Disclosure of Invention
The invention mainly solves the technical problem of reducing the energy consumption of the fan and improving the user experience.
According to a first aspect, the present invention provides an electrical device control method based on artificial intelligence, including: an electrical equipment control method based on artificial intelligence is characterized by comprising the following steps: acquiring a room thermal imaging map; determining population distribution information in the room using a population distribution model based on the room thermal imaging map; determining the rotation range of the fan by using a rotation range determining model based on the information of the number of people in the room and the position of the fan; determining the wind resistance degree based on the ambient wind speed and the air information; processing the number distribution information of the people in the room, the rotation range of the fan, the ventilation degree of the wall and the wind resistance degree based on a fan rotation speed determining model to determine a plurality of sub-rotation ranges of the fan and the fan rotation speed corresponding to each of the plurality of sub-rotation ranges of the fan; and controlling the fan to work based on the plurality of sub-rotation ranges of the fan and the fan rotating speed corresponding to each sub-rotation range of the plurality of sub-rotation ranges of the fan.
Further, the position of the fan is determined by processing the video shot by the room, and the processing and determining process for the video shot by the room comprises the following steps: acquiring a room shooting video; processing the shot video by using a shot video processing model based on the room shot video to determine room position information, window position information, room door position information, socket position information and furniture position information; and determining the position of the fan by using a graph neural network model based on the room position information, the window position information, the room door position information, the socket position information and the furniture position information.
Still further, the input of the graph neural network model includes a plurality of nodes and a plurality of edges, the plurality of nodes are room nodes, window nodes, room door nodes, socket nodes and furniture nodes, the plurality of edges are distances and directions among the plurality of nodes, wherein each of the plurality of nodes includes a plurality of node features, the node features of the room nodes include room position information, the node features of the window nodes include window position information, the node features of the room door nodes include room door position information, the node features of the socket nodes include socket position information, the node features of the furniture nodes include furniture position information, and the output of the graph neural network model is a position where the fan is located.
Further, the shot video processing model is a long-short period neural network model.
According to a second aspect, the present invention provides an artificial intelligence based electrical device control system, comprising: the acquisition module is used for acquiring a room thermal imaging image; the people number distribution determining module is used for determining the people number distribution information in the room by using a people number distribution model based on the room thermal imaging graph; the rotation range determining module is used for determining the rotation range of the fan by using a rotation range determining model based on the number distribution information of people in the room and the position of the fan; the wind resistance degree determining module is used for determining the wind resistance degree based on the ambient wind speed and the air information; the fan rotating speed determining module is used for processing the number distribution information of people in the room, the rotating range of the fan, the wall ventilation degree and the wind resistance degree based on a fan rotating speed determining model to determine a plurality of sub-rotating ranges of the fan and the fan rotating speed corresponding to each of the plurality of sub-rotating ranges of the fan; and the control module is used for controlling the fan to work based on the plurality of sub-rotation ranges of the fan and the fan rotating speed corresponding to each sub-rotation range of the plurality of sub-rotation ranges of the fan.
Still further, the system further comprises a location determination module, the location determination module further configured to: acquiring a room shooting video; processing the shot video by using a shot video processing model based on the room shot video to determine room position information, window position information, room door position information, socket position information and furniture position information; and determining the position of the fan by using a graph neural network model based on the room position information, the window position information, the room door position information, the socket position information and the furniture position information.
Still further, the input of the graph neural network model includes a plurality of nodes and a plurality of edges, the plurality of nodes are room nodes, window nodes, room door nodes, socket nodes and furniture nodes, the plurality of edges are distances and directions among the plurality of nodes, wherein each of the plurality of nodes includes a plurality of node features, the node features of the room nodes include room position information, the node features of the window nodes include window position information, the node features of the room door nodes include room door position information, the node features of the socket nodes include socket position information, the node features of the furniture nodes include furniture position information, and the output of the graph neural network model is a position where the fan is located.
Further, the shot video processing model is a long-short period neural network model.
According to a third aspect, the present invention provides an electronic device comprising: a memory; a processor; a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method described above.
According to a fourth aspect, the present invention provides a computer readable storage medium having stored thereon a computer program which when executed by a processor implements a method as in any of the above aspects.
The invention provides an electrical equipment control method, a system, equipment and a medium based on artificial intelligence, the method comprises the steps of determining personnel distribution information in a room by using a personnel distribution model based on a room thermal imaging diagram, determining a rotation range of a fan by using a rotation range determination model based on the personnel distribution information in the room and the position of the fan, determining wind resistance degree based on ambient wind speed and air information, processing the personnel distribution information in the room, the rotation range of the fan, the ventilation degree of a wall and the wind resistance degree based on the fan rotation speed determination model, determining fan rotation speeds corresponding to a plurality of sub-rotation ranges of the fan and each sub-rotation range of the fan, and controlling the fan to work based on the fan rotation speeds corresponding to the plurality of sub-rotation ranges of the fan and each sub-rotation range of the fan.
Detailed Description
The invention will be described in further detail below with reference to the drawings by means of specific embodiments. Wherein like elements in different embodiments are numbered alike in association. In the following embodiments, numerous specific details are set forth in order to provide a better understanding of the present invention.
In the embodiment of the invention, an electrical equipment control method based on artificial intelligence is provided as shown in fig. 1, and the electrical equipment control method based on artificial intelligence comprises the following steps of S1-S6:
step S1, acquiring a room thermal imaging diagram.
In some embodiments, the thermal image of the room may be captured by a thermal imaging camera above the fan. The room thermography may reflect the population distribution in the room. In some embodiments, a thermal imaging camera may be mounted above the fan. In some embodiments, the fan includes a control device that is communicatively connectable to the external processing device and controls the fan.
And step S2, determining the number of people distribution information in the room by using a number of people distribution model based on the room thermal imaging graph.
The population distribution model is a convolutional neural network model, and the convolutional neural network model comprises a convolutional neural network. Convolutional neural network models are one implementation of artificial intelligence. The input of the people number distribution model comprises the room thermal imaging diagram, and the output of the people number distribution model is the information of the people number distribution in the room. The information on the distribution of the number of people in the room includes the position of each person in the room and the position coordinates of the room. In some embodiments, the number of people distribution information in the room may also include the height, weight, body shape, etc. of each person in the room. In some embodiments, the number of people distribution information in the room may include a set of vectors composed of location coordinates of each person distribution in the room.
The population distribution model may be trained by a plurality of training samples, and in some embodiments, the population distribution model may be trained by a gradient descent method to obtain a trained population distribution model.
And step S3, determining the rotation range of the fan by using a rotation range determination model based on the number distribution information of people in the room and the position of the fan.
The rotation range determination model is a deep neural network model, and the deep neural network model comprises a deep neural network (Deep Neural Networks, DNN). The deep neural network model is one implementation of artificial intelligence. The deep neural network may include a recurrent neural network (Recurrent Neural Network, RNN), a convolutional neural network (Convolutional Neural Networks, CNN), a generating countermeasure network (Generative Adversarial Networks, GAN), and so on. The input of the rotation range determining model is the distribution information of the number of people in the room and the position of the fan, and the output of the rotation range determining model is the rotation range of the fan.
The fan location indicates the location of the fan in the room. In some embodiments, the fan may be located at a position that is optionally placed by a person. In some embodiments, the location of the fan may also be determined in advance by processing the room shot video, and the processing of the room shot video may be determined as described in fig. 2 and the related description.
The rotation range determination model can be obtained through training of training samples, input of the training samples comprises sample population distribution information in a room and positions of sample fans, and output labels of the training samples are rotation ranges of the sample fans. The label of training sample can be obtained through manual mark of staff, and in manual mark in-process, when staff can annotate the rotation scope of fan, can make the rotation scope of fan cover the position of people in the room just, has both covered the people in the room like this, has also avoided the fan to move at unmanned department. The range of rotation of the fan output by the training-completed range determination model also happens to cover the position of the person in the room due to the relative guidance at the time of labeling.
The rotation range of the fan indicates the rotation angle covered when the fan rotates, the fan can be the fan which can rotate by 360 degrees, and after the rotation range of the fan is determined by the rotation range determination model, the fan only rotates in the rotation range, so that the energy waste is avoided, the waiting time of a user when blowing air is reduced, and the user experience is improved. For example, if the rotation range of a plurality of persons in a room relative to the fan is 0 ° -123 °, the rotation range determination model outputs the rotation range of the fan by processing the distribution information of the number of persons in the room and the position of the fan. The rotation range of the fan output by the training rotation range determination model can cover the positions of people in the room, so that the people in the room can be covered by the rotation range of the fan, and the fan is prevented from running at an unmanned place.
And S4, determining the wind resistance degree based on the ambient wind speed and the air information.
The ambient wind speed represents a wind speed in the current environment, including a wind speed direction and a wind speed magnitude, and in some embodiments, the wind speed in the current environment may be obtained by a wind speed sensor. The air information includes air composition, proportion of each component in the air composition, air temperature, humidity, etc. Both ambient wind speed and air information affect wind drag.
The wind resistance degree represents the resistance to the wind blown out by the fan, the wind resistance degree can be a value between 0 and 1, and the greater the wind resistance degree is, the greater the resistance to the wind in the current environment is. In some embodiments, the environmental wind speed and the air information can be constructed as a vector to be matched, and the reference wind resistance degree corresponding to the reference vector with the distance smaller than the threshold value is determined as the current wind resistance degree by calculating the distance between the vector to be matched and each reference vector in the database. The database comprises reference vectors and reference wind resistance degrees corresponding to the reference vectors, the database is pre-constructed, the reference vectors are constructed based on the environmental wind speed and air information in the historical data, and the reference wind resistance degrees corresponding to the reference vectors are determined wind resistance degrees in the historical data. In some embodiments, the wind resistance degree can be determined by processing the ambient wind speed and the air information through a deep neural network.
And S5, processing the number distribution information of the people in the room, the rotation range of the fan, the wall ventilation degree and the wind resistance degree based on a fan rotation speed determination model to determine a plurality of sub-rotation ranges of the fan and the fan rotation speed corresponding to each of the plurality of sub-rotation ranges of the fan.
The degree of wall ventilation may be used to describe the ability of a wall to ventilate a flowing wind, the greater the degree of wall ventilation, the less the wall's ability to block the wind. The greater the wall ventilation degree, the more easily the air outside the wall can penetrate through the wall to enter the room, and resists with the wind sent by the fan, so as to reduce the effect of the wind sent by the fan.
In some embodiments, the degree of wall ventilation may be detected in advance by detecting the wall, for example, blowing the wind of a fan against the wall, and then measuring the wind speed of the wall ventilation on the other side of the wall, and determining the degree of wall ventilation based on the wind speed of the fan and the wind speed of the wall ventilation.
In some embodiments, the flatness, material and thickness of the wall can be determined by transmitting ultrasonic waves to the wall and receiving reflected waves of the wall, by the characteristics of sound velocity, amplitude, frequency, waveform and the like of the reflected waves, and then determining the ventilation degree of the wall through a ventilation degree determination model based on the flatness, material and thickness of the wall, wherein the input of the ventilation degree determination model is the flatness, material and thickness of the wall, the output of the ventilation degree determination model is the ventilation degree of the wall, and the ventilation degree determination model is a deep neural network model. The determination of the flatness, material, thickness of the wall by means of ultrasound is prior art and will not be described here in detail. The ventilation degree of the wall can be influenced by the flatness, the material and the thickness of the wall, so the ventilation degree of the wall can be determined by processing the flatness, the material and the thickness of the wall. The ventilation degree determination model can be obtained through training, the input of a training sample comprises sample flatness, sample material and sample thickness of the wall, and the output label of the training sample is the ventilation degree of the sample wall. The label of the training sample can be obtained through manual labeling of a staff member.
The plurality of sub-rotation ranges of the fan are obtained by dividing the rotation range of the fan. The sum of the ranges of the plurality of sub-rotation ranges of the fan is equal to the rotation range of the fan. For example, the fans may rotate in a range of 0 ° -180 °, and the sub-rotation ranges of the plurality of fans may be 0 ° -25 °,25 ° -60 °,60 ° -90 °, 90 ° -120 °, 120 ° -150 °, 150 ° -180 °, respectively.
The fan speed corresponding to each of the plurality of sub-rotation ranges of the fan indicates that each of the sub-rotation ranges corresponds to one fan speed, and the fan speeds of each of the sub-rotation ranges may be the same or different. For example, the plurality of fans may have a sub-rotation range of 0 ° -25 °,25 ° -60 °,60 ° -90 °, 90 ° -120 °, 120 ° -150 °, 150 ° -180 °. The fan speed corresponding to 0 ° -25 ° is 700 rpm, the fan speed corresponding to 25 ° -60 ° is 800 rpm, the fan speed corresponding to 60 ° -90 ° is 600 rpm, the fan speed corresponding to 90 ° -120 ° is 500 rpm, the fan speed corresponding to 120 ° -150 ° is 900 rpm, and the fan speed corresponding to 150 ° -180 ° is 500 rpm.
Because the number distribution in the room is disordered and irregular, each sub-rotation range of the fan can correspond to different numbers of people, and if the number of people corresponding to each sub-rotation range of the fan is larger, the fan rotation speed corresponding to the sub-rotation range of the fan is higher. For example, the sub-rotation ranges of the plurality of fans are 0 ° -25 °,25 ° -60 °,60 ° -90 °, the number of people corresponding to 0 ° -25 ° is 5, the corresponding fan rotation speed is 500 rotations per minute, the number of people corresponding to 25 ° -60 ° is 7, the corresponding fan rotation speed is 800 rotations per minute, the number of people corresponding to 60 ° -90 ° is 9, and the corresponding fan rotation speed is 900 rotations per minute. In some embodiments, if the positions of the plurality of persons corresponding to each sub-rotation range of the fan are closer to the fan position, the fan rotation speed corresponding to the sub-rotation range of the fan is lower.
In some embodiments, the fan rotation range may be divided by a fan rotation speed determination model to obtain a plurality of sub-rotation ranges of the fan and a fan rotation speed corresponding to each of the plurality of sub-rotation ranges of the fan.
The fan rotation speed determination model is a deep neural network model, the fan rotation speed determination model can be obtained through training, the input of a training sample comprises sample population distribution information in a room, a sample rotation range of a fan, a wall sample ventilation degree and a sample wind resistance degree, and the output label of the training sample is a plurality of sample rotation ranges of the fan and sample fan rotation speeds corresponding to each sample rotation range in the plurality of sample rotation ranges of the fan. The label of training sample can be obtained through staff manual annotation, and staff can consider sample number distribution information, the sample rotation range of fan, wall sample ventilation degree and sample windage degree when manually annotating to divide into a plurality of sample rotation ranges of fan with the sample rotation range of fan, and annotate the sample fan rotational speed that each sample rotation range corresponds. The input of the fan rotation speed determination model after training is the distribution information of the number of people in the room, the rotation range of the fan, the ventilation degree of the wall and the wind resistance degree, and the output of the fan rotation speed determination model after training is the fan rotation speed corresponding to each of a plurality of sub-rotation ranges of the fan.
And S6, controlling the fan to work based on the plurality of sub-rotation ranges of the fan and the fan rotating speed corresponding to each sub-rotation range of the plurality of sub-rotation ranges of the fan.
The control device of the fan may be based on a plurality of sub-rotation ranges of the fan and a fan rotation speed corresponding to each of the plurality of sub-rotation ranges of the fan.
Fig. 2 is a schematic flow chart of determining a position of a fan by processing a video shot by a room, where the processing of the video shot by the room to determine the position of the fan includes steps S21 to S23:
step S21, capturing a room shot video.
In some embodiments, a room shot video taken by a user may be acquired. The room shooting video contains room position information, window position information, room door position information, socket position information and furniture position information. In some embodiments, each place of the room may be photographed in advance to obtain a photographed video of each place, and the photographed videos of each place are synthesized to obtain a photographed video of the room.
And step S22, processing and determining room position information, window position information, room door position information, socket position information and furniture position information by using a shot video processing model based on the room shot video.
The room shot video can be processed through the shot video processing model to determine room position information, window position information, room door position information, socket position information and furniture position information. The shot video processing model is a long-short-period neural network model. The long-term neural network model is one implementation of artificial intelligence. The Long and Short Term neural network model includes a Long and Short Term neural network (LSTM), which is one of RNNs (Recurrent Neural Network, recurrent neural networks). The long-term and short-term neural network model can process sequence data with any length, capture sequence information and output results based on the association relationship of front data and rear data in the sequence. The long-short-term neural network model is used for processing the room shooting videos in the continuous time period, so that the characteristics of the association relationship among the room shooting videos comprehensively considered at each time point can be output, and the output characteristics are more accurate and comprehensive.
The input of the shot video processing model comprises the room shot video, and the output of the shot video processing model is room position information, window position information, room door position information, socket position information and furniture position information.
The room position information includes the length, width, height of the room, the center position coordinates of the room, the coordinate information of a plurality of vertices of the room, and the like. The window position information includes length, width, height of a window, center position coordinates of a window, coordinate information of a plurality of vertices of a window, etc., the room door position information includes length, width, height of a room door, center position coordinates of a room door, coordinate information of a plurality of vertices of a room door, etc., the socket position information includes coordinate information of a plurality of vertices of a socket, etc., the furniture position information includes length, width, height of furniture, center position coordinates of furniture, coordinate information of a plurality of vertices of furniture, etc.
Step S23, determining the position of the fan by using a neural network model based on the room position information, the window position information, the room door position information, the socket position information and the furniture position information.
The graph neural network model includes a graph neural network (Graph Neural Network, GNN) and a full connectivity layer. A graph neural network is a neural network that acts directly on a graph, which is a data structure made up of two parts, nodes and edges. The graph neural network is one implementation of artificial intelligence. The graph neural network model is based on an information propagation mechanism, and each node updates its own node state by exchanging information with each other until a certain stable value is reached.
The input of the graph neural network model comprises a plurality of nodes and a plurality of edges, the nodes are room nodes, window nodes, room door nodes, socket nodes and furniture nodes, and the edges are distances and directions among the nodes. In some embodiments, the plurality of edges are distances and directions between a plurality of nodes.
Wherein each node of the plurality of nodes comprises a plurality of node features, the node features of the room node comprise room position information, the node features of the window node comprise window position information, the node features of the room door node comprise room door position information, the node features of the socket node comprise socket position information, the node features of the furniture node comprise furniture position information, and the output of the graph neural network model is the position of the fan.
The graph neural network model can be obtained through training of training samples. The input of the training sample comprises a plurality of nodes and a plurality of edges, the nodes are room nodes, window nodes, room door nodes, socket nodes and furniture nodes, the edges are distances and directions among the nodes, and the output label of the training sample is the position of the fan. The label of training sample can be obtained through manual mark of staff, in manual mark in-process, the staff can consider each node, relation between each node, also can consider the factor of air convection nature, socket distance, fan area of blowing, house to the shielding of wind, mark the most suitable fan position, because the relevant direction has when the mark, so the picture neural network model after training can output the most suitable fan position when the output for the fan position is more scientific, and the fan is better to user's experience in the in-service use.
In some embodiments, the graph neural network model may be trained by a gradient descent method to obtain a trained graph neural network model.
Based on the same inventive concept, fig. 3 is a schematic diagram of an electrical equipment control system based on artificial intelligence according to an embodiment of the present invention, where the electrical equipment control system based on artificial intelligence includes:
an acquisition module 31 for acquiring a room thermal imaging map;
a population distribution determination module 32 for determining population distribution information in the room using a population distribution model based on the room thermal imaging map;
a rotation range determination module 33 for determining a rotation range of the fan using a rotation range determination model based on the information of the number of people in the room and the position of the fan;
a windage determination module 34 for determining a windage level based on ambient wind speed and air information;
a fan rotation speed determination module 35 that determines a plurality of sub-rotation ranges of the fan and a fan rotation speed corresponding to each of the plurality of sub-rotation ranges of the fan by processing the distribution information of the number of people in the room, the rotation range of the fan, the wall ventilation degree, and the wind resistance degree based on a fan rotation speed determination model;
the control module 36 is configured to control the fan to operate based on a plurality of sub-rotation ranges of the fan and a fan rotation speed corresponding to each of the plurality of sub-rotation ranges of the fan.
Based on the same inventive concept, an embodiment of the present invention provides an electronic device, as shown in fig. 4, including:
comprising the following steps: a processor 41; a memory 42; a computer program; wherein the computer program is stored in the memory 42 and configured to be executed by the processor 41 to implement the artificial intelligence based appliance control method as provided above, the method comprising: acquiring a room thermal imaging map; determining population distribution information in the room using a population distribution model based on the room thermal imaging map; determining the rotation range of the fan by using a rotation range determining model based on the information of the number of people in the room and the position of the fan; determining the wind resistance degree based on the ambient wind speed and the air information; processing the number distribution information of the people in the room, the rotation range of the fan, the ventilation degree of the wall and the wind resistance degree based on a fan rotation speed determining model to determine a plurality of sub-rotation ranges of the fan and the fan rotation speed corresponding to each of the plurality of sub-rotation ranges of the fan; and controlling the fan to work based on the plurality of sub-rotation ranges of the fan and the fan rotating speed corresponding to each sub-rotation range of the plurality of sub-rotation ranges of the fan.
Based on the same inventive concept, the present embodiment provides a computer-readable storage medium having stored thereon a computer program which, when executed by the processor 41, implements the artificial intelligence-based electrical device control method provided above, the method comprising acquiring a room thermal imaging map; determining population distribution information in the room using a population distribution model based on the room thermal imaging map; determining the rotation range of the fan by using a rotation range determining model based on the information of the number of people in the room and the position of the fan; determining the wind resistance degree based on the ambient wind speed and the air information; processing the number distribution information of the people in the room, the rotation range of the fan, the ventilation degree of the wall and the wind resistance degree based on a fan rotation speed determining model to determine a plurality of sub-rotation ranges of the fan and the fan rotation speed corresponding to each of the plurality of sub-rotation ranges of the fan; and controlling the fan to work based on the plurality of sub-rotation ranges of the fan and the fan rotating speed corresponding to each sub-rotation range of the plurality of sub-rotation ranges of the fan.
Finally, it should be understood that the embodiments described in this specification are merely illustrative of the principles of the embodiments of this specification. Other variations are possible within the scope of this description. Thus, by way of example, and not limitation, alternative configurations of embodiments of the present specification may be considered as consistent with the teachings of the present specification. Accordingly, the embodiments of the present specification are not limited to only the embodiments explicitly described and depicted in the present specification.