Sitting posture recognition method, vehicle controller and vehicle

Abstract

The invention relates to the technical field of vehicles, in particular to a sitting posture recognition method, a vehicle controller and a vehicle. The method comprises the following steps: acquiring an actual pressure value acquired by a preset pressure acquisition device which is arranged on a seat of a vehicle and is used for sitting posture recognition; the actual pressure values are analyzed, and coordinate information of each actual pressure value and the key points of the human body is determined; based on the coordinate information, determining a skeleton vector corresponding to each target skeleton; wherein the target skeleton comprises at least any one of a left thigh skeleton, a right thigh skeleton, a shoulder skeleton and a spine skeleton; and performing data classification based on the coordinate information, the skeleton vector and the actual pressure value, determining the actual probability of each sitting posture classification, and determining a sitting posture recognition result based on the actual probability.

Description

Technical Field

[0001] The present disclosure relates to the field of vehicle technology, and in particular to a sitting posture recognition method, a vehicle controller and a vehicle. Background Art

[0002] With the popularity of smart cars, more and more people choose to travel by car. When riding for a long time, the driver is likely to have an abnormal sitting posture due to fatigue, which increases the risk of bone and muscle diseases and easily leads to various health problems.

[0003] Therefore, how to accurately understand the driver's sitting posture has become a problem that needs to be solved urgently. Summary of the invention

[0004] The present disclosure provides a sitting posture recognition method, a vehicle controller and a vehicle, which are used to solve the problem of how to accurately recognize the sitting posture of a driver.

[0005] In a first aspect, the present application provides a sitting posture recognition method, comprising: obtaining actual pressure values ​​collected by a preset pressure collection device for sitting posture recognition set on a seat of a vehicle; analyzing the actual pressure values ​​to determine the coordinate information of each actual pressure value and a key point of the human body; based on the coordinate information, determining the bone vector corresponding to each target bone; wherein the target bones include at least one of the left thigh bone, the right thigh bone, the shoulder bone and the spinal bone; performing data classification based on the coordinate information, the bone vector and the actual pressure values ​​to determine the actual probability of each sitting posture classification; wherein the sitting posture classification includes normal sitting posture and abnormal sitting posture; based on the actual probability, determining the sitting posture recognition result.

[0006] In some feasible examples, before obtaining the actual pressure value collected by the preset pressure collection device for sitting posture recognition set on the seat of the vehicle, the sitting posture recognition method provided by the present application also includes: obtaining first training sample data and a first marking result of the first training sample data; wherein the first training sample data includes at least one group of historical pressure values ​​collected by the preset pressure collection device for sitting posture recognition set on the seat of the vehicle, and the first marking result includes the human body key point corresponding to each historical pressure value in each group of historical pressure values, and the coordinate information of the human body key point; inputting the first training sample data into the detection model based on the visual algorithm for learning, and obtaining the first prediction result of the detection model based on the visual algorithm for the first training sample data; adjusting the network parameters of the detection model based on the visual algorithm based on the first prediction result and the first marking result until the detection model based on the visual algorithm converges to obtain the key point detection model; analyzing the actual pressure values ​​to determine the coordinate information of each actual pressure value and the human body key point, including: inputting the actual pressure values ​​into the key point detection model for analysis, and determining the coordinate information of each actual pressure value and the human body key point.

[0007] In some implementable examples, the key points of the human body include at least a left shoulder key point, a right shoulder key point, a left hip key point, a right hip key point, a left thigh key point and a right thigh key point; based on the coordinate information, the bone vector corresponding to each target bone is determined, including: determining the bone vector corresponding to the left thigh bone based on the line connecting the coordinate point corresponding to the coordinate information of the left hip key point and the coordinate point corresponding to the coordinate information of the left thigh key point; determining the bone vector corresponding to the right thigh bone based on the line connecting the coordinate point corresponding to the coordinate information of the right hip key point and the coordinate point corresponding to the coordinate information of the right thigh key point; determining the bone vector corresponding to the shoulder bone based on the line connecting the coordinate point corresponding to the coordinate information of the left shoulder key point and the coordinate point corresponding to the coordinate information of the right shoulder key point; determining the bone vector corresponding to the spinal bone based on the line connecting the coordinate point corresponding to the average value of the coordinate information of the left shoulder key point and the coordinate information of the right shoulder key point and the coordinate point corresponding to the average value of the coordinate information of the left hip key point and the coordinate information of the right hip key point.

[0008] In some feasible examples, before obtaining the actual pressure value collected by the preset pressure collection device for sitting posture recognition set on the seat of the vehicle, the sitting posture recognition method provided by the present application also includes: obtaining second training sample data and second labeling results of the second training sample data; wherein the second training sample data includes historical bone vectors, historical pressure values ​​and coordinate information of each key point of the human body corresponding to different sitting postures when the driver sits on the seat of the vehicle, and the second labeling results include any one of normal sitting posture and abnormal sitting posture; inputting the second training sample data into the fully connected neural network for learning, and obtaining the learning results of the fully connected neural network for the second training sample data; using the normalized indicator function to classify the learning results, and obtaining the predicted probability of the learning results in each sitting posture classification; based on the cross entropy loss function, calculating the loss value of the sitting posture classification corresponding to the maximum predicted probability and the second labeling result; when the loss value is less than the target value, determining that the fully connected neural network and the normalized indicator function converge, and obtaining the sitting posture analysis model;

[0009] Data classification is performed based on coordinate information, bone vectors and actual pressure values ​​to determine the actual probability of each sitting posture classification, including: determining sitting posture analysis data based on coordinate information, bone vectors and actual pressure values; normalizing the sitting posture analysis data to obtain normalized data; inputting the normalized data into a sitting posture analysis model for data classification to determine the actual probability of the normalized data in each sitting posture classification.

[0010] In some feasible examples, the sitting posture recognition result is determined based on the actual probability, including: when the sitting posture corresponding to the maximum actual probability is classified as an abnormal sitting posture, determining the sitting posture recognition result as an abnormal sitting posture.

[0011] In some feasible examples, the sitting posture recognition method provided in the present application also includes: when the sitting posture recognition result is an abnormal sitting posture, generating prompt information; wherein the prompt information is used for at least the user to adjust the current sitting posture.

[0012] In a second aspect, the present application provides a vehicle controller, comprising: an acquisition module, used to acquire actual pressure values ​​collected by a preset pressure collection device for sitting posture recognition set on a seat of the vehicle; a processing module, used to analyze the actual pressure values ​​acquired by the acquisition module, and determine the coordinate information of each actual pressure value and a key point of the human body; the processing module is also used to determine the bone vector corresponding to each target bone based on the coordinate information; wherein the target bones include at least one of the left thigh bone, the right thigh bone, the shoulder bone and the spinal bone; the processing module is also used to perform data classification based on the coordinate information, the bone vector and the actual pressure values ​​acquired by the acquisition module, and determine the actual probability of each sitting posture classification; wherein the sitting posture classification includes normal sitting posture and abnormal sitting posture; the processing module is also used to determine the sitting posture recognition result based on the actual probability.

[0013] In some feasible examples, the acquisition module is also used to acquire first training sample data and a first labeling result of the first training sample data; wherein the first training sample data includes at least one group of historical pressure values ​​collected by a preset pressure collection device for sitting posture recognition arranged on a vehicle seat, and the first labeling result includes a human key point corresponding to each historical pressure value in each group of historical pressure values, and coordinate information of the human key point; the processing module is also used to input the first training sample data acquired by the acquisition module into a detection model based on a visual algorithm for learning, and obtain a first prediction result of the detection model based on the visual algorithm for the first training sample data; the processing module is also used to adjust the network parameters of the detection model based on the visual algorithm based on the first prediction result and the first labeling result acquired by the acquisition module until the detection model based on the visual algorithm converges to obtain a key point detection model; the processing module is specifically used to input the actual pressure value acquired by the acquisition module into the key point detection model for analysis, and determine the coordinate information of each actual pressure value and the human key point.

[0014] In a third aspect, the present application provides a vehicle, the vehicle comprising the vehicle controller as described above.

[0015] In a fourth aspect, the present application provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the above method.

[0016] Compared with the prior art, the technical solution provided by the present invention has the following advantages:

[0017] The sitting posture device method provided by the present disclosure obtains the actual pressure value collected by the preset pressure collection device for sitting posture recognition set on the seat of the vehicle; analyzes the actual pressure value to determine the coordinate information of the key points of the human body for each actual pressure value; determines the bone vector corresponding to each target bone based on the coordinate information; performs data classification based on the coordinate information, the bone vector and the actual pressure value to determine the actual probability of each sitting posture classification; then, determines the sitting posture recognition result based on the actual probability. In this way, the vehicle controller can determine the sitting posture recognition result of the driver based on the actual pressure value collected by the pressure collection device installed on the vehicle seat. When the sitting posture classification corresponding to the largest actual probability is an abnormal sitting posture, the sitting posture recognition result is determined to be an abnormal sitting posture, thereby identifying the driver's sitting posture, solving the problem of how to accurately identify the driver's sitting posture. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0019] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0020] Figure 1 exemplarily shows one of the flow charts of a sitting posture recognition method provided in the first embodiment;

[0021] Figure 2 exemplarily shows one of the scene schematic diagrams of a sitting posture recognition method provided in the first embodiment;

[0022] Figure 3 FIG. 2 exemplarily shows a second scene diagram of a sitting posture recognition method provided in the first embodiment;

[0023] Figure 4 FIG2 exemplarily shows a second flow chart of a sitting posture recognition method provided in the first embodiment;

[0024] Figure 5 FIG3 is a flowchart of a sitting posture recognition method provided in the first embodiment;

[0025] Figure 6 FIG4 is a flowchart of a sitting posture recognition method provided by the first embodiment;

[0026] Figure 7FIG. 5 exemplarily shows a flowchart of a sitting posture recognition method provided in the first embodiment;

[0027] Figure 8 6 is a flowchart of a sitting posture recognition method provided in the first embodiment;

[0028] Fig. 9 One of the structural diagrams of the vehicle controller provided in the second embodiment is exemplarily shown in FIG.

[0029] Fig.10 FIG. 2 exemplarily shows a second structural diagram of the vehicle controller provided in the second embodiment. DETAILED DESCRIPTION

[0030] In order to more clearly understand the above-mentioned objectives, features and advantages of the present disclosure, the scheme of the present disclosure will be further described below. It should be noted that the embodiments of the present disclosure and the features in the embodiments can be combined with each other without conflict.

[0031] In the following description, many specific details are set forth to facilitate a full understanding of the present disclosure, but the present disclosure may also be implemented in other ways different from those described herein; it is obvious that the embodiments in the specification are only part of the embodiments of the present disclosure, rather than all of the embodiments.

[0032] It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0033] YOLOv8-Pose in the disclosed embodiment is a variant of YOLOv8, which is specially designed for human pose estimation. YOLO (You Only Look Once) is a popular real-time target detection algorithm, known for its speed and accuracy.

[0034] Softmax in the embodiments of the present disclosure is a mathematical function, which is generally used to convert a set of arbitrary real numbers into real numbers representing probability distribution.

[0035] Embodiment 1

[0036] Figure 1 A flowchart of a sitting posture recognition method is shown in FIG. 1 . The execution subject of this example may be a vehicle controller, such as Figure 1 As shown, the method includes:

[0037] S11, obtaining an actual pressure value collected by a preset pressure collection device for sitting posture recognition disposed on a seat of a vehicle.

[0038] In some examples, in order to facilitate the analysis of the driver's sitting posture, at least two pressure collection devices are provided on the vehicle seat, so that the driver's sitting posture can be analyzed based on the actual pressure values ​​collected by the pressure collection devices.

[0039] In some examples, the pressure acquisition device may be a pressure sensor.

[0040] In some examples, in order to understand the driver's sitting posture, it is necessary to mark the key points of the driver's body in different sitting postures. At the same time, in order to facilitate the positioning of the key points of the body, a coordinate system needs to be established, so that the coordinate information of the key points of the body can be determined based on the coordinate system.

[0041] Exemplarily, the seat 1 of the vehicle is as follows Figure 2 As shown, it includes a backrest portion 1-1 and a seat cushion portion 1-2. The seat cushion portion 1-2 includes four vertices, namely, an upper left vertex a-1, a lower left vertex a-2, an upper right vertex b-1, and an upper right vertex b-2. With the upper right vertex b-2 as the origin, the direction perpendicular to the plane where the seat cushion portion 1-2 is located as the y direction, the direction parallel to the plane where the seat cushion portion 1-2 is located as the x direction, and the direction perpendicular to the x direction and parallel to the plane where the seat cushion portion 1-2 is located as the -y direction, the following is established: Figure 3 In this way, the coordinate information of each actual pressure value and key points of the human body can be determined based on the established coordinate system.

[0042] S12. Analyze the actual pressure values ​​to determine the coordinate information of each actual pressure value and key points of the human body.

[0043] In some examples, when the driver sits on the seat of the vehicle, the pressure collection device set on the seat in the car will report the corresponding actual pressure value, so that the driver's current sitting posture and the coordinate information of each key point of the driver's body can be determined based on the actual pressure value, such as: all the actual pressure values ​​are input into the key point detection model for analysis to determine the coordinate information of each actual pressure value and the key point of the human body. Or, based on the sum of all the actual pressure values, the total pressure value is determined. Afterwards, based on the total pressure value, the corresponding relationship pre-stored in the memory of the vehicle is queried to determine the theoretical sitting posture corresponding to the total pressure value. Afterwards, based on the theoretical sitting posture and the actual pressure set (including all the actual pressure values), a query is performed in the data set pre-stored in the memory of the vehicle to determine the theoretical pressure set that matches the actual pressure set under the theoretical sitting posture, and the coordinate information of each key point of the human body in the coordinate information set corresponding to the theoretical pressure set is used as the coordinate information of each key point of the human body of the driver.

[0044] For example, taking the theoretical pressure set containing 20 pressure values ​​and the human body key points containing 11 key points (1-left shoulder key point, 2-right shoulder key point, 3-left back key point, 4-right back key point, 5-left waist key point, 6-right waist key point, 7-left hip key point, 8-right hip key point, 9-left thigh key point, 10-right thigh key point, 11-neck key point) as an example, the correspondence between the theoretical sitting posture, the theoretical pressure set and the coordinate information of the human body key points is shown in Table 1.

[0045] Table 1

[0046]

[0047] When the actual pressure set is [10,20,30,50,40,0,0,0,0,150,180,220,230,280,100,33,44,55,66,77], based on the actual pressure set query Table 1, it can be known that the pressure set matching the actual pressure set is the theoretical pressure set 1. Therefore, it can be determined that the coordinate information of each human body key point corresponding to the actual pressure set is the coordinate information of each human body key point contained in the coordinate information set 1 corresponding to the theoretical pressure set 1.

[0048] Specifically, the theoretical pressure set that matches the actual pressure set can be determined based on the sum of all pressure values ​​in the actual pressure set. For example, if the sum of all pressure values ​​in the actual pressure set is 400, the set in which the sum of all pressure values ​​in the theoretical pressure set is equal to 400 is used as the theoretical pressure set that matches the actual pressure set.

[0049] Alternatively, based on the sum of all pressure values ​​in the actual pressure set and the position of each actual pressure value in the actual pressure set, a matching theoretical pressure set is determined, such as: the sum of all pressure values ​​in the actual pressure set is 400, the sum of all pressure values ​​in theoretical pressure set 1 is equal to 400, and the sum of all pressure values ​​in theoretical pressure set 2 is determined to be equal to 400. Since the pressure value at each position in theoretical pressure set 2 is the same as the actual pressure value at each position in the actual pressure set, theoretical pressure set 2 is determined to be the theoretical pressure set that matches the actual pressure set.

[0050] S13, based on the coordinate information, determining the bone vector corresponding to each target bone, wherein the target bones include at least one of the left thigh bone, the right thigh bone, the shoulder bone and the spine bone.

[0051] In some examples, in order to facilitate the analysis of the driver's sitting posture, it is necessary to determine the position of the driver's target bone. Therefore, the bone vector corresponding to each target bone can be determined based on the coordinate information, such as: based on the line connecting the coordinate point corresponding to the coordinate information of the left span key point and the coordinate point corresponding to the coordinate information of the left thigh key point, the bone vector corresponding to the left thigh bone is determined; based on the line connecting the coordinate point corresponding to the coordinate information of the right span key point and the coordinate point corresponding to the coordinate information of the right thigh key point, the bone vector corresponding to the right thigh bone is determined; based on the line connecting the coordinate point corresponding to the coordinate information of the left shoulder key point and the coordinate point corresponding to the coordinate information of the right shoulder key point, the bone vector corresponding to the shoulder bone is determined; based on the line connecting the coordinate point corresponding to the average value of the coordinate information of the left shoulder key point and the coordinate information of the right shoulder key point, and the coordinate point corresponding to the average value of the coordinate information of the left span key point and the coordinate information of the right span key point, the bone vector corresponding to the spinal bone is determined.

[0052] S14, classify data based on coordinate information, bone vectors and actual pressure values, and determine the actual probability of each sitting posture classification. The sitting posture classification includes normal sitting posture and abnormal sitting posture.

[0053] In some examples, sitting posture analysis data is determined based on coordinate information, bone vectors, and actual pressure values; then, the sitting posture analysis data is normalized to obtain normalized data; and data classification is performed on the normalized data to determine the actual probability of each sitting posture classification in the normalized data.

[0054] S15. Determine the sitting posture recognition result based on the actual probability.

[0055] In some examples, in order to prevent the driver from moving while driving the vehicle, which may lead to the driver's sitting posture being misidentified as an abnormal sitting posture, the sitting posture recognition method provided in the embodiment of the present disclosure can prevent the possible misdetection of single-frame detection when the sitting posture recognition results are abnormal sitting postures for n consecutive times (such as 5 times). When an abnormal sitting posture is detected, an early warning can be issued to the people in the car to provide prompt information and guide the relevant people to correct their sitting posture. After completing an early warning, the actual pressure value collected by the pressure collection device at each moment is re-analyzed at an interval of m (such as 20) minutes to avoid frequent prompts to the driver and distracting the driver's attention.

[0056] From the above, it can be seen that in the sitting posture device method provided by the embodiment of the present disclosure, the vehicle controller can determine the driver's sitting posture recognition result based on the actual pressure value collected by the pressure collection device installed on the vehicle seat. When the sitting posture corresponding to the largest actual probability is classified as an abnormal sitting posture, the sitting posture recognition result is determined to be an abnormal sitting posture, thereby identifying the driver's sitting posture, solving the problem of how to accurately identify the driver's sitting posture.

[0057] In some possible implementation examples, combined with Figure 1 ,like Figure 4 As shown, before executing S11, the sitting posture recognition method provided by the embodiment of the present disclosure further includes S16-S18, and the above S12 can be specifically implemented by the following S120.

[0058] S16, obtaining first training sample data and a first labeling result of the first training sample data; wherein the first training sample data includes at least one set of historical pressure values ​​collected by a preset pressure collection device for sitting posture recognition disposed on a vehicle seat, and the first labeling result includes a human body key point corresponding to each historical pressure value in each set of historical pressure values, and coordinate information of the human body key point;

[0059] In some examples, when a person sits on a seat, the pressure collection device transmits the actual pressure value in real time, and the actual pressure value is divided into two parts: the backrest and the cushion (i.e., the actual pressure value collected by the pressure sensor arranged on the backrest part and the actual pressure value collected by the pressure collection device arranged on the cushion part). The actual pressure values ​​collected by the pressure collection device at the same time are taken as a set of actual pressure sets, and the human body key corresponding to the driver at that time and the coordinate information corresponding to the human body key points are marked, so as to obtain the mapping relationship between the actual pressure set, the human body key, and the coordinate information corresponding to the human body key points. Afterwards, the mapping relationships corresponding to different drivers at different times can be statistically analyzed to obtain the first training sample data and the first labeling result of the first training sample data.

[0060] For example, the backrest part of the vehicle seat is provided with 7×7 pressure collection devices, and the cushion part of the vehicle seat is provided with 6×7 pressure collection devices. At this time, each group of actual pressure sets contains 13×7 actual pressure values.

[0061]

[0062] Among them, K n represents the actual pressure set corresponding to time n, x m Indicates the horizontal coordinate corresponding to the mth human key point, y m Represents the ordinate corresponding to the mth human key point.

[0063] In some examples, when collecting actual pressure sets, data from as many different people as possible is required, and data from different abnormal sitting postures and normal sitting postures are collected to ensure sufficient data volume. Statistics are performed on the mapping relationship, and the data in each set of actual pressure sets is upsampled to 416×224 using the bilinear difference method, so that more training sample data can be obtained.

[0064] The reason for this is that the input of the deep learning model we will use later needs to be a multiple of 32, and secondly, it can enrich the details of the pressure matrix so that the model can better learn the features. If the coordinates corresponding to some key points are outside the pressure sensor, the point is marked as empty.

[0065] S17, inputting the first training sample data into the detection model based on the visual algorithm for learning, and obtaining a first prediction result of the detection model based on the visual algorithm for the first training sample data;

[0066] In some examples, the vision algorithm based detection model can be a YOLOv8-Pose model.

[0067] S18. Adjust the network parameters of the detection model based on the visual algorithm based on the first prediction result and the first marking result until the detection model based on the visual algorithm converges to obtain a key point detection model.

[0068] S120, inputting the actual pressure value into the key point detection model for analysis, and determining the coordinate information of each actual pressure value and the key point of the human body.

[0069] From the above, it can be seen that by training a detection model based on a visual algorithm, the coordinate information of the key points of the human body can be predicted based on the actual pressure value collected by the pressure acquisition device, and then the driver's sitting posture can be analyzed based on the coordinate information to determine whether the driver's current sitting posture is abnormal, thereby improving the user experience.

[0070] In some feasible examples, the key points of the human body include at least a left shoulder key point, a right shoulder key point, a left hip key point, a right hip key point, a left thigh key point and a right thigh key point; Figure 1 ,like Figure 5 As shown, the above S13 can be specifically implemented through the following S130-S133.

[0071] S130, determining a bone vector corresponding to the left thigh bone based on a line connecting a coordinate point corresponding to the coordinate information of the left span key point and a coordinate point corresponding to the coordinate information of the left thigh key point;

[0072] S131, determining a bone vector corresponding to the right thigh bone based on a line connecting a coordinate point corresponding to the coordinate information of the right span key point and a coordinate point corresponding to the coordinate information of the right thigh key point;

[0073] S132, determining a bone vector corresponding to the shoulder bone based on a line connecting a coordinate point corresponding to the coordinate information of the left shoulder key point and a coordinate point corresponding to the coordinate information of the right shoulder key point;

[0074] S133. Determine the bone vector corresponding to the spinal bone based on the coordinate point corresponding to the average value of the coordinate information of the left shoulder key point and the coordinate information of the right shoulder key point, and the line connecting the coordinate point corresponding to the average value of the coordinate information of the left span key point and the coordinate information of the right span key point.

[0075] From the above, it can be seen that in order to facilitate the analysis of the driver's current sitting posture, by analyzing the key points of the human body, the bone vector corresponding to the target bone can be determined, and then the driver's sitting posture can be analyzed based on the bone vector to determine whether the driver's current sitting posture is abnormal and improve the user experience.

[0076] In some possible implementation examples, combined with Figure 1 ,like Figure 6 As shown, before executing S11, the sitting posture recognition method provided by the embodiment of the present disclosure also includes S19-S23, and the above S14 can be specifically implemented through the following S140-S142.

[0077] S19. Obtain second training sample data and a second labeling result of the second training sample data; wherein the second training sample data includes historical bone vectors, historical pressure values, and coordinate information of each key point of the human body corresponding to different sitting postures when the driver sits on the seat of the vehicle, and the second labeling result includes any one of a normal sitting posture and an abnormal sitting posture.

[0078] In some examples, there are four abnormal sitting postures, namely, leaning to the left, leaning to the right, crossing the left leg, and crossing the right leg. Since drivers have various sitting postures and many actions are difficult to distinguish, the sitting posture recognition method provided in the embodiment of the present invention analyzes the key points of the human body in different sitting postures, extracts the abnormal sitting posture features of the key points of the human body in combination with the key points of the human body, and distinguishes different sitting postures. We found that under normal circumstances, the abnormal sitting posture of the human body is highly correlated with the angles of the limbs, spine, and shoulders.

[0079] In the above step S12, we obtain the position information of the key points of the human body, and then we can obtain the bone vectors corresponding to the left thigh, right thigh, shoulder, spine, etc. according to the position information, and then we can calculate the local bone angle of the abnormal sitting posture of the human body according to the bone vector. Abnormal sitting posture is usually related to the angle of the human limbs. We calculate the bone vectors of the left thigh, right thigh, shoulder, and spine according to the key points, as shown in Table 2 below.

[0080] Table 2

[0081]

[0082] Let the x-axis and y-axis unit vectors be u1 and u2 respectively. We can calculate the angles between each bone vector and the x-axis and y-axis at time n based on the bone vector.

[0083] Among them (i=1,2,3,4;j=1,2).

[0084] According to the 8 angles calculated above, construct the angle set:

[0085] θ={θ ij |i=1,2,3,4;j=1,2};

[0086] At the same time, considering the integrity of the features, the 13×7 pressure value is pulled into a 1×91 vector p, and the horizontal coordinates corresponding to all the key points of the human body in the actual pressure set corresponding to the time n are combined into the horizontal coordinate vector K x , the ordinates corresponding to all the key points of the human body in the actual pressure set corresponding to the time n are combined into a ordinate vector K y , and the θ calculated above are used as sitting posture analysis data at the same time. The specific formula is as follows: X=(θ,p,K x ,K y ).

[0087] At the same time, the sitting posture analysis data corresponding to the time n is marked with the driver's sitting posture, so as to obtain the marking results (normal sitting posture and abnormal sitting posture) corresponding to the sitting posture analysis data.

[0088] In this way, the sitting posture analysis data corresponding to time n and the marking results corresponding to the sitting posture analysis data are obtained. Afterwards, the sitting posture analysis data corresponding to different drivers at different times and the marking results corresponding to the sitting posture analysis data can be statistically analyzed to obtain the second training sample data and the marking results of the second training sample data.

[0089] S20: Input the second training sample data into the fully connected neural network for learning, and obtain a learning result of the fully connected neural network on the second training sample data.

[0090] In some examples, when the second training sample data is input into the fully connected neural network for learning, the second training sample data needs to be normalized, that is, the second training sample data needs to be scaled to between 0 and 1, so as to facilitate model convergence.

[0091] S21, classifying the learning results using a normalized indicator function to obtain the predicted probability of the learning results in each sitting posture classification;

[0092] In some examples, the normalized indicator function can be a Softmax function.

[0093] S22. Based on the cross entropy loss function, calculate the loss value of the sitting posture classification corresponding to the maximum prediction probability and the second labeling result.

[0094] In some examples, when the sitting posture corresponding to the maximum predicted probability is classified as an abnormal sitting posture and the second marking result is a normal sitting posture, the loss value at this time is 1; when the sitting posture corresponding to the maximum predicted probability is classified as an abnormal sitting posture and the second marking result is an abnormal sitting posture, the loss value at this time is 0.

[0095] S23. When the loss value is less than the target value, determine that the fully connected neural network and the normalized indicator function converge to obtain a sitting posture analysis model;

[0096] S140, determining sitting posture analysis data based on the coordinate information, the bone vector and the actual pressure value;

[0097] S141, normalizing the sitting posture analysis data to obtain normalized data;

[0098] S142, inputting the normalized data into the sitting posture analysis model for data classification, and determining the actual probability of the normalized data in each sitting posture classification.

[0099] From the above, it can be seen that by training the fully connected neural network using the normalized indicator function, the driver's sitting posture can be analyzed based on the actual pressure value collected by the pressure acquisition device, the coordinate information of the key points of the human body, and the bone vector of the target bone, so as to determine whether the driver's current sitting posture is abnormal and improve the user experience.

[0100] In some possible implementation examples, combined with Figure 1 ,like Figure 7 As shown, the above S15 can be specifically implemented through the following S150.

[0101] S150. When the sitting posture corresponding to the maximum actual probability is classified as an abnormal sitting posture, determine that the sitting posture recognition result is an abnormal sitting posture.

[0102] In some examples, when the sitting posture classification corresponding to the maximum actual probability is a normal sitting posture, the sitting posture recognition result is determined to be a normal sitting posture. Since the sitting posture recognition result of the driver is a normal sitting posture, there is no need to generate prompt information.

[0103] From the above, it can be seen that by analyzing the actual probability corresponding to each sitting posture classification, the driver's current sitting posture can be determined. For example, the sitting posture classification corresponding to the largest actual probability is an abnormal sitting posture, which means that there is an abnormality in the driver's current seat. Then, based on the sitting posture recognition result, the driver can be prompted to correct the sitting posture to improve the user experience.

[0104] In some possible implementation examples, combined with Figure 1 ,like Figure 8 As shown, the sitting posture recognition method provided in the above S15 embodiment of the present disclosure also includes: S24.

[0105] S24. When the sitting posture recognition result is an abnormal sitting posture, generate prompt information; wherein the prompt information is used for at least the user to adjust the current sitting posture.

[0106] From the above, it can be seen that by analyzing the actual probability corresponding to each sitting posture classification, the driver's current sitting posture can be determined. For example, the sitting posture classification corresponding to the largest actual probability is an abnormal sitting posture, which means that there is an abnormality in the driver's current seat. Then, based on the sitting posture recognition result, the driver can be prompted to correct the sitting posture to improve the user experience.

[0107] Embodiment 2

[0108] Fig. 9 The structural diagram of the vehicle controller provided in the second embodiment of the present application is exemplarily shown in FIG. Fig. 9 As shown, the vehicle controller includes: an acquisition module 81 and a processing module 82.

[0109] An acquisition module 81 is used to acquire an actual pressure value collected by a preset pressure collection device for sitting posture recognition provided on a seat of a vehicle;

[0110] A processing module 82 is used to analyze the actual pressure values ​​obtained by the obtaining module 81 to determine the coordinate information of each actual pressure value and a key point of the human body;

[0111] The processing module 82 is further used to determine the bone vector corresponding to each target bone based on the coordinate information; wherein the target bone includes at least one of a left thigh bone, a right thigh bone, a shoulder bone and a spine bone;

[0112] The processing module 82 is further used to classify data based on the coordinate information, the bone vector and the actual pressure value obtained by the acquisition module 81, and determine the actual probability of each sitting posture classification; wherein the sitting posture classification includes a normal sitting posture and an abnormal sitting posture;

[0113] The processing module 82 is further used to determine the sitting posture recognition result based on the actual probability.

[0114] In some feasible examples, the acquisition module 81 is also used to acquire first training sample data and a first labeling result of the first training sample data; wherein the first training sample data includes at least one group of historical pressure values ​​collected by a preset pressure collection device for sitting posture recognition arranged on a vehicle seat, and the first labeling result includes a human key point corresponding to each historical pressure value in each group of historical pressure values, and coordinate information of the human key point; the processing module 82 is also used to input the first training sample data acquired by the acquisition module 81 into a detection model based on a visual algorithm for learning, and obtain a first prediction result of the detection model based on the visual algorithm for the first training sample data; the processing module 82 is also used to adjust the network parameters of the detection model based on the visual algorithm based on the first prediction result and the first labeling result acquired by the acquisition module 81, until the detection model based on the visual algorithm converges to obtain a key point detection model; the processing module 82 is specifically used to input the actual pressure value acquired by the acquisition module 81 into the key point detection model for analysis, and determine the coordinate information of each actual pressure value and the human key point.

[0115] In some implementable examples, the key points of the human body include at least a left shoulder key point, a right shoulder key point, a left hip key point, a right hip key point, a left thigh key point and a right thigh key point; the processing module 82 is specifically used to determine the bone vector corresponding to the left thigh bone based on the line connecting the coordinate point corresponding to the coordinate information of the left hip key point and the coordinate point corresponding to the coordinate information of the left thigh key point; the processing module 82 is specifically used to determine the bone vector corresponding to the right thigh bone based on the line connecting the coordinate point corresponding to the coordinate information of the right hip key point and the coordinate point corresponding to the coordinate information of the right thigh key point; the processing module 82 is specifically used to determine the bone vector corresponding to the shoulder bone based on the line connecting the coordinate point corresponding to the coordinate information of the left shoulder key point and the coordinate point corresponding to the coordinate information of the right shoulder key point; the processing module 82 is specifically used to determine the bone vector corresponding to the spinal bone based on the line connecting the coordinate point corresponding to the average value of the coordinate information of the left shoulder key point and the coordinate information of the right shoulder key point and the coordinate point corresponding to the average value of the coordinate information of the left hip key point and the coordinate information of the right hip key point.

[0116] In some feasible examples, the acquisition module 81 is further used to acquire the second training sample data and the second labeling result of the second training sample data; wherein the second training sample data includes the historical bone vectors, historical pressure values ​​and coordinate information of each key point of the human body corresponding to different sitting postures when the driver sits on the seat of the vehicle, and the second labeling result includes any one of a normal sitting posture and an abnormal sitting posture; the processing module 82 is further used to input the second training sample data acquired by the acquisition module 81 into the fully connected neural network for learning, and obtain the learning result of the fully connected neural network for the second training sample data; the processing module 82 is further used to classify the learning result using a normalized indicator function, and obtain the learning result in each sitting posture classification. Prediction probability; processing module 82 is also used to calculate the loss value of the sitting posture classification and the second labeling result corresponding to the maximum prediction probability based on the cross entropy loss function; processing module 82 is also used to determine the convergence of the fully connected neural network and the normalized indicator function when the loss value is less than the target value, so as to obtain the sitting posture analysis model; processing module 82 is specifically used to determine the sitting posture analysis data based on the coordinate information, the bone vector and the actual pressure value obtained by the acquisition module 81; processing module 82 is specifically used to normalize the sitting posture analysis data to obtain normalized data; processing module 82 is specifically used to input the normalized data into the sitting posture analysis model for data classification, and determine the actual probability of each sitting posture classification in the normalized data.

[0117] In some implementable examples, the processing module 82 is specifically configured to determine that the sitting posture recognition result is an abnormal sitting posture when the sitting posture corresponding to the maximum actual probability is classified as an abnormal sitting posture.

[0118] In some feasible examples, the processing module 82 is further used to generate prompt information when the sitting posture recognition result is an abnormal sitting posture; wherein the prompt information is used to at least prompt the user to adjust the current sitting posture.

[0119] Of course, the vehicle controller provided by the embodiment of the present invention includes but is not limited to the above modules, for example, the vehicle controller may also include a storage module 83. The storage module 83 may be used to store the program code of the vehicle controller, and may also be used to store data generated during the operation of the vehicle controller, such as diagnostic data.

[0120] Fig.10 A schematic diagram of the structure of a vehicle controller provided by an embodiment of the present invention is shown in FIG. Fig.10 As shown, the vehicle controller may include: at least one processor 51 , a memory 52 , a communication interface 53 and a communication bus 54 .

[0121] Combine the following Fig.10 A detailed introduction to the various components of the vehicle controller:

[0122] The processor 51 is the control center of the vehicle controller, and may be a processor or a general term for multiple processing elements. For example, the processor 51 is a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention, such as one or more DSPs, or one or more field programmable gate arrays (FPGAs).

[0123] In a specific implementation, as an embodiment, the processor 51 may include one or more CPUs, such as Fig.10 Also, as an embodiment, the vehicle controller may include multiple processors, such as Fig.10 51 and 55 are shown in FIG. Each of these processors may be a single-core processor (Single-CPU) or a multi-core processor (Multi-CPU). The processor here may refer to one or more devices, circuits, and / or processing cores for processing data (e.g., computer program instructions).

[0124] The memory 52 may be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 52 may exist independently and be connected to the processor 51 via a communication bus 54. The memory 52 may also be integrated with the processor 51.

[0125] In a specific implementation, the memory 52 is used to store the data of the present invention and execute the software program of the present invention. The processor 51 can execute various functions of the air conditioner by running or executing the software program stored in the memory 52 and calling the data stored in the memory 52.

[0126] The communication interface 53 uses any transceiver-like device to communicate with other devices or communication networks, such as Radio Access Network (RAN), Wireless Local Area Networks (WLAN), terminals, the cloud, etc. The communication interface 53 may include an acquisition module 81 to implement the acquisition function.

[0127] The communication bus 54 may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Fig.10 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0128] As an example, combining Fig. 9The functions implemented by the acquisition module 81 of the vehicle controller are similar to Fig.10 The function of the communication interface 53 in the vehicle controller is the same as that of the processing module 82 in the vehicle controller. Fig.10 The functions of the processor 51 in the vehicle controller are the same as those of the storage module 83 in the vehicle controller. Fig.10 The function of the memory 52 in is the same.

[0129] An embodiment of the present application also provides a vehicle, which may include the vehicle controller in any embodiment.

[0130] An embodiment of the present application further provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the method in any embodiment.

[0131] The above description is only a specific embodiment of the present disclosure, so that those skilled in the art can understand or implement the present disclosure. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present disclosure. Therefore, the present disclosure will not be limited to the embodiments described herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A sitting posture recognition method, characterized in that: include: Acquiring an actual pressure value collected by a preset pressure collection device for sitting posture recognition disposed on a seat of the vehicle; Analyze the actual pressure values ​​to determine coordinate information of each actual pressure value and key points of the human body; Based on the coordinate information, determine the bone vector corresponding to each target bone; wherein the target bones include at least one of the left thigh bone, the right thigh bone, the shoulder bone and the spine bone; Perform data classification based on the coordinate information, the bone vector and the actual pressure value to determine the actual probability of each sitting posture classification; wherein the sitting posture classification includes a normal sitting posture and an abnormal sitting posture; Based on the actual probability, a sitting posture recognition result is determined.

2. The sitting posture recognition method according to claim 1, characterized in that: Before obtaining the actual pressure value collected by a preset pressure collection device for sitting posture recognition disposed on the seat of the vehicle, the method further includes: Acquire first training sample data and a first labeling result of the first training sample data; wherein the first training sample data includes at least one set of historical pressure values ​​collected by a preset pressure collection device for sitting posture recognition set on a vehicle seat, and the first labeling result includes a human body key point corresponding to each historical pressure value in each set of historical pressure values, and coordinate information of the human body key point; Inputting the first training sample data into a detection model based on a visual algorithm for learning, and obtaining a first prediction result of the detection model based on the visual algorithm for the first training sample data; Adjusting the network parameters of the detection model based on the visual algorithm based on the first prediction result and the first marking result until the detection model based on the visual algorithm converges to obtain a key point detection model; The analyzing the actual pressure values ​​to determine the coordinate information of each actual pressure value and a key point of the human body includes: The actual pressure value is input into the key point detection model for analysis to determine the coordinate information of each actual pressure value and the key point of the human body.

3. The sitting posture recognition method according to claim 1, characterized in that: The human body key points include at least a left shoulder key point, a right shoulder key point, a left hip key point, a right hip key point, a left thigh key point and a right thigh key point; The step of determining a bone vector corresponding to each target bone based on the coordinate information includes: Determine the bone vector corresponding to the left thigh bone based on the line connecting the coordinate point corresponding to the coordinate information of the left span key point and the coordinate point corresponding to the coordinate information of the left thigh key point; Determine the bone vector corresponding to the right thigh bone based on the line connecting the coordinate point corresponding to the coordinate information of the right span key point and the coordinate point corresponding to the coordinate information of the right thigh key point; Determine a bone vector corresponding to the shoulder bone based on a line connecting a coordinate point corresponding to the coordinate information of the left shoulder key point and a coordinate point corresponding to the coordinate information of the right shoulder key point; Based on the coordinate point corresponding to the average value of the coordinate information of the left shoulder key point and the coordinate information of the right shoulder key point, and the line connecting the coordinate point corresponding to the average value of the coordinate information of the left span key point and the coordinate information of the right span key point, the bone vector corresponding to the spinal bone is determined.

4. The sitting posture recognition method according to claim 1, characterized in that: Before obtaining the actual pressure value collected by a preset pressure collection device for sitting posture recognition disposed on the seat of the vehicle, the method further includes: Acquire second training sample data and a second labeling result of the second training sample data; wherein the second training sample data includes historical bone vectors, historical pressure values, and coordinate information of each key point of the human body corresponding to different sitting postures when the driver sits on the seat of the vehicle, and the second labeling result includes any one of a normal sitting posture and an abnormal sitting posture; Inputting the second training sample data into a fully connected neural network for learning, and obtaining a learning result of the fully connected neural network on the second training sample data; Using a normalized indicator function to classify the learning result, and obtaining a predicted probability of the learning result in each sitting posture classification; Based on the cross entropy loss function, calculating the loss value of the sitting posture classification corresponding to the maximum predicted probability and the second labeling result; When the loss value is less than the target value, determining that the fully connected neural network and the normalized indicator function converge, and obtaining a sitting posture analysis model; The data classification based on the coordinate information, the bone vector and the actual pressure value to determine the actual probability of each sitting posture classification includes: Determine sitting posture analysis data based on the coordinate information, the bone vector and the actual pressure value; Normalizing the sitting posture analysis data to obtain normalized data; The normalized data is input into the sitting posture analysis model for data classification, and the actual probability of each sitting posture classification in the normalized data is determined.

5. The sitting posture recognition method according to claim 1, characterized in that: The determining of the sitting posture recognition result based on the actual probability includes: When the sitting posture corresponding to the maximum actual probability is classified as an abnormal sitting posture, the sitting posture recognition result is determined to be an abnormal sitting posture.

6. The sitting posture recognition method according to claim 5, characterized in that: The method further comprises: When the sitting posture recognition result is an abnormal sitting posture, a prompt message is generated; wherein, the prompt message is used for at least the user to adjust the current sitting posture.

7. A vehicle controller, characterized in that: include: An acquisition module, used to acquire an actual pressure value acquired by a preset pressure acquisition device for sitting posture recognition arranged on a seat of the vehicle; A processing module, used for analyzing the actual pressure values ​​obtained by the obtaining module, and determining the coordinate information of each actual pressure value and a key point of the human body; The processing module is further used to determine the bone vector corresponding to each target bone based on the coordinate information; wherein the target bone includes at least one of a left thigh bone, a right thigh bone, a shoulder bone and a spinal bone; The processing module is further used to perform data classification based on the coordinate information, the bone vector and the actual pressure value obtained by the acquisition module to determine the actual probability of each sitting posture classification; wherein the sitting posture classification includes normal sitting posture and abnormal sitting posture; The processing module is further used to determine a sitting posture recognition result based on the actual probability.

8. The vehicle controller according to claim 7, characterized in that: The acquisition module is further used to acquire first training sample data and a first labeling result of the first training sample data; wherein the first training sample data includes at least one group of historical pressure values ​​collected by a preset pressure collection device for sitting posture recognition set on a vehicle seat, and the first labeling result includes a human body key point corresponding to each historical pressure value in each group of historical pressure values, and coordinate information of the human body key point; The processing module is further used to input the first training sample data acquired by the acquisition module into the detection model based on the visual algorithm for learning, so as to obtain a first prediction result of the detection model based on the visual algorithm for the first training sample data; The processing module is further used to adjust the network parameters of the detection model based on the visual algorithm based on the first prediction result and the first marking result obtained by the acquisition module until the detection model based on the visual algorithm converges to obtain a key point detection model; The processing module is specifically used to input the actual pressure value obtained by the obtaining module into the key point detection model for analysis, and determine the coordinate information of each actual pressure value and the key point of the human body.

9. A vehicle, characterized in that: The vehicle comprises a vehicle controller as claimed in any one of claims 7 or 8.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 6 when executed by a processor.

Images