KR20180051729A - Air conditioner and control method thereof - Google Patents

Abstract

An air conditioner according to one aspect of the present invention comprises: a camera obtaining an exterior image; an object recognition module for recognizing an object from the image obtained by the camera; and a control portion for controlling an air flow based on the location and a texture of the recognized object, thereby capable of improving a cooling efficiency.

Description

[0001] The present invention relates to an air conditioner and a control method thereof,

The present invention relates to an air conditioner and a control method thereof, and more particularly, to an air conditioner for controlling the airflow by determining the position of an object and a control method thereof.

The air conditioner is installed to provide a comfortable indoor environment for humans by discharging cold air to the room to adjust the room temperature and purify the room air to create a pleasant indoor environment.

Generally, the air conditioner includes an indoor unit which is constituted by a heat exchanger and installed in a room, and an outdoor unit which is constituted by a compressor, a heat exchanger and the like and supplies the refrigerant to the indoor unit.

The air conditioner is operated by being controlled separately by an indoor unit constituted by a heat exchanger, an outdoor unit constituted by a compressor, a heat exchanger and the like, and supplied to a compressor or a heat exchanger. Also, at least one indoor unit may be connected to the outdoor unit, and the air conditioner is operated in the cooling or heating mode by supplying the refrigerant to the indoor unit according to the requested operation state.

However, there is a problem that a uniformly discharged airflow can not reach the indoor space, such as an area where the airflow actually discharged does not reach depending on the position or space where the indoor unit is installed.

Prior Art 1 (Laid-Open Patent Publication No. 10-2008-0048866) is an invention relating to an air conditioner and its airflow control method, and it controls the airflow direction by determining the position of occupant.

However, in the prior art 1, since the object to be sensed is limited to the human body, rapid cooling effect can not be expected due to various heat sources.

In addition, since the airflow is directly controlled toward the human body, there is a problem that the efficiency of haptic sense may vary depending on the user.

Background Art [0002] Conventional art 2 (Japanese Unexamined Patent Application Publication No. 10-2009-0050434) relates to an air conditioner for generating a thermal image and a method for generating a thermal image, wherein the position of the occupant is determined using a thermal camera, .

However, since the sensing object is limited to the human body in the prior art 2, rapid cooling effect can not be expected due to existence of various heat sources and objects.

In addition, since the production cost is increased by adopting the thermal imaging camera and the airflow is directly controlled toward the human body, there is a problem in that the user's sensory perception efficiency may be different.

On the other hand, interest in machine learning such as artificial intelligence and deep running has greatly increased recently.

Conventional machine learning was centered on statistical based classification, regression, and cluster models. In particular, in the learning of classification and regression model learning, a learning model that distinguishes the characteristics of learning data and new data based on these characteristics was defined by a person in advance. Deep learning, on the other hand, is one in which a computer identifies and identifies its own characteristics.

One of the factors that accelerated the development of deep learning is the deep-running framework that is provided as open source. With the release of deep learning frameworks, in addition to deep learning algorithms, the extraction and selection of data used in learning processes, learning methods, and learning are becoming more important for effective learning and perception.

Also, researches are being conducted to utilize artificial intelligence and machine learning in various products and services.

When the air conditioner is operated, it is impossible to expect a quick and uniform cooling effect due to various heat sources present in the home / business. Accordingly, the present invention proposes a control method of increasing the cooling efficiency and the user convenience by recognizing the object corresponding to the occupant and the heat source using the previously learned image-based object detector and changing the cooling pattern according to the situation.

1. Published Patent Application No. 10-2008-0048866 (Published on June 3, 2008) 2. Published Patent Application No. 10-2009-0050434 (Published on May 20, 2009)

An object of the present invention is to provide an air conditioner having improved cooling efficiency through object recognition and a control method thereof.

An object of the present invention is to provide an air conditioner and a control method thereof that can prevent the cooling efficiency from being lowered due to a heat generating object and a high specific heat.

An object of the present invention is to provide an air conditioner and a control method thereof capable of recognizing a heat source based on machine learning and efficiently performing a cooling function.

According to an aspect of the present invention, there is provided an air conditioner comprising: a camera for acquiring an external image; an object recognition module for recognizing an object in an image acquired by the camera; The cooling efficiency can be improved.

According to another aspect of the present invention, there is provided a method of controlling an air conditioner, including: obtaining an external image through a camera; recognizing an object in an image acquired by the camera; And controlling the air flow based on the position of the object.

According to at least one of the embodiments of the present invention, it is possible to provide an air conditioner and its control method with improved cooling efficiency through object recognition.

According to at least one of the embodiments of the present invention, it is possible to prevent cooling efficiency deterioration caused by a heat generating object and a high-specific object.

Further, according to at least one of the embodiments of the present invention, the heat source can be recognized based on the machine learning and the cooling function can be efficiently performed.

Meanwhile, various other effects will be directly or implicitly disclosed in the detailed description according to the embodiment of the present invention to be described later.

1 is a schematic diagram of an air conditioner according to an embodiment of the present invention.
2 is a block diagram illustrating the control relationship between the main components of the air conditioner according to the embodiment of the present invention.
3 is a block diagram illustrating the control relationship between the main components of the air conditioner according to the embodiment of the present invention.
4 is a flowchart illustrating a method of controlling an air conditioner according to an embodiment of the present invention.
5 to 7 are diagrams referred to in explanation of Deep Learning.
FIG. 8 is a diagram referred to the description of a texture recognizer according to an embodiment of the present invention.
9 is a diagram referred to in description of texture recognition according to an embodiment of the present invention.
10 is a diagram referred to the description of an object recognizer according to an embodiment of the present invention.
FIGS. 11 and 12 are views referred to the description of object recognition according to an embodiment of the present invention.
13 is an internal block diagram of a server according to an embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. However, it is needless to say that the present invention is not limited to these embodiments and can be modified into various forms.

In the drawings, the same reference numerals are used for the same or similar parts throughout the specification.

The suffix "module" and " part "for components used in the following description are given merely for convenience of description and do not give special significance or role in themselves. Accordingly, the terms "module" and "part" may be used interchangeably.

FIG. 1 is a simplified conceptual diagram of an air conditioner according to an embodiment of the present invention, and FIG. 2 is a block diagram illustrating a control relationship between main configurations of an air conditioner according to an embodiment of the present invention.

1 and 2, an air conditioner 100 according to an exemplary embodiment of the present invention includes a camera 110 for acquiring an external image, an object such as a heat source in an image acquired by the camera 110, And a control unit 140 for controlling the air flow based on the position and the texture of the recognized object.

Referring to FIG. 1, the indoor unit of the air conditioner may have a discharge port (not shown) and a discharge port 190 so that the air sucked through the suction port may be discharged to the room again.

The discharge port 190 is provided with a rover (not shown) that opens and closes the discharge port 190 and adjusts the airflow direction of the discharged air. The controller 140 drives the motor, You can switch the direction.

Accordingly, the airflow direction of the air discharged through the discharge port 190 can be adjusted.

Meanwhile, in the air conditioner, there is provided a fan for controlling the flow of the indoor air sucked by the suction port to the indoor space through the discharge port 190. The fan is controlled to rotate by the fan motor, The operation of the fan motor is controlled by the control unit 140.

Accordingly, the controller 140 controls the rover and the fan motor to control the air flow (air flow) discharged from the air conditioner. The control unit 140 controls the speed of the fan motor to control the amount and speed of the airflow, and controls the direction of the airflow by controlling the rover.

Also, the controller 140 can control the direction and speed of the airflow so that the area where the specific object is located can be cooled quickly.

In particular, the controller 140 can determine the specific heat of the object by recognizing the texture of the object, and adjust the cooling rate according to the specific heat of the object.

For example, when the seat made of iron and the sofa made of leather are recognized, the control unit 140 can control the airflow so that the cooling intensity can be further increased for the area where the sofa made of leather is located.

In addition, the controller 140 recognizes the texture of the object, determines the specific heat of the object based on the texture of the object, and controls the concentrated cooling for the region where the object having a high specific heat is located.

The air conditioner according to an embodiment of the present invention can perform object recognition based on machine learning. Here, the object recognition module 150 may use an image obtained through the camera 110 as input data of an artificial neural network learned by machine learning, The object can be recognized. The machine learning and the deep learning will be described later with reference to Figs. 5 to 7.

Meanwhile, the control unit 140 may control the airflow according to the object recognition result of the object recognition module 150. For example, the control unit 140 may control the direction and intensity of the airflow so that the corresponding area can be cooled quickly based on the position of the recognized object.

That is, the object recognition module 150 recognizes the distribution of the object such as the heat source in the room based on the machine learning, and the control unit 140 controls the fan motor and the rover according to the recognized object distribution to perform the optimized airflow control .

In addition, the control unit 140 may control the airflow control to be repeatedly performed.

The object recognition module 150 includes an object recognizer 152 that recognizes the type of object based on data learned through machine learning from the image acquired by the camera 110, And a texture recognizer 151 for recognizing the texture of the object based on the data learned through machine learning.

That is, the object recognizer 152 identifies the object, and the texture recognizer 151 identifies the texture of the object, so that even the same object can identify an object having a high specific heat.

In addition, the control unit 140 may adjust the driving speed and the angle of the fan motor based on the recognized texture of the object and the object.

According to an embodiment of the present invention, the object recognition module 150 may further include a signal generation unit 153 for generating a control signal based on the type and texture of the recognized object, Can control the fan motor and the rover according to the control signal generated by the signal generation unit 153. [

According to an embodiment, the object recognition module 150 may not be provided separately from the controller 140, and may be configured as a convex portion of the controller 140. [

FIG. 3 is a block diagram illustrating a control relationship between the main components of the air conditioner according to an embodiment of the present invention. In FIG. 3, an object recognition module 145 is provided in the controller 140.

In this case, the object recognition module 145 may include an object recognizer that recognizes the type of the object based on data learned through machine learning from the image acquired by the camera 110, And a texture recognizer that recognizes the texture of the object based on data previously learned from the image acquired by the machine learning.

The object recognition module 145 may further include a signal generator for generating a control signal based on the type and texture of the recognized object.

Meanwhile, the object recognition module 150 and 145 can recognize the type and texture of the object based on the data learned by machine learning in the image acquired by the camera.

In addition, the controller 140 may generate a control signal based on the type and texture of the recognized object, and may control the fan motor and the rover based on the type and texture of the recognized object.

Alternatively, the object recognition module (150, 145) recognizes the type and texture of the object based on the data learned through machine learning in the image acquired by the camera, The control signal can be generated based on the texture. The control unit 140 may control the fan motor and the rover according to the generated control signal.

Meanwhile, the object recognition module 150 or 145 may extract a sub window by sliding a window in a predetermined pixel unit in an image acquired by the camera, And the type of the object can be recognized by using it as input data of the learned artificial neural network (Artificial Neural Network).

In addition, the object recognition module 150 or 145 may extract a sub window on a predetermined pixel basis from the image acquired by the camera, and extract the extracted sub window using an artificial neural network The texture of the object can be recognized by using it as input data.

On the other hand, the air conditioner may include a memory (not shown), a communication unit 170, and the like.

The memory 145 may store data for controlling the operation of the air conditioner and data received through the communication unit 170.

On the other hand, the air conditioner may include a buffer for temporarily storing data, and the buffer may be included in the controller 140 or memory.

The control unit 140 may process various data received through the communication unit 170 to update the memory. For example, when the data input through the communication unit 170 is update data for an operation program pre-stored in the memory, the update data is used to update the data in the memory. If the input data is a new operation program, . ≪ / RTI >

The control unit 140 controls the flow of data input to or output from the air conditioner, and generates and applies a control command.

The communication unit 170 may include one or more communication modules and may be communicatively coupled to an external device.

In addition, the communication unit 170 may transmit at least a part of the images obtained by the camera 110 to a predetermined server, and receive data related to machine learning from the predetermined server.

In addition, the object recognition module 150 and 145 may be updated based on data related to machine learning received from the predetermined server.

4 is a flowchart illustrating a method of controlling an air conditioner according to an embodiment of the present invention.

Referring to FIG. 4, an external image can be acquired through the camera 110 (S410).

The object recognition modules 150 and 145 may recognize the object in the image acquired by the camera 110 (S420). For example, the object recognition modules 150 and 145 can recognize the heat source based on the data learned through the machine learning (S420).

The object recognition module 150 or 145 can recognize the type and texture of an object included in the image acquired by the camera 110. [

The control unit 140 may control the airflow based on the position of the recognized object (S430). For example, the control unit 140 may control the fan motor and the rover based on the type and texture of the recognized object.

The signal generator 153 or the controller 140 of the object recognition module 150 or 145 may generate a control signal based on the object recognition result at step S430.

According to an embodiment, at least a part of the images obtained through the camera 110 may be transmitted to a predetermined server. In this case, upon receipt of data related to machine learning from the predetermined server, the pre-learned data may be updated based on data related to machine learning received from the predetermined server.

5 to 7 are diagrams referred to in explanation of Deep Learning.

Deep Learning, a type of machine learning, is a multi-level, deep learning process based on data.

Deep learning can represent a set of machine learning algorithms that extract key data from multiple sets of data as they step up.

The deep learning structure may include an artificial neural network (ANN). For example, the deep learning structure may include a deep neural network (DNN) such as CNN (Convolutional Neural Network), RNN (Recurrent Neural Network), DBN Be able to

Referring to FIG. 5, the ANN may include an input layer, a hidden layer, and an output layer. Each layer contains a plurality of nodes, and each layer is associated with the next layer. The nodes between adjacent layers can be connected to each other with a weight.

Referring to FIG. 6, the computer (machine) finds a certain pattern from the input data 610 and forms a feature map. The computer (machine) can extract from the low level feature 620, the intermediate level feature 630, and the high level feature 640, recognize the object, and output the result 650.

The artificial neural network can be abstracted into higher-level features as it goes to the next layer.

Referring to FIGS. 5 and 6, each node may operate based on an activation model, and an output value corresponding to an input value may be determined according to an activation model.

The output value of any node, e.g., lower level feature 620, may be input to the next layer associated with that node, e.g., a node of the intermediate level feature 630. A node of the next layer, for example, a node of the intermediate level feature 630, may receive values output from a plurality of nodes of the lower level feature 620.

In this case, the input value of each node may be a value to which a weight is applied to the output value of the node of the previous layer. The weight can mean the link strength between nodes.

Also, the deep running process can be seen as a process of finding an appropriate weight.

On the other hand, the output value of any node, e. G., The intermediate level feature 630, may be input to the node of the next layer associated with that node, e. G. A node of the next layer, e.g., a node of the higher level feature 640, may receive values output from a plurality of nodes of the intermediate level feature 630.

The artificial neural network can extract feature information corresponding to each level using a learned layer corresponding to each level. The artificial neural network is sequentially abstracted, and a predetermined object can be recognized using the feature information of the highest level.

For example, if we look at the face recognition process by deep learning, the computer distinguishes between bright pixels and dark pixels according to the brightness of a pixel from the input image, distinguishes simple forms such as a border and an edge, You can distinguish things. Finally, the computer can grasp the form that defines the human face.

The deep running structure according to the present invention can utilize various known structures. For example, the deep learning structure according to the present invention may be a Convolutional Neural Network (CNN), a Recurrent Neural Network (RNN), a Deep Belief Network (DBN), or the like.

Recurrent Neural Network (RNN) is widely used in natural language processing, and it is possible to construct an artificial neural network structure by stacking layers at each moment with an effective structure for time-series data processing that varies with time .

Deep Belief Network (DBN) is a deep-run structure consisting of multiple layers of deep-running RBM (Restricted Boltzman Machine). The Restricted Boltzman Machine (RBM) learning is repeated, and a DBN (Deep Belief Network) having a corresponding number of layers can be constituted by a certain number of layers.

CNN (Convolutional Neural Network) is a structure widely used in the field of object recognition, and will be described with reference to FIG.

CNN (Convolutional Neural Network) is a model that simulates a person's brain function based on the assumption that when a person recognizes an object, it extracts the basic features of the object, then undergoes complicated calculations in the brain and recognizes the object based on the result to be.

7 is a view showing a CNN (Convolutional Neural Network) structure.

The CNN (Convolutional Neural Network) may also include an input layer, a hidden layer, and an output layer.

A predetermined image 710 is input to an input layer.

Referring to FIG. 7, the hidden layer is composed of a plurality of layers, and may include a convolution layer and a sub-sampling layer.

In the CNN (Convolutional Neural Network), various filters for extracting features of images through convolution operations and pooling or non-linear activation functions for adding nonlinear characteristics Is used.

Convolution is mainly used in filter processing in image processing and is used to implement a filter for extracting features from an image.

For example, if the convolution operation is repeatedly performed on the whole image while moving a window of a predetermined size, an appropriate result can be obtained according to the weight value of the window.

The convolution layer performs a convolution operation on the input image data 710 and 712 using a convolution filter and generates feature maps 711 and 713 in which features of the input image 710 are expressed do.

As a result of the convolutional filtering, filtering images corresponding to the number of filters can be generated according to the number of filters included in the convolution layer. The convolution layer may consist of the nodes included in the filtered images.

In addition, a sub-sampling layer paired with a convolution layer may include the same number of feature maps as a convolution layer that forms a pair.

The sub-sampling layer reduces the dimension of the feature maps 711, 713 through sampling or pooling.

The output layer recognizes the input image 710 by combining the various features represented in the feature map 714.

The object recognition module of the air conditioner according to the present invention can use the above-described various deep-running structures. For example, although the present invention is not limited, a CNN (Convolutional Neural Network) structure widely used in object recognition in an image can be used.

On the other hand, the learning of the artificial neural network can be achieved by adjusting the weight of the inter-node connection line so that the desired output is obtained for a given input. Also, the artificial neural network can continuously update the weight value by learning. Back propagation can be used for artificial neural network learning.

FIG. 8 is a drawing referred to in explaining a texture recognizer according to an embodiment of the present invention, and is divided into a learning process and a recognition process.

9 is a diagram referred to in description of texture recognition according to an embodiment of the present invention.

Referring to FIG. 8, the texture data is collected to construct a texture image database (DB), and the parameters (weight and bias) of the neural network are determined through artificial neural network learning of the texture recognizer, ) To estimate the model (dotted line flow).

Thereafter, the texture of the object included in the newly acquired image acquired by the camera 110 can be recognized using the learned model parameters (solid line flow).

9, a sub window 910 is extracted on the basis of a pixel in an image recognized through the camera 110, and texture recognition is performed on a pixel-by-pixel basis using the learned text recognizer 151 Perform, and classify the texture of the entire image.

The learned texture recognizer 151 can output the pixel unit texture recognition result 920 such as 'wood'.

Even objects of the same kind vary in specific heat depending on the texture of the material. Accordingly, the present invention recognizes the texture of an object, discriminates an object having a high specific heat, and adjusts the direction and intensity of an air flow to the area, thereby improving cooling performance.

Meanwhile, the texture recognizer 151 performs a machine learning based recognizer learning and recognition process, and each of the machine learning techniques such as SNM (Support Vector Machine), Adaboost, and Deep Learning may be used.

In FIG. 8, the learning process indicated by a dotted line is a process of extracting all the features of the texture image database collected before learning and learning the recognizer. At this time, the image database is an image that knows which category belongs.

The recognition process indicated by the solid line in FIG. 8 is a process of performing a texture recognition process after extracting features of an input image using a recognizer learned through a learning process.

FIG. 10 is a diagram referred to a description of an object recognizer according to an embodiment of the present invention, and FIGS. 11 and 12 are views referencing a description of object recognition according to an embodiment of the present invention.

10, the object data is collected and an object image database DB is constructed. After extracting features, parameters of the neural network (weight and bias) are determined through artificial neural network learning of the object recognizer. ) To estimate the model (dotted line flow).

Thereafter, the type and position of the objects included in the newly acquired image acquired by the camera 110 can be recognized using the learned model parameters (solid line flow).

Referring to FIG. 11, a sliding window method is used in which a sub window 1110 is extracted and sliced to extract an object 1111 from a recognized image through a camera 110, The type and position of the object in the image can be estimated through the object recognizer 152 previously learned for the sub window 1110.

The learned object recognizer 152 can output the object recognition result 1120 such as 'TV'.

Also, as shown in FIG. 12, it is possible to recognize a plurality of objects such as 'TV' and 'Chair' in the same image and output the result.

On the other hand, the object recognizer 152 performs a machine learning based recognizer learning and recognition process, and a machine learning technique such as a support vector machine (SNM), an adaboost, and a deep learning may be used.

In FIG. 10, a learning process indicated by a dotted line is a process of extracting all the features of the object image database collected before learning and learning the recognizer. At this time, the image database is an image that knows which category belongs.

The recognition process indicated by the solid line in FIG. 10 is a process of performing the object recognition process after the feature extraction for the input image using the recognizer learned through the learning process.

On the other hand, the controller 140 can control the air flow by adjusting the number of revolutions of the fan motor, the angle of the rover, the angle of the guide, or the like based on the type, position, and texture of the recognized object.

In addition, the air conditioner performs object detection and texture recognition using the learned deep artificial neural network (DNN) to detect images of occupants and objects and objects with a high specific heat . By controlling the airflow control and the number of revolutions of the fan motor installed in the main body, the angle of the rover, the angle of the guide, etc., it is possible to prevent the cooling efficiency from being lowered due to the heat generating object and the object having a high specific heat, .

Meanwhile, learning and updating of the texture recognizer 151 and the object recognizer 152 according to the present invention may be performed through an external server.

13 is an internal block diagram of a server according to an embodiment of the present invention.

13, the server 1300 may include a communication unit 1320, a storage unit 1330, a learning module 1340, and a processor 1310.

The processor 1310 may control the overall operation of the server 1300. [

Meanwhile, the server 1300 may be a server operated by a home appliance manufacturer such as an air conditioner or a server operated by a service provider, or may be a kind of a cloud server.

The communication unit 1320 can receive various kinds of data such as state information, operation information, and operation information from a portable appliance, an air conditioner, and the like, a home appliance, a gateway, and the like.

The communication unit 1320 can transmit data corresponding to the received various information to a home appliance such as a portable terminal, an air conditioner, a gateway, and the like.

To this end, the communication unit 1320 may include one or more communication modules such as an Internet module and a mobile communication module.

The storage unit 1330 may store received information and may include data for generating result information corresponding thereto.

In addition, the storage unit 1330 may store data used for machine learning, result data, and the like.

The learning module 1340 may serve as a learning device of the home appliances such as the air conditioner.

The learning module 1340 may include a Deep Neural Network (DNN) such as an artificial neural network, for example, a Convolutional Neural Network (RNN), a Recurrent Neural Network (RNN) Neural networks can be learned.

As the learning method of the learning module 1340, both unsupervised learning and supervised learning can be used.

Meanwhile, the controller 1310 may control the artificial neural network structure of the home appliance such as the air conditioner to be updated to the learned artificial neural network structure after learning according to the setting.

In order to minimize the inconvenience of the user by efficiently operating the air conditioner, the room temperature should be controlled by accurately extracting the position of the object such as the heat source and the object having a high specific heat and performing the concentrated cooling operation.

Accordingly, the present invention provides a method for detecting an object of an indoor object in an image input through a camera. Texture recognition is performed to detect the position of an object having a high heat source / specific heat and to control the cooling to be a top priority, thereby improving cooling efficiency and providing high convenience.

The present invention also uses machine learning algorithms to perform object and texture recognition. Deep Learning based Artificial Neural Network, which is one of the machine learning technologies, can be used at this time.

According to at least one of the embodiments of the present invention, it is possible to provide an air conditioner and its control method with improved cooling efficiency through object recognition.

According to at least one of the embodiments of the present invention, it is possible to prevent cooling efficiency deterioration caused by a heat generating object and a high-specific object.

Further, according to at least one of the embodiments of the present invention, the heat source can be recognized based on the machine learning and the cooling function can be efficiently performed.

The air conditioner according to the present invention is not limited to the configuration and method of the embodiments described above, but the embodiments may be modified so that all or some of the embodiments are selectively combined .

Meanwhile, the control method of the air conditioner according to the embodiment of the present invention can be implemented as a code that can be read by a processor on a recording medium readable by the processor. The processor-readable recording medium includes all kinds of recording apparatuses in which data that can be read by the processor is stored. Examples of the recording medium that can be read by the processor include a ROM, a RAM, a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, and the like, and may also be implemented in the form of a carrier wave such as transmission over the Internet . In addition, the processor-readable recording medium may be distributed over network-connected computer systems so that code readable by the processor in a distributed fashion can be stored and executed.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments, but, on the contrary, It should be understood that various modifications may be made by those skilled in the art without departing from the spirit and scope of the present invention.

Camera: 110
Control section: 140
Object recognition module: 150
Communication department: 170

Claims (15)

A camera for acquiring an external image;
An object recognition module for recognizing an object in an image acquired by the camera; And
And controlling the airflow based on the position and the texture of the recognized object. The method according to claim 1,
Wherein the object recognition module comprises:
An object recognizer for recognizing the type of the object based on data learned through machine learning in an image acquired by the camera;
And a texture recognizer that recognizes the texture of the object based on data previously learned by machine learning in an image acquired by the camera. 3. The method of claim 2,
Wherein the object recognition module comprises:
Further comprising a signal generator for generating a control signal based on the type and texture of the recognized object. The method according to claim 1,
Wherein the object recognition module comprises:
Wherein the type of the object and the texture of the object are recognized based on data previously learned by machine learning in an image acquired by the camera. 5. The method of claim 4,
Wherein the control unit generates a control signal based on the type and texture of the recognized object. The method according to claim 1,
Wherein the control unit controls the fan motor and the rover based on the type and texture of the recognized object. The method according to claim 1,
The object recognition module extracts a sub window by sliding a window in a predetermined pixel unit in an image acquired by the camera and extracts the extracted sub window from the learned artificial neural network Neural Network), and recognizes the type of the object. The method according to claim 1,
The object recognition module extracts a sub window on a predetermined pixel basis from the image acquired by the camera and uses the extracted sub window as input data of the learned artificial neural network, Wherein the air conditioner recognizes the texture. The method according to claim 1,
And a communication unit for transmitting at least a part of the obtained images to a predetermined server and receiving data related to machine learning from the predetermined server. 10. The method of claim 9,
And updates the object recognition module based on data related to machine learning received from the predetermined server. Acquiring an external image through a camera;
Recognizing an object in an image acquired by the camera; And
And controlling the airflow based on the position and the texture of the recognized object. 12. The method of claim 11,
Wherein the object recognition step comprises:
Wherein the type and the texture of the object are recognized based on data learned by machine learning in the image acquired by the camera. 12. The method of claim 11,
Wherein the airflow control step includes:
And controlling the fan motor and the rover based on the type and texture of the recognized object. 12. The method of claim 11,
Transmitting at least a part of the acquired images to a predetermined server;
And receiving data related to machine learning from the predetermined server. 15. The method of claim 14,
And updating the pre-learned data based on data related to machine learning received from the predetermined server.

Images