CN114183881B - Intelligent thermal comfort control method based on visual assistance - Google Patents

Intelligent thermal comfort control method based on visual assistance Download PDF

Info

Publication number
CN114183881B
CN114183881B CN202210132584.2A CN202210132584A CN114183881B CN 114183881 B CN114183881 B CN 114183881B CN 202210132584 A CN202210132584 A CN 202210132584A CN 114183881 B CN114183881 B CN 114183881B
Authority
CN
China
Prior art keywords
person
thermal comfort
personnel
thermal
clothing
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202210132584.2A
Other languages
Chinese (zh)
Other versions
CN114183881A (en
Inventor
戴昌志
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Jiangsu Hengweizhi Information Technology Co ltd Changzhou Economic Development Branch
Original Assignee
Jiangsu Hengweizhi Information Technology Co ltd Changzhou Economic Development Branch
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Jiangsu Hengweizhi Information Technology Co ltd Changzhou Economic Development Branch filed Critical Jiangsu Hengweizhi Information Technology Co ltd Changzhou Economic Development Branch
Priority to CN202210132584.2A priority Critical patent/CN114183881B/en
Publication of CN114183881A publication Critical patent/CN114183881A/en
Application granted granted Critical
Publication of CN114183881B publication Critical patent/CN114183881B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F24—HEATING; RANGES; VENTILATING
    • F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
    • F24F11/00—Control or safety arrangements
    • F24F11/30—Control or safety arrangements for purposes related to the operation of the system, e.g. for safety or monitoring
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F24—HEATING; RANGES; VENTILATING
    • F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
    • F24F11/00—Control or safety arrangements
    • F24F11/50—Control or safety arrangements characterised by user interfaces or communication
    • F24F11/56—Remote control
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F24—HEATING; RANGES; VENTILATING
    • F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
    • F24F11/00—Control or safety arrangements
    • F24F11/62—Control or safety arrangements characterised by the type of control or by internal processing, e.g. using fuzzy logic, adaptive control or estimation of values
    • F24F11/63—Electronic processing
    • F24F11/64—Electronic processing using pre-stored data
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F24—HEATING; RANGES; VENTILATING
    • F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
    • F24F11/00—Control or safety arrangements
    • F24F11/70—Control systems characterised by their outputs; Constructional details thereof
    • F24F11/72—Control systems characterised by their outputs; Constructional details thereof for controlling the supply of treated air, e.g. its pressure
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F24—HEATING; RANGES; VENTILATING
    • F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
    • F24F11/00—Control or safety arrangements
    • F24F11/70—Control systems characterised by their outputs; Constructional details thereof
    • F24F11/80—Control systems characterised by their outputs; Constructional details thereof for controlling the temperature of the supplied air
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F24—HEATING; RANGES; VENTILATING
    • F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
    • F24F11/00—Control or safety arrangements
    • F24F11/88—Electrical aspects, e.g. circuits
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F24—HEATING; RANGES; VENTILATING
    • F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
    • F24F2110/00—Control inputs relating to air properties
    • F24F2110/10—Temperature
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F24—HEATING; RANGES; VENTILATING
    • F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
    • F24F2110/00—Control inputs relating to air properties
    • F24F2110/20—Humidity
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F24—HEATING; RANGES; VENTILATING
    • F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
    • F24F2120/00—Control inputs relating to users or occupants
    • F24F2120/10—Occupancy
    • F24F2120/12—Position of occupants
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F24—HEATING; RANGES; VENTILATING
    • F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
    • F24F2120/00—Control inputs relating to users or occupants
    • F24F2120/20—Feedback from users

Abstract

The invention discloses a thermal comfort intelligent control method based on visual assistance, which comprises the following steps: s1, acquiring the current activity area of the person X and acquiring the image information of the person X through a visual sensor; s2, acquiring a personnel positioning frame and a personnel label of a personnel X through a personnel detector; s3, obtaining the activity type and the metabolic equivalent met corresponding to the person X; s4, obtaining clothing thermal resistance clo of the person X; s5, determining a classification model corresponding to the person X according to the identity information; s6, acquiring environmental parameters of the activity area; s7, inputting the metabolic equivalent met, the clothing thermal resistance clo and the environmental parameters of the person X into a classification model, and outputting the thermal comfort level of the person X by the classification model so as to adjust the air conditioning system. The invention can intelligently adjust the air conditioning system of the human living environment in real time, improve the accuracy of prediction and meet the individual requirements of personnel through online learning.

Description

Intelligent thermal comfort control method based on visual assistance
Technical Field
The invention relates to the technical field of thermal environment control, in particular to an intelligent thermal comfort control method based on visual assistance.
Background
The indoor hot and wet environment refers to a physical environment formed by the most intuitive indoor temperature and relative humidity sensed by people, the feeling of people to the indoor hot and wet environment is called as the indoor hot and wet environment thermal comfort level (thermal comfort level for short), and the quantitative standard of the thermal comfort level is called as a thermal comfort index.
In the prior art, the thermal comfort degree is judged by mainly utilizing a thermal comfort degree index PMV equation to calculate a PMV value, but due to the complexity, nonlinearity and time lag existing in the calculation of the thermal comfort degree index PMV equation, accurate evaluation cannot be carried out, and further accurate control on the thermal comfort degree index PMV cannot be carried out.
For example, chinese patent No. CN201711124386.7 discloses a PMV control method for centralized air conditioning thermal comfort fused with image information, which utilizes computer vision technology to analyze and process people in an indoor environment, obtains load change, fresh air volume demand change and people dressing situation caused by the change of the number of people in a building space, establishes a PMV model in a dynamic environment based on these factors, and is used for controlling an air conditioner, so as to quickly meet the requirements of people in the indoor environment on thermal comfort. However, this method has at least the following drawbacks: (1) in the scheme, the image information is only used for extracting the crowd density, and more parameters influencing the thermal comfort degree are not obtained; (2) according to the scheme, the clothing thermal resistance calculation estimates the proportion of a rectangular frame through a skin complexion model, misjudgment is easy to occur on the conditions that the clothing color of the skin color of a person is similar, and meanwhile, the error of the estimation of the area of a human body by the rectangular frame is large.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the control method aims to solve the technical problem that the accuracy rate of judging the thermal comfort level of the human body is low in the control method in the prior art. The invention provides a thermal comfort intelligent control method based on visual assistance, which is characterized in that more information influencing human thermal sensation is acquired in real time through the visual assistance to serve as the input of a classification model, and the deviation of the classification model is corrected in an online learning mode, so that the accuracy of thermal comfort prediction can be improved, and the personalized requirements can be met.
The technical scheme adopted by the invention for solving the technical problems is as follows: a thermal comfort intelligent control method based on visual assistance comprises the following steps: s1, acquiring the current activity area of the person X through a visual sensor and acquiring the image information of the person X; s2, acquiring a personnel positioning frame and a personnel label of the personnel X through a personnel detector; s3, acquiring an activity type and a metabolic equivalent met corresponding to the person X through the person positioning frame and the image information of the person X; s4, acquiring the clothing thermal resistance clo of the person X through the person positioning frame and the image information of the person X; s5, acquiring identity information of the person X through the image information of the person X, and determining a classification model corresponding to the person X according to the identity information; s6, acquiring the environmental parameters of the activity area; s7, inputting the personnel label, the metabolic equivalent met, the clothing thermal resistance clo and the environmental parameters of the personnel X into the classification model, outputting the thermal comfort level of the personnel X by the classification model, and setting air conditioning parameters according to the thermal comfort level.
According to the invention, the activity area of a person is detected in real time through the visual sensor, the video image of the person is collected, the video image is processed and analyzed, the identity information, the person label, the metabolic equivalent met and the clothing thermal resistance clo of the person can be obtained, the cold and hot requirements of the person can be effectively predicted by combining the environmental parameters, the prediction accuracy of the classification model is improved through an efficient online learning mode, and the individual requirements of the person are met.
Further, the training process of the classification model comprises the following steps:
s10, using data output by the existing thermal comfort prediction model as training data;
s20, pre-training the classifier by using the training data to obtain a first classification model;
s30, inputting a personnel label, a metabolic equivalent met, a clothing thermal resistance clo and environmental parameters of the personnel X into the first classification model, wherein the first classification model outputs a first thermal comfort level of the personnel X;
s40, if the personnel X adjust the air-conditioning temperature control system, the fact that the real thermal comfort level of the personnel X has deviation from the first thermal comfort level is indicated, and the real thermal comfort level of the personnel X is recorded;
s50, using the personnel label, the metabolic equivalent met, the clothing thermal resistance clo, the environmental parameter and the real thermal comfort level of the personnel X as a group of real samples; respectively adding the metabolic equivalent met, the clothing thermal resistance clo and the environmental parameters to carry out positive or negative random disturbance, and generating a batch of new training samples;
and S60, retraining the first classification model by using the new training sample to obtain the classification model of the person X.
Further, if there is a deviation between the real thermal comfort level of the person X and the thermal comfort level outputted from the classification model obtained in step S60, steps S40-S60 are repeated.
Further, the step S50 of generating a batch of new training samples specifically includes:
the environmental parameters at least comprise temperature T1, radiation temperature T2 and wind speed F,
respectively acquiring data corresponding to the metabolic equivalent met, the clothing thermal resistance clo, the temperature T1, the radiation temperature T2 and the wind speed F;
if the real thermal comfort level of the person X is higher than the first thermal comfort level, respectively adding positive random disturbance to the metabolic equivalent met, the clothing thermal resistance clo, the temperature T1 and the radiation temperature T2, and adding negative random disturbance to the wind speed F to obtain a batch of new training samples;
if the real thermal comfort level of the person X is lower than the first thermal comfort level, adding negative random disturbance to the metabolic equivalent met, the clothing thermal resistance clo, the temperature T1 and the radiation temperature T2, and adding positive random disturbance to the wind speed F to obtain a batch of new training samples.
Further, the new training sample includes a plurality of sets of new training data, each set of new training data includes the metabolic equivalent met, the clothing thermal resistance clo, and the environmental parameter, and the classification result of each set of new training data corresponds to the real thermal comfort level of the person X.
Further, the people tags include gender, BMI rating, and age rating of person X.
Further, if the personnel detector does not detect personnel activity for n minutes continuously, the personnel detector outputs a control signal to turn off the air conditioner, wherein n is an integer larger than 15.
Further, in step S3, obtaining the activity type and the metabolic equivalent met corresponding to the person X specifically includes:
establishing a metabolic data set of a mapping relation between the activity type of the person and the metabolic level;
inputting the image information of the personnel positioning frame and the personnel X into a SLOWFAST neural network model, and outputting the activity type of the personnel X;
and matching the activity type of the person X with the data in the metabolic data set to obtain the metabolic equivalent met of the person X.
Further, in step S4, acquiring the clothing thermal resistance clo of the person X specifically includes:
establishing a thermal resistance data set of a mapping relation between the clothing of the person and the thermal resistance of the clothing;
acquiring dresses of the person X, namely head and neck dresses, upper body dresses, lower body dresses and foot dresses, according to the person positioning frame and the image information of the person X;
matching the head and neck dressing, the upper body dressing, the lower body dressing and the foot dressing with the data in the thermal resistance data set respectively to obtain four thermal resistance data corresponding to the head and neck dressing, the upper body dressing, the lower body dressing and the foot dressing;
and adding and calculating the four thermal resistance data to obtain the clothing thermal resistance clo of the person X.
The beneficial effect of the invention is that,
according to the intelligent thermal comfort control method based on visual assistance, the activity area of a person and the video image of the person can be collected in real time through the visual sensor, the person label, the metabolic equivalent met, the clothing thermal resistance clo and the identity information of the person can be obtained through analysis and processing of the image information, the personalized classification model belonging to the person can be taken out according to the identity information, the environmental parameter of the current activity area is obtained, and then the person label, the metabolic equivalent met, the clothing thermal resistance clo and the environmental parameter are input into the classification model together, so that the current thermal comfort level of the person can be predicted, the air conditioner is intelligently controlled, and the cold and heat requirements of different persons are met. In addition, the classification model can be updated in real time through the personalized online training of the classification model, and the prediction accuracy of the classification model is improved.
Drawings
The invention is further illustrated with reference to the following figures and examples.
Fig. 1 is a flow chart of the intelligent thermal comfort control method based on visual assistance of the invention.
FIG. 2 is a flow chart of the classification model training process of the present invention.
Detailed Description
The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams each illustrating the basic structure of the present invention only in a schematic manner, and thus show only the constitution related to the present invention.
In the description of the present invention, it is to be understood that the terms "central," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," "circumferential," and the like are used in the orientations and positional relationships indicated in the drawings for convenience in describing the invention and to simplify the description, and are not intended to indicate or imply that the referenced devices or elements must have a particular orientation, be constructed and operated in a particular orientation, and are therefore not to be considered limiting of the invention. Furthermore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of the present invention, "a plurality" means two or more unless otherwise specified.
In the description of the present invention, it should be noted that, unless otherwise explicitly specified or limited, the terms "mounted," "connected," and "connected" are to be construed broadly, e.g., as meaning either a fixed connection, a removable connection, or an integral connection; can be mechanically or electrically connected; they may be connected directly or indirectly through intervening media, or they may be interconnected between two elements. The specific meanings of the above terms in the present invention can be understood in specific cases to those skilled in the art.
As shown in fig. 1, a thermal comfort intelligent control method based on visual assistance includes the following steps.
And S1, acquiring the current activity area of the person X through the vision sensor and acquiring the image information of the person X.
Specifically, the visual sensor may be, for example, a camera, a 3D camera, or the like, and the visual sensor may monitor an activity area where a person is currently located, and the activity area may be, for example, an office, a living space, a vehicle, or the like, where cooling and heating adjustment is required. The visual sensor can also collect image information of people, including human face features, clothing features, limb features and the like. Notably, the vision sensors are monitored and acquired in real time.
And S2, acquiring the personnel positioning frame and the personnel label of the personnel X through the personnel detector.
Specifically, a single frame of image collected by the vision sensor is input into the people detector, and the people detector can output a corresponding people positioning frame to position people in the video or the image, in other words, the people detector can capture people in the video image. Meanwhile, the person detector may output a person tag corresponding to the person according to the single frame image, the person tag including sex (male, female), BMI rating (low, medium, high), and age rating (child, adult, old man) of the person. For example, the person tag for person X is "female, middle, adult". Because different sexes, different BMI indexes, people of different ages are different to the impression of heat comfort level, and this embodiment is through carrying out the classification label to personnel, carries out the prediction of heat comfort level to different personnel again, not only can improve the accuracy of prediction, can also satisfy different people's individualized demand. If the human detector does not detect human activity for n minutes continuously, the control signal is output to turn off the air conditioner, wherein n is an integer larger than 15, such as 20 minutes, 30 minutes and the like, so that energy waste can be avoided.
Further, the personnel tags can be classified through a fasterncn network model, the fasterncn network model is a model trained in advance by using the marked image data, and three classification heads are set for output, and correspond to gender, BMI and age respectively. Alternatively, other network models may be used to separately train the classifications of gender, BMI, and age.
S3, obtaining the activity type and the metabolic equivalent met corresponding to the person X through the person positioning frame and the image information of the person X.
The method specifically comprises the following steps: establishing a metabolic data set of a mapping relation between the activity type of the person and the metabolic level; inputting the image information of the personnel positioning frame and the personnel X into a SLOWFAST neural network model, and outputting the activity type of the personnel X; and matching the activity type of the person X with the data in the metabolic data set to obtain the metabolic equivalent met of the person X.
It should be noted that the human metabolic level is closely related to the current activity state of a person, for example, the metabolic capacity during exercise is increased several times to dozens times than during rest, the metabolic capacity during walking is increased three times to five times than during rest, and the metabolic capacity during running is increased ten times to two hundred times than during rest. Thus, the metabolic levels of a person in different activity types are different, so that the sensation of thermal comfort is also different. Therefore, it is necessary to establish a metabolic data set of mapping relationship between the activity type and the metabolic level of the person, for example, 0.95METs for sleeping, 1.3METs for watching tv while sitting, 3.8METs for exercising, and so on. In other words, the metabolic data set contains metabolic equivalents corresponding to human activities.
And inputting the person positioning frame and the video image output by the person detector into a SLOWFAST network model, wherein the SLOWFAST network model can output an activity type corresponding to the person. The Slow channel may be used for analyzing static content in the video, and the Fast channel may be used for analyzing dynamic content in the video, so that on one hand, the processing accuracy may be improved, and on the other hand, the processing speed may be improved. After obtaining the activity type corresponding to the person, matching the activity type with the data in the metabolic data set, and obtaining the metabolic equivalent met of the person, for example, if the activity type of person X is sleeping, the metabolic equivalent met of person X is 0.95 METs.
And S4, acquiring the clothing thermal resistance clo of the person X through the person positioning frame and the image information of the person X.
The method specifically comprises the following steps: establishing a thermal resistance data set of a mapping relation between the clothing of the person and the thermal resistance of the clothing; acquiring dresses of the person X into head and neck dresses, upper body dresses, lower body dresses and foot dresses according to the person positioning frame and the image information of the person X; matching the head and neck dressing, the upper body dressing, the lower body dressing and the foot dressing with data in the thermal resistance data set respectively to obtain four thermal resistance data corresponding to the head and neck dressing, the upper body dressing, the lower body dressing and the foot dressing; and adding and calculating the four thermal resistance data to obtain the clothing thermal resistance clo of the person X.
It should be noted that the thermal resistance of the garment affects the evaporation of the skin surface of the human body. On the one hand, the diffusion of water vapor from the surface of the human skin by the garment creates an additional resistance, and on the other hand, the garment absorbs a portion of the perspiration, resulting in only a portion of the perspiration being able to evaporate and thereby lower the temperature of the skin surface. The thermal resistance of the clothing is an index of the heat preservation performance of the clothing. The thermal comfort of a person is affected by the difference in thermal resistance of the person's clothing. In this embodiment, a thermal resistance data set of a mapping relationship between the thermal resistances of the person's clothing and the clothing is first established, for example, the thermal resistance corresponding to shorts is 0.06clo, the thermal resistance corresponding to jacket is 0.35clo, the thermal resistance corresponding to short-sleeved shirt is 0.09clo, the thermal resistance corresponding to boot is 0.05clo, the thermal resistance corresponding to long-sleeve thick-sleeve high-collar sweater is 0.37clo, and the like. Then, the image information of the person positioning frame and the person is input into a RESNET or DARKNET network model, wherein the RESNET or DARKNET network model of the embodiment is a pre-trained model. It should be noted that the image information input into the RESNET or the DARKNET network model is a clipped image, for example, a complete image is divided into four small images according to the head and neck, the upper body, the lower body and the feet, and then the four small images are input into the RESNET or the DARKNET network model, the RESNET or the DARKNET network model can output four dressing classifications (head and neck dressing, upper body dressing, lower body dressing and foot dressing, respectively), for example, the output dressing classification is: the head is provided with a hat, the upper body is provided with a shirt, the lower body is provided with trousers, and the feet are provided with slippers. And then, the output head neck dresses, upper body dresses, lower body dresses and foot dresses are respectively matched with the data in the thermal resistance data set to obtain the segmented heat (namely four thermal resistances) of the person dresses. And finally, adding the obtained segmented thermal resistances to obtain the clothing thermal resistance clo of the person. In the embodiment, the personnel image is cut and segmented according to the body part, the dressing thermal resistances of different parts are respectively output, and then the dressing thermal resistances of different parts are added to obtain the total clothing thermal resistance clo, so that the accuracy of judging the clothing thermal resistance can be improved, and the misjudgment can be prevented.
S5, obtaining the identity information of the person X through the image information of the person X, and determining the classification model corresponding to the person X according to the identity information.
Specifically, firstly, a person image is subjected to face marking, then a face part is cut out, the cut face image is preprocessed, then face features are extracted through CNN, then the extracted face features are compared with data in a local feature database, if the comparison is successful, an identity ID corresponding to the person is output, and if the comparison is unsuccessful, the person is indicated to be a new user, the face features of the person need to be stored in the local feature database, and the identity ID belonging to the person is configured. After obtaining the person ID, the classification model for the person may be recalled from local storage. The classification model is a model that has been trained.
And S6, acquiring the environmental parameters of the activity area.
Specifically, the environmental parameters may include temperature, radiation temperature, wind speed, humidity, and the like, and the relevant environmental parameters may be obtained by different types of sensors. The thermal comfort of a person is also affected by the environment in which they are currently located. For example, heat is easily felt in summer, heat is more likely to be felt in rainy days, and cold is easily felt in winter, so environmental parameters are important factors for predicting the thermal comfort level of people.
S7, inputting the metabolic equivalent met of the person X, the clothing thermal resistance clo and the environmental parameters into a classification model, outputting the thermal comfort level of the person X by the classification model, and setting air-conditioning parameters according to the thermal comfort level.
Specifically, the thermal comfort level may be divided into three levels, i.e., a cold level, a heat neutral level and a heat level, but is not limited thereto, and may also be divided into five levels, i.e., a cold level, a micro-heat level, a heat neutral level, a very hot level and a hot level, which may be set according to actual requirements. After the classification model outputs the thermal comfort level of the person X, the thermal comfort level can be sent to a corresponding control system, and the control system sets air conditioner parameters (namely the temperature, the wind speed, the wind direction and the like of an air conditioner) according to the thermal comfort level. According to the embodiment, the thermal comfort degree of the person is comprehensively predicted according to the metabolic equivalent met, the clothing thermal resistance clo and the environmental parameters, so that the accuracy of the cold and hot requirements of the person can be improved, and the individual requirements of different persons can be met.
As shown in fig. 2, in the present embodiment, the training process of the classification model includes the following steps.
S10, using the data output from the existing thermal comfort prediction model as training data.
Specifically, data of the existing PMV thermal comfort model may be used as initial training data. PMV takes the basic equation of human body heat balance and the grade of psychophysiological subjective thermal sensation as a starting point, and takes the comprehensive evaluation indexes of a plurality of relevant factors of human body thermal comfort into consideration. However, the prediction results of the PMV thermal comfort model sometimes deviate greatly. Therefore, this embodiment only takes the data of the PMV thermal comfort model as the initial training data.
And S20, pre-training the classifier by using the training data to obtain a first classification model.
Specifically, initial training data is utilized to pre-train a classifier, the classifier can be an SVM classifier or an MLP classifier and the like, the initial training data is randomly divided into a training set and a test set, the classifier is trained by the training set, and then the prediction accuracy of the classifier is tested by the test set, so that a first classification model can be obtained. It should be noted that the first classification model is not the classification model that is finally predicted in the present embodiment.
S30, inputting the metabolic equivalent met of the person X, the clothing thermal resistance clo and the environmental parameters into a first classification model, wherein the first classification model outputs a first thermal comfort level of the person X.
Specifically, the metabolic equivalent met, the clothing thermal resistance clo and the environmental parameters of the person X obtained in real time in the embodiment are input into the first classification model, and the first classification model can predict the first thermal comfort level for the person X. For example, the first thermal comfort level of person X is predicted to be thermal.
S40, if the person X adjusts the air temperature control system (indicating that the person reflects the heat demand condition through the temperature control system adjustment action), indicating that the real thermal comfort level of the person X has deviation from the first thermal comfort level, and recording the real thermal comfort level of the person X.
Specifically, because the prediction result of the PMV thermal comfort model sometimes has a deviation, if the person X feels inappropriate for the predicted first thermal comfort level and the person X actively adjusts the parameter of the air conditioner, the true thermal comfort level of the person X is recorded, for example, the true thermal comfort level of the person X is thermally neutral.
S50, taking the metabolic equivalent met of the person X, the clothing thermal resistance clo, the environmental parameters and the real thermal comfort level as a group of real samples; and respectively adding positive or negative random disturbance to the metabolic equivalent met, the clothing thermal resistance clo and the environmental parameters to generate a batch of new training samples.
Specifically, the environmental parameters at least include temperature T1, radiation temperature T2 and wind speed F, and data corresponding to the metabolic equivalent met, the clothing thermal resistance clo, the temperature T1, the radiation temperature T2 and the wind speed F are respectively obtained, and it should be noted that the data input into the first classification model are data obtained by standardizing the parameters of the metabolic equivalent met, the clothing thermal resistance clo, the temperature T1, the radiation temperature T2, the wind speed F and the like, and are values between 0 and 1, so that errors caused by different dimensions of different parameters can be reduced. If the real thermal comfort level of the person X is higher than the first thermal comfort level, positive random disturbance is added to the metabolic equivalent met, the clothing thermal resistance clo, the temperature T1 and the radiation temperature T2, negative random disturbance is added to the wind speed F, and a batch of new training samples are obtained. For example, if the first thermal comfort level is thermal neutral and the real thermal comfort level is thermal, the classification result corresponding to the data obtained by adding positive random disturbance to the metabolic equivalent met, the clothing thermal resistance clo, the temperature T1, and the radiation temperature T2 and adding negative random disturbance to the wind speed F also meets the real thermal comfort level of the person X, and the obtained new training sample also meets the real feeling of the person X. If the real thermal comfort level of the person X is lower than the first thermal comfort level, negative random disturbance is added to the metabolic equivalent met, the clothing thermal resistance clo, the temperature T1 and the radiation temperature T2, positive random disturbance is added to the wind speed F, and a batch of new training samples are obtained. For example, if the first thermal comfort level is thermal and the real thermal comfort level is thermal neutral, the classification result corresponding to the data obtained by adding negative random disturbance to the metabolic equivalent met, the clothing thermal resistance clo, the temperature T1, and the radiation temperature T2 and adding positive random disturbance to the wind speed F also meets the real thermal comfort level of the person X, and the obtained new training sample also meets the real feeling of the person X. Specifically, the variation range of the numerical value may be 1 to 1.2 times of the original real data when the random disturbance in the positive direction is performed, and the variation range of the numerical value may be 0.8 to 1 time of the original real data when the random disturbance in the negative direction is performed. In other words, the present embodiment may derive more training data according to the obtained set of real samples, generate a plurality of variation values within the boundary range of the real data, and obtain more training samples. The new training sample comprises a plurality of new sets of training data, each set of new training data comprises the metabolic equivalent met, the clothing thermal resistance clo and the environmental parameter, and the classification result of each set of new training data corresponds to the real thermal comfort level of the person X, for example, the classification result corresponding to a set of data obtained by adding negative random disturbance to the metabolic equivalent met, the clothing thermal resistance clo, the temperature T1, the radiation temperature T2 and positive random disturbance to the wind speed F is still the real thermal comfort level of the person X. In this embodiment, the range of values is expanded by positive or negative disturbance to the true value, in other words, more possibilities are increased, so that the training samples are richer, and the accuracy and diversity of classification model prediction can be improved.
And S60, retraining the first classification model by using the new training sample to obtain the classification model of the person X.
Specifically, if the real thermal comfort level of the person X is still deviated from the thermal comfort level output by the classification model obtained in step S60, steps S40-S60 are repeated, and training is performed again to update the first classification model. In other words, the classification model of the embodiment can be trained and updated in real time all the time, so that the prediction result of the classification model can be more accurate, and the personalized requirements of the personnel can be met.
In this embodiment, an individualized training method is adopted for a classification model, and a first classification model is obtained by training data related to an initial model (i.e., a PMV thermal comfort model), then a real sample of a person X adjusting an air conditioner each time is recorded, then more data samples are generated by performing positive or negative disturbance on the real sample, and then the first classification model is retrained again by using the real sample and derived data samples, so that a more accurate and individualized prediction model (i.e., a classification model) for the person can be obtained. In other words, the training method of the embodiment does not singly use a group of real samples to retrain the first classification model, but actively expands more data according to the group of real samples, and particularly, the data fluctuate up and down at the boundary of the numerical value, so that more variation possibilities are increased, and the classification model obtained by retraining can be more accurate, comprehensive and personalized aiming at the thermal comfort prediction result of the person.
In summary, the intelligent control method for thermal comfort based on visual assistance of the invention can acquire the activity area of a person and the video image of the person in real time through the visual sensor, can respectively obtain the personnel label, the metabolic equivalent met, the clothing thermal resistance clo and the identity information of the person through the analysis and the processing of the image information, can adjust and take out the personalized classification model belonging to the person according to the identity information, then obtains the environmental parameter of the current activity area, and then inputs the metabolic equivalent met, the clothing thermal resistance clo and the environmental parameter into the classification model together, can predict the current thermal comfort level of the person, thereby intelligently controlling the air conditioner and meeting the cold and heat requirements of different persons.
In light of the foregoing description of the preferred embodiment of the present invention, many modifications and variations will be apparent to those skilled in the art without departing from the spirit and scope of the invention. The technical scope of the present invention is not limited to the contents of the specification, and must be determined by the scope of the claims.

Claims (6)

1. A thermal comfort intelligent control method based on visual assistance is characterized by comprising the following steps:
s1, acquiring the current activity area of the person X through a visual sensor and acquiring the image information of the person X;
s2, acquiring a personnel positioning frame and a personnel label of the personnel X through a personnel detector;
s3, acquiring an activity type and a metabolic equivalent met corresponding to the person X through the person positioning frame and the image information of the person X;
the obtaining of the activity type and the metabolic equivalent met corresponding to the person X specifically includes:
establishing a metabolic data set of a mapping relation between the activity type of the person and the metabolic level;
inputting the image information of the personnel positioning frame and the personnel X into a SLOWFAST neural network model, and outputting the activity type of the personnel X;
matching the activity type of the person X with the data in the metabolic data set to obtain the metabolic equivalent met of the person X;
s4, acquiring the clothing thermal resistance clo of the person X through the person positioning frame and the image information of the person X;
s5, acquiring identity information of the person X through the image information of the person X, and determining a classification model corresponding to the person X according to the identity information;
s6, acquiring the environmental parameters of the activity area;
s7, inputting the metabolic equivalent met, the clothing thermal resistance clo and the environmental parameters of the person X into the classification model, outputting the thermal comfort level of the person X by the classification model, and setting air conditioning parameters according to the thermal comfort level;
the training process of the classification model comprises the following steps:
s10, using data output by the existing thermal comfort prediction model as training data;
s20, pre-training the classifier by using the training data to obtain a first classification model;
s30, inputting the metabolic equivalent met, the clothing thermal resistance clo and the environmental parameters of the person X into the first classification model, wherein the first classification model outputs a first thermal comfort level of the person X;
s40, if the personnel X adjust the air-conditioning temperature control system, indicating that the real thermal comfort level of the personnel X has deviation from the first thermal comfort level, and recording the real thermal comfort level of the personnel X;
s50, taking the metabolic equivalent met of the person X, the clothing thermal resistance clo, the environmental parameters and the real thermal comfort level as a group of real samples; respectively adding positive or negative random disturbance to the metabolic equivalent met, the clothing thermal resistance clo and the environmental parameters to generate a batch of new training samples;
s60, retraining the first classification model by using the new training sample to obtain a classification model of the person X;
wherein, the generating of a batch of new training samples in step S50 specifically includes:
the environmental parameters at least comprise temperature T1, radiation temperature T2 and wind speed F,
respectively acquiring data corresponding to the metabolic equivalent met, the clothing thermal resistance clo, the temperature T1, the radiation temperature T2 and the wind speed F;
if the real thermal comfort level of the person X is higher than the first thermal comfort level, respectively adding positive random disturbance to the metabolic equivalent met, the clothing thermal resistance clo, the temperature T1 and the radiation temperature T2, and adding negative random disturbance to the wind speed F to obtain a batch of new training samples;
if the real thermal comfort level of the person X is lower than the first thermal comfort level, adding negative random disturbance to the metabolic equivalent met, the clothing thermal resistance clo, the temperature T1 and the radiation temperature T2, and adding positive random disturbance to the wind speed F to obtain a batch of new training samples.
2. The intelligent thermal comfort control method based on visual aid according to claim 1, wherein if there is still a deviation between the true thermal comfort level of person X and the thermal comfort level outputted by the classification model obtained in step S60, repeating steps S40-S60.
3. The intelligent thermal comfort control method based on visual aid according to claim 1, wherein the new training samples comprise a plurality of new sets of training data, each set of training data comprises metabolic equivalent met, thermal clothing resistance clo and environmental parameters, and the classification result of each set of training data corresponds to the real thermal comfort level of the person X.
4. The intelligent thermal comfort control method based on visual aids of claim 1, wherein the personnel tags include gender, BMI rating, and age rating of personnel X.
5. The intelligent thermal comfort control method based on visual assistance according to claim 1, wherein if the human detector does not detect human activities for n minutes continuously, a control signal is output to turn off the air conditioner, wherein n is an integer greater than 15.
6. The intelligent thermal comfort control method based on visual assistance according to claim 1, wherein in step S4, obtaining the clothing thermal resistance clo of the person X specifically includes:
establishing a thermal resistance data set of a mapping relation between the clothing of the person and the thermal resistance of the clothing;
acquiring dresses of the person X, namely head and neck dresses, upper body dresses, lower body dresses and foot dresses, according to the person positioning frame and the image information of the person X;
matching the head and neck dressing, the upper body dressing, the lower body dressing and the foot dressing with the data in the thermal resistance data set respectively to obtain four thermal resistance data corresponding to the head and neck dressing, the upper body dressing, the lower body dressing and the foot dressing;
and adding and calculating the four thermal resistance data to obtain the clothing thermal resistance clo of the person X.
CN202210132584.2A 2022-02-14 2022-02-14 Intelligent thermal comfort control method based on visual assistance Active CN114183881B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210132584.2A CN114183881B (en) 2022-02-14 2022-02-14 Intelligent thermal comfort control method based on visual assistance

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210132584.2A CN114183881B (en) 2022-02-14 2022-02-14 Intelligent thermal comfort control method based on visual assistance

Publications (2)

Publication Number Publication Date
CN114183881A CN114183881A (en) 2022-03-15
CN114183881B true CN114183881B (en) 2022-05-24

Family

ID=80545843

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210132584.2A Active CN114183881B (en) 2022-02-14 2022-02-14 Intelligent thermal comfort control method based on visual assistance

Country Status (1)

Country Link
CN (1) CN114183881B (en)

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104456841B (en) * 2014-11-13 2017-01-25 重庆大学 Thermal and humid environment integrated control air-conditioning system and method based on thermal comfort evaluation
CN107917507B (en) * 2017-11-14 2020-05-12 西安建筑科技大学 PMV (Power management Unit) control method for thermal comfort of centralized air conditioner by fusing image information
CN110377936B (en) * 2019-06-06 2021-01-19 西安交通大学 System and method for intelligent dynamic perception of building personnel personalized thermal comfort
CN113627448A (en) * 2020-05-07 2021-11-09 香港大学浙江科学技术研究院 Method for constructing human body thermal comfort prediction model, prediction method and system
CN111626210B (en) * 2020-05-27 2023-09-22 上海科技大学 Personnel dressing detection method, processing terminal and storage medium
CN113963315A (en) * 2021-11-16 2022-01-21 重庆邮电大学 Real-time video multi-user behavior recognition method and system in complex scene

Also Published As

Publication number Publication date
CN114183881A (en) 2022-03-15

Similar Documents

Publication Publication Date Title
Aryal et al. A comparative study of predicting individual thermal sensation and satisfaction using wrist-worn temperature sensor, thermal camera and ambient temperature sensor
EP3750012B1 (en) System and method for controlling operation
CN110377936B (en) System and method for intelligent dynamic perception of building personnel personalized thermal comfort
Cheng et al. NIDL: A pilot study of contactless measurement of skin temperature for intelligent building
WO2018210124A1 (en) Clothing recommendation method and clothing recommendation device
Feng et al. Data-driven personal thermal comfort prediction: A literature review
US20170123442A1 (en) System and Method of Smart and Energy-Saving Environmental Control
WO2020151732A1 (en) Non-invasive ai sensing method for human thermal comfort
US20220090811A1 (en) Detecting presence and estimating thermal comfort of one or more human occupants in a built space in real-time using one or more thermographic cameras and one or more rgb-d sensors
CN112862145A (en) Occupant thermal comfort inference using body shape information
Burzo et al. Using infrared thermography and biosensors to detect thermal discomfort in a building’s inhabitants
CN112303861A (en) Air conditioner temperature adjusting method and system based on human body thermal adaptability behavior
CN111444943A (en) Device and method for adaptive personalized thermal comfort prediction based on human body similarity
CN108426349B (en) Air conditioner personalized health management method based on complex network and image recognition
CN114183881B (en) Intelligent thermal comfort control method based on visual assistance
Fan et al. Real-time machine learning-based recognition of human thermal comfort-related activities using inertial measurement unit data
CN113627448A (en) Method for constructing human body thermal comfort prediction model, prediction method and system
Liu et al. Vision-based individual factors acquisition for thermal comfort assessment in a built environment
CN110659594B (en) Thermal comfort attitude estimation method based on AlphaPose
Bucarelli et al. Deep learning approach for recognizing cold and warm thermal discomfort cues from videos
Xu et al. Action-based personalized dynamic thermal demand prediction with video cameras
Yang et al. Macro pose based non-invasive thermal comfort perception for energy efficiency
KR102233157B1 (en) Method and system for calculating occupant activity using occupant pose classification based on deep learning
CN111612746A (en) Dynamic detection method of functional brain network central node based on graph theory
Cosma Real-time Individual Thermal Preferences Prediction Using Visual Sensors

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant