patent Document 1, since sorting is mainly performed based on image data, it may be difficult to sort objects having similar appearances. For example, when an object has an outer shell and contents, it may be difficult to identify the contents and whether or not there is damage inside the object.

An object of the present invention is to identify the type of an object with high accuracy in an identification device, an identification method, and a program.

The first aspect of the present invention is
An identification device that identifies the type of object
A measuring device having a vibration application unit that applies vibration to the object and an acceleration measurement unit that measures the acceleration of the object.
A transmission characteristic calculation unit that calculates the transmission characteristics from the vibration applied by the vibration application unit to the acceleration measured by the acceleration measurement unit with respect to the object.
The transmission characteristics calculated by the transmission characteristic calculation unit for the object whose type is known are used as teacher input data, the type of the object is used as teacher output data, and the teacher input data and the teacher output data. With a classifier that has learned the correlation with
It is provided with a determination unit for determining the type of the object of unknown type by using the classifier as input data for determination using the transmission characteristic calculated by the transmission characteristic calculation unit for the object of unknown type. , Provide an identification device.

According to this configuration, vibration is applied to the object from the vibration application unit, the acceleration of the object is measured by the acceleration measurement unit, and the transmission characteristic from vibration to acceleration is calculated by the transmission characteristic calculation unit. Since the discriminator has learned the correlation between the types of various objects and the transmission characteristics, when the transmission characteristics for an object of unknown type are given, the target is based on the learned correlation. You can determine the type of object. Therefore, even if the object is difficult to discriminate in appearance, the type of the object can be identified with high accuracy. Here, the type of an object is a broad concept including not only a general type of an article but also a material, an internal state, a degree of damage, etc., and refers to any feature that can be affected by the above-mentioned transmission characteristics. ..

The measuring device includes a robot hand having a first finger member and a second finger member.
The vibration applying portion is attached to the first finger member.
The acceleration measuring unit may be attached to the second finger member.

According to this configuration, the type of object can be identified by touching various objects with a robot hand. In particular, it is possible to identify dangerous objects when touched by humans and in environments where it is difficult for humans to act.

The measuring device includes a first fitting and a second fitting to be worn on a human hand.
The vibration application portion is attached to the first fitting.
The acceleration measuring unit may be attached to the second attachment.

According to this configuration, when a human touches an object and wants to identify its type, it can be operated intuitively. That is, the type of the object can be easily identified by wearing the first attachment and the second attachment on the hand and touching the object to be identified. For example, the mode of the first wearing tool and the second wearing tool may be a ring type worn on a fingertip, a glove type worn on the palm, or the like.

The measuring device includes a pressing unit that presses the object and a force displacement measuring unit that measures the pressing force of the pressing unit and the displacement amount of the object.
An identification device further comprising an elastic modulus calculation unit that calculates the elastic modulus of the object as a relationship between the pressing force measured by the force displacement measuring unit and the displacement amount.
The classifier has already learned the elastic modulus calculated by the elastic modulus calculation unit for the object of known type as the input data for the teacher.
The determination unit may further include the elastic modulus calculated by the elastic modulus calculation unit for the object of unknown type as the input data for determination.

According to this configuration, a pressing force is applied to the object from the pressing portion, the pressing force of the pressing portion and the displacement amount of the object are measured by the force displacement measuring unit, and the pressing force and the displacement amount are determined by the elastic modulus calculation unit. The elastic modulus is calculated as a relationship. Since the discriminator has learned the correlation between the elastic modulus and the type of the object as teacher data, when the elastic modulus for the object of unknown type is given, it is based on the learned correlation. The type of object can be determined with high accuracy.

An imaging unit that images the object,
An identification device further comprising an image feature extraction unit that extracts image features of the object from the image data of the object captured by the imaging unit.
The classifier has further learned the image features extracted by the image feature extraction unit from the object of known type as the input data for the teacher.
The determination unit may further include the image feature extracted by the image feature extraction unit for the object of unknown type as the determination input data.

According to this configuration, the image pickup unit images the object to generate image data, and the image feature extraction unit extracts the image features of the object from the image data. Since the discriminator has learned the correlation between the image feature and the type of the object as teacher data, when the image feature is given to the object of unknown type, the learned correlation is obtained. Based on this, the type of object can be determined with high accuracy.

The vibration applying portion may include a vibrator that expands and contracts due to an electric signal, and a first support portion that comes into contact with the object together with the vibrator around the vibrator.

According to this configuration, when the vibrator expands and contracts and applies vibration to the object, the first support portion can receive a shearing force that may be applied to the vibrator from the object. In other words, it is possible to realize the application of vibration by the vibrator while supporting the object by the first support portion. Therefore, the influence of the shearing direction on the application of vibration in the expansion / contraction direction of the vibrator can be suppressed, and the vibration can be stably applied, so that stable transmission characteristics can be obtained. Further, since it is possible to suppress the application of shearing force to the vibrator, damage to the vibrator can also be suppressed.

The vibration application portion includes a mass body that holds the vibrator and the first support portion, and a first attachment that urges the vibrator and the first support portion toward the object via the mass body. It may be provided with a force unit.

According to this configuration, since the vibrator and the first support portion are held by the mass body, the vibrator and the first support portion can obtain a reaction force due to inertia from the mass body. That is, vibration can be stably applied to the object. Further, since the vibrator and the first support portion are urged toward the object by the first urging portion via the mass body, the vibrator and the first support portion can be surely brought into contact with the object.

The acceleration measuring unit may include an acceleration sensor that measures the acceleration and a second support unit that comes into contact with the object together with the acceleration sensor around the acceleration sensor.

According to this configuration, when vibration is applied to the object by the vibration applying portion, the second support portion can receive the shearing force applied to the acceleration sensor from the object. In other words, it is possible to measure the acceleration by the acceleration sensor while supporting the object by the second support portion. Therefore, the acceleration of the object can be measured stably, and stable transmission characteristics can be obtained. Further, since it is possible to suppress the application of shearing force to the acceleration sensor, damage to the acceleration sensor can also be suppressed.

The acceleration measuring unit may include a second urging unit that urges the acceleration sensor toward the object.

According to this configuration, since the acceleration sensor is urged toward the object by the second urging portion, the acceleration sensor can be surely brought into contact with the object.

A second aspect of the present invention is
An identification method that identifies the type of object
Vibration is applied to the object to measure the acceleration of the object, and the object is measured.
The transmission characteristic to the acceleration measured from the vibration applied to the object was calculated.
For the object of known type, the transmission characteristic is used as teacher input data, the type of the object is used as teacher output data, and the correlation between the teacher input data and the teacher output data has been learned. By using the discriminator of the above, there is provided an identification method including determining the type of the object of unknown type by using the transmission characteristic as input data for determination with respect to the object of unknown type.

A third aspect of the present invention is
A program for identifying the type of object
Vibration is applied to the object and the acceleration of the object is measured.
The transmission characteristic to the acceleration measured from the vibration applied to the object is calculated.
For the object whose type is known, the transmission characteristic is used as the input data for the teacher, the type of the object is used as the output data for the teacher, and the correlation between the input data for the teacher and the output data for the teacher has been learned. Provided is a program that makes a computer function so as to determine the type of the object of unknown type by using the transmission characteristic as input data for determination for the object of unknown type by using the classifier of the above. do.

According to the present invention, the type of an object can be identified with high accuracy in an identification device, an identification method, and a program.

The schematic block diagram of the identification apparatus which concerns on 1st Embodiment of this invention. The partial cross-sectional view which shows the detail of the vibration application part. FIG. 2 is a cross-sectional view taken along the line III-III of FIG. A partial cross-sectional view showing the details of the acceleration measuring unit. FIG. 4 is a cross-sectional view taken along the line VV of FIG. The block diagram of the identification device of FIG. A flowchart showing an identification method. The flowchart which shows the detail of the calculation of the transmission characteristic of FIG. The flowchart which shows the detail of the calculation of the elastic modulus of FIG. The flowchart which shows the detail of the extraction of the image feature of FIG. Perspective view of five objects. The graph which shows the transmission characteristic of the 1st object of FIG. The graph which shows the transmission characteristic of the 2nd object of FIG. The graph which shows the transmission characteristic of the 3rd object of FIG. The graph which shows the transmission characteristic of the 4th object of FIG. The graph which shows the transmission characteristic of the 5th object of FIG. FIG. 2 is a partial cross-sectional view showing a modified example of the vibration application portion of FIG. The schematic block diagram of the identification apparatus which concerns on 2nd Embodiment.

Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.

(First Embodiment)
FIG. 1 shows a schematic configuration diagram of the identification device 10 according to the first embodiment of the present invention.

The identification device 10 identifies the type of the object T by coming into contact with the object T. Here, the type of the object T is a broad concept including not only the general type of the article but also the material, the internal state, the degree of damage, and the like, and any feature that can be influenced by the above-mentioned transmission characteristics. say.

The identification device 10 of the present embodiment includes a measuring device 100 and a controller unit 200.

The measuring device 100 has a robot hand 110. The robot hand 110 has a first finger member 120 and a second finger member 130 for contacting the object T, and a base 140 connecting the first finger member 120 and the second finger member 130. ing.

A vibration application portion 121 is attached to the tip end (opposite end to the base portion 140) of the first finger member 120. Further, the first finger member 120 is provided with three hinge portions 123a to 123c so as to support and connect the three links 122a to 122c. The three links 122a to 122c and the three hinge portions 123a to 123c are driven by an electric motor (not shown) or an actuator by pneumatic pressure or the like. As a result, the vibration application unit 121 can be moved to a desired position.

FIG. 2 is a partial cross-sectional view showing the details of the vibration application unit 121. In FIG. 2, the first support portion 121b, which will be described later, is shown with hatching as a cross section. Further, FIG. 3 is a cross-sectional view taken along the line III-III of FIG.

The vibration application unit 121 has a vibrator 121a that expands and contracts by an electric signal, and a first support portion 121b that comes into contact with the object T together with the vibrator 121a around the vibrator 121a.

The vibrator 121a is, for example, a columnar type, and is a piezoelectric ceramic type that expands and contracts in the axial direction of the columnar shape (direction of arrow A1 in FIG. 2). The amplitude of the piezoelectric ceramic type vibrator 121a of the present embodiment has stable vibration characteristics up to a high band of several tens of kHz. Therefore, it is possible to apply vibration having a wide band frequency corresponding to various types of objects T.

The first support portion 121b is, for example, an annular shape and is arranged around the vibrator 121a. A thin film non-slip member 121c is attached to the tip of the first support portion 121b. The non-slip member 121c is made of a material having a high coefficient of friction, such as rubber or resin, so that it can receive a force in the shearing direction when it comes into contact with the object T.

Further, the vibration applying portion 121 urges the vibrator 121a and the first support portion 121b toward the object T via the mass body 121d holding the vibrator 121a and the first support portion 121b, and the mass body 121d. It has a first urging unit 121e and a first urging unit 121e.

The mass body 121d is arranged on the proximal end side (opposite side of the object T) with respect to the vibrator 121a and the first support portion 121b. The mass body 121d is a member that is considerably heavier than the vibrator 121a so that a reaction force due to inertia can be applied to the vibration of the vibrator 121a.

The first urging portion 121e is, for example, a coil spring. The first urging portion 121e is arranged on the proximal end side with respect to the mass body 121d. The first urging portion 121e is arranged between the connecting tool 121f attached to the hinge portion 123c and the mass body 121d. Specifically, one end of the first urging portion 121e is attached to the mass body 121d, and the other end is attached to the connector 121f.

Further, at the center of the first urging portion 121e, a slide mechanism 121g for mechanically connecting the mass body 121d and the connecting tool 121f is provided. The slide mechanism 121g makes the distance between the mass body 121d and the connector 121f variable, and substantially limits the displacement direction of the first urging portion 121e to the central axis direction (arrow A1 direction in FIG. 2). As a result, the object T can be accurately grasped by the robot hand 110 (see FIG. 1). Alternatively, a plurality of slide mechanisms 121g may be provided around the first urging portion 121e at equal intervals.

With reference to FIG. 1 again, an acceleration measuring unit 131 is attached to the tip of the second finger member 130 (the end opposite to the base 140). Similar to the first finger member 120, the second finger member 130 is provided with three hinge portions 133a to 133c so as to support and connect the three links 132a to 132c. The three links 132a to 132c and the three hinge portions 133a to 133c are driven by an electric motor (not shown) or an actuator by pneumatic pressure or the like. As a result, the acceleration measuring unit 131 can be moved to a desired position.

FIG. 4 is a partial cross-sectional view showing the details of the acceleration measuring unit 131. In FIG. 4, the second support portion 131b, which will be described later, is shown with hatching as a cross section. Further, FIG. 5 is a cross-sectional view taken along the line VV of FIG.

The acceleration measuring unit 131 has an acceleration sensor 131a that measures the acceleration, and a second support unit 131b that comes into contact with the object T together with the acceleration sensor 131a around the acceleration sensor 131a.

The acceleration sensor 131a has, for example, a columnar shape, and measures the acceleration in the columnar axial center direction (direction of arrow A2 in FIG. 4). In the present embodiment, since the object T is sandwiched and gripped by the vibration applying unit 121 of FIG. 2 and the acceleration measuring unit 131 of FIG. 3, the A1 direction of FIG. 2 and the A2 direction of FIG. 4 coincide with each other.

The second support portion 131b is, for example, an annular shape and is arranged around the acceleration sensor 131a. A thin film non-slip member 131c is attached to the tip of the second support portion 131b. The non-slip member 131c is made of a material having a high coefficient of friction, such as rubber or resin, so that it can receive a force in the shearing direction when it comes into contact with the object T. The non-slip member 131c may be the same member as the non-slip member 121c of the vibration application unit 121. The base end (opposite end to the object T) of the second support portion 131b is attached to the connector 131d attached to the hinge portion 133c.

Further, the acceleration measuring unit 131 has a second urging unit 131e that urges the acceleration sensor 131a toward the object T.

The second urging portion 131e is, for example, a sponge-like cushioning material. The second urging portion 131e is attached to the proximal end side with respect to the acceleration sensor 131a. The second urging portion 131e is attached to the connector 131d together with the second support portion 131b.

Further, the measuring device 100 in the present embodiment includes a pressing unit 121 (see FIG. 2) that presses the object T, and a force displacement measuring unit 170 that measures the pressing pressure of the pressing unit 121 and the displacement amount of the object T. Have.

Since the vibration applying portion 121 is used as the pressing portion 121, the same reference numerals are given to both portions. When the vibration application unit 121 is used as the pressing unit 121, it is not necessary to apply vibration to the object T because the displacement is measured as described later.

The force displacement measuring unit 170 includes a displacement measuring unit (not shown) capable of measuring the amount of movement of the pressing unit 121 (that is, the displacement amount of the object T) and a pressing force measuring unit 171 capable of measuring the pressing force from the pressing unit 121. And have. Although the details are not shown, the displacement measuring unit is not particularly limited as long as it can measure the displacement of the pressing unit 121, and a known displacement meter can be used. The pressing force measuring unit 171 can be, for example, a load cell for measuring a load. The pressing force measuring unit 171 is arranged in, for example, the connector 131d, but its position and mode are not particularly limited.

With reference to FIG. 1 again, the measuring device 100 in the present embodiment has an imaging unit 180 for imaging the object T.

The image pickup unit 180 is, for example, a digital camera. The image pickup unit 180 captures a still image of the object T.

FIG. 6 shows a block diagram including the controller unit 200 of the identification device 10 of the present embodiment.

The controller unit 200 has a control unit 210, a storage unit 220, and a display unit 230.

The control unit 210 executes the identification process and controls the entire identification device 10. The control unit 210 includes, for example, a CPU (Central Processing Unit) or an MPU (Micro Processing Unit) that realizes the above functions in cooperation with software. The control unit 210 realizes a predetermined function by reading data and programs stored in the storage unit 220 and performing various arithmetic processes. Further, the program executed by the control unit 210 may be provided by a communication unit or the like that communicates according to a predetermined communication standard, or may be stored in a portable recording medium.

Alternatively, the control unit 210 may be composed of a dedicated electronic circuit designed to realize a predetermined function, or a hardware circuit such as a reconfigurable electronic circuit. Further, the control unit 210 may be composed of various semiconductor integrated circuits. Examples of various semiconductor integrated circuits include, in addition to CPUs and MPUs, microcomputers, DSPs (Digital Signal Processors), FPGAs (Field Programmable Gate Arrays), ASICs (Application Specific Integrated Circuits), and the like.

The control unit 210 has an operation management unit 211, a calculation unit 212, and a determination unit 213 as functional configurations. These are realized by the collaboration between the processor or the like, which is the above-mentioned hardware resource, and the program, which is the software stored in the storage unit 220 or the like.

The operation management unit 211 is a part that manages various operations of the identification device 10. The operation management unit 211 includes a vibration control unit 211a, a drive control unit 211b, and an imaging control unit 211c.

The vibration control unit 211a is a part that controls an electric signal given to the vibrator 121a. Further, the vibration control unit 211a includes a signal generation unit 211a1 and a signal amplification unit 211a2. The signal generation unit 211a1 is a portion that generates an electric signal given to the vibrator 121a. In the present embodiment, the signal generation unit 211a1 generates white noise as an electric signal given to the vibrator 121a. The signal amplification unit 211a2 amplifies the electric signal generated by the signal generation unit 211a1 and transmits it to the vibrator 121a. The vibrator 121a receives the electric signal amplified by the signal amplification unit 211a2, and repeats expansion and contraction so as to apply vibration to the object T.

The drive control unit 211b drives and controls the first finger member 120 and the second finger member 130, respectively. As a result, the robot hand 110 can grip the object T and press the object T.

The image pickup control unit 211c is a part that controls the image pickup unit 180. The image pickup control unit 211c controls the image pickup unit 180 to image the object T, and acquires the image data of the object T.

The calculation unit 212 is a part that executes various calculations. The calculation unit 212 includes a transmission characteristic calculation unit 212a, an elastic modulus calculation unit 212b, and an image feature extraction unit 212c.

The transmission characteristic calculation unit 212a is a portion that calculates the transmission characteristic from the vibration applied by the vibration application unit 121 to the object T to the acceleration measured by the acceleration measurement unit 131.

The elastic modulus calculation unit 212b is a portion that calculates the elastic modulus of the object T as the relationship between the pressing force measured by the force displacement measuring unit 170 and the displacement amount.

The image feature extraction unit 212c is a portion that extracts the image feature of the object T from the image data of the object T imaged by the image pickup unit 180. The image feature indicates an appearance feature such as the size, shape, and color of the object T that can be read from the image data.

The determination unit 213 is a portion that executes determination of the type of the object T. The determination unit 213 has a classifier 213a for identifying the type of the object T. As the classifier 213a, for example, an arbitrary learning device such as a neural network or an SVM (Support Vector Machine) can be used. Of course, the classifier 213a may be one to which so-called deep learning is applied. For example, the configuration of a neural network is the same as that of a known technique. That is, when a predetermined input data is given to the input layer, the neural network propagates (calculates) it to the intermediate layer and the output layer in order, so that the "type of object T" is selected as the output data from the output layer. Output.

In the present embodiment, the classifier 213a has the following features for the object T whose type is known as the input data for the teacher, the type of the object T as the output data for the teacher, and the input data for the teacher and the output data for the teacher. The correlation of is learned. The input data for the teacher includes the transmission characteristics calculated by the transmission characteristic calculation unit 212a, the elastic modulus calculated by the elastic modulus calculation unit 212b, and the image features extracted by the image feature extraction unit 212c for the object T of a known type. including.

The learned classifier 213a identifies the type of the object T by inputting the following input data for determination to the object T whose type is unknown. The input data for determination includes the transmission characteristics calculated by the transmission characteristic calculation unit 212a, the elastic modulus calculated by the elastic modulus calculation unit 212b, and the image features extracted by the image feature extraction unit 212c for the object T of unknown type. including.

The storage unit 220 records data necessary for identification processing such as the program operated by the control unit 210 and the various teacher data described above.

The display unit 230 is a portion that displays the determination result of the determination unit 213. The display unit 230 is composed of, for example, a liquid crystal display, an organic EL display, a plasma display, or the like.

FIG. 7 is a flowchart showing the identification method of the present embodiment. FIG. 8 is a flowchart showing the details of the calculation of the transmission characteristic (step S7-1) in FIG. 7. FIG. 9 is a flowchart showing the details of the calculation of the elastic modulus in FIG. 7 (step S7-2). FIG. 10 is a flowchart showing details of image feature extraction (step S7-3).

When the identification process is started, the transfer characteristics are calculated (step S7-1), the elastic modulus is calculated (step S7-2), and the image features are extracted (step S7-3) in this order.

In the calculation of the transmission characteristic (step S7-1), the vibration control unit 211a controls the vibration application unit 121 to apply vibration to the object T (step S8-1). At that time, the acceleration measurement unit 131 measures the acceleration of the object T (step S8-2). The transmission characteristic is calculated by the transmission characteristic calculation unit 212a from the vibration and the acceleration (step S8-3). Details will be described later with an example, but in the present embodiment, the object T is pressed for a predetermined time with a predetermined pressing force, a fast Fourier transform is performed on the measured acceleration, and the obtained results are added and averaged. As a result, transmission characteristics can be obtained. That is, a graph of the gain value for each frequency is obtained as the transmission characteristic. The transmission characteristics differ depending on the type of the object T, such as the magnitude of the peak of the gain value, the frequency at which the peak of the gain value is taken, and the tendency of the gain value to increase or decrease with respect to the frequency change.

In the calculation of the elastic modulus (step S7-2), the pressing unit 121 is controlled by the drive control unit 211b, and the pressing force is applied from the pressing unit 121 to the object T (step S9-1). At this time, the pressing force measured by the pressing force measuring unit 171 is pressed until it reaches a predetermined load, and the moving amount of the pressing unit 121 (that is, the displacement amount of the object T) at that time is measured by the displacement measuring unit. (Step S9-2). Then, the elastic modulus calculation unit 212b calculates the elastic modulus of the object T as the relationship between the pressing force and the displacement amount (step S9-3).

In the extraction of image features (step S7-3), the image pickup unit 180 is controlled by the image pickup control unit 211c, and the object T is imaged by the image pickup unit 180 (step S10-1). Then, the image feature extraction unit 212c extracts the image feature of the object T from the image data of the object T imaged by the image pickup unit 180 (step S10-2).

When the transfer characteristics, elastic modulus, and image features are acquired as described above (steps S7-1 to S7-3), the type of the object is unknown using the classifier 213a as input data for determination. The type of the object T is determined (step S7-4). Since the classifier 213a has already learned the correlation between the types and transmission characteristics of various objects T, the elastic modulus, and the image feature, the classifier 213a is provided with the transfer characteristic, the elastic modulus, and the image feature. By inputting as the input data for determination, the type of the object T can be identified with high accuracy. The type of the object T obtained here is displayed on the display unit 230 (step S7-5).

The present embodiment may be realized by a program that causes a computer to execute each step constituting the above method. By executing the program, the same effect as the above method can be obtained. In other words, it can be said that it uses the above method.

Further, in the above configuration and method, an example of using three input data (transmission characteristic, elastic modulus, and image feature) has been described, but the elastic modulus and image feature may be omitted if necessary. In this case, the processing of steps S7-2 and S7-3 and the configurations and methods related thereto may be omitted.

Hereinafter, as an application example of the present embodiment, a case where five objects T1 to T5 (see FIG. 11) are identified will be described.

FIG. 11 shows a perspective view of five egg-shaped objects T1 to T5. Objects T1 to T5 are similar in appearance, but are plastic eggs, wooden eggs, rubber eggs, shelled boiled eggs, and shellless boiled eggs, respectively.

First, the difference due to the transmission characteristics of the objects T1 to T5 is confirmed. Specifically, while the pressing unit 121 presses the objects T1 to T5 until the pressing pressure reaches 0.5 N, the vibration applying unit 121 applies a wideband frequency vibration for 24 seconds, and the acceleration measuring unit 131 applies the target. The acceleration of objects T1 to T5 was measured. High-speed Fourier transforms of these measurement results are added and averaged, and the obtained transfer characteristics are shown in FIGS. 12 to 16.

In FIGS. 12 to 16, the horizontal axis represents the frequency (Hz) and the vertical axis represents the gain value (dB). FIG. 12 shows the transmission characteristics of the plastic egg (object T1), FIG. 13 shows the transmission characteristics of the wooden egg (object T2), and FIG. 14 shows the transmission characteristics of the rubber egg (object T3). FIG. 15 shows the transmission characteristics of a boiled egg with a shell (object T4), and FIG. 16 shows the transmission characteristics of a boiled egg without a shell (object T5).

With reference to FIGS. 12 to 16, it can be confirmed that the magnitude of the peak value, the frequency at which the peak value is taken, the increasing / decreasing tendency of the gain value with respect to the frequency change, and the like are different depending on the types of the objects T1 to T5. Therefore, by using the classifier 213a that has already learned the relationship between the transmission characteristics and the objects T1 to T5 as teacher data, the objects T1 to T5 can be identified based on the transmission characteristics. In particular, by using the classifier 213a using such machine learning, highly accurate discrimination may be possible even when transmission characteristics that may be difficult for humans to discriminate are input as input data for judgment. ..

Specific application examples of the present invention include a cleanup robot that identifies various cluttered articles and stores them in a predetermined place, an inspection robot that checks for damage to structures such as an outer wall of a building or an inner wall of a tunnel, and an inspection robot. Examples include planetary surface exploration robots that can explore the internal state of rocks such as laminated structures, cracks, and voids. In particular, under such conditions where the identification target is limited, the teacher data used is also limited, so that more accurate identification may be possible.

Further, in the present embodiment, an example in which the robot hand 110 has two finger members 120 and 130 is described, but the number of such finger members may be three or more. For example, the robot hand 110 may have one finger member to which the vibration application unit 121 is attached and two or more finger members to which the acceleration measurement unit 131 is attached. Further, although the vibration applying portion 121 and the pressing portion 121 are integrally formed, different finger members may be prepared for each.

According to the above embodiment, the following effects are exhibited.

Vibration is applied to the object T from the vibration application unit 121, the acceleration of the object T is measured by the acceleration measurement unit 131, and the transmission characteristic from vibration to acceleration is calculated by the transmission characteristic calculation unit 212a. Since the classifier 213a has learned the correlation between various types of the object T and the transmission characteristics, when the transmission characteristics for the object T of unknown type are given, the learned correlation is obtained. The type of the object T can be determined based on the above. As described above, in the present embodiment, the type of the object T can be identified with high accuracy even if the object T is difficult to discriminate in appearance.

Further, in the present embodiment, the robot hand 110 can identify various types of objects. In particular, it is possible to identify an object T that is dangerous when touched by a human, or in an environment where it is difficult for a human to act.

Further, a pressing force is applied from the pressing unit 121 to the object T, the pressing force of the pressing unit 121 and the displacement amount of the object T are measured by the force displacement measuring unit 170, and the pressing force and the displacement amount are measured by the elastic modulus calculation unit 212b. The elastic modulus is calculated as the relationship with. Since the classifier 213a has learned the correlation between the elastic modulus and the type of the object T as teacher data, the learned correlation is given when the elastic modulus with respect to the object T of unknown type is given. The type of object can be determined with high accuracy based on.

Further, the image pickup unit 180 images the object T to generate image data, and the image feature extraction unit 212c extracts the image feature of the object T from the image data. Since the classifier 213a has learned the correlation between the image feature and the type of the object as teacher data, when the image feature is given to the object T whose type is unknown, the learned correlation is given. The type of the object T can be determined with high accuracy based on the relationship.

Further, when the vibrator 121a expands and contracts to apply vibration to the object T, the first support portion 121b can receive a shearing force that may be applied to the vibrator 121a from the object T. In other words, it is possible to realize the application of vibration by the vibrator 121a while supporting the object T by the first support portion 121b. Therefore, the influence of the shearing direction on the application of vibration in the expansion / contraction direction of the vibrator 121a can be suppressed, and the vibration can be stably applied, so that stable transmission characteristics can be obtained. Further, since it is possible to suppress the application of a shearing force to the vibrator 121a, damage to the vibrator 121a can also be suppressed.

Further, since the vibrator 121a and the first support portion 121b are held by the mass body 121d, the vibrator 121a and the first support portion 121b can obtain a reaction force due to inertia from the mass body 121d. That is, vibration can be stably applied to the object T. Further, since the vibrator 121a and the first support portion 121b are urged by the first biasing portion 121e in the direction of the object T via the mass body 121d, the vibrator 121a and the first support portion 121b are used as the object T. It can be reliably contacted.

FIG. 17 is a partial cross-sectional view showing a modified example of the vibration applying portion 121 of FIG. The first urging portion 121e of the vibration applying portion 121 is attached to the hinge portion 123b as shown in FIG. 17 instead of the mode of the coil spring as shown in FIG. 2, and is a bumper that makes the drive of the hinge portion 123b elastic. May be the embodiment of. Furthermore, the first urging portion 121e may be replaced by flexibly torque-controlling an actuator (not shown) of each hinge portion.

Further, when vibration is applied to the object T by the vibration application unit 121, the second support unit 131b can receive the shearing force applied from the object T to the acceleration sensor 131a. In other words, it is possible to measure the acceleration by the acceleration sensor 131a while supporting the object T by the second support portion 131b. Therefore, the acceleration of the object T can be measured stably, and stable transmission characteristics can be obtained. Further, since it is possible to suppress the application of shearing force to the acceleration sensor 131a, damage to the acceleration sensor 131a can also be suppressed.

Further, since the acceleration sensor 131a is urged in the T direction of the object by the second urging portion 131e, the acceleration sensor 131a can be surely brought into contact with the object.

(Second Embodiment)
The identification device 10 of the second embodiment shown in FIG. 18 uses a measuring device 100 worn on a human hand. Except for the configuration related to the measuring device 100, the configuration is substantially the same as the configuration of the identification device 10 of the first embodiment. Therefore, the same parts as those shown in the first embodiment may be designated by the same reference numerals and the description thereof may be omitted.

In the present embodiment, the measuring device 100 is attached to a human hand so that it can come into contact with the object T. The measuring device 100 has a first attachment 125 attached to the index finger, a second attachment 135 attached to the thumb, and a wrist holder 141 attached to the wrist.

A vibration application unit 121 is attached to the first attachment 125, and an acceleration measurement unit 131 is attached to the second attachment 135. The vibration application unit 121 and the acceleration measurement unit 131 are attached to the ventral side of the finger, facing each other, and are attached so that the object T can be sandwiched. However, also in this embodiment, it is not necessary to grip the object T, and the object T may be brought into contact with the object T in a mode other than gripping.

The first fitting 125, the second fitting 135, and the wrist holder 141 are mechanically connected via links 142a to 142c and hinge portions 143a to 143c. Angle detectors (not shown) are provided on the hinge portions 143a to 143c so that the movement amounts of the vibration application unit 121 and the acceleration measurement unit 131 can be measured.

The operation and effect of the identification device 10 of the present embodiment are substantially the same as those of the first embodiment. However, in the present embodiment, when a human wants to identify the type of the object T, it can be operated intuitively. That is, the type of the object T can be easily identified by wearing the first attachment 125 and the second attachment 135 in the hand and touching the object T to be identified.

As a modification of the present embodiment, although details are not shown, a glove-shaped measuring device 100 worn on the palm may be used. That is, even if the vibration application unit 121 and the acceleration measurement unit 131 are not arranged at the fingertips as in the first attachment 125 and the second attachment 135, but are configured to be located on the palm of the hand. good.

Although specific embodiments of the present invention and variations thereof have been described above, the present invention is not limited to the above embodiments, and various modifications can be made within the scope of the present invention.

10 Identification device 100 Measuring device 110 Robot hand 120 First finger member 121 Vibration application part (pressing part)
121a Oscillator 121b First support 121c Non-slip member 121d Mass body 121e First urging part 121f Connector 121g Slide mechanism 122a to 122c Link 123a to 123c Hing part 125 First attachment 130 Second finger member 131 Acceleration Measuring unit 131a Accelerometer 131b Second support 131c Non-slip member 131d Connector 131e Second bias 132a to 132c Link 133a to 133c Hinge 135 Second attachment 140 Base 141 Wrist holder 142a to 142c Link 143a to 143c Hing part 170 Force displacement measurement unit 171 Push pressure measurement unit 180 Imaging unit 200 Controller unit 210 Control unit 211 Operation control unit 211a Vibration control unit 211a1 Signal generation unit 211a2 Signal amplification unit 211b Drive control unit 211c Imaging control unit 212 Calculation unit 212a Transmission Characteristic calculation unit 212b Elasticity calculation unit 212c Image feature extraction unit 213 Judgment unit 213a Discriminator 220 Storage unit 230 Display unit

Claims (11)

An identification device that identifies the type of object
A measuring device having a vibration application unit that applies vibration to the object and an acceleration measurement unit that measures the acceleration of the object.
A transmission characteristic calculation unit that calculates the transmission characteristics from the vibration applied by the vibration application unit to the acceleration measured by the acceleration measurement unit with respect to the object.
The transmission characteristics calculated by the transmission characteristic calculation unit for the object whose type is known are used as teacher input data, the type of the object is used as teacher output data, and the teacher input data and the teacher output data. With a classifier that has learned the correlation with
It is provided with a determination unit for determining the type of the object of unknown type by using the classifier as input data for determination using the transmission characteristic calculated by the transmission characteristic calculation unit for the object of unknown type. , Identification device. The measuring device includes a robot hand having a first finger member and a second finger member.
The vibration applying portion is attached to the first finger member.
The identification device according to claim 1, wherein the acceleration measuring unit is attached to the second finger member. The measuring device includes a first fitting and a second fitting to be worn on a human hand.
The vibration application portion is attached to the first fitting.
The identification device according to claim 1, wherein the acceleration measuring unit is attached to the second attachment. The measuring device includes a pressing unit that presses the object and a force displacement measuring unit that measures the pressing force of the pressing unit and the displacement amount of the object.
An identification device further comprising an elastic modulus calculation unit that calculates the elastic modulus of the object as a relationship between the pressing force measured by the force displacement measuring unit and the displacement amount.
The classifier has already learned the elastic modulus calculated by the elastic modulus calculation unit for the object of known type as the input data for the teacher.
The item according to any one of claims 1 to 3, wherein the determination unit further includes the elastic modulus calculated by the elastic modulus calculation unit for the object of unknown type as input data for determination. Identification device. An imaging unit that images the object,
An identification device further comprising an image feature extraction unit that extracts image features of the object from the image data of the object captured by the imaging unit.
The classifier has further learned the image features extracted by the image feature extraction unit from the object of known type as the input data for the teacher.
The item according to any one of claims 1 to 4, wherein the determination unit further includes the image feature extracted by the image feature extraction unit for the object of unknown type as input data for determination. Identification device. Any one of claims 1 to 5, wherein the vibration applying portion includes a vibrator that expands and contracts by an electric signal, and a first support portion that comes into contact with the object together with the vibrator around the vibrator. The identification device described in the section. The vibration application portion includes a mass body that holds the vibrator and the first support portion, and a first attachment that urges the vibrator and the first support portion toward the object via the mass body. The identification device according to claim 6, further comprising a force unit. Any one of claims 1 to 7, wherein the acceleration measuring unit includes an acceleration sensor that measures acceleration and a second support unit that comes into contact with the object together with the acceleration sensor around the acceleration sensor. The identification device described in. The identification device according to claim 8, wherein the acceleration measuring unit includes a second urging unit that urges the acceleration sensor toward the object. An identification method that identifies the type of object
Vibration is applied to the object to measure the acceleration of the object, and the object is measured.
The transmission characteristic to the acceleration measured from the vibration applied to the object was calculated.
For the object of known type, the transmission characteristic is used as teacher input data, the type of the object is used as teacher output data, and the correlation between the teacher input data and the teacher output data has been learned. A discrimination method including determining the type of the object of unknown type by using the discriminator of the above as input data for determination of the transmission characteristic for the object of unknown type. A program for identifying the type of object
Vibration is applied to the object and the acceleration of the object is measured.
The transmission characteristic to the acceleration measured from the vibration applied to the object is calculated.
For the object whose type is known, the transmission characteristic is used as the input data for the teacher, the type of the object is used as the output data for the teacher, and the correlation between the input data for the teacher and the output data for the teacher has been learned. A program that makes a computer function so as to determine the type of the object of unknown type by using the transmission characteristic as input data for determination for the object of unknown type by using the classifier of the above.

Images