Seat temperature control system based on thermal management

Abstract

The invention belongs to the technical field of temperature control management, and discloses a seat temperature control system based on thermal management, which comprises a database module used for collecting and associating body shape characteristic data, internal and external environment parameters and seat temperature control parameters of a user so as to generate a user data set and a user tag corresponding to the user data set; the recognition module is used for collecting body shape feature data of the current user and inputting the body shape feature data into the pre-trained user recognition model for recognition to obtain a current user tag; according to the method, the user can be identified according to the body shape feature data, more suitable seat temperature control parameters are provided for the user, the personalized preference of the user for the seat temperature control parameters can be met, the overall satisfaction degree and comfort degree of the user are improved, the regulation and control process of the user for the seat temperature control parameters can be reduced, and time is saved.

Description

Technical Field

[0001] The present invention relates to the technical field of temperature control management, and more specifically, to a seat temperature control system based on thermal management. Background Art

[0002] Thermal management refers to the process of regulating and controlling the temperature or temperature difference of a specific object by heating or cooling according to its requirements. Thermal management consumes energy to create a temperature difference between the object and the outside world, and achieves temperature regulation of the object through heat transfer. With the development of the modern automobile industry, the comfort of vehicles has become one of the important factors for consumers to purchase. In the thermal management of seats in vehicles, the temperature control of seats, as parts that directly contact the human body, directly affects the comfort experience of users (drivers or passengers). Traditional seat heating systems usually increase the seat temperature by embedding heating elements, while seat ventilation systems reduce the seat temperature through vents or fans.

[0003] However, when the temperature of existing seats is adjusted, the user is required to adjust the seat temperature control parameters every time, and in many cases the user cannot adjust the seat temperature control parameters to the desired level at one time, which makes the user's adjustment process cumbersome and time-consuming.

[0004] In view of this, the present invention proposes a seat temperature control system based on thermal management to solve the above problems. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solution: a seat temperature control system based on thermal management, comprising:

[0006] A database module is used to collect and associate the user's body feature data, internal and external environment parameters, and seat temperature control parameters, and then generate a user data set and its corresponding user tag;

[0007] The recognition module is used to collect the body feature data of the current user and input it into the pre-trained user recognition model for recognition to obtain the current user label;

[0008] A control module is used to collect current internal and external environmental parameters, and input the current internal and external environmental parameters and the current user label into a pre-trained seat temperature control parameter prediction model to obtain and apply the current seat temperature control parameters;

[0009] The updating module is used to update the user identification model and the seat temperature control parameter prediction model, and apply the updated user identification model and the seat temperature control parameter prediction model.

[0010] Furthermore, the step of collecting and associating the user's body feature data, internal and external environment parameters, and seat temperature control parameters to generate a user data set and its corresponding user tag includes:

[0011] Collect the body shape characteristic data, internal and external environmental parameters and seat temperature control parameters of the same user in the seat during the use of the seat belt;

[0012] The collected body feature data of the same user are associated with the internal and external environment parameters and the seat temperature control parameters to obtain user data. The user data of f pieces of the same user constitute a user data set. The user data set is labeled to obtain a user label.

[0013] The body shape feature data includes k body shape skeleton matrices, the internal and external environmental parameters include the temperature inside the vehicle, the humidity inside the vehicle, the temperature outside the vehicle and the humidity outside the vehicle, and the temperature control parameters include the ventilation temperature and the ventilation wind speed;

[0014] Collect user data sets of different users and their corresponding user tags to form a database.

[0015] Furthermore, the step of obtaining the k-body skeleton matrix comprises:

[0016] An m×n pressure sensor array is installed on the seat to obtain k pressure distribution matrices of the user during the use of the seat belt, wherein the pressure distribution matrix is ​​composed of pressure values ​​measured by the m×n pressure sensor array;

[0017] For each pressure distribution matrix, nonlinear mapping is used to convert the pressure distribution matrix into a grayscale image, and a bilateral filter is used to smooth the grayscale image. The Sobel operator is used to calculate the gradient intensity and direction of each pixel in the grayscale image to generate a gradient image. In the gradient image, for each pixel, two neighboring pixels in the gradient direction are selected for comparison according to its gradient direction. If the gradient intensity of the current pixel is greater than or equal to the intensity of the two neighboring pixels in the gradient direction, the current pixel is retained as an edge point. Otherwise, only the local maximum value of the current pixel along the gradient direction is retained. Subsequently, the Canny edge detection algorithm is applied to define a high threshold and a low threshold. Pixels above the high threshold are considered to be strong edges, and pixels below the low threshold are considered to be non-edges. Finally, the detected edge pixels are connected to obtain a basic contour image. The basic contour image is converted into a binary image by the Otsu algorithm, and the binary image is morphologically preprocessed and a skeleton extraction algorithm is applied to obtain a body skeleton matrix.

[0018] Furthermore, the conversion formula for converting the pressure distribution matrix into a grayscale image is:

[0019] P gamma(x,y)=(P(x,y)+ε) γ ,

[0020] Where P gamma (x, y) represents the grayscale value at the coordinate (x, y) in the grayscale image, P(x, y) represents the pressure value at the coordinate (x, y) in the pressure distribution matrix, ε is a small smoothing constant, and γ is the gamma parameter.

[0021] Furthermore, the step of performing morphological preprocessing on the binary image and applying a skeleton extraction algorithm to obtain a body shape skeleton matrix includes:

[0022] Create J circular structural elements, the radius of each of the J circular structural elements increases in equal proportion, perform opening and closing operations on the basic contour map through the J circular structural elements, and obtain J basic contour maps after morphological preprocessing;

[0023] Arrange the J preprocessed basic contour images from large to small according to the radius of the corresponding circular structure elements to obtain a sorting sequence, and apply the improved Guo-Hal l thinning algorithm to each preprocessed basic contour image according to the sorting sequence;

[0024] The process of applying the improved Guo-Hall thinning algorithm is:

[0025] Initialize the preprocessed basic contour image, mark the image boundary points as 1, mark the internal points as 0, and iterate each image boundary point of the basic contour image repeatedly until no image boundary point is deleted, that is, J skeleton extraction images are obtained;

[0026] Subsequently, the skeleton extracted image at the first position in the sequence of the J skeleton extracted images is used as the basic skeleton, and starting from the skeleton extracted image at the second position in the sequence, the skeleton extracted images are used as the current skeleton one by one according to the sorting sequence, and are merged with the basic skeleton in sequence, and the merging process is: calculating the similarity between the current skeleton and the basic skeleton, if the similarity is greater than a preset first threshold, merging the different parts of the current skeleton and the basic skeleton into the basic skeleton, if the similarity is less than or equal to the first threshold, multiplying the current skeleton with a weight matrix, and then merging with the basic skeleton, wherein the weight matrix is ​​calculated by the distance transformation from each point in the current skeleton to the nearest non-zero point;

[0027] The branches whose pixel lengths in the merged basic skeleton are less than a preset second threshold are removed, and the merged basic skeleton is smoothed by mean filtering, that is, the body shape skeleton matrix is ​​output.

[0028] Furthermore, the training process of the user identification model is:

[0029] Extract all different user labels and their corresponding body shape feature data from the database as the first sample set, construct a feature vector G1 with the body shape feature data corresponding to each user label, use keras to define the input layer and output layer and build a model framework; use the feature vector G1 as input data and the user label as output data;

[0030] The first sample set is divided into a training set and a test set in a ratio of 70% and 30%, a user recognition model is constructed and trained, the user recognition model is tested using the test set, and a user recognition model that meets a preset accuracy is output.

[0031] Furthermore, the training process of the seat temperature control parameter prediction model is as follows:

[0032] All user tags and their corresponding internal and external environment parameters and temperature control parameters are used as the second sample set, and the temperature control parameters are numbered;

[0033] Construct feature vector G2 with user labels and their corresponding internal and external environment parameters, use TensorFlow to define input and output layers and build a model framework; use feature vector G2 as input data and temperature control parameter number as output data;

[0034] The second sample set is divided into a training set and a test set in a ratio of 70% and 30%, and a seat temperature control parameter prediction model is constructed and trained. The seat temperature control parameter prediction model is tested using the test set, and a seat temperature control parameter prediction model that meets the preset accuracy is output.

[0035] Furthermore, the step of updating the user identification model and the seat temperature control parameter prediction model and applying the updated user identification model and the seat temperature control parameter prediction model comprises:

[0036] Determine whether the current user's body feature data is recognized;

[0037] If the current user's body feature data is not recognized by the user recognition model, the current user's body feature data and the corresponding internal and external environment parameters and temperature control parameters are used as a new user data set and a new user tag is generated, and then the new user data set and the new user tag are added to the existing database;

[0038] The user identification model is trained and updated using the new user tag and the corresponding body shape feature data, thereby obtaining an updated user identification model;

[0039] The seat temperature control parameter prediction model is trained and updated using the internal and external environment parameters and temperature control parameters corresponding to the new user label, thereby obtaining an updated seat temperature control parameter prediction model;

[0040] The updated user identification model and the updated seat temperature control parameter prediction model are applied.

[0041] Furthermore, the updating of the user identification model and the seat temperature control parameter prediction model also includes:

[0042] After the current user's body shape feature data is identified, other body shape skeleton matrices existing in the same seat belt usage time are added to the current user's body shape feature data, and then the user identification model is trained and updated;

[0043] The seat temperature control parameter prediction model is trained and updated using new internal and external environmental parameters and new temperature control parameters corresponding to the current user's body feature data.

[0044] Furthermore, the method of determining whether the current user's body shape feature data is recognized is:

[0045] If the user identification model can output the user tag, it is determined that the body feature data of the current user is identified; otherwise, it is determined that the body feature data of the current user is not identified.

[0046] The technical effects and advantages of a seat temperature control system based on thermal management of the present invention are as follows:

[0047] 1. The Otsu algorithm and nonlinear gamma transform can adaptively determine the optimal threshold and perform pressure mapping, and can accurately distinguish different pressure areas; by merging different skeleton extraction images one by one, the accuracy of the body skeleton matrix can be improved; thus, the user can be identified according to the body feature data, and more suitable seat temperature control parameters can be provided for the user, which can also meet the user's personalized preferences for seat temperature control parameters, improve the user's overall satisfaction and comfort, and reduce the user's adjustment process of seat temperature control parameters, which is more time-saving;

[0048] 2. By judging whether the body feature data of the current user is recognized and updating the user recognition model, the body feature data of new users can be effectively processed, thereby realizing self-learning and expansion of user recognition; by associating new user tags with internal and external environmental parameters and seat temperature control parameters and updating the seat temperature control parameter prediction model, the seat temperature control parameters can be adjusted according to the personalized needs of each user, thereby making the response capability of the present invention faster and greatly improving the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a structural schematic diagram of a seat temperature control system based on thermal management of the present invention;

[0050] Figure 2 It is a flow chart of the improved Guo-Hall thinning algorithm applied in the present invention. DETAILED DESCRIPTION

[0051] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 creative work are within the scope of protection of the present invention.

[0052] Example 1

[0053] See also Figure 1 to Figure 2 As shown, the present embodiment provides a seat temperature control system based on thermal management, including:

[0054] A database module is used to collect and associate the user's body feature data, internal and external environment parameters, and seat temperature control parameters, and then generate a user data set and its corresponding user tag;

[0055] The recognition module is used to collect the body feature data of the current user and input it into the pre-trained user recognition model for recognition to obtain the current user label;

[0056] The control module is used to collect the current internal and external environmental parameters, and input the current internal and external environmental parameters and the current user labels into the pre-trained seat temperature control parameter prediction model to obtain the current seat temperature control parameters and apply them.

[0057] In one embodiment of the present invention, the step of collecting and associating the user's body feature data, internal and external environment parameters, and seat temperature control parameters, and then generating a user data set and its corresponding user tag includes:

[0058] Collect the body shape feature data, internal and external environment parameters and seat temperature control parameters of the same user in the seat during the seat belt usage time (from buckling the seat belt to unbuckling the seat belt);

[0059] The collected body feature data of the same user are associated with the internal and external environment parameters and the seat temperature control parameters to obtain user data. The user data of f pieces of the same user constitute a user data set. The user data set is labeled to obtain a user label.

[0060] Among them, the body shape feature data includes k body shape skeleton matrices, the internal and external environmental parameters include the temperature inside the vehicle, the humidity inside the vehicle, the temperature outside the vehicle and the humidity outside the vehicle, and the temperature control parameters include the ventilation temperature and the ventilation wind speed;

[0061] In the user data, the internal and external environmental parameters are the average in-car temperature, average in-car humidity, average outside temperature and average outside humidity corresponding to the user's use of the seat belt. The temperature control parameters are the ventilation temperature and ventilation speed last adjusted by the user during the use of the seat belt.

[0062] Collect user data sets of different users and their corresponding user tags to form a database.

[0063] In one embodiment of the present invention, the step of obtaining the k-body skeleton matrix includes:

[0064] An m×n pressure sensor array (or pressure sensor pad) is installed on the seat to obtain k pressure distribution matrices of the user during the use of the seat belt, wherein the pressure distribution matrix is ​​composed of the pressure values ​​measured by the m×n pressure sensor array;

[0065] For each pressure distribution matrix, a nonlinear mapping is used to convert the pressure distribution matrix into a grayscale image, wherein each pressure value is mapped to the grayscale value of the grayscale image, a bilateral filter is used to smooth the grayscale image, and a Sobel operator is used to calculate the gradient intensity and direction of each pixel in the grayscale image to generate a gradient image. In the gradient image, for each pixel, according to its gradient direction, two neighboring pixels in the gradient direction are selected for comparison. If the gradient intensity of the current pixel is greater than or equal to the intensity of the two neighboring pixels in the gradient direction, the current pixel is retained as an edge point, otherwise, the current pixel is suppressed (i.e., Its value is set to 0), only the local maximum value of the current pixel along the gradient direction is retained, and then the Canny edge detection algorithm is applied to define the high threshold and the low threshold. Pixels above the high threshold are considered to be strong edges, and pixels below the low threshold are considered to be non-edges. For pixels between the high threshold and the low threshold, it is determined whether they are connected to the strong edge. If connected, they are retained as edge points. Finally, the detected edge pixels are connected to obtain the basic contour image; the basic contour image is converted into a binary image through the Otsu algorithm, the binary image is morphologically preprocessed and the skeleton extraction algorithm is applied to obtain the body skeleton matrix.

[0066] Specifically, the process of converting the basic contour image into a binary image by the Otsu algorithm is as follows: the number of occurrences of each gray value (0-255) in the basic contour image is counted to generate a gray histogram, and the optimal binarization threshold is determined by maximizing the inter-class variance and applied to obtain a binary image, in which the pressure area is white (the area with pressure is regarded as the pressure area) and the background (the area without pressure) is black.

[0067] In one embodiment of the present invention, the conversion formula for converting the pressure distribution matrix into a grayscale image is:

[0068] P gamma (x,y)=(P(x,y)+ε) γ ,

[0069] Where P gamma (x,y) represents the gray value (pixel gray value) at the coordinate (x,y) in the gray image, P(x,y) represents the pressure value at the coordinate (x,y) in the pressure distribution matrix, ε is a small smoothing constant, generally set to 0.01 or less, γ is the gamma parameter, which represents the degree of nonlinearity of the control change;

[0070] Among them, if P median -P min <P max -P median ,but otherwise, Where P min is the minimum pressure value in the pressure distribution matrix, P max is the maximum pressure value in the pressure distribution matrix, P median is the median of the pressure values ​​in the pressure distribution matrix, k is the first scaling factor, which can be set experimentally to adjust the contraction degree of the gamma value (k is generally set to 1 or 0.5), m is the second scaling factor, which can be set experimentally to adjust the enhancement degree of the gamma value (m is generally set to 2), e% is the threshold percentage, which is obtained by relevant personnel based on experiments and is generally set to 5;

[0071] In one embodiment of the present invention, the step of performing morphological preprocessing on the binary image and applying a skeleton extraction algorithm to obtain a body skeleton matrix includes:

[0072] Create J circular structural elements, the radius of each of the J circular structural elements increases in the same size, and perform opening and closing operations on the basic contour map through the J circular structural elements to obtain J basic contour maps after morphological preprocessing.

[0073] Specifically, the minimum radius of the J circular structural elements is less than or equal to the minimum area length in the basic contour image, the maximum radius of the J circular structural elements is greater than the minimum area length and less than the maximum area length, and the area length is the geometric size (width or length, etc.) of the pressure area in the basic contour image; for example, 2R, 4R, 6R, R is the radius length, which is set by relevant personnel; the opening operation uses the central element of the circular structural element as a moving point to traverse the basic contour image to first erode and then expand, and the closing operation process uses the central element of the circular structural element as a moving point to traverse the basic contour image to first expand and then erode;

[0074] Arrange the J preprocessed basic contour images from large to small according to the radius of their corresponding circular structure elements to obtain a sorting sequence, and apply the improved Guo-Hall thinning algorithm to each preprocessed basic contour image according to the sorting sequence;

[0075] The process of applying the improved Guo-Hall thinning algorithm is:

[0076] Initialize the preprocessed basic contour image, mark the image boundary points as 1, mark the internal points as 0, and iterate each image boundary point of the basic contour image repeatedly until no image boundary point is deleted, that is, J skeleton extraction images are obtained;

[0077] Specifically, the iterative process is:

[0078] First subiteration: Delete points P1 that satisfy all of the following conditions:

[0079] 2≤B(P1)≤6, A(P1)=1, P2*P4*P6=0, P4*P6*P8=0,

[0080] Second sub-iteration: Delete points P1 that satisfy all of the following conditions:

[0081] 2≤B(P1)≤6, A(P1)=1, P2*P4*P8=0, P2*P6*P8=0

[0082] Where B(P1) is the number of pixels with a value of 1 in the 8-neighborhood of P1, and A(P1) is the number of 0-1 or 1-0 transitions in the 8-neighborhood of P1;

[0083] Subsequently, the skeleton extracted image at the first position in the sequence of the J skeleton extracted images is used as the basic skeleton S, and starting from the skeleton extracted image at the second position in the sequence, the skeleton extracted images are used as the current skeleton Si one by one according to the sorting sequence, and are merged with the basic skeleton S in sequence. The merging process is: calculating the similarity between the current skeleton Si and the basic skeleton S, wherein the similarity is defined as the number of pixels of the intersection of the two skeletons divided by the number of pixels of their union. If the similarity is greater than a preset first threshold, the part of the current skeleton that is different from the basic skeleton is directly merged into the basic skeleton. If the similarity is less than or equal to the first threshold, the current skeleton is multiplied by a weight matrix and then merged with the basic skeleton, wherein the weight matrix is ​​calculated by the distance transformation from each point in the current skeleton to the nearest non-zero point.

[0084] Specifically, the first skeleton extraction image in the sequence is used as the basic skeleton S;

[0085] Starting from the second skeleton-extracted image in the sequence, process the remaining J-1 skeleton-extracted images one by one;

[0086] Calculate the similarity sim between the current skeleton S i and the base skeleton S: Sim = (S i∩S) / (S i∪S);

[0087] If Sim is greater than a preset first threshold (such as 0.7), then S = S∪(SiS) to merge different parts, otherwise S = S∪(Si*W), where W is a weight matrix, which is calculated based on the distance transformation;

[0088] The branches whose pixel lengths in the merged basic skeleton S are less than a preset second threshold are removed, and the merged basic skeleton S is smoothed by mean filtering, that is, the body shape skeleton matrix is ​​output.

[0089] It should be noted that through bilateral filtering, Sobe l operator, Canny edge detection algorithm, Otsu algorithm and morphological preprocessing can effectively extract the key contours of body feature data and ensure that the extracted body skeleton matrix has high accuracy and consistency; by converting the pressure distribution matrix into a grayscale image and performing skeleton extraction processing, a representative body skeleton matrix can be extracted from the pressure distribution matrix, reducing the amount of complex data, which is helpful to generate highly simplified and representative body feature data for subsequent analysis and training of user recognition models; through the Otsu algorithm and nonlinear gamma transform, the optimal threshold and pressure mapping can be adaptively determined, so that different pressure areas can be accurately distinguished in the image; through the opening and closing operations in the morphological preprocessing, small noise points and non-connected areas can be further removed to ensure the continuity and integrity of skeleton extraction; in the skeleton extraction process, by merging different skeleton extraction images layer by layer and using the weight matrix for fine adjustment, errors can be reduced and key features can be strengthened, thereby improving the accuracy of the final body skeleton matrix; by removing redundant branches and applying mean filtering to smooth the basic skeleton, it is beneficial to improve the simplicity of the final body skeleton, cope with the changes in the body feature data of different users, and maintain consistency and accuracy when generating the body skeleton matrix.

[0090] In one embodiment of the present invention, the step of collecting the body feature data of the current user and inputting it into a pre-trained user recognition model for recognition to obtain the current user label includes:

[0091] Set the collection period of body shape feature data, which is generally collected after the seat belt is fastened. The length of the collection period is set by relevant personnel according to actual conditions. Collect the body shape feature data of the current user and input the body shape feature data of the current user into the user recognition model, that is, output the current user label;

[0092] The training process of the user identification model is:

[0093] Extract all different user labels and their corresponding body shape feature data from the database as the first sample set, construct a feature vector G1 with the body shape feature data corresponding to each user label, use keras to define the input layer and output layer and build a model framework; use the feature vector G1 as input data and the user label as output data;

[0094] The first sample set is divided into a training set and a test set in a ratio of 70% and 30%, a user recognition model is constructed and trained, the user recognition model is tested using the test set, and a user recognition model that meets a preset accuracy is output. The user recognition model can be a convolutional neural network.

[0095] It should be noted that by effectively matching body feature data with user labels and using a trained model for real-time recognition, users can be identified based on their body feature data, thereby improving the intelligence level of the system so that the seat temperature control parameter prediction model can be used to provide users with more efficient, more suitable and more personalized temperature control parameters.

[0096] In one embodiment of the present invention, the steps of collecting current internal and external environmental parameters, inputting the current internal and external environmental parameters and the current user tags into a pre-trained seat temperature control parameter prediction model, obtaining the current seat temperature control parameters and applying them include:

[0097] Collect current internal and external environmental parameters, input the current internal and external environmental parameters into the seat temperature control parameter prediction model, and output the current seat temperature control parameters;

[0098] The training process of the seat temperature control parameter prediction model is as follows:

[0099] All user tags and their corresponding internal and external environment parameters and temperature control parameters are used as the second sample set, and the temperature control parameters are numbered;

[0100] Construct feature vector G2 with user labels and their corresponding internal and external environment parameters, use TensorFlow to define input and output layers and build a model framework; use feature vector G2 as input data and temperature control parameter number as output data;

[0101] The second sample set is divided into a training set and a test set in a ratio of 70% and 30%, and a seat temperature control parameter prediction model is constructed and trained. The seat temperature control parameter prediction model is tested using the test set, and a seat temperature control parameter prediction model that meets a preset accuracy is output. The seat temperature control parameter prediction model is one of a logistic regression model, a naive Bayes model, or a support vector machine model.

[0102] It should be noted that by accurately predicting and regulating the temperature control parameters of the current seat, more suitable seat temperature control parameters can be provided to users, which meets the user's personalized preferences for seat temperature control parameters and improves the user's overall satisfaction and comfort.

[0103] This embodiment can extract the user's body skeleton matrix, ensure the high accuracy and continuity of the body skeleton matrix, and can adaptively determine the optimal threshold and perform pressure mapping through the Otsu algorithm and nonlinear gamma transform, and can accurately distinguish different pressure areas; by merging different skeleton extraction images one by one, the accuracy of the body skeleton matrix can be improved; and then the user can be identified according to the body feature data, and more suitable seat temperature control parameters can be provided for the user, which meets the user's personalized preference for seat temperature control parameters, improves the user's overall satisfaction and comfort, and can also reduce the user's adjustment process of the seat temperature control parameters, which is more time-saving.

[0104] Example 2

[0105] See also Figure 1 As shown, the present embodiment provides a seat temperature control system based on thermal management, including:

[0106] The updating module is used to update the user identification model and the seat temperature control parameter prediction model, and apply the updated user identification model and the seat temperature control parameter prediction model.

[0107] In one embodiment of the present invention, the steps of updating the user identification model and the seat temperature control parameter prediction model and applying the updated user identification model and the seat temperature control parameter prediction model include:

[0108] Determine whether the current user's body feature data is recognized;

[0109] If the current user's body feature data is not recognized by the user recognition model, the current user's body feature data and the corresponding internal and external environment parameters and temperature control parameters are used as a new user data set and a new user tag is generated, and then the new user data set and the new user tag are added to the existing database;

[0110] The user identification model is trained and updated using the new user tag and the corresponding body shape feature data, thereby obtaining an updated user identification model;

[0111] The seat temperature control parameter prediction model is trained and updated using the internal and external environment parameters and temperature control parameters corresponding to the new user tags, thereby obtaining an updated seat temperature control parameter prediction model;

[0112] The updated user identification model and the updated seat temperature control parameter prediction model are applied.

[0113] In one embodiment of the present invention, the updating of the user identification model and the seat temperature control parameter prediction model further includes:

[0114] After the current user's body shape feature data is identified, other body shape skeleton matrices existing in the same seat belt usage time are added to the current user's body shape feature data, and then the user identification model is trained and updated;

[0115] The seat temperature control parameter prediction model is trained and updated using new internal and external environment parameters and new temperature control parameters corresponding to the current user's body shape feature data;

[0116] Among them, the training frequency and update frequency of the user recognition model and the seat temperature control parameter prediction model are set by relevant personnel according to actual conditions. For example, the user recognition model can be trained and updated when two new users or other body skeleton matrices are added, and the seat temperature control parameter prediction model can also be trained and updated when five new temperature control parameters and new internal and external environment parameters are added.

[0117] In one embodiment of the present invention, the method of determining whether the current user's body shape feature data is recognized is:

[0118] If the user recognition model can output the user tag within the preset time, it is determined that the body feature data of the current user is recognized; otherwise, it is determined that the body feature data of the current user is not recognized.

[0119] In this embodiment, by determining whether the body feature data of the current user is recognized and updating the user recognition model, it is possible to effectively process the body feature data of new users, thereby achieving self-learning and expansion of user recognition, so that the accuracy of user recognition continues to improve with the growth of the user data set; by associating new user tags with internal and external environmental parameters and seat temperature control parameters and updating the seat temperature control parameter prediction model, the seat temperature control parameters can be dynamically adjusted according to the personalized needs of each user, and the body feature data of users in different postures can be processed, thereby enhancing the stability and accuracy of user recognition, and being able to adapt to changes in user needs, thereby providing users with more accurate and comfortable seat temperature control parameters; it helps the system to quickly identify users and apply the adjusted temperature control parameters, so that the response capability of the system is fast and the user experience is greatly improved.

[0120] The basic principles of the present application are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present application. In addition, the specific details disclosed above are only for the purpose of illustration and ease of understanding, not for limitation, and the above details do not limit the present application to being implemented by adopting the above specific details.

[0121] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0122] The above are only preferred embodiments of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technical users in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A seat temperature control system based on thermal management, characterized in that: include: The database module is used to collect and associate the user's body feature data, internal and external environmental parameters, and seat temperature control parameters, and then generate a user data set and its corresponding user tag; The recognition module is used to collect the current user's body feature data and input it into the pre-trained user recognition model to perform recognition and obtain the current user label; The control module is used to collect current internal and external environmental parameters and input the current internal and external environmental parameters and the current user label into the pre-trained seat temperature control parameter prediction model to obtain and apply the current seat temperature control parameters; The updating module is used to update the user identification model and the seat temperature control parameter prediction model, and apply the updated user identification model and the seat temperature control parameter prediction model.

2. A seat temperature control system based on thermal management according to claim 1, characterized in that: The step of collecting and associating the user's body feature data, internal and external environment parameters, and seat temperature control parameters to generate a user data set and its corresponding user tag includes: Collect body shape data, internal and external environmental parameters, and seat temperature control parameters of the same user during the time the seat belt is in use; The collected body feature data of the same user is associated with the internal and external environment parameters and the seat temperature control parameters to obtain user data. The user data of f pieces of the same user constitute a user dataset. The user dataset is labeled to obtain a user label. The body shape feature data includes k body shape skeleton matrices, the internal and external environmental parameters include vehicle interior temperature, vehicle interior humidity, vehicle exterior temperature and vehicle exterior humidity, and the temperature control parameters include ventilation temperature and ventilation wind speed; Collect user data sets of different users and corresponding user tags to form a database.

3. The seat temperature control system based on thermal management according to claim 2, characterized in that: The step of obtaining the k-body skeleton matrix includes: Install an m×n pressure sensor array on the seat to obtain k pressure distribution matrices during the user's seat belt usage time, where the pressure distribution matrix is ​​composed of pressure values ​​measured by the m×n pressure sensor array; For each pressure distribution matrix, a nonlinear mapping is used to convert the pressure distribution matrix into a grayscale image, and a bilateral filter is used to smooth the grayscale image. The Sobel operator is used to calculate the gradient intensity and direction of each pixel in the grayscale image to generate a gradient image. In the gradient image, for each pixel, two neighboring pixels in the gradient direction are selected for comparison according to its gradient direction. If the gradient intensity of the current pixel is greater than or equal to the intensity of the two neighboring pixels in the gradient direction, the current pixel is retained as an edge point. Otherwise, only the local maximum value of the current pixel along the gradient direction is retained. Subsequently, the Canny edge detection algorithm is applied, and a high threshold and a low threshold are defined. Pixels above the high threshold are considered to be strong edges, and pixels below the low threshold are considered to be non-edges. Finally, the detected edge pixels are connected to obtain a basic contour image; the basic contour image is converted into a binary image using the Otsu algorithm, and the binary image is morphologically preprocessed and a skeleton extraction algorithm is applied to obtain a body skeleton matrix.

4. The seat temperature control system based on thermal management according to claim 3, characterized in that: The conversion formula for converting the pressure distribution matrix into a grayscale image is: Pgamm(a x,y)=(P(x,y)+ε) γ , Where Pgamm(ax,y) represents the grayscale value at the coordinate (x,y) in the grayscale image, P(x,y) represents the pressure value at the coordinate (x,y) in the pressure distribution matrix, ε is a small smoothing constant, and γ is the gamma parameter.

5. The seat temperature control system based on thermal management according to claim 3, characterized in that: The step of performing morphological preprocessing on the binary image and applying a skeleton extraction algorithm to obtain a body skeleton matrix includes: Create J circular structuring elements, where the radius of each circular structuring element increases in an equal manner, and perform opening and closing operations on the basic contour image through the J circular structuring elements to obtain J basic contour images after morphological preprocessing; Arrange the J preprocessed basic contour images from large to small according to the radius of their corresponding circular structure elements to obtain a sorted sequence, and apply the improved Guo-Hall thinning algorithm to each preprocessed basic contour image according to the sorted sequence; The process of applying the improved Guo-Hall thinning algorithm is: Initialize the preprocessed basic contour image, mark the image boundary points as 1, mark the internal points as 0, and iterate each image boundary point of the basic contour image repeatedly until no image boundary points are deleted, thus obtaining J skeleton extraction images; Subsequently, the skeleton extracted image at the first position in the sequence of the J skeleton extracted images is used as the base skeleton. Starting from the skeleton extracted image at the second position in the sequence, the skeleton extracted images are used one by one as the current skeleton in the sorted sequence, and are merged with the base skeleton in sequence. The merging process is as follows: the similarity between the current skeleton and the base skeleton is calculated. If the similarity is greater than a preset first threshold, the parts that are different from the current skeleton and the base skeleton are merged into the base skeleton. If the similarity is less than or equal to the first threshold, the current skeleton is multiplied by a weight matrix and then merged with the base skeleton. The weight matrix is ​​calculated by the distance transformation from each point in the current skeleton to the nearest non-zero point. The branches whose pixel lengths in the merged basic skeleton are less than a preset second threshold are removed, and the merged basic skeleton is smoothed by mean filtering, that is, the body skeleton matrix is ​​output.

6. The seat temperature control system based on thermal management according to claim 2, characterized in that: The training process of the user identification model is as follows: Extract all different user labels and their corresponding body shape feature data from the database as the first sample set, construct a feature vector G1 based on the body shape feature data corresponding to each user label, use Keras to define the input layer and output layer, and build the model framework; Take the feature vector G1 as input data and the user label as output data; The first sample set is divided into a training set and a test set in a ratio of 70% and 30%, a user recognition model is constructed and trained, the user recognition model is tested using the test set, and a user recognition model that meets the preset accuracy is output.

7. The seat temperature control system based on thermal management according to claim 1, characterized in that: The training process of the seat temperature control parameter prediction model is as follows: All user tags and their corresponding internal and external environment parameters and temperature control parameters are used as the second sample set, and the temperature control parameters are numbered; Construct the feature vector G2 from the user label and its corresponding internal and external environment parameters, use TensorFlow to define the input and output layers, and build the model framework; The eigenvector G2 is used as input data and the temperature control parameter number is used as output data; The second sample set is divided into a training set and a test set in a ratio of 70% and 30%, respectively. A seat temperature control parameter prediction model is constructed and trained. The seat temperature control parameter prediction model is tested using the test set, and a seat temperature control parameter prediction model that meets the preset accuracy is output.

8. The seat temperature control system based on thermal management according to claim 2, characterized in that: The step of updating the user identification model and the seat temperature control parameter prediction model and applying the updated user identification model and the seat temperature control parameter prediction model includes: Determine whether the current user's body feature data is recognized; If the current user's body feature data is not recognized by the user recognition model, the current user's body feature data and the corresponding internal and external environment parameters and temperature control parameters are used as a new user data set and a new user label is generated. The new user data set and the new user label are then added to the existing database; The user identification model is trained and updated using the new user label and the corresponding body feature data, thereby obtaining an updated user identification model; The seat temperature control parameter prediction model is trained and updated using the internal and external environmental parameters and temperature control parameters corresponding to the new user tags, thereby obtaining an updated seat temperature control parameter prediction model. The updated user identification model and the updated seat temperature control parameter prediction model are applied.

9. The seat temperature control system based on thermal management according to claim 8, characterized in that: The updating of the user identification model and the seat temperature control parameter prediction model also includes: After the current user's body shape feature data is identified, the other body shape skeleton matrices that existed during the same seat belt usage time are added to the current user's body shape feature data, and then the user recognition model is trained and updated; The seat temperature control parameter prediction model is trained and updated using the new internal and external environmental parameters and new temperature control parameters corresponding to the current user's body feature data.

10. The seat temperature control system based on thermal management according to claim 9, characterized in that: The method for determining whether the current user's body shape feature data is recognized is: If the user recognition model can output the user label, it is determined that the body feature data of the current user is recognized; otherwise, it is determined that the body feature data of the current user is not recognized.

Images