Specification
Title of Invention: Parameter Determining Device, Parameter Determining Method and Recording Medium
Technical field
[0001]
The present invention relates to the technical field of a parameter determination device, a parameter determination method, and a recording medium for determining image generation parameters used to generate an input image to be input to a recognition device for recognizing an input image.
Background technology
[0002]
A technique for automatically calculating (in other words, determining) parameters for image processing applied to an image is known (see Patent Document 1, for example). Also, there is known a technique for efficiently finding conditions for performing information processing using an image with high accuracy (see, for example, Patent Document 2). In addition, Patent Document 3 can be cited as a prior art document related to the present invention.
prior art documents
patent literature
[0003]
Patent document 1: JP 2012-198680 A
Patent Document 2: International Publication No. 2014/002398 Pamphlet
Patent Document 3: JP 2017-130794 A
SUMMARY OF THE INVENTION
Problems to be Solved by the Invention
[0004]
An example of information processing using images is recognition processing for recognizing images. For example, face authentication processing for recognizing a person's face in an image and authenticating the person based on the recognized face is an example of recognition processing. In this case, an image is input from a photographing device such as a camera to the recognition device that recognizes the image.
[0005]
Here, a photographing device that photographs an image usually outputs an image that is easy for humans to see (that is, easy for human eyes to see). This is because images captured by the imaging device are generally used for viewing by humans. Therefore, the parameters for determining the optical characteristics of the imaging device and the parameters for determining the content of image processing performed inside the imaging device are determined so as to satisfy the condition that the imaging device outputs an image that is easy for humans to see. On the other hand, an image that is easy for humans to see is not necessarily an image that is easy for the recognition device to recognize. This is because the recognition device treats the image as digital data. Therefore, there is a possibility that the recognition device cannot appropriately recognize the image only by performing the recognition processing using the image output from the photographing device as it is.
[0006]
An object of the present invention is to provide a parameter determination device, a parameter determination method, and a recording medium that can solve the technical problems described above. As an example, the present invention provides parameters capable of determining image generation parameters used to generate an image to be subjected to a recognition operation so that the recognition apparatus can perform a recognition operation on an image that is easily recognized by the recognition apparatus. An object of the present invention is to provide a determination device, a parameter determination method, and a recording medium.
Means to solve problems
[0007]
According to one aspect of the parameter determination device, calculation means for calculating an evaluation value for evaluating the recognition results based on the recognition results of the plurality of recognition target images by a recognition device that performs a recognition operation on the recognition target images; and determining means for determining an image generation parameter used to generate the recognition target image based on the evaluation value.
[0008]
One aspect of the parameter determination method includes: calculating an evaluation value for evaluating the recognition results based on recognition results of a plurality of the recognition target images by a recognition device that performs a recognition operation on the recognition target images; determining image generation parameters used to generate the recognition target image based on the evaluation value.
[0009]
One aspect of a recording medium is a non-temporary recording medium in which a computer program for causing a computer to execute a parameter determination method is recorded, wherein the parameter determination method is performed by a recognition device that performs a recognition operation on a recognition target image. calculating an evaluation value for evaluating the recognition result based on the recognition result of the recognition target image of and setting an image generation parameter used to generate the recognition target image based on the evaluation value determining.
Effect of the invention
[0010]
According to one aspect of each of the parameter determination device, the parameter determination method, and the recording medium described above, an image to be subjected to a recognition operation is selected so that the recognition device can perform the recognition operation on an image that the recognition device can easily recognize. Image generation parameters used to generate can be determined.
Brief description of the drawing
[0011]
1] Fig. 1 is a block diagram showing the overall configuration of a recognition system according to a first embodiment. [Fig.
[Fig. 2] Fig. 2 is a block diagram showing the configuration of an imaging device according to the first embodiment.
3] Fig. 3 is a block diagram showing the configuration of the recognition device of the first embodiment. [Fig.
4] Fig. 4 is a block diagram showing the configuration of the parameter determination device of the first embodiment. [Fig.
5] Fig. 5 is a flowchart showing a flow of parameter determination operation performed by the parameter determination device of the first embodiment. [Fig.
6] Fig. 6 is a block diagram showing the configuration of a recognition device according to a second embodiment. [Fig.
7] FIG. 7 is a flow chart showing the flow of parameter determination operation performed by the parameter determination device of the second embodiment. [FIG.
8] Fig. 8 is a block diagram showing the configuration of a parameter determination device according to a third embodiment. [Fig.
9] Fig. 9 is a plan view showing candidate values ​​of image generation parameters to be set in the image capturing device or the recognition device in the parameter determination operation. [Fig.
MODE FOR CARRYING OUT THE INVENTION
[0012]
Embodiments of a parameter determination device, a parameter determination method, and a recording medium will be described below with reference to the drawings. A recognition system SYS to which the embodiments of the parameter determination device, the parameter determination method, and the recording medium are applied will be described below. The recognition system SYS is a system for recognizing a recognition target image (specifically, recognizing an object reflected in the recognition target image). In the following, for convenience of explanation, a system (so-called An example of a face authentication system) will be described. However, the recognition system SYS is not limited to a system that recognizes the face of a person reflected in the recognition target image 100 and authenticates the person reflected in the recognition target image 100 using the recognized face.
[0013]
(1) Recognition system SYS of the first embodiment
First, the recognition system SYS of the first embodiment will be described. Hereinafter, the recognition system SYS of the first embodiment will be referred to as "recognition system SYSa".
[0014]
(1-1) Configuration of recognition system SYSa
(1-1-1) Overall configuration of recognition system SYSa
First, the overall configuration of the recognition system SYSa of the first embodiment will be described with reference to FIG. FIG. 1 is a block diagram showing the overall configuration of the recognition system SYSa of the first embodiment.
[0015]
As shown in FIG. 1, the recognition system SYSa includes an imaging device 1, a recognition device 2, and a parameter determination device 3. The photographing device 1, the recognition device 2, and the parameter determination device 3 are connected via a communication network 4 so as to be able to communicate with each other. The communication network 4 may include a wired network, or may include a wireless network.
[0016]
The photographing device 1 is a device that generates a recognition target image 100 in which a person is captured by photographing the person. The imaging device 1 transmits (in other words, inputs) the generated recognition target image 100 to the recognition device 2 via the communication network 4 .
[0017]
The recognition device 2 acquires (in other words, receives) the recognition target image 100 generated by the imaging device 1 via the communication network 4 . The recognition device 2 recognizes the face of the person reflected in the recognition target image 100 based on the acquired recognition target image 100, and authenticates the person reflected in the recognition target image 100 using the recognized face. perform recognition operations for
[0018]
The parameter determination device 3 executes a parameter determination operation for determining the image generation parameters 300 (specifically, determining the values ​​of the image generation parameters 300). The image generation parameters 300 are used (in other words, referred to) by the imaging device 1 to generate the recognition target image 100 . The image generation parameter 300 defines the operation content of the imaging device 1 that generates the recognition target image 100 . Therefore, the photographing device 1 generates the recognition target image 100 based on the image generation parameters determined by the parameter determination device 3 .
[0019]
(1-1-2) Configuration of imaging device 1
Next, the configuration of the photographing device 1 will be described with reference to FIG. FIG. 2 is a block diagram showing the configuration of the photographing device 1. As shown in FIG.
[0020]
As shown in FIG. 2, the photographing device 1 includes a camera 11, an arithmetic device 12, and a communication device 13. The camera 11 , arithmetic device 12 and communication device 13 are connected via a data bus 14 .
[0021]
By photographing a person, the camera 11 generates a photographed image 101 in which the person is reflected. The optical properties of camera 11 are defined by optical parameters 301 , which are an example of imaging parameters 300 . In this case, the camera 11 becomes a camera having optical properties defined by the optical parameters 301 set (in other words, applied, reflected, or registered) in the camera 11 . Accordingly, the characteristics of the captured image 101 generated by the camera 11 are defined by the optical parameters 301 . That is, the camera 11 generates the captured image 101 based on the optical parameters 301 . As an example of the optical characteristics of the camera 11 defined by the optical parameter 301, at least one of the aperture value of the camera 11, the focus position of the camera 11 (in other words, the focus position), the shutter speed of the camera 11, and the sensitivity of the camera 11 is can give. Therefore, the optical parameters 301 include a parameter that defines the aperture value of the camera 11, a parameter that defines the focus position (in other words, focus position) of the camera 11, a parameter that defines the shutter speed of the camera 11, and the sensitivity of the camera 11. It may contain at least one of the parameters to be specified.
[0022]
The computing device 12 includes, for example, at least one of a CPU (Central Processing Unit) and a GPU (Graphic Processing Unit). Arithmetic device 12 reads a computer program. For example, the computing device 12 may read a computer program stored in a storage device (not shown) included in the photographing device 1 . For example, the computing device 12 may read a computer program stored in a computer-readable and non-transitory recording medium using a recording medium reading device (not shown). The computing device 12 may acquire (that is, download or read) a computer program from a device (not shown) arranged outside the imaging device 1 via the communication device 13 .
[0023]
The computing device 12 executes the read computer program. As a result, an image processing unit 121 for executing image processing on the captured image 101 is realized as a logical functional block within the arithmetic unit 12 . That is, the arithmetic device 21 can function as a controller for realizing the image processing section 121 .
[0024]
The image processing unit 121 generates the recognition target image 100 by performing predetermined image processing on the captured image 101 generated by the camera 11 . The details of the image processing executed by the image processing unit 121 are defined by processing parameters 302 that are an example of the image generation parameters 300 . In this case, the image processing unit 121 generates the recognition target image 100 by performing image processing whose content is specified by the processing parameter 302 on the captured image 101 . As an example of image processing executed by the image processing unit 121, white balance correction processing for correcting the white balance of the captured image 101;1, contrast correction processing for correcting the contrast of the photographed image 101, dehaze processing for improving the image quality of the photographed image 101 whose visibility has deteriorated due to the effect of haze, photographing HDR (High Dynamic Range) processing for adjusting the dynamic range of the image 101 to improve the image quality of the captured image 101, denoising processing for improving the image quality of the captured image 101 whose visibility has deteriorated due to noise, and , and skeleton texture separation processing for separating the captured image 101 into a skeleton image and a texture image. Therefore, the processing parameters 302 include a parameter that defines the details of white balance correction processing, a parameter that defines the details of brightness correction processing, a parameter that defines the details of contrast correction processing, a parameter that defines the details of dehaze processing, and a parameter that defines the details of dehaze processing. may include at least one of a parameter that defines the content of the denoising process, a parameter that defines the content of the skeleton texture separation process, and a parameter that defines the content of the denoising process. The parameters that define the content of the white balance correction process are, for example, parameters that define whether to execute the white balance correction process, parameters that define the intensity of the white balance correction process, parameters that define the limit value of the correction amount, At least one of a parameter that defines the limit value of the correction amount of the G (Green) component with respect to the R (Red) component and a parameter that defines the limit value of the B (Blue) component correction amount with respect to the R component may contain The parameters that define the content of the luminance correction process include, for example, a parameter that defines whether or not to execute the luminance correction process, a parameter that defines the intensity of the luminance correction process, a parameter that defines the target value of luminance, and a correction amount limit. It may contain at least one of the parameters defining the value. The parameters that define the details of the contrast correction process are, for example, parameters that define whether or not to execute the contrast correction process, parameters that define the strength of the contrast correction process, and limits on the amount of correction for relatively dark areas. and at least one of a parameter defining a limit value of the amount of correction for a relatively bright area. The parameters that define the details of the dehaze process are, for example, parameters that define whether to execute the dehaze process, parameters that define the strength of the dehaze process, and parameters that define the limit value of the correction amount of the captured image 101 by the dehaze process. may include at least one of The parameters that define the details of HDR processing are, for example, parameters that define whether HDR processing is to be executed, parameters that define the intensity of HDR processing, parameters that define target values ​​of brightness for HDR processing, It may include at least one of parameters defining a limit value for the correction amount of the image 101 and a threshold value for identifying black areas. The parameters that define the details of the denoising process may include, for example, at least one of a parameter that defines whether to execute the denoising process and a parameter that defines the strength of the denoising process. The parameters that define the details of the skeletal texture separation process may include, for example, at least one of a parameter that defines whether to execute the skeletal texture separation process and a parameter that defines the strength of the skeletal texture separation process.
[0025]
The communication device 13 can communicate with the recognition device 2 and the parameter determination device 3 via the communication network 4. In the first embodiment, the communication device 13 can transmit the recognition target image 100 to the recognition device 2 via the communication network 4 . Also, the communication device 13 can receive the image generation parameters 300 (specifically, the optical parameters 301 and the processing parameters 302) determined by the parameter determination device 3 via the communication network 4. FIG. The optical parameters 301 received by the communication device 13 are applied to the camera 11 . Therefore, the camera 11 becomes a camera having optical characteristics defined by the optical parameters 301 received by the communication device 13 . The processing parameters 302 received by the communication device 13 are applied to the image processing unit 121 . Therefore, the image processing unit 121 performs image processing on the captured image 101 , the content of which is defined by the processing parameters 302 received by the communication device 13 .
[0026]
(1-1-3) Configuration of recognition device 2
Next, the configuration of the recognition device 2 will be described with reference to FIG. FIG. 3 is a block diagram showing the configuration of the recognition device 2. As shown in FIG.
[0027]
As shown in FIG. 3, the recognition device 2 includes an arithmetic device 21, a storage device 22, and a communication device 23. Arithmetic device 21 , storage device 22 and communication device 23 are connected via data bus 24 .
[0028]
The computing device 21 includes, for example, at least one of a CPU and a GPU. Arithmetic device 21 reads a computer program. For example, arithmetic device 21 may read a computer program stored in storage device 22 . For example, the computing device 21 may read a computer program stored in a computer-readable non-temporary recording medium using a recording medium reading device (not shown). The computing device 21 may acquire (that is, download or read) a computer program from a device (not shown) arranged outside the recognition device 2 via the communication device 23 . Arithmetic device 21 executes the read computer program. As a result, logical functional blocks for executing the operation (specifically, the above-described recognition operation) to be performed by the recognition device 2 are implemented in the arithmetic unit 21 . That is, the arithmetic unit 21 can function as a controller for implementing logical functional blocks for executing recognition operations.
[0029]
FIG. 3 shows an example of logical functional blocks implemented within the arithmetic unit 21 to perform recognition operations. As shown in FIG. 3, a recognition unit 211 is realized as a logical functional block within the arithmetic unit 21 . The recognition unit 211 recognizes (more specifically, detects) the face of a person reflected in the recognition target image 100 based on the recognition target image 100 transmitted from the photographing device 1 . Note that the recognition unit 211 recognizes (detects) the face of the person reflected in the recognition target image 100 using an existing method for recognizing (detecting) the face of the person reflected in the image. good too. Further, the recognition unit 211 authenticates the person appearing in the recognition target image 100 using the recognized face. The recognition unit 211 may authenticate a person using an existing method for authenticating a person based on the person's face (that is, an existing face authentication method). An example of a method for authenticating a person based on the person's face will be briefly described below. The recognition unit 211 searches the face authentication DB (DataBase) 220 for a record that satisfies face authentication conditions determined according to the feature amount of the recognized face. The face authentication DB 220 includes a plurality of records in which facial features of one person and identification information for uniquely identifying the person are associated with each other. In this case, the recognition unit 211 searches for a record that satisfies the face authentication condition by matching the feature amount of the recognized person's face with the feature amount included in the face authentication DB 220 . For example, the recognition unit 211 may search the face authentication DB 220 for a record that satisfies the face authentication condition that the feature amount is the same as the feature amount of the recognized face. If a record that satisfies the face authentication condition exists in the face authentication DB 220, the recognition unit 211 recognizes that the person appearing in the recognition target image 100 is the person specified by the identification information included in the record that satisfies the face authentication condition. authenticate as If there is no record that satisfies the face authentication condition in the face authentication DB 220, the recognition unit 211 determines that the person appearing in the recognition target image 100 cannot be authenticated. That is, the recognition unit 211 determines that the person appearing in the recognition target image 100 cannot be authenticated.
[0030]
The storage device 22 can store desired data. For example, the storage device 22 may temporarily store computer programs executed by the arithmetic device 21 . The storage device 22 may temporarily store data temporarily used by the arithmetic device 21 while the arithmetic device 21 is executing a computer program. The storage device 22 may store data that the recognition device 2 saves over a long period of time. In the first embodiment, the storage device 22 can store the face DB 220 described above. Furthermore, the storage device 22 can store an image DB (DataBase) 221 for accumulating (that is, storing, recording, or storing) the recognition target image 100 transmitted from the photographing device 1 . Further, the storage device 22 stores recognition result information indicating the result of the recognition operation by the recognition unit 211 (for example, information regarding the recognition result of the person appearing in the recognition target image 100 and information regarding the authentication result of the recognized person). can be stored in the recognition result DB 222 for accumulating. The storage device 22 may include at least one of RAM (Random Access Memory), ROM (Read Only Memory), hard disk device, magneto-optical disk device, SSD (Solid State Drive), and disk array device. good. That is, the storage device 22 may include non-transitory recording media.
[0031]
The communication device 23 can communicate with the imaging device 1 and the parameter determination device 3 via the communication network 4. In the first embodiment, the communication device 23 can acquire (that is, receive) the recognition target image 100 transmitted from the imaging device 1 via the communication network 4 . Furthermore, the communication device 23 can transmit recognition result information accumulated in the recognition result DB 222 to the parameter determination device 3 via the communication network 4 . The parameter determination device 3 determines image generation parameters 300 based on the recognition result information.
[0032]
(1-1-4) Configuration of parameter determination device 3
Next, the configuration of the parameter determining device 3 will be described with reference to FIG. FIG. 4 is a block diagram showing the configuration of the parameter determination device 3. As shown in FIG.
[0033]
As shown in FIG. 4, the parameter determination device 3 includes an arithmetic device 31, a storage device 32, and a communication device 33. Arithmetic device 31 , storage device 32 , and communication device 33 are connected via data bus 34 .
[0034]
The computing device 31 includes, for example, at least one of a CPU and a GPU. Arithmetic device 31 reads a computer program. For example, arithmetic device 31 may read a computer program stored in storage device 32 . For example, the computing device 31 may read a computer program stored in a computer-readable non-temporary recording medium using a recording medium reading device (not shown). The computing device 31 may acquire (that is, download or read) a computer program from a device (not shown) arranged outside the parameter determination device 3 via the communication device 23 . Arithmetic device 31 executes the read computer program. As a result, logical functional blocks for executing the operations to be performed by the parameter determination device 3 (specifically, the above-described parameter determination operation) are realized in the arithmetic device 31 . That is, the arithmetic device 31 can function as a controller for implementing logical functional blocks for executing parameter determination operations.
[0035]
FIG. 4 shows an example of logical functional blocks implemented within the computing device 31 for executing the parameter determination operation. As shown in FIG. 4, a parameter setting unit 311, a recognition result acquisition unit 312, an evaluation unit 313, and a parameter determination unit 314 are implemented as logical functional blocks in the arithmetic unit 31. FIG. Note that the parameter setting unit 311, the recognition result acquisition unit 312, the evaluationDetails of the operations of the section 313 and the parameter determination section 314 will be described later with reference to FIG. The parameter setting unit 311 sets the image generation parameters 300 of the imaging device 1 by transmitting the image generation parameters 300 to be set in the imaging device 1 . The imaging device 1 generates the recognition target image 100 using the image generation parameters 300 set by the parameter setting unit 311 . The recognition result acquisition unit 312 acquires recognition result information indicating the result of the recognition operation (that is, recognition result information accumulated in the recognition result DB 222 of the storage device 22) from the recognition device 2. FIG. Based on the recognition result information, the evaluation unit 313 determines the image generation parameters 300 used by the imaging device 1 to generate the recognition target image 100 (that is, the image generation parameters 300 actually set in the imaging device 1). An evaluation value for evaluating whether or not is appropriate is calculated. The parameter determination unit 314 determines (in other words, calculates) the values ​​of the image generation parameters 300 to be set in the imaging device 1 based on the evaluation values ​​calculated by the evaluation unit 313 .
[0036]
The storage device 32 can store desired data. For example, the storage device 32 may temporarily store computer programs executed by the arithmetic device 31 . The storage device 32 may temporarily store data temporarily used by the arithmetic device 31 while the arithmetic device 31 is executing a computer program. The storage device 32 may store data that the parameter determination device 3 saves for a long time. In the first embodiment, the storage device 32 can store a parameter DB 321 for accumulating (that is, storing, recording, or storing) information regarding the image generation parameters 300 determined by the parameter determining section 314 . Furthermore, the storage device 32 can store a correct answer DB 322 that accumulates correct answer data to be compared with the recognition result information in order to calculate the aforementioned evaluation value (for example, the F value described later). The storage device 32 may include at least one of RAM (Random Access Memory), ROM (Read Only Memory), hard disk device, magneto-optical disk device, SSD (Solid State Drive), and disk array device. good. That is, the storage device 32 may include non-transitory recording media.
[0037]
The communication device 33 can communicate with the imaging device 1 and the recognition device 2 via the communication network 4. In the first embodiment, the communication device 33 can acquire (that is, receive) recognition result information transmitted from the recognition device 2 via the communication network 4 . Furthermore, under the control of the parameter setting unit 311 , the communication device 33 can transmit the image generation parameters 300 to be set in the image capturing device 1 to the image capturing device 1 via the communication network 4 . The image generation parameters 300 transmitted by the communication device 33 are applied to the imaging device 1 . That is, the imaging device 1 generates the recognition target image 100 using the image generation parameters 300 transmitted by the communication device 33 .
[0038]
(1-2) Operation of recognition system SYSa
Next, the operation of the recognition system SYSa of the first embodiment will be described with reference to FIG. In particular, the parameter determination operation performed by the parameter determination device 3 will be described below. FIG. 5 is a flow chart showing the flow of the parameter determination operation performed by the parameter determination device 3. As shown in FIG.
[0039]
As shown in FIG. 5, the parameter setting unit 311 first initializes the variable r to 0 (step S301). After that, the parameter setting unit 311 sets the image generation parameters 300 of the photographing device 1 (specifically, the optical parameters 301 of the camera 11 and the processing parameters 302 of the image processing unit 121) to initial values ​​(step S310). As the initial value of the image generation parameter 300, any one of a plurality of candidate values ​​of the image generation parameter 300 to be set in the imaging device 1 in the parameter determination operation may be used. A plurality of candidate values ​​may include recommended values ​​preset in the imaging device 1 .
[0040]
After that, the parameter determining unit 314 controls the photographing device 1 so as to generate a predetermined number or more of recognition target images 100 based on the image generation parameters 300 set in step S310 (step S311). That is, the parameter determination unit 314 controls the camera 11 having the optical characteristics defined by the optical parameters 301 set in step S310 so as to capture (that is, generate) a predetermined number of captured images 101 or more. Further, the parameter determining unit 314 generates a predetermined number or more of recognition target images 100 by applying image processing whose content is defined by the processing parameter 302 set in step S310 to a predetermined number or more of the captured images 101. The image processing unit 121 is controlled so as to The recognition device 2 acquires the recognition target image 100 generated by the imaging device 1 using the communication device 23 and accumulates it in the image DB 221 of the storage device 22 .
[0041]
In addition, in the parameter determination operation, the camera 11 may photograph the person under the condition that the person is actually standing in front of the camera 11 . Alternatively, the camera 11 may photograph the image or the model under a situation in which an image or model imitating a person is placed in front of the camera 11 . Furthermore, information about a person photographed by the camera 11 (or a person modeled by an image photographed by the camera 11 or a model) is accumulated in the correct answer DB 322 as correct answer data. That is, in the correct answer DB 322, correct answer data (for example, identification information for uniquely identifying a person) indicating a person appearing in each recognition target image 100 transmitted from the photographing device 1 to the recognition device 2 is stored as a parameter. The number of recognition target images 100 used in the determination operation is accumulated.
[0042]
After the predetermined number or more of the recognition target images 100 have been generated, the parameter determining unit 314 performs the above recognition operation on each of the predetermined number or more of the recognition target images 100 generated in step S311. The recognition device 2 (in particular, the recognition unit 211) is controlled (step S312). The recognition device 2 accumulates recognition result information indicating the result of the recognition operation in the recognition result DB 222 of the storage device 22 .
[0043]
After the recognition operation for the predetermined number or more of the recognition target images 100 is completed, the recognition result acquisition unit 312 acquires recognition result information indicating the result of the recognition operation in step S312 from the recognition device 2 via the communication device 33. (Step S313).
[0044]
After that, the evaluation unit 313 determines that the image generation parameters 300 actually set in the photographing device 1 in step S310 are appropriate based on the recognition result information acquired in step S313 and the correct answer data stored in the correct answer DB 322. An evaluation value for evaluating whether or not is calculated (step S314). That is, the evaluation unit 313 calculates an evaluation value for evaluating the recognition action performed on the recognition target image 100 generated based on the image generation parameters 300 set in step S310.
[0045]
For example, the evaluation unit 313 may calculate an F value (F-scale: F-measure) as an evaluation value. The F value is an evaluation value determined based on precision and recall. Specifically, the F value is an evaluation value defined by the formula F value=(2×relevance×recall)/(relevance+recall). The matching rate indicates the ratio of the number of recognition target images 100 in which the person has been successfully authenticated to the number of recognition target images 100 in which the person's face has been detected. The recall rate is the total number of recognition target images 100 in which a person is reflected (that is, the total number of recognition target images 100 for which a person should be authenticated, and the total number of correct data corresponding to the recognition target image 100). The ratio of the number of successful recognition target images 100 is shown.
[0046]
The “recognition target image 100 in which the person has been successfully authenticated” in the present embodiment means “a recognition target image in which the person reflected in the authentication target image 100 has been correctly authenticated by the recognition unit 211 as the person. 100”. In other words, “the recognition target image 100 in which the recognition unit 211 has authenticated the person reflected in the authentication target image 100 as a different person (that is, the recognition target image 100 in which the recognition unit 211 has erroneously authenticated 100)” is not included in the “number of recognition target images 100 for which person authentication has succeeded”. Specifically, the recognition result information for one recognition target image 100 in which one person is reflected is authentication that "the person reflected in one recognition target image 100 has been authenticated as one person". When the results are shown, one recognition target image 100 is included in "recognition target images 100 for which person authentication has succeeded". On the other hand, the recognition result information for one recognition target image 100 in which one person is reflected is "the person reflected in one recognition target image 100 has been authenticated as another person different from the one person". If the authentication result indicates that the person has been successfully authenticated, the one recognition target image 100 is not included in the "recognition target images 100 whose person has been successfully authenticated." For this reason, the correct answer DB 322 referred to for calculating the evaluation value accumulates correct answer data indicating that the person reflected in one recognition target image 100 is one person. In this case, the evaluation unit 313 can use the evaluation result information and the correct answer DB 322 to calculate “the number of recognition target images 100 for which person authentication has succeeded”, and as a result, the F value (that is, the evaluation value) can be calculated appropriately.
[0047]
Alternatively, the evaluation unit 313 may calculate a value other than the F value as the evaluation value. For example, the evaluation unit 313 may calculate the above-described relevance rate itself as the evaluation value. For example, the evaluation unit 313 may calculate the recall itself described above as the evaluation value. For example, the evaluation unit 313 may calculate an evaluation value determined based on the relevance rate described above. For example, the evaluation unit 313 may calculate an evaluation value determined based on the recall rate described above. Depending on the calculated evaluation value, the correct answer DB 322 may not be used. In this case, the storage device 32 may not store the correct answer DB 322 .
[0048]
After that, the parameter determination unit 314 determines whether or not the variable r is 0 (step S315).
[0049]
As a result of the determination in step S315, when it is determined that the variable r is 0 (step S315: Yes), the parameter determination unit 314 stores information about the image generation parameters 300 set in step S310 in the parameter DB 321. Record (step S317). The information about the image generation parameter 300 may include, for example, information indicating the value of the image generation parameter 300 set in step S310 and information indicating the evaluation value calculated in step S314.
[0050]
The information about the image generation parameters 300 recorded in the parameter DB 321 is the image generation parameters 300 to be used by the photographing device 1 to generate the recognition target image 100 (that is, the image generation parameters 300 to be actually set in the photographing device 1). ). Therefore, after the parameter determination operation shown in FIG. 5 is completed, the values ​​of the image generation parameters 300 recorded in the parameter DB 321 are actually set in the photographing apparatus 1 . That is, after the parameter determination operation shown in FIG. 5 is completed, the imaging device 1 generates the recognition target image 100 based on the image generation parameters 300 recorded in the parameter DB 321. FIG.
[0051]
After that, the parameter setting unit 311 determines whether or not all the candidate values ​​of the image generation parameters 300 to be set in the imaging device 1 in the parameter determination operation have actually been set in the imaging device 1 (step S318). ). For example, in a situation where the first candidate value to the fifth candidate value should be set in the imaging apparatus 1 as the image generation parameter 300 in the parameter determination operation, the parameter setting unit 311 selects the first candidate value to the fifth candidate value. Each of the values ​​is actually set in the imaging device 1 as the image generation parameter 300.Determine whether or not the If at least one of the first to fifth candidate values ​​has not yet been set in the imaging device 1 as the image generation parameter 300, the parameter setting unit 311 sets it in the imaging device 1 in the parameter determination operation. It is determined that all the candidate values ​​of the imaging parameters 300 to be set are not actually set for the imaging device 1 .
[0052]
As a result of the determination in step S318, if it is determined that all the candidate values ​​of the image generation parameters 300 to be set in the imaging device 1 in the parameter determination operation have not yet been set in the imaging device 1 (step S318: No), the parameter setting unit 311 increments the variable r by 1 (step S319), and sets the imaging device 1 to the image generation parameter 300 having a new value (that is, a new candidate value for the image generation parameter 300). is set (step S310). Henceforth, the process after step S311 is repeated.
[0053]
On the other hand, as a result of the determination in step S318, if it is determined that all the candidate values ​​of the image generation parameters 300 to be set in the imaging device 1 in the parameter determination operation have already been set in the imaging device 1 ( Step S318: Yes), the parameter determination device 3 ends the parameter determination operation shown in FIG.
[0054]
On the other hand, as a result of the determination in step S315, if it is determined that the variable r is not 0 (step S315: No), information regarding the image generation parameters 300 has already been recorded in the parameter DB321. In this case, the parameter determining unit 314 determines whether or not the evaluation value newly calculated in step S314 is improved with respect to the evaluation value recorded in the parameter DB 321 in association with the information indicating the image generation parameter 300. Determine (step S316). That is, the parameter determining unit 314 determines whether the evaluation value corresponding to the image generation parameter 300 newly set in step S301 is improved with respect to the evaluation value corresponding to the image generation parameter 300 recorded in the parameter DB 321. determine whether or not
[0055]
Note that the evaluation value is an index value that improves as the recognition operation improves (for example, the recognition accuracy of a person's face improves and/or the authentication accuracy of a person improves). Therefore, in step S316, the parameter determination unit 314 determines that the result of the recognition operation using the recognition target image 100 generated based on the image generation parameter 300 newly set in step S301 is recorded in the parameter DB 321. It can be said that it is determined whether or not the result of the recognition operation using the recognition target image 100 generated based on the image generation parameters 300 is improved. For example, the parameter determination unit 314 performs image generation in which the accuracy of human face recognition using the recognition target image 100 generated based on the image generation parameters 300 newly set in step S301 is recorded in the parameter DB 321. It can be said that it is determined whether or not the recognition accuracy of the person's face using the recognition target image 100 generated based on the parameters 300 is improved. For example, the parameter determination unit 314 determines the accuracy of person authentication using the recognition target image 100 generated based on the image generation parameter 300 newly set in step S301, and the image generation parameter 300 recorded in the parameter DB 321. It can also be said that determination is made as to whether or not the accuracy of person authentication using the recognition target image 100 generated based on is improved.
[0056]
As a result of the determination in step S316, if it is determined that the evaluation value is improved (step S316: Yes), the recognition target image 100 generated based on the image generation parameter 300 newly set in step S310 is better than the result of the recognition operation using the recognition target image 100 generated based on the image generation parameters 300 recorded in the parameter DB 321 . For example, the recognition accuracy of a person's face using the recognition target image 100 generated based on the image generation parameter 300 newly set in step S310 is higher than that based on the image generation parameter 300 recorded in the parameter DB 321. It is estimated that the recognition accuracy of a person's face using the recognition target image 100 generated by the method described above is better than the recognition accuracy of the person's face. For example, the accuracy of person authentication using the recognition target image 100 generated based on the image generation parameter 300 newly set in step S310 is higher than that generated based on the image generation parameter 300 recorded in the parameter DB 321. is estimated to be better than the accuracy of person authentication using the recognition target image 100 that has been obtained. Therefore, for the purpose of executing an appropriate recognition operation (for example, a highly accurate recognition operation), the image generation parameters 300 newly set in step S310 are better than the image generation parameters recorded in the parameter DB 321. Estimated to be better than 300. In this case, the parameter determination unit 314 newly records information on the image generation parameter 300 newly set in step S310 in the parameter DB 321 (step S317). That is, the parameter determining unit 314 rewrites (that is, updates) the information about the image generation parameters 300 recorded in the parameter DB 321 with the information about the image generation parameters 300 newly set in step S310.
[0057]
On the other hand, as a result of the determination in step S316, if it is determined that the evaluation value has not improved (step S316: No), the recognition generated based on the image generation parameters 300 newly set in step S310 It is estimated that the result of the recognition operation using the recognition target image 100 generated based on the image generation parameters 300 recorded in the parameter DB 321 is better than the result of the recognition operation using the target image 100. be. For example, rather than the recognition accuracy of a person's face using the recognition target image 100 generated based on the image generation parameters 300 newly set in step S310, the recognition accuracy based on the image generation parameters 300 recorded in the parameter DB 321 It is estimated that the human face recognition accuracy using the generated recognition target image 100 is better. For example, rather than the accuracy of person authentication using the recognition target image 100 generated based on the image generation parameters 300 newly set in step S310, the It is estimated that the accuracy of person authentication using the recognition target image 100 is better. Therefore, for the purpose of executing an appropriate recognition operation (for example, a highly accurate recognition operation), the image generation parameters 300 recorded in the parameter DB 321 are preferred over the image generation parameters 300 newly set in step S310. is presumed to be more suitable. In this case, the parameter determination unit 314 does not newly record information on the image generation parameter 300 newly set in step S310 in the parameter DB 321 . That is, the information about the image generation parameters 300 recorded in the parameter DB 321 continues to be recorded in the parameter DB 321 as it is.
[0058]
(1-3) Technical effects of the recognition system SYSa
As described above, the recognition system SYSa (in particular, the parameter determination device 3) of the first embodiment can determine the image generation parameters 300 using recognition result information indicating recognition results of recognition actions. That is, the parameter determination device 3 can determine the values ​​of the image generation parameters 300 to be set in the imaging device 1 . Therefore, the parameter determination device 3 can determine the values ​​of the image generation parameters 300 that enable an appropriate recognition operation (for example, a recognition operation that improves recognition accuracy and/or authentication accuracy). That is, the parameter determination device 3 can determine the values ​​of the image generation parameters 300 that enable the recognition device 2 to easily recognize the recognition target image 100 . The parameter determination device 3 determines the image generation parameters 300 used to generate the recognition target image 100 so that the recognition device 2 can perform a recognition operation using the recognition target image 100 that the recognition device 2 can easily recognize. value can be determined. Therefore, when the parameter determination operation is performed, the photographing device 1 can generate the recognition target image 100 that is easier for the recognition device 2 to recognize than when the parameter determination operation is not performed. Furthermore, when the parameter determination operation is performed, the recognition device 2 performs an appropriate recognition operation using the recognition target image 100 that is easily recognized by the recognition device 2, compared to when the parameter determination operation is not performed. can be executed.
[0059]
It should be noted that the parameter determination device 3 may perform the parameter determination operation before the operation of the recognition system SYSa is started. Alternatively, the parameter determination device 3 may perform the parameter determination operation after the operation of the recognition system SYSa is started.
[0060]
Also, as described above, the image generation parameters 300 include multiple types of parameters. In this case, the parameter determination device 3 may determine multiple types of parameters in order. For example, the parameter determination device 3 determines a first type of parameter (for example, a parameter that defines the aperture value of the camera 11), and then determines a second type of parameter (for example, defines the content of white balance correction processing). parameter) may be determined.
[0061]
Further, the parameter determination device 3 may determine the processing parameters 302 of the image processing unit 121 after determining the optical parameters 301 of the camera 11 . Since the image processing unit 121 performs image processing on the captured image 101 captured by the camera 11, when the value of the optical parameter 301 is changed, the value of the processing parameter 302 also needs to be changed. may come out. Therefore, if the processing parameters 302 are determined after the determination of the optical parameters 301 is completed, the parameter determining device 3 can determine the processing parameters 302 under the condition that the values ​​of the optical parameters 301 are fixed. Therefore, the parameter determination device 3 can determine the image generation parameters 300 relatively efficiently.
[0062]
(2) Recognition system SYS of the second embodiment
Next, the recognition system SYS of the second embodiment will be described. Hereinafter, the recognition system SYS of the second embodiment will be referred to as "recognition system SYSb".
[0063]
(2-1) Configuration of recognition system SYSb
First, the configuration of the recognition system SYSb of the second embodiment will be described. The recognition system SYSb of the second embodiment differs from the recognition system SYSa of the first embodiment in that a recognition device 2b is provided instead of the recognition device 2. FIG. Other features of the recognition system SYSb may be identical to other features of the recognition system SYSa. Therefore, the configuration of the recognition device 2b of the second embodiment will be described below with reference to FIG. FIG. 6 is a block diagram showing the configuration of the recognition device 2b of the second embodiment. Constituent elements that have already been explained are denoted by the same reference numerals, and detailed explanation thereof will be omitted.
[0064]
As shown in FIG. 6, the recognition device 2b differs from the recognition device 2 described above in that an image processing unit 212b is implemented in the arithmetic device 21. Other features of recognition device 2 b may be identical to other features of recognition device 2 .
[0065]
The image processing unit 212b generates the recognition target image 200 by performing predetermined image processing on the recognition target image 100 acquired from the imaging device 1 . The recognition target image 200 is used for the recognition operation by the recognition unit 211 . That is, in the second embodiment, the recognition unit 211 performs the recognition operation using the recognition target image 200 instead of the recognition target image 100 . That is, the recognition unit 211 recognizes the recognition target image 20The face of the person reflected in 0 is recognized, and the person reflected in the recognition target image 200 is authenticated based on the recognized face.
[0066]
The content of image processing executed by the image processing unit 212b is defined by a processing parameter 303b, which is an example of the image generation parameter 300. In this case, the image processing unit 212b generates the recognition target image 200 by performing image processing, the content of which is defined by the processing parameter 303b, on the recognition target image 100. FIG. The image processing unit 212b may perform the same type of image processing as the image processing performed by the image processing unit 121 included in the photographing device 1 described above. For example, the image processing unit 212b may perform at least one of white balance correction processing, luminance correction processing, contrast correction processing, dehaze processing, HDR processing, denoising processing, and skeleton texture separation processing. In this case, similarly to the processing parameter 302 described above, the processing parameter 303b includes a parameter that defines the content of the white balance correction processing, a parameter that defines the content of the brightness correction processing, a parameter that defines the content of the contrast correction processing, and a dehaze processing. may include at least one of a parameter that defines the content of HDR processing, a parameter that defines the content of denoising processing, and a parameter that defines the content of skeletal texture separation processing. Alternatively, the image processing unit 212b may perform a different type of image processing from the image processing performed by the image processing unit 121 included in the photographing device 1 described above.
[0067]
(2-2) Operation of recognition system SYSb
Next, the operation of the recognition system SYSb of the second embodiment will be described. In particular, the parameter determination operation performed by the parameter determination device 3 will be described below. In the second embodiment, as the image generation parameters 300, in addition to the parameters (specifically, the optical parameters 301 and the processing parameters 302) related to the imaging device 1 described in the first embodiment, the parameters related to the recognition device 2 (specifically , the processing parameter 303b) is used. For this reason, the parameter determination device 3 determines the optical parameters 301 and the processing parameters 302 related to the imaging device 1 by performing the parameter determination operation of the above-described first embodiment. Determine parameters 302 . Accordingly, the parameter determination operation for determining the processing parameters 302 for the recognition device 2 will now be described with reference to FIG. FIG. 7 is a flow chart showing the flow of the parameter determination operation for determining the processing parameters 302 for the recognition device 2. As shown in FIG. In the following description, unless otherwise specified, the image generation parameter 300 means the processing parameter 303b.
[0068]
As shown in FIG. 7, the parameter setting unit 311 initializes the variable r to 0 (step S302). After that, the parameter determination unit 314 controls the photographing device 1 so as to generate a predetermined number or more of recognition target images 100 (step S321). Note that the process of step S321 may be the same as the process of step S311 described above. However, in step S321, if the optical parameter 301 and the processing parameter 302 have already been determined, the parameter determination unit 314 selects a predetermined number or more of recognition target images based on the already determined optical parameter 301 and processing parameter 302. The imaging device 1 may be controlled to generate 100 . Alternatively, when the optical parameters 301 and the processing parameters 302 have not been determined, the parameter determination unit 314 performs shooting so as to generate a predetermined number or more of the recognition target images 100 based on the default optical parameters 301 and processing parameters 302. The device 1 may be controlled.
[0069]
After the predetermined number or more of recognition target images 100 are generated, the parameter determination unit 314 sets the image generation parameter 300 of the recognition device 2 (specifically, the processing parameter 303b of the image processing unit 212b) to the initial value. (step S320). As the initial value of the image generation parameter 300, any one of a plurality of candidate values ​​of the image generation parameter 300 to be set in the recognition device 2 in the parameter determination operation may be used. A plurality of candidate values ​​may include recommended values ​​preset in the recognition device 2 .
[0070]
After that, the parameter determination unit 314 performs image processing whose contents are defined by the image generation parameters 300 set in step S320 on each of the predetermined number or more of the recognition target images 100 generated in step S321. Then, the image processing unit 212b of the recognition device 2 is controlled (step S331). As a result, the image processing unit 212b generates a predetermined number of recognition target images 200 or more.
[0071]
After that, the parameter determination unit 314 controls the recognition device 2 (particularly, the recognition unit 211) so as to execute the recognition operation described above for each of the predetermined number or more of the recognition target images 200 generated in step S331. (Step S322). Recognition result information indicating the result of the recognition operation is accumulated in the recognition result DB 222 of the storage device 22 .
[0072]
After the recognition operation for the predetermined number or more of the recognition target images 200 is completed, the recognition result acquisition unit 312 acquires recognition result information indicating the result of the recognition operation in step S322 from the recognition device 2 via the communication device 33. (Step S323).
[0073]
Thereafter, the evaluation unit 313 evaluates whether or not the image generation parameters 300 actually set (in other words, applied) to the recognition device 2 in step S322 are appropriate based on the recognition result information acquired in step S323. An evaluation value for is calculated (step S324). That is, the evaluation unit 313 calculates an evaluation value for evaluating the recognition action performed on the recognition target image 200 generated based on the image generation parameters 300 set in step S320. Note that the evaluation values ​​used in the second embodiment may be the same as the evaluation values ​​used in the first embodiment, so detailed description thereof will be omitted. That is, the process of step S324 may be the same as the process of step S314 described above. However, in the second embodiment, since the recognition operation is performed on the authentication target image 200 instead of the authentication target image 100, the matching rate is the number of recognition target images 200 in which a person's face is detected. The recall rate is the total number of recognition target images 200 in which a person is captured (that is, the total number of recognition target images 200 for which a person should be authenticated, and the recognition target It shows the ratio of the number of recognition target images 200 for which person authentication has succeeded to the total number of correct data corresponding to the recognition target image 100 that is the basis of the image 200 .
[0074]
After that, the parameter determination unit 314 determines whether or not the variable r is 0 (step S325). As a result of the determination in step S325, when it is determined that the variable r is 0 (step S325: Yes), the parameter determination unit 314 stores information about the image generation parameters 300 set in step S320 in the parameter DB 321. Record (step S327). The information about the image generation parameter 300 may include, for example, information indicating the value of the image generation parameter 300 set in step S320 and information indicating the evaluation value calculated in step S324.
[0075]
Information about the image generation parameters 300 recorded in the parameter DB 321 is the image generation parameters 300 to be used by the recognition device 2 to generate the recognition target image 200 (that is, the image generation parameters 300 to be actually set in the recognition device 2). ). Therefore, the image generation parameters 300 recorded in the parameter DB 321 are actually set in the recognition device 2 after the parameter determination operation shown in FIG. 7 is completed. That is, after the parameter determination operation shown in FIG. 7 is completed, the recognition device 2 generates the recognition target image 200 based on the image generation parameters 300 recorded in the parameter DB 321. FIG.

The scope of the claims
[Claim 1]
a calculation means for calculating an evaluation value for evaluating the recognition results based on the recognition results of the plurality of recognition target images by a recognition device that performs a recognition operation on the recognition target images;
a determining means for determining an image generation parameter used to generate the recognition target image based on the evaluation value;
A parameter determination device comprising
[Claim 2]
The recognition target image is generated by subjecting an image input from a photographing device equipped with a camera to the recognition device to the first image processing by image processing means provided in the recognition device,
The image generation parameters include first processing parameters that define the content of the first image processing
The parameter determination device according to claim 1.
[Claim 3]
　The recognition target image is input to the recognition device from an imaging device equipped with a camera,
　The image generation parameters include optical parameters that define the optical characteristics of the camera
The parameter determination device according to claim 1 or 2.
[Claim 4]
The recognition device receives the recognition target image from a photographing device including a camera and image processing means for generating the recognition target image by performing a second image processing on the image captured by the camera. is,
The image generation parameters include second processing parameters that define the content of the second image processing
The parameter determination device according to any one of claims 1 to 3.
[Claim 5]
The image generation parameters further include optical parameters that define optical characteristics of the camera,
The determining means determines the second processing parameter after determining the optical parameter
The parameter determination device according to claim 4.
[Claim 6]
The recognition device receives the recognition target image from a photographing device including a camera and image processing means for generating the recognition target image by performing a second image processing on the image captured by the camera. is,
The recognition device performs first image processing on the recognition target image input from the photographing device using image processing means included in the recognition device, and the recognition target image that has been subjected to the first image processing Recognize the target image,
The image generation parameters include a first processing parameter that defines the content of the first image processing and a second processing parameter that defines the content of the second image processing,
The determining means determines the first processing parameter after determining the second processing parameter
A parameter determination device according to any one of claims 1 to 5.
[Claim 7]
　Equipped with scene identification means for identifying the scene of the recognition target image,
The determining means determines at least one of the plurality of image generation parameters to be selected based on the identified scene, and at least one of the plurality of image generation parameters that is not selected. do not decide
A parameter determination device according to any one of claims 1 to 6.
[Claim 8]
The image generation parameters define the content of image processing to be performed to generate the recognition target image,
　Equipped with scene identification means for identifying the scene of the recognition target image,
The determining means determines the image generation parameter so that the intensity of the image processing is an intensity corresponding to the specified scene
A parameter determination device according to any one of claims 1 to 7.
[Claim 9]
calculating an evaluation value for evaluating the recognition results based on the recognition results of the plurality of recognition target images by a recognition device for recognizing the recognition target images;
Determining an image generation parameter used to generate the recognition target image based on the evaluation value;
Parameter determination method including
[Claim 10]
A non-temporary recording medium in which a computer program that causes a computer to execute a parameter determination method is recorded,
The parameter determination method is
calculating an evaluation value for evaluating the recognition results based on the recognition results of the plurality of recognition target images by a recognition device for recognizing the recognition target images;
Determining an image generation parameter used to generate the recognition target image based on the evaluation value;
A recording medium containing