Air conditioner and control method and device thereof, storage medium and computer program product

Abstract

The invention provides an air conditioner and a control method and device thereof, a storage medium and a computer program product, and the method comprises the following steps: detecting whether a person is in a room or not, and detecting whether a pet exists in the room or not when it is detected that no person is in the room; if it is detected that the pet exists, the air conditioner is controlled to be switched to a preset mode; collecting an image of the pet; determining whether the pet has a preset behavior or not according to the collected image of the pet; if it is determined that the pet has the preset behavior, an air conditioner and / or an air purifier in the air conditioning system are / is controlled; and / or, acquiring hair humidity detection data of the pet; according to the acquired hair humidity detection data of the pet, judging the dry and wet state of the hair of the pet; and controlling an air conditioner and / or a dehumidifier and / or a humidifier in the air conditioning system according to the judged dry and wet state of the hair of the pet. According to the scheme provided by the invention, the comfort level of the pet in indoor life can be improved.

Description

Technical Field

[0001] The present invention relates to the field of control, and in particular to an air conditioner and a control method, device, storage medium and computer program product thereof. Background Art

[0002] Smart home technologies in related art are gradually being applied to daily life, especially in smart devices such as air conditioners, air purifiers, and dehumidifiers, which have realized the functions of intelligent sensing and automatic environmental adjustment. However, most air conditioning systems in related art can only be adjusted according to the presence of humans or environmental conditions in the room, and do not provide special solutions for the special needs of pets, such as environmental temperature, humidity, and air quality requirements. Summary of the invention

[0003] The main purpose of the present invention is to overcome the defects of the above-mentioned related technologies and to provide an air conditioner and its control method, device, storage medium and computer program product to solve the problem in the related technologies that the air conditioning system can only be adjusted according to the presence of humans in the room or the environmental conditions, and no special solutions are provided for the special needs of pets.

[0004] On one hand, the present invention provides a control method for an air-conditioning system, the air-conditioning system comprising: at least one of an air conditioner, an air purifier, a dehumidifier and a humidifier; the control method comprises: detecting whether there is anyone in a room, and when it is detected that no one is in the room, detecting whether there is a pet in the room; if it is detected that there is a pet in the room, controlling the air conditioner to switch to a preset mode; in the preset mode, collecting an image of the pet; determining whether the pet has a preset behavior based on the collected image of the pet; if it is determined that the pet has the preset behavior, controlling the air conditioner and / or air purifier in the air-conditioning system; and / or, in the preset mode, obtaining hair humidity detection data of the pet; judging the dryness and wetness state of the pet's hair based on the obtained hair humidity detection data of the pet; and controlling the air conditioner and / or dehumidifier and / or humidifier in the air-conditioning system based on the judged dryness and wetness state of the pet's hair.

[0005] Optionally, determining whether the pet has a preset behavior based on the collected image of the pet includes: uploading the collected image of the pet to a server, and the server determining whether the pet has a preset behavior based on the collected image of the pet; and / or judging the dryness and wetness state of the pet's hair based on acquired hair humidity detection data of the pet, including: uploading the acquired hair humidity detection data of the pet to a server, and the server judging the dryness and wetness state of the pet's hair based on the acquired hair humidity detection data of the pet.

[0006] Optionally, based on the captured images of the pet, determining whether the pet has a preset behavior includes: preprocessing the captured images of the pet; extracting key point posture vectors of the pet's body based on the preprocessed images of the pet; converting the key point posture vectors extracted from multiple frames of continuously captured images of the pet into a time series; and inputting the time series into a pre-trained behavior recognition model to identify whether the pet has a preset behavior.

[0007] Optionally, the pet image is preprocessed, including: denoising the pet image; extracting a dynamic foreground in the pet image by a frame difference method; and performing morphological operations on the extracted dynamic foreground.

[0008] Optionally, based on the preprocessed image of the pet, the key point posture vector of the pet's body is extracted, including: inputting the preprocessed image of the pet into a deep convolutional neural network to extract a feature map containing structural information; mapping the extracted feature map to a heat map space to generate a heat map of each key point of the pet's body; extracting the coordinates of each joint point from the generated heat map of each key point of the pet's body; and converting the extracted coordinates of each key point into a key point posture vector of the pet's body.

[0009] Optionally, if it is determined that the pet has a preset behavior, the air conditioner and / or air purifier in the air-conditioning system is controlled, including: if it is determined that the pet has a preset behavior, the fresh air system of the air conditioner is started and the air purifier is turned on; the indoor air quality is detected by a sensor, and the running time of the fresh air system and the air purifier is controlled according to the detected air quality; and / or the wet and dry state of the pet's hair includes: a wet state and a dry state; according to the determined wet and dry state of the pet's hair, the air conditioner and / or the dehumidifier and / or the humidifier in the air-conditioning system is controlled, including: if the pet's hair is determined to be dry or wet, the humidifier in the air-conditioning system is controlled to work, and / or the air conditioner is controlled to perform environmental humidity adjustment, and / or the dehumidifier in the air-conditioning system is controlled to shut down; if the pet's hair is determined to be wet, the air conditioner in the air-conditioning system is controlled to turn on the dehumidification mode, and / or the dehumidifier in the air-conditioning system is controlled to start working, and / or the dehumidifier in the air-conditioning system is controlled to increase the dehumidification intensity.

[0010] Another aspect of the present invention provides a control device for an air conditioning system, the air conditioning system comprising: at least one of an air conditioner, an air purifier, a dehumidifier and a humidifier; the control device comprising: a detection unit for detecting whether there is anyone in the room, and when it is detected that there is no one in the room, detecting whether there is a pet in the room; a control unit for controlling the air conditioner to switch to a preset mode if the detection unit detects that there is a pet in the room; a collection unit for collecting an image of the pet in the preset mode; a determination unit for determining whether the pet has a preset behavior based on the image of the pet collected by the collection unit; the control unit is further used to: if the determination unit determines that the pet has the preset behavior, control the air conditioner and / or the air purifier in the air conditioning system; and / or, an acquisition unit for acquiring the pet's hair humidity detection data in the preset mode; a judgment unit for judging the pet's hair dryness and wetness state based on the pet's hair humidity detection data acquired by the acquisition unit; the control unit is further used to: according to the pet's hair dryness and wetness state judged by the judgment unit, control the air conditioner and / or the dehumidifier and / or the humidifier in the air conditioning system.

[0011] Optionally, the determination unit determines whether the pet has a preset behavior based on the collected image of the pet, including: uploading the collected image of the pet to a server, and the server determines whether the pet has a preset behavior based on the collected image of the pet; and / or the judgment unit judges the dryness and wetness status of the pet's hair based on the acquired hair humidity detection data of the pet, including: uploading the acquired hair humidity detection data of the pet to a server, and the server judges the dryness and wetness status of the pet's hair based on the acquired hair humidity detection data of the pet.

[0012] Optionally, the determination unit determines whether the pet has a preset behavior based on the captured images of the pet, including: preprocessing the captured images of the pet; extracting key point posture vectors of the pet's body based on the preprocessed images of the pet; converting the key point posture vectors extracted from multiple frames of continuously captured images of the pet into a time series; and inputting the time series into a pre-trained behavior recognition model to identify whether the pet has a preset behavior.

[0013] Optionally, the determination unit pre-processes the image of the pet, including: denoising the image of the pet; extracting a dynamic foreground in the image of the pet by a frame difference method; and performing morphological operations on the extracted dynamic foreground.

[0014] Optionally, the determination unit extracts the key point posture vector of the pet's body based on the preprocessed image of the pet, including: inputting the preprocessed image of the pet into a deep convolutional neural network to extract a feature map containing structural information; mapping the extracted feature map to a heat map space to generate a heat map of each key point of the pet's body; extracting the coordinates of each joint point from the generated heat map of each key point of the pet's body; and converting the extracted coordinates of each key point into a key point posture vector of the pet's body.

[0015] Optionally, if the determination unit determines that the pet has a preset behavior, the control unit controls the air conditioner and / or air purifier in the air conditioning system, including: if it is determined that the pet has a preset behavior, the fresh air system of the air conditioner is started and the air purifier is turned on; the indoor air quality is detected by a sensor, and the running time of the fresh air system and the air purifier is controlled according to the detected air quality; and / or the wet and dry state of the pet's hair includes: a wet state and a dry state; the judgment unit controls the air conditioner and / or the dehumidifier and / or the humidifier in the air conditioning system according to the judged wet and dry state of the pet's hair, including: if it is judged that the pet's hair is in a dry state, the humidifier in the air conditioning system is controlled to work, and / or the air conditioner is controlled to perform environmental humidity adjustment, and / or the dehumidifier in the air conditioning system is controlled to shut down; if it is judged that the pet's hair is in a wet state, the air conditioner in the air conditioning system is controlled to turn on the dehumidification mode, and / or the dehumidifier in the air conditioning system is controlled to start working, and / or the dehumidifier in the air conditioning system is controlled to increase the dehumidification intensity.

[0016] Another aspect of the present invention provides a storage medium having a computer program stored thereon, wherein the program implements the steps of any of the aforementioned methods when executed by a processor.

[0017] In another aspect, the present invention provides an air conditioner, comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any of the aforementioned methods when executing the program.

[0018] In another aspect, the present invention provides an air conditioner, comprising any one of the control devices described above.

[0019] In yet another aspect, the present invention provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the steps of any of the aforementioned methods are implemented.

[0020] According to the technical solution of the present invention, the needs of pets can be intelligently identified, and the air quality and temperature and humidity can be adjusted accordingly according to the pet's behavior and hair condition, thereby improving the comfort and health protection of the pet's indoor life.

[0021] According to the technical solution of the present invention, when the air conditioner detects that there is no one indoors but there is a pet, it automatically switches to pet mode. The air conditioning system monitors the behavior of the pet and controls the air conditioning system according to the behavior of the pet. When the pet is detected to have excretion behavior, the air conditioning system starts the fresh air deodorization function; according to the detected dry and wet state of the pet, the air conditioning system intelligently adjusts the ambient humidity. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0023] Figure 1 is a method schematic diagram of an embodiment of a method for controlling an air conditioning system provided by the present invention;

[0024] Figure 2 A flowchart showing a specific implementation of the step of determining whether the pet has a preset behavior based on the collected image of the pet according to the present invention;

[0025] Figure 3 A flowchart showing a specific implementation of the step of extracting the key point posture vector of the pet's body based on the preprocessed image of the pet according to the present invention;

[0026] Figure 4 is a method schematic diagram of a specific embodiment of the control method of the air conditioning system provided by the present invention;

[0027] Figure 5 It is a structural block diagram of an embodiment of a control device for an air conditioning system provided by the present invention. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0029] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0030] Most air conditioning systems in the related art can only be adjusted according to the presence of humans or environmental conditions in the room, and do not provide special solutions for the special needs of pets, such as ambient temperature, humidity, and air quality requirements. In particular, when pets produce odors or skin health problems due to excrement, the air conditioning systems in the related art lack timely and effective response measures.

[0031] The smart air conditioning system in the related art can usually detect whether there is someone in the room through sensors, and adjust the temperature and humidity according to the indoor environment. However, when there is no one in the room and there is a pet, the air conditioning system usually cannot effectively sense the presence of the pet, and there is no function to automatically switch to the pet-specific mode. In addition, although smart air purifiers and dehumidifiers can process indoor air and humidity, they lack the ability to link with pet behavior equipment and cannot perform targeted environmental adjustments when pet excretion or skin problems are detected.

[0032] In summary, the air conditioning system in the related art lacks the ability to recognize the presence of pets indoors and automatically switch modes, and cannot provide a suitable environment for pets when no one is around. Smart air purifiers and dehumidifiers also do not have the function of linking with pet smart devices, and cannot automatically deal with pet excrement odor or skin moisture problems.

[0033] The present invention provides a control method for an air conditioning system. The air conditioning system comprises: at least one of an air conditioner, an air purifier, a dehumidifier and a humidifier. The air conditioning system can be controlled in conjunction with a pet breeding device arranged indoors. The control method can be implemented in a controller of the air conditioner, for example, in a main control unit of the air conditioner controller.

[0034] The air conditioner includes but is not limited to an edge computing unit, a perception unit, an image acquisition unit (including an image acquisition device such as a camera), a main control unit, and a wireless communication module. The edge computing unit processes sensor data in real time and calculates temperature sensitivity through an optimized algorithm; the perception unit senses whether there are people in the environment; the image acquisition unit is turned on when the air conditioner enters the pet mode, and is mainly used to monitor the behavior of pets when they are not active in the smart house, and the pet behavior data is uploaded to the server in real time; the main control unit is responsible for logical scheduling and control; the wireless communication module unit is responsible for data transmission, communication, etc.

[0035] The air conditioner can communicate with the air purifier, dehumidifier and humidifier through a wireless network or through short-range wireless communication, such as Bluetooth or WiFi. The air conditioner can also communicate with the pet breeding device, such as through a wireless network or through short-range wireless communication.

[0036] The pet breeding device (specifically, a pet smart house) has an image acquisition device (such as a camera) and / or a humidity sensor. The image acquisition device (such as a camera) captures real-time images and video data of the pet inside the pet breeding device for analyzing the pet's behavior, such as excretion behavior, eating behavior, etc.; the humidity sensor detects the environmental humidity inside the pet breeding device (inside the pet smart house) in real time.

[0037] Figure 1 It is a method schematic diagram of an embodiment of a method for controlling an air-conditioning system provided by the present invention.

[0038] like Figure 1 As shown, according to one embodiment of the present invention, the control method of the air-conditioning system includes at least step S110, step S120, and also includes step S130, step S140 and step S150, and / or also includes step S160, step S170 and step S180.

[0039] Step S110, detecting whether there is anyone in the room, and when it is detected that there is no one in the room, detecting whether there is a pet in the room.

[0040] Specifically, whether there is someone in the room can be detected by the sensing module unit of the air conditioning system.

[0041] The perception module unit may specifically include a millimeter wave radar sensor. For example, when a 4D millimeter wave radar sensor detects whether there is someone in the room, it senses changes in the surrounding environment based on the reflected signals of the emitted and received electromagnetic waves. The data processing module of the millimeter wave sensor transmits the processed information to the main control unit of the air conditioning system, and the main control unit analyzes the detection data to determine whether there is someone in the room. The 4D millimeter wave radar measures the height data of the measured target through the ability to measure the pitch angle, so that it can distinguish whether it is a user or a pet by the height.

[0042] When it is detected that no one is in the room, the sensor of the pet breeding device is further used to detect whether there is a pet in the room. Specifically, the sensor of the pet breeding device (such as a pet smart house) can be used to specifically detect whether there is a pet in the room. The sensor is, for example, a millimeter wave radar sensor, which is similar to the method of detecting whether there is a person in the room in the above embodiment, and distinguishes that the detected one is a pet by height. The sensor is, for example, an image sensor, such as a camera, which recognizes whether there is a pet in the room by image.

[0043] Step S120: If the presence of a pet in the room is detected, the air conditioner is controlled to switch to a preset mode.

[0044] The preset mode may specifically be a pet mode. For example, when the sensor system integrated in the air conditioner detects that no one is in the room, the pet smart house sensor further detects whether there is a pet in the room. If there is a pet, the air conditioner system switches to the pet mode.

[0045] Step S130: collecting an image of the pet in the preset mode.

[0046] For example, the camera of the air-conditioning device in the air-conditioning system can be used to collect indoor images outside the pet breeding device (such as a pet smart house). The camera of the air-conditioning device is turned on to monitor the pet's home situation in real time. At the same time, the camera of the pet breeding device is also started to monitor the pet's activities in real time. If the pet is outside the pet breeding device, the camera of the air-conditioning device in the air-conditioning system collects indoor images outside the pet breeding device; if the pet is inside the pet breeding device, the camera of the pet breeding device collects images inside the pet smart house.

[0047] Step S140, determining whether the pet has a preset behavior based on the collected image of the pet.

[0048] Preferably, the collected image of the pet can be uploaded to a server, and the server determines whether the pet has a preset behavior based on the collected image of the pet.

[0049] Figure 2A flowchart showing a specific implementation of the step of determining whether the pet has a preset behavior based on the collected image of the pet according to the present invention.

[0050] like Figure 2 As shown, in a specific implementation, step S140 includes steps S141 to S144.

[0051] Step S141, pre-processing the captured image of the pet.

[0052] In a specific implementation, preprocessing the pet image may include the following processing steps:

[0053] (1) De-noising the image of the pet.

[0054] Specifically, the noise in the pet image can be removed by using a Gaussian filter.

[0055] For image I(x,y), the processing formula of Gaussian filter is:

[0056]

[0057] Among them, G(i,j) is a two-dimensional Gaussian kernel, which is used to smooth the image and eliminate small noise interference.

[0058] I filtered (x,y) represents the pixel value of the image at coordinate (x,y) after Gaussian filtering; I(x+i,y+j) represents the pixel value of the input image at coordinate (x+i,y+j), which means the original pixel value of the area around the pixel participating in the operation under the action of the Gaussian kernel. G(i,j) is a two-dimensional Gaussian kernel function, which represents the weight of the Gaussian distribution on pixel (i,j).

[0059] The weights are defined by the Gaussian distribution formula: Among them, σ is the standard deviation of the Gaussian distribution, which controls the smoothing strength of the filter; (x, y) represents the coordinates of the current pixel, that is, the pixel being processed; i, j represent the offset relative to the current pixel in the filter window; the value range of i and j is [-k, k], that is, the window coverage area; k represents the radius of the filter window. Usually the window size is (2k+1)×(2k+1), for example, k=1 corresponds to a 3×3 filter window, and k=2 corresponds to a 5×5 filter window.

[0060] (2) Extracting the dynamic foreground in the image of the pet by frame difference method.

[0061] Specifically, the frame difference method is used to extract the dynamic foreground (i.e., the pet in motion) and eliminate background interference. The frame difference method detects moving objects based on the difference in pixels between consecutive frames. The formula is:

[0062] D t (x,y)=|I t (x,y)-I t-1 (x,y)

[0063] Among them, D t (x, y) is the pixel difference map between the tth frame and the t-1th frame; I t (x, y) represents the pixel value at coordinate (x, y) in the tth frame of the image sequence; I t-1 (x,y) represents the pixel value at coordinate (x,y) in the t-1th frame in the image sequence.

[0064] (3) Performing morphological operations on the extracted dynamic foreground.

[0065] Specifically, morphological operations are used to process the extracted foreground, eliminate small noise blocks, and maintain the integrity of the object.

[0066] Step S142: extracting key point posture vectors of the pet's body based on the preprocessed image of the pet.

[0067] Specifically, the pet's body posture can be estimated by extracting multiple key points (such as head, trunk, limbs, tail). Assume that the posture of each pet can be represented by a set of 2D (two-dimensional) coordinates P = {p1, p2, p3, ..., p n}, where p i =(x i ,y i ) are the pixel coordinates of the keypoints.

[0068] Figure 3 A flowchart of a specific implementation of the step of extracting the key point posture vector of the pet's body based on the preprocessed image of the pet according to the present invention is shown. Figure 3 As shown, in a specific implementation, step S142 includes steps S1421 to S1424.

[0069] Step S1421, input the preprocessed image of the pet into a deep convolutional neural network to extract a feature map containing structural information.

[0070] Specifically, the preprocessed image is input into a deep convolutional neural network (CNN), such as ResNet, high-level features are extracted from the input image, the image is gradually downsampled, and key structural information, such as contour and / or shape features, is extracted to obtain multiple feature maps.

[0071] Step S1422, mapping the extracted feature map into the heat map space to generate a heat map of each key point of the pet's body.

[0072] Heatmaps are used to represent the possible locations of each key point in an image. The heatmap of each key point is a grayscale image, and the grayscale value of each pixel in the image represents the probability or confidence that the key point appears at that location. The higher the grayscale value, the more likely the location is to be the key point.

[0073] Specifically, the feature map extracted by the convolutional neural network generates multiple heat maps through layer-by-layer convolution, pooling and upsampling. The heat map of each key point will be represented in a separate channel. Assuming there are n key points, n heat maps will be obtained, and each heat map specifically corresponds to a key point in the image.

[0074] In order to generate the heat map of each key point, a Gaussian distribution is applied to the actual position of the key point in the image, and the area around the key point position is also assigned a high confidence value. The heat map generated in this way can help the network to accurately locate in the local area. For example, assuming that the real position of a key point is at the image coordinate (x, y), on the generated heat map, (x, y) can be used as the center of the Gaussian distribution, and then a two-dimensional Gaussian distribution is generated, forming a heat area centered on (x, y). The grayscale value in this area is the highest, and gradually decreases as the distance from (x, y) increases.

[0075] The goal of network training is to continuously optimize the accuracy of each key point heat map through layer-by-layer convolution operations until it is close to the actual key point location. The loss function during training can be calculated based on the difference between the generated heat map and the actual heat map, for example,

[0076] It is usually calculated based on the mean square error (MSE) between the predicted heat map and the true heat map:

[0077]

[0078] Among them, H i is the true heat map of the i-th key point, is the predicted heat map, L is the loss value, which represents the average error between the model predicted heat map and the real heat map. The optimization goal is to minimize L, and n is the number of key points, which represents the total number of all key points in the image that need to be detected. For example, if a pet posture contains 12 key points, then n = 12. The squared difference in pixel values ​​between the predicted heatmap and the true heatmap for the i-th keypoint (squared Euclidean distance). The error of all n key points is averaged to make the loss value independent of the number of key points, which is convenient for training and evaluation.

[0079] Step S1423, extracting the coordinates of each joint point from the generated heat map of each key point of the pet's body.

[0080] In a specific implementation, non-maximum suppression (NMS) can be used to extract the coordinates of each key point from the heat map of each key point. Specifically, after generating the heat map of each joint point, the exact position (coordinates) of each key point can be extracted from the corresponding heat map by non-maximum suppression (NMS).

[0081] The process of extracting the coordinates of each key point from the heat map of each key point through non-maximum suppression (NMS) can specifically include: finding the pixel with the highest grayscale value (confidence) in the heat map as the coordinate of the key point. Applying suppression around this position ensures that only this position is retained as the only peak to avoid misidentification. In this way, the coordinates of each key point can be accurately located in the heat map.

[0082] Assume that the posture of a pet can be represented by a set of 12 key points (including head, torso, limbs, tail). In this case, the model will generate 12 heat maps, each corresponding to a key point (such as head, left front foot, right front foot, etc.). Finally, the model determines the precise coordinates of each key point by extracting the peak position from these heat maps.

[0083] Step S1424, converting the coordinates of each extracted key point into a key point posture vector of the pet's body.

[0084] The posture vector includes: the angle and distance between every two adjacent key points.

[0085] Specifically, in order to recognize the pet’s posture, the extracted key points are converted into a posture vector v, which is the angle and distance between every two adjacent key points. Assume there are two key points p i =(x i ,y i ) and p j =(x j ,y j ), then the Euclidean distance between them is d ij for:

[0086]

[0087] The angle θ they form ij for:

[0088]

[0089] These posture vectors are used as input features for action recognition.

[0090] Step S143, converting key point posture vectors extracted from the continuously acquired multiple frames of images of the pet into a time series.

[0091] The time series is used to represent the key point positions and changes of the pet in multiple continuous time frames. Specifically, the key point posture vectors extracted from the multiple frames of the pet image collected continuously can be spliced ​​into a time series matrix.

[0092] For example, the data of collected T frames are spliced ​​into a time series matrix:

[0093]

[0094] Among them, T is the length of the time series (number of frames), and n is the number of key points.

[0095] Step S144, inputting the time series into a pre-trained behavior recognition model to identify whether the pet has a preset behavior.

[0096] Specifically, the time series of key point posture vectors is input into a pre-trained behavior recognition model to obtain the behavior classification of the pet, so as to identify whether the pet has a preset behavior. The recognition model can identify "excretion behavior" or "non-excretion behavior", that is, the behavior recognition model can identify whether the pet has excretion behavior.

[0097] The behavior recognition model may specifically be a long short-term memory network (LSTM) model. A pre-labeled training data set may be obtained to perform long short-term memory network (LSTM) model training to obtain a behavior recognition model.

[0098] Each training sample includes the key point information of the pet's posture in multiple time frames. Specifically, it includes: the key point coordinates (x, y) of the pet's body extracted from the video or image sequence. For example, if the pet's body has 12 key points, the input vector dimension of each frame is 12×2=24, and T consecutive frames constitute an input time series with an input dimension of T×24.

[0099] Each input sample needs to have a corresponding behavior label: for example, 1 represents "excretion behavior" and 0 represents "non-excretion behavior". Labels can be generated by manually annotating video data.

[0100] The training data is a time series. We can extract continuous frames from the video captured by the camera in real time, and extract the key point posture vector p of the pet in each frame. i =(x i ,y i), concatenate the data of T frames into a time series matrix:

[0101]

[0102] Among them, T is the length of the time series (number of frames), n is the number of key points, and each time series S corresponds to a behavior label Y. Y∈{0,1}, 0 represents non-excretion behavior; 1 represents excretion behavior.

[0103] Since the defecation behavior of pets is a dynamic process, it usually requires multiple frames of data for analysis. Therefore, LSTM is used to capture the temporal dependency of sequence data. LSTM can handle long-term dependency problems through memory units and gating mechanisms.

[0104] Given an input sequence {v t}, the update process of LSTM is:

[0105] Forget Gate:

[0106] f t =σ(W f ·[h t-1 ,v t ]+b f )

[0107] Input Gate:

[0108] i t =σ(W i ·[h t-1 ,v t ]+b i )

[0109] Candidate status:

[0110]

[0111] Output Gate:

[0112] o t =σ(W o ·[h t-1 ,v t ]+b o )

[0113] Update hidden state:

[0114]

[0115] h t =o t *tanh(C t )

[0116] Among them, h t is the hidden state at time step t, Ct is the memory cell state, and σ is the sigmoid function.

[0117] The time series features h extracted by LSTM t , and finally sent to a fully connected layer, and the Softmax classifier is used to distinguish the behavior. Assuming that the classifier output category is y, the output probability is:

[0118]

[0119] Among them, W c and b c are the parameters of the Softmax classifier.

[0120] By classifying different action sequences, it is possible to determine whether the pet is in the state of excretion. Usually, excretion behavior includes the following features: bending over, lowering the head, squatting on the hind legs, etc. After the model learns these patterns in multiple training data sets, it can recognize the behavior in real-time images or video streams with high accuracy.

[0121] The performance of the image recognition algorithm is evaluated by accuracy, precision, recall and F1 value.

[0122] Accuracy:

[0123]

[0124] Accuracy:

[0125]

[0126] Recall:

[0127]

[0128] F1 value:

[0129]

[0130] Among them, TP, FP, TN, and FN represent the number of true positives, false positives, true negatives, and false negatives, respectively. Through the verification of a large number of test samples, the algorithm can achieve an accuracy rate of more than 90% and can effectively detect the excretion behavior of pets.

[0131] The above-mentioned image recognition algorithm can efficiently and accurately identify the excretion behavior of pets through image preprocessing, key point detection, posture estimation and temporal behavior classification.

[0132] Step S150: If it is determined that the pet has a preset behavior, the air conditioner and / or air purifier in the air conditioning system is controlled.

[0133] In a specific implementation, if it is determined that the pet has a preset behavior, the fresh air system of the air conditioner is started and the air purifier is turned on; the indoor air quality is detected by a sensor, and the running time of the fresh air system and the air purifier is controlled according to the detected air quality. Optionally, the server can determine whether the pet has a preset behavior based on the collected image of the pet. If it is determined that the pet has a preset behavior, the server sends a control instruction to the air conditioner to control the air conditioner and / or air purifier in the air conditioning system.

[0134] For example, when the server detects that the pet has excreted through the behavior recognition algorithm, the server notifies the air conditioner and sends the corresponding action command. After receiving the signal, the air conditioner starts the fresh air system and introduces outdoor air into the room through the ventilation device. At the same time, the air purifier is turned on or the air purification function of the air purifier is increased. At the same time, the air sensor of the air purifier detects the air quality of the environment, determines the persistence of indoor odors, and gradually adjusts the operation time of the fresh air system and air purifier through sensor feedback to ensure fresh air.

[0135] Step S160, obtaining the pet's hair humidity detection data in the preset mode.

[0136] In a specific embodiment, the pet's hair humidity can be detected by a sensor, for example, a sensor is worn by the pet to collect hair humidity detection data. The sensor has a micro humidity sensing circuit that can reflect the humidity of an object by measuring the resistance change on the surface of the object, and the hair humidity detection data can specifically be resistance change data collected by the sensor. For example, the pet wears a flexible collar sensor, the sensor fits the pet's hair and senses the change in resistance, so that the pet's hair humidity can be determined based on the resistance change data.

[0137] For example, the sensor includes a micro humidity sensing circuit, a main control unit, and a wireless communication module. The micro humidity sensing circuit includes a circuit for measuring resistance changes, which can collect data in real time and transmit it to the main control unit. The main control unit transmits the collected data to a pet breeding device (such as a pet smart house) or an air-conditioning system through a wireless communication module such as Bluetooth or Wi-Fi.

[0138] Step S170, judging the dryness or wetness state of the pet's hair according to the acquired hair humidity detection data of the pet.

[0139] Preferably, the obtained pet's hair humidity detection data can be uploaded to a server, and the server determines the pet's hair dryness and wetness state based on the obtained pet's hair humidity. For example, the obtained pet's hair humidity is uploaded to a server, and the server-side hair humidity state detection algorithm calculates the pet's hair dryness and wetness state. The pet's hair dryness and wetness state can specifically include: a wet state and a dry state.

[0140] Specifically, the resistance change data collected by the sensor is uploaded to the server, which analyzes and processes it. When the measured resistance value is lower than a certain value or decreases, it is judged that the pet's hair is in a wet state. When the measured resistance value is higher than a certain value or increases, it is judged that the pet's hair is in a dry state.

[0141] The sensor includes, for example, a hygroresistor, which is a sensor that generates resistance changes based on humidity changes. It is usually composed of a layer of humidity sensitive material and two electrodes. When the humidity changes, the resistance value of the sensor will change, thereby achieving humidity measurement.

[0142] Step S180, controlling the air conditioner and / or dehumidifier and / or humidifier in the air conditioning system according to the determined dryness or wetness state of the pet's hair.

[0143] In a specific embodiment, if the pet's hair is judged to be in a dry state, the humidifier in the air-conditioning system is controlled to operate, and / or the air conditioner is controlled to perform environmental humidity adjustment, and / or the dehumidifier in the air-conditioning system is controlled to turn off; if the pet's hair is judged to be in a wet state, the air conditioner in the air-conditioning system is controlled to turn on the dehumidification mode, and / or the dehumidifier in the air-conditioning system is controlled to start working, and / or the dehumidifier in the air-conditioning system is controlled to increase the dehumidification intensity.

[0144] Optionally, the obtained hair humidity of the pet can be uploaded to the server, and the server determines the dryness and wetness state of the pet's hair based on the obtained hair humidity of the pet. The server controls the air conditioner and / or dehumidifier and / or humidifier in the air conditioning system based on the determined dryness and wetness state of the pet's hair. For example, the pet's physical condition data will also be reported to the server in real time. When the server-side hair humidity state detection algorithm detects the dryness and wetness state of the pet's hair, the server sends an instruction to notify the air conditioner to increase the air humidity, and the air conditioner performs environmental humidity adjustment. At the same time, the air conditioner notifies the dehumidifier to shut down. When the hair is wet, the server sends an instruction to notify the air conditioner to turn on the dehumidification mode, and simultaneously notifies the dehumidifier to start, or increases the dehumidification intensity of the dehumidifier to reduce the humidity in the air and ensure the comfort of the pet.

[0145] In order to clearly illustrate the technical solution of the present invention, the execution process of the control method of the air-conditioning system provided by the present invention is described below with reference to a specific embodiment.

[0146] Figure 4 FIG. 1 is a schematic diagram of a method for controlling an air conditioning system according to a specific embodiment of the present invention. Figure 4 As shown, the air conditioner detects that there is no one in the room and switches to pet mode. The air conditioning system camera and the pet smart house camera monitor in real time. The server performs image recognition, behavior analysis, and body condition analysis. After identifying the pet's behavior, the server generates relevant executions and transmits them to the air conditioning equipment through the network. The air conditioner turns on the fresh air system and links the air purifier. The server analyzes the hair status, detects the dryness or wetness of the hair through an algorithm, and generates corresponding instructions. The air conditioning dehumidification function is adjusted and linked to the dehumidifier control.

[0147] The present invention also provides a control device for an air conditioning system. The air conditioning system includes: at least one of an air conditioner, an air purifier, a dehumidifier and a humidifier. The air conditioning system can be controlled in conjunction with a pet breeding device arranged indoors. The control device can be implemented in the controller of the air conditioner, for example, in the main control unit of the air conditioner controller. The air conditioner includes but is not limited to an edge computing unit, a perception unit, an image acquisition unit (including an image acquisition device such as a camera), a main control unit, and a wireless communication module. The edge computing unit processes sensor data in real time and calculates temperature sensitivity through an optimized algorithm; the perception unit senses whether there are people in the environment; the image acquisition unit is turned on when the air conditioner enters the pet mode, and is mainly used to monitor the behavior of pets that are not active in the smart house, and the pet behavior data is uploaded to the server in real time; the main control unit is responsible for logical scheduling and control; the wireless communication module unit is responsible for data transmission, communication, etc.

[0148] The air conditioner can communicate with the air purifier, dehumidifier and humidifier through a wireless network or through short-range wireless communication, such as Bluetooth or WiFi. The air conditioner can also communicate with the pet breeding device, such as through a wireless network or through short-range wireless communication.

[0149] The pet breeding device may be, for example, a pet smart house, and may have an image acquisition device (e.g., a camera) and / or a humidity sensor. The image acquisition device (e.g., a camera) captures real-time images and video data of the pet inside the pet breeding device for analyzing the pet's behavior, such as excretion behavior, eating behavior, etc.; the humidity sensor detects the environmental humidity inside the pet breeding device (inside the pet smart house) in real time.

[0150] Figure 5 FIG. 1 is a structural block diagram of an embodiment of a control device for an air conditioning system provided by the present invention. Figure 5As shown, the control device 100 includes: a detection unit 110 , a control unit 120 , and also includes: a collection unit 130 and a determination unit 140 , and / or also includes: an acquisition unit 160 and a judgment unit 170 .

[0151] The detection unit 110 is used to detect whether there is anyone in the room, and when it is detected that there is no one in the room, detect whether there is a pet in the room.

[0152] Specifically, whether there is someone in the room can be detected by the sensing module unit of the air conditioning system.

[0153] The perception module unit may specifically include a millimeter wave radar sensor. For example, when a 4D millimeter wave radar sensor detects whether there is someone in the room, it senses changes in the surrounding environment based on the reflected signals of the emitted and received electromagnetic waves. The data processing module of the millimeter wave sensor transmits the processed information to the main control unit of the air conditioning system, and the main control unit analyzes the detection data to determine whether there is someone in the room. The 4D millimeter wave radar measures the height data of the measured target through the ability to measure the pitch angle, so that it can distinguish whether it is a user or a pet by the height.

[0154] When it is detected that no one is in the room, the sensor of the pet breeding device is further used to detect whether there is a pet in the room. Specifically, the sensor of the pet breeding device (such as a pet smart house) can be used to specifically detect whether there is a pet in the room. The sensor is, for example, a millimeter wave radar sensor, which is similar to the method of detecting whether there is a person in the room in the above embodiment, and distinguishes that the detected one is a pet by height. The sensor is, for example, an image sensor, such as a camera, which recognizes whether there is a pet in the room by image.

[0155] The control unit 120 is used to control the air conditioner to switch to a preset mode if the detection unit detects that a pet is present indoors.

[0156] The preset mode may specifically be a pet mode. For example, when the sensor system integrated in the air conditioner detects that no one is in the room, the pet smart house sensor further detects whether there is a pet in the room. If there is a pet, the air conditioner system switches to the pet mode.

[0157] The acquisition unit 130 is used to acquire the image of the pet in the preset mode.

[0158] For example, the camera of the air-conditioning device in the air-conditioning system can be used to collect indoor images outside the pet breeding device (such as a pet smart house). The camera of the air-conditioning device is turned on to monitor the pet's home situation in real time. At the same time, the camera of the pet breeding device is also started to monitor the pet's activities in real time. If the pet is outside the pet breeding device, the camera of the air-conditioning device in the air-conditioning system collects indoor images outside the pet breeding device; if the pet is inside the pet breeding device, the camera of the pet breeding device collects images inside the pet smart house.

[0159] The determination unit 140 determines whether the pet has a preset behavior according to the image of the pet collected by the collection unit.

[0160] Preferably, the determination unit 140 may upload the collected image of the pet to a server, and the server may determine whether the pet has a preset behavior based on the collected image of the pet.

[0161] like Figure 2 As shown, in a specific implementation, the step of the determination unit 140 determining whether the pet has a preset behavior based on the collected image of the pet may specifically include steps S141 to S144.

[0162] Step S141, pre-processing the captured image of the pet.

[0163] In a specific implementation, preprocessing the pet image may include the following processing steps:

[0164] (1) De-noising the image of the pet.

[0165] Specifically, the noise in the pet image can be removed by using a Gaussian filter.

[0166] For image I(x,y), the processing formula of Gaussian filter is:

[0167]

[0168] Among them, G(i,j) is a two-dimensional Gaussian kernel, which is used to smooth the image and eliminate small noise interference.

[0169] I filtered (x,y) represents the pixel value of the image at coordinate (x,y) after Gaussian filtering; I(x+i,y+j) represents the pixel value of the input image at coordinate (x+i,y+j), which means the original pixel value of the area around the pixel participating in the operation under the action of the Gaussian kernel. G(i,j) is a two-dimensional Gaussian kernel function, which represents the weight of the Gaussian distribution on pixel (i,j).

[0170] The weights are defined by the Gaussian distribution formula: Among them, σ is the standard deviation of the Gaussian distribution, which controls the smoothing strength of the filter; (x, y) represents the coordinates of the current pixel, that is, the pixel being processed; i, j represent the offset relative to the current pixel in the filter window; the value range of i and j is [-k, k], that is, the window coverage area; k represents the radius of the filter window. Usually the window size is (2k+1)×(2k+1), for example, k=1 corresponds to a 3×3 filter window, and k=2 corresponds to a 5×5 filter window.

[0171] (2) Extracting the dynamic foreground in the image of the pet by frame difference method.

[0172] Specifically, the frame difference method is used to extract the dynamic foreground (i.e., the pet in motion) and eliminate background interference. The frame difference method detects moving objects based on the difference in pixels between consecutive frames. The formula is:

[0173] D t (x,y)=|I t (x,y)-I t-1 (x,y)

[0174] Among them, D t (x, y) is the pixel difference map between the tth frame and the t-1th frame; I t (x, y) represents the pixel value at coordinate (x, y) in the tth frame of the image sequence; I t-1 (x,y) represents the pixel value at coordinate (x,y) in the t-1th frame in the image sequence.

[0175] (3) Performing morphological operations on the extracted dynamic foreground.

[0176] Specifically, morphological operations are used to process the extracted foreground, eliminate small noise blocks, and maintain the integrity of the object.

[0177] Step S142: extracting key point posture vectors of the pet's body based on the preprocessed image of the pet.

[0178] Specifically, the pet's body posture can be estimated by extracting multiple key points (such as head, trunk, limbs, tail). Assume that the posture of each pet can be represented by a set of 2D coordinates P = {p1, p2, p3, ..., p n}, where p i =(x i ,y i ) are the pixel coordinates of the keypoints.

[0179] like Figure 3As shown, in a specific implementation, the determination unit 140 extracts the key point posture vector of the pet's body based on the preprocessed image of the pet, which may specifically include steps S1421 to S1424.

[0180] Step S1421, input the preprocessed image of the pet into a deep convolutional neural network to extract a feature map containing structural information.

[0181] Specifically, the preprocessed image is input into a deep convolutional neural network (CNN), such as ResNet, high-level features are extracted from the input image, the image is gradually downsampled, and key structural information, such as contour and / or shape features, is extracted to obtain multiple feature maps.

[0182] Step S1422, mapping the extracted feature map into the heat map space to generate a heat map of each key point of the pet's body.

[0183] Heatmaps are used to represent the possible locations of each key point in an image. The heatmap of each key point is a grayscale image, and the grayscale value of each pixel in the image represents the probability or confidence that the key point appears at that location. The higher the grayscale value, the more likely the location is to be the key point.

[0184] Specifically, the feature map extracted by the convolutional neural network generates multiple heat maps through layer-by-layer convolution, pooling and upsampling. The heat map of each key point will be represented in a separate channel. Assuming there are n key points, n heat maps will be obtained, and each heat map specifically corresponds to a key point in the image.

[0185] In order to generate the heat map of each key point, a Gaussian distribution is applied to the actual position of the key point in the image, and the area around the key point position is also assigned a high confidence value. The heat map generated in this way can help the network to accurately locate in the local area. For example, assuming that the real position of a key point is at the image coordinate (x, y), on the generated heat map, (x, y) can be used as the center of the Gaussian distribution, and then a two-dimensional Gaussian distribution is generated, forming a heat area centered on (x, y). The grayscale value in this area is the highest, and gradually decreases as the distance from (x, y) increases.

[0186] The goal of network training is to continuously optimize the accuracy of each key point heat map through layer-by-layer convolution operations until it is close to the actual key point location. The loss function during training can be calculated based on the difference between the generated heat map and the actual heat map, for example,

[0187] It is usually calculated based on the mean square error (MSE) between the predicted heat map and the true heat map:

[0188]

[0189] Among them, H i is the true heat map of the i-th key point, is the predicted heat map, L is the loss value, which represents the average error between the model predicted heat map and the real heat map. The optimization goal is to minimize L, and n is the number of key points, which represents the total number of all key points in the image that need to be detected. For example, if a pet posture contains 12 key points, then n = 12. The squared difference in pixel values ​​between the predicted heatmap and the true heatmap for the i-th keypoint (squared Euclidean distance). The error of all n key points is averaged to make the loss value independent of the number of key points, which is convenient for training and evaluation.

[0190] Step S1423, extracting the coordinates of each joint point from the generated heat map of each key point of the pet's body;

[0191] In a specific implementation, non-maximum suppression (NMS) can be used to extract the coordinates of each key point from the heat map of each key point. Specifically, after generating the heat map of each joint point, the exact position (coordinates) of each key point can be extracted from the corresponding heat map by non-maximum suppression (NMS).

[0192] The process of extracting the coordinates of each key point from the heat map of each key point through non-maximum suppression (NMS) can specifically include: finding the pixel with the highest grayscale value (confidence) in the heat map as the coordinate of the key point. Applying suppression around this position ensures that only this position is retained as the only peak to avoid misidentification. In this way, the coordinates of each key point can be accurately located in the heat map.

[0193] Assume that the posture of a pet can be represented by a set of 12 key points (including head, torso, limbs, tail). In this case, the model will generate 12 heat maps, each corresponding to a key point (such as head, left front foot, right front foot, etc.). Finally, the model determines the precise coordinates of each key point by extracting the peak position from these heat maps.

[0194] Step S1424, converting the coordinates of each extracted key point into a key point posture vector of the pet's body.

[0195] The posture vector includes: the angle and distance between every two adjacent key points.

[0196] Specifically, in order to recognize the pet’s posture, the extracted key points are converted into a posture vector v, which is the angle and distance between every two adjacent key points. Assume there are two key points p i =(x i ,y i ) and p j =(x j ,y j ), then the Euclidean distance between them is d ij for:

[0197]

[0198] The angle θ they form ij for:

[0199]

[0200] These posture vectors are used as input features for action recognition.

[0201] Step S143, converting key point posture vectors extracted from the continuously acquired multiple frames of images of the pet into a time series.

[0202] The time series is used to represent the key point positions and changes of the pet in multiple continuous time frames. Specifically, the key point posture vectors extracted from the multiple frames of the pet image collected continuously can be spliced ​​into a time series matrix.

[0203] For example, the data of collected T frames are spliced ​​into a time series matrix:

[0204]

[0205] Among them, T is the length of the time series (number of frames), and n is the number of key points.

[0206] Step S144, inputting the time series into a pre-trained behavior recognition model to identify whether the pet has a preset behavior.

[0207] Specifically, the time series of key point posture vectors is input into a pre-trained behavior recognition model to obtain the behavior classification of the pet, so as to identify whether the pet has a preset behavior. The recognition model can identify "excretion behavior" or "non-excretion behavior", that is, the behavior recognition model can identify whether the pet has excretion behavior.

[0208] The behavior recognition model may specifically be a long short-term memory network (LSTM) model. A pre-labeled training data set may be obtained to perform long short-term memory network (LSTM) model training to obtain a behavior recognition model.

[0209] Each training sample includes the key point information of the pet's posture in multiple time frames. Specifically, it includes: the key point coordinates (x, y) of the pet's body extracted from the video or image sequence. For example, if the pet's body has 12 key points, the input vector dimension of each frame is 12×2=24, and T consecutive frames constitute an input time series with an input dimension of T×24.

[0210] Each input sample needs to have a corresponding behavior label: for example, 1 represents "excretion behavior" and 0 represents "non-excretion behavior". Labels can be generated by manually annotating video data.

[0211] The training data is a time series. We can extract continuous frames from the video captured by the camera in real time, and extract the key point posture vector p of the pet in each frame. i =(x i ,y i ), concatenate the data of T frames into a time series matrix:

[0212]

[0213] Among them, T is the length of the time series (number of frames), n is the number of key points, and each time series S corresponds to a behavior label Y. Y∈{0,1}, 0 represents non-excretion behavior; 1 represents excretion behavior.

[0214] Since the defecation behavior of pets is a dynamic process, it usually requires multiple frames of data for analysis. Therefore, LSTM is used to capture the temporal dependency of sequence data. LSTM can handle long-term dependency problems through memory units and gating mechanisms.

[0215] The image recognition algorithm can efficiently and accurately identify the excretion behavior of pets through image preprocessing, key point detection, posture estimation and temporal behavior classification.

[0216] The control unit 120 is further configured to: if the determination unit determines that the pet has a preset behavior, control the air conditioner and / or air purifier in the air conditioning system.

[0217] In a specific implementation, if it is determined that the pet has a preset behavior, the fresh air system of the air conditioner is started and the air purifier is turned on; the indoor air quality is detected by a sensor, and the running time of the fresh air system and the air purifier is controlled according to the detected air quality. Optionally, the server can determine whether the pet has a preset behavior based on the collected image of the pet. If it is determined that the pet has a preset behavior, the server sends a control instruction to the air conditioner to control the air conditioner and / or air purifier in the air conditioning system.

[0218] For example, when the server detects that the pet has excreted through the behavior recognition algorithm, the server notifies the air conditioner and sends the corresponding action command. After receiving the signal, the air conditioner starts the fresh air system and introduces outdoor air into the room through the ventilation device. At the same time, the air purifier is turned on or the air purification function of the air purifier is increased. At the same time, the air sensor of the air purifier detects the air quality of the environment, determines the persistence of indoor odors, and gradually adjusts the operation time of the fresh air system and air purifier through sensor feedback to ensure fresh air.

[0219] The acquisition unit 160 is used to acquire the pet's hair humidity detection data in the preset mode.

[0220] In a specific embodiment, the pet's hair humidity can be detected by a sensor, for example, a sensor is worn by the pet to collect hair humidity detection data. The sensor has a micro humidity sensing circuit that can reflect the humidity of an object by measuring the resistance change on the surface of the object, and the hair humidity detection data can specifically be resistance change data collected by the sensor. For example, the pet wears a flexible collar sensor, the sensor fits the pet's hair and senses the change in resistance, so that the pet's hair humidity can be determined based on the resistance change data.

[0221] For example, the sensor includes a micro humidity sensing circuit unit, a main control unit, and a wireless communication module. The micro humidity sensing circuit includes a circuit for measuring resistance changes, which can collect data in real time and transmit it to the main control unit. The main control unit transmits the collected data to a pet breeding device (such as a pet smart house) or an air-conditioning system through a wireless communication module such as Bluetooth or Wi-Fi.

[0222] The judging unit 170 judges the dryness or wetness state of the pet's hair according to the pet's hair wetness detection data acquired by the acquiring unit.

[0223] Preferably, the obtained hair humidity of the pet can be uploaded to a server, and the server determines the dryness and wetness state of the pet's hair according to the obtained hair humidity of the pet. For example, the obtained hair humidity of the pet is uploaded to a server, and the hair humidity state detection algorithm on the server side calculates the dryness and wetness state of the pet's hair. The dryness and wetness state of the pet's hair can specifically include: a wet state and a dry state.

[0224] Specifically, the resistance change data collected by the sensor is uploaded to the server, which analyzes and processes it. When the measured resistance value is lower than a certain value or decreases, it is judged that the pet's hair is in a wet state. When the measured resistance value is higher than a certain value or increases, it is judged that the pet's hair is in a dry state.

[0225] The sensor includes, for example, a hygroresistor, which is a sensor that generates resistance changes based on humidity changes. It is usually composed of a layer of humidity sensitive material and two electrodes. When the humidity changes, the resistance value of the sensor will change, thereby achieving humidity measurement.

[0226] The control unit 120 is further used to control the air conditioner and / or dehumidifier and / or humidifier in the air conditioning system according to the wetness and dryness state of the pet's hair determined by the determination unit 170.

[0227] In a specific embodiment, if the pet's hair is judged to be in a dry state, the humidifier in the air-conditioning system is controlled to operate, and / or the air conditioner is controlled to perform environmental humidity adjustment, and / or the dehumidifier in the air-conditioning system is controlled to turn off; if the pet's hair is judged to be in a wet state, the air conditioner in the air-conditioning system is controlled to turn on the dehumidification mode, and / or the dehumidifier in the air-conditioning system is controlled to start working, and / or the dehumidifier in the air-conditioning system is controlled to increase the dehumidification intensity.

[0228] Optionally, the obtained hair humidity of the pet can be uploaded to the server, and the server determines the dryness and wetness state of the pet's hair based on the obtained hair humidity of the pet. The server controls the air conditioner and / or dehumidifier and / or humidifier in the air conditioning system based on the determined dryness and wetness state of the pet's hair. For example, the pet's physical condition data will also be reported to the server in real time. When the server-side hair humidity state detection algorithm detects the dryness and wetness state of the pet's hair, the server sends an instruction to notify the air conditioner to increase the air humidity, and the air conditioner performs environmental humidity adjustment. At the same time, the air conditioner notifies the dehumidifier to shut down. When the hair is wet, the server sends an instruction to notify the air conditioner to turn on the dehumidification mode, and simultaneously notifies the dehumidifier to start, or increases the dehumidification intensity of the dehumidifier to reduce the humidity in the air and ensure the comfort of the pet.

[0229] The present invention also provides a storage medium corresponding to the control method of the air-conditioning system, on which a computer program is stored, and when the program is executed by a processor, the steps of any of the aforementioned methods are implemented.

[0230] The present invention also provides an air conditioner corresponding to the control method of the air conditioning system, comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any of the aforementioned methods when executing the program.

[0231] The present invention also provides an air conditioner corresponding to the control device of the air conditioning system, comprising any of the control devices described above.

[0232] The present invention also provides a computer program product corresponding to the control method of the air-conditioning system, comprising a computer program, and when the computer program is executed by a processor, the steps of any of the aforementioned methods are implemented.

[0233] Based on this, the solution provided by the present invention can intelligently identify the needs of pets, and adjust the air quality and temperature and humidity accordingly according to the pet's behavior and hair condition, thereby improving the comfort and health protection of the pet's indoor life.

[0234] According to the technical solution of the present invention, when the air conditioner detects that there is no one indoors but there is a pet, it automatically switches to pet mode. The air conditioning system monitors the behavior of the pet and controls the air conditioning system according to the behavior of the pet. When the pet is detected to have excretion behavior, the air conditioning system starts the fresh air deodorization function; according to the detected dry and wet state of the pet, the air conditioning system intelligently adjusts the ambient humidity.

[0235] The functions described herein may be implemented in hardware, software executed by a processor, firmware, or any combination thereof. If implemented in software executed by a processor, the functions may be stored as one or more instructions or codes on a computer-readable medium or transmitted via a computer-readable medium. Other examples and implementations are within the scope and spirit of the present invention and the appended claims. For example, due to the nature of software, the functions described above may be implemented using software executed by a processor, hardware, firmware, hard wiring, or a combination of any of these. In addition, each functional unit may be integrated into a processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit.

[0236] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0237] The units described as separate components may or may not be physically separated, and the components of the control device may or may not be physical units, that is, they may be located in one place or distributed in multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0238] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the relevant technology or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk and other media that can store program codes.

[0239] The above description is only an embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention shall be included in the scope of the claims of the present invention.

Claims

1. A control method for an air conditioning system, characterized in that: The air conditioning system comprises: at least one of an air conditioner, an air purifier, a dehumidifier and a humidifier; the control method comprises: Detect whether there is anyone in the room. If no one is detected in the room, detect whether there is a pet in the room; If the presence of a pet is detected indoors, the air conditioner is controlled to switch to a preset mode; In the preset mode, collecting an image of the pet; Determining whether the pet has a preset behavior based on the collected image of the pet; If it is determined that the pet has a preset behavior, controlling the air conditioner and / or air purifier in the air conditioning system; and / or, In the preset mode, obtaining the pet's hair humidity detection data; Determining the dryness and wetness state of the pet's hair according to the acquired hair humidity detection data of the pet; The air conditioner and / or dehumidifier and / or humidifier in the air conditioning system are controlled according to the determined dryness and wetness state of the pet's hair.

2. The method according to claim 1, characterized in that Determining whether the pet has a preset behavior based on the collected image of the pet includes: Uploading the collected image of the pet to a server, and the server determining whether the pet has a preset behavior based on the collected image of the pet; and / or, Judging the dryness and wetness state of the pet's hair according to the acquired hair humidity detection data of the pet includes: The acquired hair humidity detection data of the pet is uploaded to a server, and the server determines the dryness or wetness state of the pet's hair according to the acquired hair humidity detection data of the pet.

3. The method according to claim 1 or 2, characterized in that: Determining whether the pet has a preset behavior based on the collected image of the pet includes: Preprocessing the captured image of the pet; Extracting key point posture vectors of the pet's body based on the preprocessed image of the pet; Converting key point pose vectors extracted from a plurality of frames of images of the pet acquired continuously into a time series; The time series is input into a pre-trained behavior recognition model to identify whether the pet has a preset behavior.

4. The method according to claim 3, characterized in that The pet image is preprocessed, including: Performing denoising on the image of the pet; Extracting the dynamic foreground in the image of the pet by using a frame difference method; Perform morphological operations on the extracted dynamic foreground.

5. The method according to claim 3, characterized in that: Based on the preprocessed image of the pet, extracting the key point posture vector of the pet's body, including: Inputting the preprocessed image of the pet into a deep convolutional neural network to extract a feature map containing structural information; Mapping the extracted feature map into a heat map space to generate a heat map of each key point of the pet's body; Extracting the coordinates of each joint point from the generated heat map of each key point of the pet's body; The coordinates of each extracted key point are converted into a key point posture vector of the pet's body.

6. The method according to claim 1, characterized in that If it is determined that the pet has a preset behavior, the air conditioner and / or air purifier in the air conditioning system is controlled, including: If it is determined that the pet has the preset behavior, the fresh air system of the air conditioner is started and the air purifier is turned on; Detecting indoor air quality through sensors, and controlling the operating time of the fresh air system and the air purifier according to the detected air quality; and / or, The wet and dry state of the pet's hair includes: a wet state and a dry state; according to the determined wet and dry state of the pet's hair, the air conditioner and / or the dehumidifier and / or the humidifier in the air conditioning system is controlled, including: If it is determined that the pet's hair is in a dry or wet state, the humidifier in the air-conditioning system is controlled to operate, and / or the air-conditioner is controlled to perform environmental humidity adjustment, and / or the dehumidifier in the air-conditioning system is controlled to shut down; If it is determined that the pet's hair is wet, the air conditioner in the air-conditioning system is controlled to turn on the dehumidification mode, and / or the dehumidifier in the air-conditioning system is controlled to start working, and / or the dehumidifier in the air-conditioning system is controlled to increase the dehumidification intensity.

7. A control device for an air conditioning system, characterized in that: The air conditioning system comprises: at least one of an air conditioner, an air purifier, a dehumidifier and a humidifier; the control device comprises: A detection unit is used to detect whether there is anyone in the room, and when it is detected that there is no one in the room, detect whether there is a pet in the room; A control unit, configured to control the air conditioner to switch to a preset mode if the detection unit detects that a pet is present indoors; A collection unit, used for collecting the image of the pet in the preset mode; a determination unit, for determining whether the pet has a preset behavior according to the image of the pet collected by the collection unit; The control unit is further configured to: control the air conditioner and / or air purifier in the air conditioning system if the determination unit determines that the pet has a preset behavior; and / or, An acquisition unit, configured to acquire the pet's hair humidity detection data in the preset mode; a judgment unit, for judging the dryness or wetness state of the pet's hair according to the pet's hair humidity detection data acquired by the acquisition unit; The control unit is further used to control the air conditioner and / or dehumidifier and / or humidifier in the air conditioning system according to the dryness and wetness status of the pet's hair determined by the determination unit.

8. A storage medium, characterized in that: A computer program is stored thereon, and when the program is executed by a processor, the steps of any method described in claims 1-6 are implemented.

9. An air conditioner, characterized in that: The method comprises a processor, a memory and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any method described in claims 1 to 6 when executing the program, or comprises the control device described in claim 7.

10. A computer program product, characterized in that The invention comprises a computer program, which implements the steps of any one of the methods of claims 1 to 6 when executed by a processor.

Images