JP7326772B2 - lighting equipment - Google Patents

Description

Embodiments of the present invention relate to lighting devices and lighting systems.

2. Description of the Related Art Conventionally, there is known a technique for making the color of an object look vivid, for example, by controlling the output of a light source according to the color components of image data of an illuminated object.

JP 2009-099510 A

However, the object is usually observed as a mixture of the spectral distribution characteristics of the illumination light and the spectral reflection characteristics of the surface of the object. Judging the color components.

For this reason, when the conventional technology is used, it is necessary to provide appropriate illumination light according to the spectral reflectance characteristics of the object itself. difficult to provide.

A problem to be solved by the present invention is to provide suitable illumination light according to an object.

A lighting device according to an example of an embodiment includes a light source unit, an image processing unit, and an output control unit. The light source section has a plurality of light sources. The image processing unit performs image processing on image data obtained by imaging an object illuminated by the light source unit so as to separate the spectral reflection characteristics of the object from the spectral distribution characteristics of the light source, Recognize the object. The output control unit controls the output of the light source according to the recognition result of the object. Further, the image processing unit recognizes the object using a learning model generated by machine learning using image data serving as a positive example, and estimates the spectral reflection characteristic as a recognition result of the object. After that, the spectral distribution characteristic is selected based on the spectral reflection characteristic so that the object can be viewed in a desired manner. Further, the output control section controls the output of the light source based on the spectral distribution characteristic selected by the image processing section.

ADVANTAGE OF THE INVENTION According to this invention, it becomes possible to provide suitable illumination light according to a target object.

1A is a diagram illustrating an example of a lighting system including a lighting device according to a first embodiment; FIG. FIG. 1B is a diagram showing an overview of image processing according to the first embodiment. FIG. 2 is a block diagram showing an example of the configuration of the lighting system according to the first embodiment. FIG. 3 is a schematic explanatory diagram of image processing executed by an image processing unit according to the first embodiment. 4 is a block diagram illustrating an example of the configuration of an image processing unit according to the first embodiment; FIG. FIG. 5A is a process explanatory diagram (part 1) of image processing executed by an image processing unit according to the first embodiment; 5B is a process explanatory diagram (part 2) of the image processing executed by the image processing unit according to the first embodiment; FIG. 5C is a process explanatory diagram (part 3) of image processing executed by the image processing unit according to the first embodiment; FIG. 5D is a process explanatory diagram (part 4) of image processing executed by the image processing unit according to the first embodiment; FIG. FIG. 6 is a flowchart illustrating a processing procedure of processing executed by the lighting device according to the first embodiment; FIG. 7 is a block diagram showing an example of the configuration of an image processing unit according to the second embodiment. FIG. 8A is a process explanatory diagram (part 1) of image processing executed by an image processing unit according to the second embodiment; FIG. 8B is a process explanatory diagram (part 2) of image processing executed by the image processing unit according to the second embodiment; FIG. 9 is a flowchart illustrating a processing procedure of processing executed by the lighting device according to the second embodiment. FIG. 10 is a block diagram showing an example of the configuration of an image processing unit according to the third embodiment; FIG. 11A is a process explanatory diagram (part 1) of image processing executed by an image processing unit according to the third embodiment; 11B is a process explanatory diagram (part 2) of image processing executed by the image processing unit according to the third embodiment; FIG. FIG. 12 is a flowchart illustrating a processing procedure of processing executed by the lighting device according to the third embodiment. FIG. 13 is a block diagram showing an example of the configuration of an image processing unit according to a modification of the third embodiment; FIG. 14 is a flowchart illustrating a processing procedure of processing executed by the lighting device according to the modification of the third embodiment; FIG. 15 is a block diagram showing an example configuration of a lighting system according to the fourth embodiment. FIG. 16 is a block diagram showing an example of the configuration of an image processing unit according to the fifth embodiment; FIG. 17A is a process explanatory diagram (part 1) of image processing executed by an image processing unit according to the fifth embodiment; 17B is a process explanatory diagram (part 2) of image processing executed by the image processing unit according to the fifth embodiment; FIG. 17C is a process explanatory diagram (part 3) of image processing executed by the image processing unit according to the fifth embodiment; FIG. FIG. 18 is a flowchart illustrating a processing procedure of processing executed by the lighting device according to the fifth embodiment; FIG. 19 is a diagram (part 1) showing a modification of each embodiment. FIG. 20 is a diagram (part 2) showing a modification of each embodiment. FIG. 21 is a diagram (part 3) showing a modification of each embodiment. FIG. 22 is a diagram (part 4) showing a modification of each embodiment.

The lighting devices according to the first to fifth embodiments described below include a light source section and an output control section. The light source section has a plurality of light sources. The output control unit outputs the light source according to the recognition result of the object illuminated by the light source unit or the object recognized based on the image data of the object before being illuminated, which is imaged by the image sensor. to control. Note that the illumination device may or may not include an image sensor. If not, for example, a portable communication device with a camera may be used as the image sensor, or an image sensor installed separately from the lighting device on the ceiling may be used.

Further, the illumination devices according to the first to fifth embodiments described below further include an image processing section. The image processing section executes predetermined image processing for recognizing the object based on the image data. Note that the illumination device may or may not include an image processing unit. If not, for example, an external device such as a cloud server may have an image processing unit, and the image processing unit may exchange information with the lighting device via a network such as the Internet.

Further, the image processing units according to the first to fifth embodiments described below recognize the target object using a learning model generated by machine learning using image data serving as a positive example.

Also, an image processing unit according to a fourth embodiment described below uses a learning model through cloud computing.

Further, the image processing unit according to the first to fifth embodiments described below separates the characteristics related to the color of the object and the characteristics related to the color of the light source from the image data obtained by imaging the object. , the above image processing is performed to recognize the object.

Further, the image processing unit according to the first and third embodiments described below estimates the spectral reflection characteristics as a recognition result of the object, and then, based on the spectral reflection characteristics, the object is displayed in a desired appearance. and the output control section controls the output of the light source based on the spectral distribution characteristic selected by the image processing section.

Further, the light source units according to the first to fifth embodiments described below have at least monochromatic light sources of red, green, and blue as light sources, and the image processing unit individually turns on the monochromatic light sources. The spectral reflectance characteristics are estimated based on a set of image data obtained by capturing an image of the object when

Further, the image processing unit according to the third embodiment described below uses the learning model to classify the objects in the image data captured by the image sensor, and then extracts the image area of the specific object. Then, the spectral reflection characteristic is estimated based on the image area.

Further, in the fifth embodiment described below, the object is a face, and the image processing unit extracts an image area corresponding to each part of the face from image data captured by an image sensor. Then, after estimating the spectral reflection characteristics of each part based on the image area, the lighting conditions of the light source including the spectral distribution characteristics are selected based on the combination of the spectral reflection characteristics, and the output control unit performs image processing. control the output of the light source based on the lighting conditions selected by the department;

Further, in the first to fifth embodiments described below, the image processing unit performs machine learning so that an evaluation value indicating the result of output control by the output control unit is included as one of the feature amounts.

Further, in the first to fifth embodiments described below, the machine learning is deep learning.

Further, the lighting system according to the first to fifth embodiments described below includes a light source section, an image processing section, and an output control section. The light source section has a plurality of light sources. The image processing unit executes predetermined image processing for recognizing an object based on image data of an object illuminated by the light source unit or an image of the object before being illuminated by the image sensor. The output control unit controls the output of the light source according to the recognition result of the object by the image processing.

Hereinafter, lighting devices according to first to fifth embodiments will be described with reference to the drawings. Components having the same function in each embodiment are denoted by the same reference numerals, and overlapping descriptions are omitted.

Also, the illumination devices described in the following embodiments are merely examples, and do not limit the present invention.

Moreover, below, the case where the target object to be illuminated is food such as omelet rice will be described as a main example. In the following description, an example of using deep learning, which is a type of machine learning, in image processing of image data obtained by imaging an object will be described.

(First embodiment)
First, an overview of a lighting system 1 including a lighting device 10 according to the first embodiment will be described with reference to FIGS. 1A and 1B. FIG. 1 is a diagram showing an example of a lighting system 1 including a lighting device 10 according to the first embodiment. FIG. 1B is a diagram showing an overview of image processing according to the first embodiment.

As shown in FIG. 1A, lighting system 1 includes lighting device 10 and learning device 20 (see FIG. 2), which will be described later. The illumination device 10 includes an image sensor 11 and a light source section 12 .

The image sensor 11 is a camera having a solid-state imaging device such as a CCD (Charge Coupled Device) device or a CMOS (Complementary Metal Oxide Semiconductor) device, and captures an image of an object illuminated by the illumination device 10 .

The light source unit 12 has a plurality of light sources having different spectral distribution characteristics, and is controlled by an output control unit 143 (see FIG. 2), which will be described later, to change the lighting state. Solid-state light sources such as LEDs (Light Emitting Diodes) can be used as the type of light source.

The plurality of light sources have, for example, red, green, and blue monochromatic light sources, and the output control unit 143 can control these independently and arbitrarily control dimming and toning. Also, the light source may have a four-color configuration including yellow in addition to the three primary colors of red, green, and blue. Other so-called categorical colors such as orange, purple, cyan, amber, magenta, and peach may also be included.

In the lighting system 1 according to the first embodiment configured as described above, as shown in FIG. Then, the omelet rice) is imaged (step S1).

Then, the illumination device 10 acquires image data captured by the image sensor 11 (step S2). Then, the illumination device 10 executes image processing using machine learning based on the acquired image data (step S3).

Here, as shown in FIG. 1B, the illumination device 10 is arranged such that the spectral distribution characteristic and the spectral reflection characteristic are separated for image data observed as the product of the spectral distribution characteristic of the light source and the spectral reflection characteristic of the object. Then, image processing is performed on the object, and the object is recognized, for example, so that the spectral reflectance characteristics of the object itself can be grasped. Note that the spectral distribution characteristic of the light source is an example of the characteristic related to the color of the light source. Therefore, it is not limited to the spectral distribution characteristic, and may be the fluorescence component of the light source, the light color (chromaticity, correlated color temperature, duv, color rendering index), or the like. Also, the spectral reflectance characteristics of the object are an example of characteristics related to the color of the object. Therefore, it is not limited to spectral reflection characteristics, and may be fluorescent components, light colors (chromaticity, correlated color temperature, duv, color rendering index), and the like. Note that both the characteristics regarding the color of the light source and the characteristics regarding the color of the object may include information on the photometric quantity (illuminance, brightness, etc.). In the following description, the spectral distribution characteristics of a light source and the spectral reflection characteristics of an object will be used as main examples.

In this image processing, the illumination device 10 according to the present embodiment uses a learning model 131 (see FIG. 2) generated by a machine learning algorithm such as deep learning to estimate the spectral reflection characteristics of the object. . The learning model 131 is generated in advance by the learning device 20, for example. Details of such image processing will be described later with reference to FIGS. 2 to 5D.

Returning to the description of FIG. 1A. Then, the illumination device 10 controls the output of the light source based on the recognition result of the target object by the image processing (step S4). For example, the illumination device 10 selects the spectral distribution characteristic of the light source suitable for the object based on the spectral reflection characteristic of the object estimated in step S3, and controls the output of each light source of the light source unit 12 accordingly.

For example, if the object is omelet rice, the illumination device 10 selects the spectral distribution characteristics of the light sources and controls the output of each light source so that the omelet rice looks more colorful and glossy.

Therefore, according to the illumination system 1 according to the present embodiment, it is possible to provide appropriate illumination light according to the object. In addition, if the object is food, for example, the object can be made to look more appetizing, thereby providing the user located in the lighting environment of the lighting device 10 with an appropriate lighting environment corresponding to the object. becomes possible.

An example of the configuration of the lighting system 1 according to the first embodiment will be described more specifically below. FIG. 2 is a block diagram showing an example of the configuration of the lighting system 1 according to the first embodiment. It should be noted that FIG. 2 shows only the constituent elements necessary for explaining the features of the embodiment, and omits the description of general constituent elements.

In other words, each component illustrated in FIG. 2 is functionally conceptual and does not necessarily need to be physically configured as illustrated. For example, the specific form of distribution/integration of each block is not limited to the one shown in the figure. It is possible to integrate and configure.

In addition, in the description using FIG. 2, the description of components that have already been described may be simplified or omitted.

As shown in FIG. 2, the lighting system 1 according to the first embodiment includes a lighting device 10 and a learning device 20. As shown in FIG. The illumination device 10 includes an image sensor 11 , a light source section 12 , a storage section 13 and a control section 14 .

Since the image sensor 11 and the light source unit 12 have already been described, descriptions thereof will be omitted here. The storage unit 13 is realized by, for example, a semiconductor memory device such as a RAM (Random Access Memory) or a flash memory, or a storage device such as a hard disk or an optical disc. In the example of FIG. do.

The learning model 131 is generated using a machine learning algorithm such as deep learning as described above, and includes, for example, an estimation model for estimating spectral reflectance characteristics of an object in an image based on image data.

The learning model 131 is generated, for example, by a learning device 20 provided separately from the lighting device 10, and set in advance in the storage unit 13 before shipment of the lighting device 10 or the like. Instead of providing the learning device 20, the control unit 14 of the illumination device 10 may be provided with a processing unit having a function corresponding to the learning device 20, for example.

The control unit 14 is a controller, and various programs stored in a storage device inside the lighting device 10 are executed by a CPU (Central Processing Unit), an MPU (Micro Processing Unit), or the like, using the RAM as a work area. It is realized by being executed. Also, the control unit 14 can be implemented by an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

The control unit 14 has an acquisition unit 141, an image processing unit 142, and an output control unit 143, and implements or executes information processing functions and actions described below.

The acquisition unit 141 acquires image data captured by the image sensor 11 . In the present embodiment, the light source unit 12 individually lights each monochromatic light source of the three primary colors of red, green, and blue, for example, and the image sensor 11 detects an object when illuminated by each of these monochromatic light sources. We will take an image.

Then, the acquiring unit 141 acquires image data of the object illuminated by each monochromatic light source. Note that the image data of the object illuminated with red light, green light, and blue light may be hereinafter referred to as a "red image," a "green image," and a "blue image," respectively.

The image processing unit 142 executes predetermined image processing for recognizing an object based on the image data acquired by the acquisition unit 141 . An example of the configuration of the image processing unit 142 according to the first embodiment will now be described more specifically with reference to FIGS. 3 to 5D.

FIG. 3 is a schematic explanatory diagram of image processing executed by the image processing unit 142 according to the first embodiment. Also, FIG. 4 is a block diagram showing an example of the configuration of the image processing unit 142 according to the first embodiment. 5A to 5D are processing explanatory diagrams (part 1) to (part 4) of image processing executed by the image processing unit 142 according to the first embodiment.

As shown in FIG. 3, the image processing unit 142 inputs the image data acquired by the acquisition unit 141 to the learning model 131, and reflects the data in the light source unit 12 based on the output value of the learning model 131 output as a result. Estimate the spectral distribution characteristics of the light source to be used.

Note that the learning model 131 can include a plurality of models NN-1, NN-2, NN-3, . . . as illustrated as a plurality of neural networks in FIG.

Specifically, as shown in FIG. 4, the image processing unit 142 has a spectral reflection characteristic estimation unit 142a and a selection unit 142b.

The model NN-1 described above is, for example, a model for estimating the spectral reflectance characteristics of an object. Estimate spectral reflectance characteristics.

The selection unit 142b selects the spectral distribution characteristic of the light source suitable for the object based on the spectral reflection characteristics estimated by the spectral reflection characteristics estimation unit 142a.

Here, the case where the model NN-1 is used as the model for estimating the spectral reflectance characteristics as described above will be described more specifically. As shown in FIG. 5A, at the time of learning when such a model NN-1 is first generated, a set of red images I_R, green images I_G, and blue images I_B of various objects, and Each data of the spectral reflectance characteristic expected as an output value when such each set is input is prepared. That is, the learning data set is a set of known data that are positive examples in so-called supervised learning.

Then, for example, the learning device 20 described above generates the model NN-1 by executing machine learning using the learning data set. As a result, the model NN-1 becomes, for example, a neural network for estimating spectral reflection characteristics corresponding to unknown image data.

Specifically, as shown in FIG. 5B, model NN-1 receives unknown image data acquired by acquisition unit 141, that is, a set of red image I_R, green image I_G, and blue image I_B, at the time of estimation. , the spectral reflectance characteristic corresponding to this is estimated and output.

Then, the spectral reflection characteristic estimating section 142a outputs the estimated spectral reflection characteristic to the selecting section 142b, and the selecting section 142b selects the spectral distribution characteristic of the light source based on this.

The appearance of the object can be derived, for example, as the product of the spectral reflection characteristics of the object and the spectral distribution characteristics of the light source. The spectral distribution characteristic of the light source may be calculated so as to obtain a desired appearance such as a more glossy appearance.

Further, the selection unit 142b may estimate the spectral distribution characteristics of the light source using the learning model 131, for example, based on the obtained spectral reflectance characteristics of the target object, and select them. Here, an estimation model for estimating the spectral distribution characteristics of a light source based on, for example, the spectral reflection characteristics of an object will be described as model NN-2.

As shown in FIG. 5C, when the model NN-2 is first generated and learned, the spectral reflectance characteristics of various objects and the respective spectral reflectance characteristics are input as the learning data set. Each data of the spectral distribution characteristic expected as an output value is prepared.

Then, for example, the learning device 20 described above generates the model NN-2 by executing machine learning using the learning data set. As a result, the model NN-2 becomes, for example, a neural network for estimating the spectral distribution characteristic corresponding to the unknown spectral reflection characteristic.

Specifically, as shown in FIG. 5D, the model NN-2 estimates the spectral distribution characteristics corresponding to the spectral reflection characteristics estimated by the spectral reflection characteristics estimation unit 142a at the time of estimation. and output.

Then, the selecting unit 142b selects the spectral distribution characteristic of the light source based on the estimated spectral distribution characteristic.

Returning to the description of FIG. The output control unit 143 controls the output of the light source according to the recognition result of the object by the image processing in the image processing unit 142 . That is, the output control unit 143 controls the output state of each light source of the light source unit 12 based on the spectral distribution characteristics of the light sources selected by the image processing unit 142 .

Specifically, the output control unit 143 inputs a PWM (Pulse Width Modulation) waveform or DC waveform current to the light source unit 12 based on the spectral distribution characteristics of the light source selected by the image processing unit 142, It controls the lighting state of each of the 12 light sources. It should be noted that the control here means not only simple dimming or toning, but also control for changing along the time axis, such as gradually brightening or darkening.

Next, a processing procedure of processing executed by the lighting device 10 according to the first embodiment will be described with reference to FIG. 6 . FIG. 6 is a flowchart showing a processing procedure of processing executed by the lighting device 10 according to the first embodiment.

As shown in FIG. 6, first, the acquisition unit 141 acquires image data from the image sensor 11 (step S101). Based on the image data, the spectral reflection characteristic estimation unit 142a estimates the spectral reflection characteristic of the object in the image using the learning model 131 (step S102).

Then, the selecting unit 142b selects the spectral distribution characteristic of the light source based on the estimated spectral reflection characteristic (step S103). Then, the output control unit 143 controls the output of the light source based on the selected spectral distribution characteristic (step S104), and ends the process.

As described above, the illumination device 10 according to the first embodiment includes the light source unit 12, the image sensor 11, the image processing unit 142, and the output control unit 143. The light source unit 12 has a plurality of light sources. The image sensor 11 images an object illuminated by the light source unit 12 . The image processing unit 142 executes predetermined image processing for recognizing the object based on image data captured by the image sensor 11 . The output control unit 143 controls the output of the light source according to the object recognition result obtained by the image processing. This makes it possible to provide suitable illumination light for the object.

(Second embodiment)
By the way, in the above-described first embodiment, the case where the learning model 131 is used to estimate the spectral reflectance characteristics of the object in the image based on the image data captured by the image sensor 11 was taken as an example. Similarly, based on the image data, the learning model 131 may be used to estimate the spectral distribution characteristics of the light source in the image.

An example of such a case will be described with reference to FIGS. 7 to 8B. FIG. 7 is a block diagram showing an example of the configuration of the image processing section 142A according to the second embodiment. 8A and 8B are process explanatory diagrams (part 1) and (part 2) of image processing executed by the image processing unit 142A according to the second embodiment.

Since FIG. 7 corresponds to FIG. 4 which has already been shown, different parts will be mainly described here.

As shown in FIG. 7, the image processing unit 142A according to the second embodiment is different from the case of FIG. 4 in that it further includes a spectral distribution characteristic estimation unit 142c.

Based on the image data acquired by the acquisition unit 141, the image processing unit 142A estimates the spectral distribution characteristics of the light source in the image using, for example, the model NN-3, which is an estimation model of the spectral distribution characteristics of the light source.

Specifically, as shown in FIG. 8A, at the time of learning when the model NN-3 is first generated, image data obtained by imaging various objects and such image data are used as learning data sets. Each data of the spectral distribution characteristic expected as an output value when is input is prepared.

Then, for example, the learning device 20 described above generates the model NN-3 by executing machine learning using the learning data set. As a result, the model NN-3 becomes, for example, a neural network for estimating the spectral distribution characteristics corresponding to unknown image data.

Specifically, as shown in FIG. 8B, when the model NN-3 receives unknown image data acquired by the acquisition unit 141 during estimation, it estimates the corresponding spectral distribution characteristic, Output.

Then, the spectral distribution characteristic estimation unit 142c outputs the estimated spectral distribution characteristics to the selection unit 142b, and the selection unit 142b outputs the spectral distribution characteristics and the spectrum of the object estimated by the spectral reflection characteristics estimation unit 142a. Based on the reflection characteristics, the spectral distribution characteristics of the light source are selected so that the object can be viewed in a desired manner.

In this way, by estimating the spectral distribution characteristics of the light source based on the image data acquired by the acquisition unit 141, it is also possible to provide appropriate illumination light according to the object.

Next, a processing procedure of processing executed by the lighting device 10A according to the second embodiment will be described with reference to FIG. FIG. 9 is a flowchart showing a processing procedure of processing executed by the lighting device 10A according to the second embodiment.

As shown in FIG. 9, first, the acquisition unit 141 acquires image data from the image sensor 11 (step S201). Based on the image data, the spectral reflection characteristic estimation unit 142a estimates the spectral reflection characteristic of the object in the image using the learning model 131 (step S202).

On the other hand, the spectral distribution characteristic estimation unit 142c estimates the spectral distribution characteristic of the light source in the image using the learning model 131 based on the same image data (step S203).

Then, the selecting unit 142b selects the spectral distribution characteristic of the light source that provides the desired appearance based on the estimated characteristics (step S204). Then, the output control unit 143 controls the output of the light source based on the selected spectral distribution characteristic (step S205), and the process ends.

As described above, the image processing unit 142A according to the second embodiment further includes the spectral distribution characteristic estimation unit 142c. The spectral distribution characteristic estimation unit 142c estimates the spectral distribution characteristic of the light source in the image based on the image data acquired by the acquisition unit 141. FIG. As a result, it is possible to select the spectral distribution characteristics of the light source that give the desired appearance of the object, based on the estimated spectral distribution characteristics of the light source in the image. It is possible to provide light.

(Third Embodiment)
By the way, in the first and second embodiments described above, the learning model 131 uses the models NN-1 and NN-2, which are estimation models for estimating the spectral reflectance characteristics of an object in an image or the spectral distribution characteristics of a light source. Examples of using these are given. Here, the learning model 131 may further include a classification model for classifying the objects in the image, and estimate the spectral reflectance characteristics of the object or the spectral distribution characteristics of the light source based on the classification result using this model. good.

An example of such a case will be described with reference to FIGS. 10 to 11B. FIG. 10 is a block diagram showing an example of the configuration of the image processing section 142B according to the third embodiment. 11A and 11B are process explanatory diagrams (part 1) and (part 2) of image processing executed by the image processing unit 142B according to the third embodiment.

Since FIG. 10 corresponds to FIGS. 4 and 7 which have already been shown, different parts will be mainly described here.

As shown in FIG. 10, the image processing unit 142B according to the third embodiment differs from the cases of FIGS. 4 and 7 in that it further includes a classification unit 142d and an extraction unit 142e.

The image processing unit 142B classifies the objects in the image based on the image data acquired by the acquisition unit 141, for example, using the model NN-4, which is a classification model for classifying the objects in the image. Then, the image processing unit 142B extracts a specific object based on the classification result, and estimates the spectral reflection characteristics of the extracted object or the spectral distribution characteristics of the light source.

Note that the model NN-4, which is a classification model, can be generated by the learning device 20 described above, for example, using an object classification algorithm such as semantic segmentation in deep learning. As a result, the model NN-4 becomes a neural network capable of classifying objects in an image, for example, in units of pixels, for unknown image data.

Then, as shown in FIG. 11A, the classification unit 142d inputs the unknown image data acquired by the acquisition unit 141 to the model NN-4 at the time of classification. Then, the model NN-4 classifies various objects reflected in the image data and outputs the classification results.

For example, in the example of FIG. 11A, various objects in the image data are classified into "table", "omelet rice", "plate", "sauce", "broccoli", and "wall". there is

Then, the extraction unit 142e extracts a specific object based on the classification result of the classification unit 142d. Here, the extracting unit 142e extracts the image area of "rice omelette", for example, if the illumination device 10B mainly illuminates the main dish on the table. Alternatively, the extraction unit 142e may simply extract, for example, an object that occupies the largest number of pixels in the image as the main specific object.

Then, based on the image area of the specific object extracted by the extracting unit 142e, the spectral reflection characteristic estimation unit 142a estimates the spectral reflection characteristics of the specific object by the method shown in FIG. 5B, for example. Also, the spectral distribution characteristic estimator 142c estimates the spectral distribution characteristic of the light source, for example, by the method shown in FIG. 5D, based on the estimated spectral reflection characteristic of the specific object. Alternatively, the spectral distribution characteristic estimation unit 142c estimates the spectral distribution characteristic of the light source, for example, by the method shown in FIG. 8B, based on the image area of the specific object extracted by the extraction unit 142e.

Then, based on the estimated spectral distribution characteristics, the selecting unit 142b selects the spectral distribution characteristics of the light source that make the specific object appear in a desired manner.

Note that the selecting unit 142b selects a spectral distribution characteristic that makes a specific object look as desired based on preset information of the spectral distribution characteristic that is set in advance for each specific object, as shown in FIG. 11B, for example. You may make it When such preset information is provided, for example, the image processing section 142B can be composed of only the classifying section 142d, the extracting section 142e, and the selecting section 142b.

In this way, by classifying the target objects in the image based on the image data acquired by the acquisition unit 141 and extracting a specific target object, it is possible to provide appropriate illumination light according to the target object. It becomes possible.

Next, a processing procedure of processing executed by the lighting device 10B according to the third embodiment will be described with reference to FIG. 12 . FIG. 12 is a flow chart showing a processing procedure of processing executed by the lighting device 10B according to the third embodiment.

As shown in FIG. 12, first, the acquisition unit 141 acquires image data from the image sensor 11 (step S301). Based on the image data, the classification unit 142d classifies the object in the image using the learning model 131 (step S302). Also, the extraction unit 142e extracts a specific object in the image based on the classification result (step S303).

Based on the extracted image area of the specific object, the spectral reflection characteristic estimation unit 142a estimates the spectral reflection characteristics of the specific object using the learning model 131 (step S304).

Then, the spectral distribution characteristic estimating unit 142c estimates the spectral distribution characteristic of the light source based on the estimated spectral reflection characteristic, and the selecting unit 142b selects the spectral distribution characteristic of the light source that gives a desired appearance based on each estimated characteristic. A distribution characteristic is selected (step S305). Then, the output control unit 143 controls the output of the light source based on the selected spectral distribution characteristic (step S306), and the process ends.

As described above, the image processing unit 142B according to the third embodiment further includes the classification unit 142d and the extraction unit 142e. The classification unit 142 d classifies objects in the image based on the image data acquired by the acquisition unit 141 . The extraction unit 142e extracts specific objects based on the classification result of the classification unit 142d. Then, the spectral reflection characteristic estimator 142a estimates the spectral reflection characteristic of the specific object based on the extraction result of the extraction section 142e. Also, the spectral distribution characteristic estimator 142c estimates the spectral distribution characteristic of the light source based on the estimation result of the spectral reflection characteristic estimator 142a. As a result, it is possible to select the spectral distribution characteristic of the light source that gives the desired appearance of the object, and thus it is possible to provide suitable illumination light according to the object.

(Modification of the third embodiment)
By the way, in the third embodiment, the case where the classifying unit 142d classifies objects in the image using the learning model 131 based on the image data acquired by the acquiring unit 141 is taken as an example. Based on the data, the learning model 131 may be used by the spectral distribution characteristic estimator 142c to estimate the spectral reflection characteristic of the light source in the image.

Such a modification will be described with reference to FIG. 13 . FIG. 13 is a block diagram showing an example configuration of an image processing unit 142B' according to a modification of the third embodiment. Since FIG. 13 corresponds to FIG. 10 already shown, different parts will be mainly described here.

As shown in FIG. 13, the image processing unit 142B' according to the modification of the third embodiment uses the learning model 131 based on the image data acquired by the acquisition unit 141 to classify the classification unit 142d into an image. 10 in that while the object is classified, the spectral distribution characteristic estimator 142c estimates the spectral distribution characteristic of the light source based on the same image data.

Note that the spectral distribution characteristic estimator 142c estimates the spectral distribution characteristic of the light source here, for example, using the method shown in FIG. 8B. Then, the selecting unit 142b selects the spectral reflectance characteristics of the specific object estimated from the classifying unit 142d through the extracting unit 142e and the spectral reflection characteristics estimating unit 142a, and the light source in the image estimated by the spectral distribution characteristics estimating unit 142c. , the spectral distribution characteristic of the light source that provides the desired appearance is selected.

In this way, by classifying the objects in the image based on the image data acquired by the acquisition unit 141, extracting a specific object, and estimating the spectral distribution characteristics of the light source based on the same image data, It is also possible to provide suitable illumination light according to the object.

Next, a processing procedure of processing executed by the illumination device 10B' according to the modified example of the third embodiment will be described with reference to FIG. FIG. 14 is a flowchart showing a processing procedure of processing executed by the lighting device 10B' according to the modification of the third embodiment.

As shown in FIG. 14, first, the acquisition unit 141 acquires image data from the image sensor 11 (step S401). Based on the image data, the classification unit 142d classifies the object in the image using the learning model 131 (step S402). Also, the extraction unit 142e extracts a specific object in the image based on the classification result (step S403).

Based on the extracted image area of the specific object, the spectral reflection characteristic estimation unit 142a estimates the spectral reflection characteristics of the specific object using the learning model 131 (step S404).

On the other hand, the spectral distribution characteristic estimation unit 142c uses the learning model 131 to estimate the spectral distribution characteristic of the light source in the image based on the image data acquired by the acquisition unit 141 (step S405).

Then, the selecting unit 142b selects the spectral distribution characteristic of the light source that provides the desired appearance based on the estimated characteristics (step S406). Then, the output control unit 143 controls the output of the light source based on the selected spectral distribution characteristic (step S407), and the process ends.

In the third embodiment, semantic segmentation is used as an object classification algorithm, but the present invention is not limited to this. Networks) etc. may be used.

(Fourth embodiment)
By the way, the case where the lighting devices 10, 10A, 10B, and 10B' have the learning model 131 has been exemplified so far. may be provided above.

An example of such a case will be described with reference to FIG. FIG. 15 is a block diagram showing an example of the configuration of a lighting system 1A according to the fourth embodiment. Note that FIG. 15 corresponds to FIG. 2, which has already been shown, and therefore different parts will be mainly described here.

As shown in FIG. 15, in a lighting system 1A according to the fourth embodiment, a lighting device 10C has a communication unit 15 and can access a cloud server 100 via a network N such as the Internet or a mobile phone network. 2 is different from the case of FIG. Also different from the case of FIG. 2 is that the cloud server 100 has the learning model 131 instead of the lighting device 10C.

The communication unit 15 is implemented by, for example, a NIC (Network Interface Card) or the like. The communication unit 15 is connected to the network N by wire or wirelessly, and transmits and receives information to and from the cloud server 100 via the network N.

Specifically, when the learning model 131 is used, the lighting device 10C transmits image data acquired by the acquisition unit 141 to the cloud server 100, for example. The cloud server 100 inputs the image data received from the lighting device 10C to the learning model 131, and returns the resulting output value to the lighting device 10C. Based on these results, the illumination device 10C estimates, for example, the spectral reflection characteristics of the object, the spectral distribution characteristics of the light source, and extracts a specific object in the image.

In addition, the cloud server 100 has a learning function corresponding to the learning device 20 described above, collects image data as appropriate from various devices (for example, the lighting device 10C) that can access the cloud server 100, and adds data based on this. The learning model 131 can be updated at any time by learning or the like.

Thus, by providing the learning model 131 in the cloud server 100 as well, it is possible to provide appropriate illumination light according to the target object. Further, the cloud server 100 updates the learning model 131 as needed based on big data of image data collected from various devices including the illumination device 10C, so that the learning model 131 can be made highly accurate.

(Fifth embodiment)
By the way, so far, the case where the object is food has been explained as an example, but the object is not limited to this, and for example, the face of a person may be the object. The impression of a person's face is influenced by skin color, hair color, eye color, etc., and there is also the effect of color illusion due to the combination of these colors. is considered to have individual differences. Also, in recent years, attention has been focused on colors that suit individuals, so-called personal colors, and it is thought that there will be an increasing need for providing illumination light under lighting conditions that allow such personal colors to shine.

Therefore, the above-described learning model 131 further includes a recognition model for recognizing a person's face in an image, and based on the recognition result using this, estimates the spectral reflection characteristics of each part of the person's face, and estimates the spectral reflection characteristics of each part of the person's face. The lighting condition of the light source may be estimated based on the combination of characteristics.

An example of such a case will be described with reference to FIGS. 16 to 17C. FIG. 16 is a block diagram showing an example configuration of an image processing unit 142C according to the fifth embodiment. 17A to 17C are process explanatory diagrams (part 1) to (part 3) of image processing executed by the image processing unit 142C according to the fifth embodiment.

Since FIG. 16 corresponds to FIG. 10 already shown, different parts will be mainly described here.

As shown in FIG. 16, an image processing unit 142C according to the fifth embodiment does not have a classification unit 142d, and has an illumination condition estimation unit 142f instead of the spectral distribution characteristics estimation unit 142c. This is different from the case of FIG.

The image processing unit 142C recognizes the face of the person in the image based on the image data acquired by the acquisition unit 141, for example, using the model NN-5, which is a recognition model for recognizing the face of the person in the image. Then, the image processing unit 142C estimates the spectral reflection characteristics based on the recognition result, and estimates the illumination condition of the light source based on the combination of the spectral reflection characteristics. The “illumination conditions” referred to here include characteristics related to the color of the light source, including the spectral distribution characteristics described above. It also includes characteristics related to the output of the light source, such as the timing of starting or ending control of the light source.

Note that the model NN-5, which is a recognition model, is preferably included in the learning model 131 provided by the cloud server 100 according to the above-described fourth embodiment, for example, from the viewpoint of high accuracy.

As a result, the model NN-5 can be generated using a face recognition algorithm using deep learning or the like, based on big data of face images appropriately collected from various devices that can access the cloud server 100. For unknown image data, it becomes a neural network capable of recognizing the face of a person in the image and even each part of the face with high accuracy.

Then, as shown in FIG. 17A, the extraction unit 142e inputs the unknown image data acquired by the acquisition unit 141 to the model NN-5 during face recognition. Then, the model NN-5 recognizes a person's face in the image, extracts the person's face and image regions corresponding to each part of the face, and outputs the extraction results.

The example of FIG. 17A shows a case where image areas corresponding to a person's “face”, “hair”, “eyes”, “skin”, and “eyebrows” are extracted from the image data. Note that an image region corresponding to lips or the like may be extracted as another part of a person's face in which color features are likely to appear.

Then, based on the image area of each part extracted by the extraction part 142e, the spectral reflection characteristic estimating part 142a estimates the spectral reflection characteristic of each part, for example, by the method shown in FIG. 5B.

Then, based on the estimated combination of spectral reflectance characteristics of each part, the illumination condition estimation unit 142f uses preset information set in advance for the combination of spectral reflectance characteristics of each part, as shown in FIG. 17B, for example. , to estimate a suitable personal color for each combination.

Then, based on the estimated personal color, the lighting condition estimation unit 142f extracts the lighting conditions that make the personal color shine, which are associated with each personal color in advance as preset information, for example, as shown in FIG. 17C.

Then, the selection unit 142b selects the illumination conditions extracted by the illumination condition estimation unit 142f, and causes the output control unit 143 to control the output of the light source based on the illumination conditions.

Note that in FIGS. 17B and 17C, the combination of the spectral reflection characteristics of each part and the illumination condition are linked via the personal color. lighting conditions may be linked as preset information.

Alternatively, instead of using the preset information, for example, the learning model 131 may be caused to estimate the optimum lighting conditions for the combination of the spectral reflectance characteristics of each part, for example, in the same manner as the method shown in FIG. 5D. good.

In this way, based on the image data acquired by the acquisition unit 141, the human face in the image is recognized, the spectral reflection characteristics are estimated based on the recognition result, and the light source is determined based on the combination of the spectral reflection characteristics. By estimating the lighting conditions, it is possible to provide suitable lighting according to the face of a person present in the lighting environment.

Note that the number of elements of the combination of spectral reflection characteristics of each part of the face may be, for example, one. In such a case, the illumination conditions of the light source may be estimated based on the spectral reflectance characteristics of a part of the person's face where the color features are most likely to appear, for example, the color of the skin.

Also, when considering the personal color mentioned above, the personal color is generally estimated based on whether the skin is yellowish or bluish. Therefore, when the image data is acquired by the acquisition unit 141, the light source unit 12 individually turns on the monochromatic light sources of yellow and blue, for example, and the image sensor 11 detects the person when illuminated by the respective monochromatic light sources. may be imaged.

Next, a processing procedure of processing executed by the lighting device 10D according to the fifth embodiment will be described with reference to FIG. FIG. 18 is a flowchart showing a processing procedure of processing executed by the lighting device 10D according to the fifth embodiment.

As shown in FIG. 18, first, the acquisition unit 141 acquires image data from the image sensor 11 (step S501). Based on the image data, the extraction unit 142e extracts the face of the person in the image using the learning model 131 (step S502). Also, the extraction unit 142e extracts each part of the face from the extracted face of the person (step S503).

Then, the spectral reflection characteristic estimation unit 142a estimates the spectral reflection characteristic of each part using the learning model 131 based on the extracted image area of each part (step S504).

Then, the illumination condition estimation unit 142f estimates illumination conditions based on the estimated combination of the spectral reflection characteristics of each part, and the selection unit 142b selects such illumination conditions (step S505). Then, the output control unit 143 controls the output of the light source based on the selected lighting conditions (step S506), and ends the process.

It should be noted that although human faces have been used as examples so far, for example, animal faces may also be used. In such a case, for example, in pet shops, zoos, etc., it is possible to provide illumination light suitable for the facial color characteristics of animals.

As described above, in the image processing unit 142C according to the fifth embodiment, the extraction unit 142e extracts an image region corresponding to each part of the face in the image based on the image data acquired by the acquisition unit 141. do. Then, the spectral reflection characteristic estimator 142a estimates the spectral reflection characteristic of each part based on the extraction result of the extraction section 142e. Also, the illumination condition estimation unit 142f estimates the illumination condition of the light source based on the estimation result of the spectral reflection characteristic estimation unit 142a. As a result, it is possible to select illumination conditions that make people, animals, etc. look good, for example, so that it is possible to provide appropriate illumination light according to the object.

In each of the embodiments described so far, the timing for analyzing the image data captured by the image sensor 11 may be at regular intervals, or at any timing (for example, when the user captures and sends an image). timing), or timing when an object moves (for example, timing when an object such as a person is replaced). Further, detailed analysis may be performed only at the timing when the object moves.

(Modified example of each embodiment)
Next, modified examples of the embodiments described so far will be described with reference to FIGS. 19 to 22. FIG. 19 to 22 are diagrams (part 1) to (part 4) showing modifications of each embodiment. In the explanation using FIGS. 19 to 22, examples of various application scenes of the embodiments explained so far are given.

First, FIG. 19 shows an example in which a lighting device 10C according to the fourth embodiment is installed in a "fitting room" such as a clothing store. As shown in FIG. 19, when the lighting device 10C is provided in the "fitting room", the lighting device 10C does not need to be equipped with the image sensor 11 as in the past. A camera or a camera-equipped mobile terminal (such as a smart phone) carried by a store clerk or a user (customer trying on clothes) may be used as an image sensor.

In this case, after wearing the clothes to be tried on in the "fitting room", the user captures an image of himself/herself reflected in the mirror using, for example, a smartphone carried by the user, and transmits the image data to the cloud server 100 described above. (step S11).

Then, the cloud server 100 analyzes the image data using the learning model 131 (step S12), and estimates the spectral reflection characteristic or the spectral distribution characteristic as before. That is, the timing at which the cloud server 100 performs such analysis is an arbitrary timing at which the user has transmitted the image data.

Note that the learning model 131 in such a case includes, for example, a recognition model that recognizes the clothing worn by the user in the image, and estimates the spectral reflection characteristics of each clothing based on the recognition result using this, The illumination condition of the light source is estimated based on the combination of such spectral reflectance characteristics.

For example, as shown in the figure, the learning model 131 recognizes each clothing item and each color such as "skirt", "dress", and "uniform", and also recognizes "casual", "chic", " Estimate the "atmosphere" such as "active". Similarly, the learning model 131 estimates "assumed scenes" such as "cafe", "party", and "stadium". For example, the user himself/herself may arbitrarily specify such "atmosphere" and "assumed scene" from a smartphone or the like via a dedicated application or the like provided by a clothing store.

Then, cloud server 100 estimates optimal lighting conditions including spectral distribution characteristics from each combination of clothing, atmosphere, and assumed scene thus recognized or estimated, and transmits the analysis results to lighting device 10C (step S13). Then, the lighting device 10C performs output control based on the received analysis result, and reproduces in the "fitting room" a lighting environment suitable for the clothes worn by the user, the atmosphere, and the assumed scene. Although FIG. 19 exemplifies a "fitting room" for clothes, etc., the same method can be applied to a photo studio or the like.

Next, FIG. 20 shows an example in which the illumination device 10D according to the fifth embodiment is installed in a cosmetic corner of a "cosmetics store". As described above, in the image processing unit 142C, the illumination device 10D extracts an image region corresponding to each part of the face in the image based on the image data acquired by the acquisition unit 141, and based on the extraction result, By estimating the spectral reflection characteristics of each part of the face and estimating the illumination condition of the light source based on the combination of the spectral reflection characteristics, appropriate illumination light corresponding to the face of the person in the lighting environment is provided.

Therefore, when the illumination device 10D is installed in a cosmetic corner or the like, the illumination range of the light source unit 12 is the face of the user (customer of the cosmetic store) who is in front of the mirror of the cosmetic corner, and the user's face is also the image sensor. 11 imaging ranges.

Then, as shown in FIG. 20, the illumination device 10D captures an image of the user's face (step S21), analyzes the image data (step S22), and performs output control under illumination conditions based on the analysis result (step S23). ) at regular intervals. Only when the image sensor 11 detects that a person is sitting in front of the mirror or that the person has changed places, the series of steps S21 to S23 is executed at least once. may

In the example of FIG. 20, the illumination device 10D estimates the illumination condition of the light source based on the combination of the spectral reflection characteristics of the colors of each part of the face, such as the user's skin, eyebrows, and lips. However, for example, depending on the type of cosmetics that the user tries, a lighting environment may be provided in which the color of the cosmetics shines.

For example, when the user tries blush, the spectral reflectance characteristics of the skin, and when the user tries lipstick, the spectral reflectance characteristics of the lips are used as main elements for extracting the lighting conditions. It is also possible to estimate a lighting condition in which the colors are bright.

Alternatively, instead of one lighting condition, for example, when trying one lipstick, it may be possible to try lighting conditions corresponding to various atmospheres and assumed scenes, as shown in FIG. Designation such as which parts are to be highlighted or what lighting conditions are to be tested may be made by the store clerk or the user via a terminal or the like provided in advance in the cosmetic corner.

In addition, although FIG. 20 exemplifies the “cosmetics store”, the same method can be applied to stage studios, beauty salons, company reception desks, and the like.

Next, FIG. 21 shows an example in which the illumination devices 10, 10A, 10B, 10B', and 10C according to the first to fourth embodiments are installed in a "conveyor-belt sushi restaurant." Note that the illumination device 10 is illustrated here.

As shown in FIG. 21, in a "conveyor-belt sushi restaurant", when a certain user (a customer of a conveyor belt sushi restaurant) picks up a plate (hereinafter referred to as "user's vicinity") is the irradiation range of the light source unit 12. The illumination device 10 is provided so that the imaging range of the image sensor 11 is in front of the irradiation range of the light source unit 12 in the conveying direction of the sushi plate.

After that, the image sensor 11 takes an image of the sushi in front of it (step S31), and the lighting device 10 analyzes the image data using the learning model 131, and detects the spectral reflection characteristic or the spectral distribution characteristic as before. to estimate

The learning model 131 in the case of such a "conveyor-belt sushi restaurant" includes a recognition model for recognizing sushi toppings such as "shrimp", "tuna", and "squid" in an image, as shown in FIG. Based on the recognition results used, the spectral reflection characteristics or spectral distribution characteristics of each piece of sushi are estimated, and the illumination conditions of the light source are estimated based on these characteristics. In the case of a "conveyor-belt sushi restaurant," such lighting conditions are, for example, lighting conditions that make each piece of sushi look fresh and delicious.

Then, the light source unit 12 illuminates the vicinity of the user based on the illumination conditions thus estimated (step S32). That is, in the example of FIG. 21, unlike the imaging range of the image sensor 11 and the irradiation range of the light source unit 12, the sushi in the image captured by the image sensor 11 is actually captured by the light source unit 12 under the estimated illumination conditions. There is a time lag until irradiation.

In the example of FIG. 21, the main object in the image is the sushi topping. may be the main object in the image. In such a case, it is possible to illuminate under lighting conditions such that the difference in color of the sushi plate can be clearly seen in the vicinity of the user.

Further, in the case of a "conveyor-belt sushi restaurant," the imaging by the image sensor 11 may be performed at a constant cycle, or may be performed each time the image sensor 11 detects that a sushi plate has entered the imaging range. good. In addition, although a "conveyor-belt sushi restaurant" is used as an example here, it is not limited to this, and may be an exhibition booth where the illuminated object moves, for example, an exhibition booth for a train model that actually moves. .

Next, FIG. 22 shows an example in which the illumination devices 10, 10A, 10B, 10B', 10C, and 10D according to the first to fifth embodiments are installed in a "supermarket" or the like. Note that the illumination device 10 is illustrated here.

As shown in FIG. 22 , the learning model 131 may be one of the elements for estimating lighting conditions by appropriately performing learning with a further value axis added, such as sales at a “supermarket”.

Specifically, in the same manner as described above, in addition to the series of procedures of imaging, analysis, and output control shown in steps S41 to S43, additional learning is performed on the learning model 131 with consideration given to sales, for example (step S44), the learning model 131 may be used to repeat the procedure from step S41 on the next business day, for example.

That is, as shown by the number of ☆ marks in the figure, for example, additional learning that takes into account the actual sales results for each spectral distribution characteristic can accumulate lighting conditions for selling the recognized product. It becomes possible. In addition, such sales are not limited to one store, and may be data that integrates the sales of multiple stores, and when the learning model 131 is additionally learned with such integrated data, the updated learning model 131 is distributed to all stores. You may make it

Here, sales are taken as an example, but the evaluation value of the applied illumination condition may be indicated by votes from users, review results, or the like, without being limited to sales. That is, machine learning may be performed so that the evaluation value indicating the result of the output control by the output control unit 143 is included as one of the feature amounts. In addition, although the case of selling products such as "supermarkets" has been taken as an example here, it does not apply to the type of business. If it is possible, it can be applied to various various scenes.

(Other embodiments)
In the first to fourth embodiments described above, the illumination devices 10, 10A, 10B, 10B', and 10C select the spectral distribution characteristics of the light sources. It may be a color deviation or the like. In addition, in the above-described fifth embodiment, the illumination device 10D selects the illumination conditions of the light source, but the illumination conditions are not limited to the spectral distribution characteristics, control start or end timing, etc. It may include color temperature, color rendering index, and the like.

Further, in each of the above-described embodiments, the learning model 131 is used based on image data, but it is not limited to image data. For example, an object and a desired appearance may be specified by a voice command or the like, and based on this, a learning model 131 may be generated and used so that the spectral distribution characteristics of the light source can be estimated.

Further, in each of the embodiments described above, deep learning is used as a machine learning algorithm, but the algorithm to be used is not limited. Therefore, the learning model 131 may be generated by performing machine learning by a regression analysis method such as support vector regression using a pattern classifier such as SVM (Support Vector Machine). Also, here, the pattern classifier is not limited to SVM, and may be AdaBoost or the like. Also, a random forest or the like may be used.

While several embodiments of the invention have been described, these embodiments have been presented by way of example and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, replacements, and modifications can be made without departing from the scope of the invention. These embodiments and their modifications are included in the scope and spirit of the invention, as well as the scope of the invention described in the claims and equivalents thereof.

Reference Signs List 1, 1A illumination system 10, 10A, 10B, 10B', 10C, 10D illumination device 11 image sensor 12 light source unit 20 learning device 100 cloud server 131 learning model 141 acquisition unit 142, 142A, 142B, 142B', 142C image processing unit 142a spectral reflection characteristic estimator 142b selector 142c spectral distribution characteristic estimator 142d classifier 142e extractor 142f illumination condition estimator 143 output controller N network NN-1 to 5 model

Claims (7)

a light source unit having a plurality of light sources;
Image processing is performed on image data obtained by imaging an object illuminated by the light source unit so as to separate the spectral reflection characteristics of the object from the spectral distribution characteristics of the light source, and the object is recognized. an image processing unit;
an output control unit that controls the output of the light source according to the recognition result of the object;
and
The image processing unit
Recognizing the object using a learning model generated by machine learning using image data that is a positive example,
after estimating the spectral reflectance characteristics as a recognition result of the object, selecting the spectral distribution characteristics that make the object appear in a desired manner based on the spectral reflectance characteristics;
The output control unit is
controlling the output of the light source based on the spectral distribution characteristic selected by the image processing unit ;
lighting device. The image processing unit
utilizing the learning model by cloud computing;
The lighting device according to claim 1 . The light source unit
Having at least red, green and blue monochromatic light sources as the light sources,
The image processing unit
estimating the spectral reflection characteristics based on a set of image data obtained by imaging the object when each of the monochromatic light sources is lit individually;
3. A lighting device according to claim 1 or 2 . The image processing unit
After classifying objects in the image data using the learning model, extracting an image area of a specific object, and estimating the spectral reflectance characteristics based on the image area;
4. A lighting device according to claim 1, 2 or 3 . the object is a face,
The image processing unit
After extracting an image area corresponding to each part of the face in the image data, estimating the spectral reflection characteristics of each part based on the image area, the spectral distribution based on the combination of the spectral reflection characteristics selecting lighting conditions for the light source, including characteristics;
The output control unit is
controlling the output of the light source based on the lighting conditions selected by the image processing unit;
The illumination device according to any one of claims 1-4 . The image processing unit
executing the machine learning so that an evaluation value indicating a result of output control by the output control unit is included as one of the feature amounts;
The lighting device according to any one of claims 1 to 5 . The machine learning is deep learning,
A lighting device according to any one of claims 1 to 6 .

Images