Behavior and posture based data processing method and apparatus, device, and storage medium

Abstract

This application relates to the field of artificial intelligence technology, and discloses a data processing method, apparatus, device, and storage medium based on behavior and posture. The method includes: acquiring a stick figure image sequence of a target object; inputting the stick figure image sequence into a trained behavior and posture recognition model and a classification model to obtain a target label; obtaining report data based on the target label; obtaining physical condition and premium adjustment strategy based on a preset evaluation database; acquiring premium data, and adjusting the premium data using the premium adjustment strategy. This method can obtain the target label by removing individual information from the stick figure image sequence, behavior and posture recognition model, and classification model, and then obtain report data based on the target label. This avoids manual observation and evaluation, thereby improving the accuracy of the report data obtained from data analysis. It enables accurate adjustment of the target object's premium data based on the report data, improving the accuracy and efficiency of insurance premium assessment.

Description

Technical Field

[0001] This invention relates to the field of behavior and posture-based data processing technology, and in particular to a behavior and posture-based data processing method, apparatus, device, and storage medium. Background Technology

[0002] With the increasing aging of the population, home-based elder care has become an important mode of elder care. In the home-based elder care environment, the behavior and posture of the elderly are of great significance for health status and risk assessment.

[0003] Traditional methods for analyzing the behavior and posture of elderly people living at home mainly rely on manual observation and assessment, which often results in inaccurate data analysis. In addition, data analysis of the behavior and posture of elderly people living at home usually involves direct analysis of real images of the elderly, leading to insufficient protection of the elderly's privacy. Summary of the Invention

[0004] Based on this, it is necessary to address the technical problems of insufficient privacy protection for the elderly in existing technologies and inaccurate data analysis of the behavior and posture of the elderly in home-based care. Therefore, a data processing method, device, computer equipment and storage medium based on behavior and posture are proposed.

[0005] Firstly, a data processing method based on behavior and posture is provided, the method comprising:

[0006] A sequence of stick figure images of a target object is obtained, wherein each stick figure image in the sequence is an image obtained by processing the image region corresponding to the human body in the image to be analyzed, and the image to be analyzed is an image obtained by a camera device capturing the target region and matching the target object;

[0007] The stick figure image sequence is input into a trained behavior and posture recognition model to extract behavior and posture features, thereby obtaining various feature data.

[0008] Each of the aforementioned feature data is input into a classification model to classify the behavior and posture labels, thereby obtaining each target label corresponding to each of the aforementioned feature data.

[0009] Data analysis is performed based on the target tags to obtain the report data corresponding to the target object;

[0010] Based on a preset assessment database, the report data is matched and queried to obtain the physical condition and premium adjustment strategy associated with the report data. The assessment database includes an association table, which records the mapping relationship between the report data, physical condition, and premium adjustment strategy.

[0011] Obtain the current premium data of the target object, and adjust the premium data according to the physical condition using the premium adjustment strategy.

[0012] Secondly, a behavior- and posture-based data processing apparatus is provided, the apparatus comprising:

[0013] The acquisition module is used to acquire a sequence of stick figure images of the target object, wherein each stick figure image in the sequence is an image obtained by processing the image to be analyzed with stick figures corresponding to the human body image region, and the image to be analyzed is an image obtained by a camera device capturing the target area and matching the target object;

[0014] The feature extraction module is used to input the stick figure image sequence into the trained behavior and posture recognition model to extract the features of behavior and posture, and obtain various feature data.

[0015] The classification module is used to input each of the feature data into the classification model to classify the behavior and posture labels, and obtain each target label corresponding to each of the feature data;

[0016] The analysis module is used to perform data analysis based on the target label to obtain report data corresponding to the target object;

[0017] The query module is used to match and query the report data according to a preset assessment database to obtain the physical condition and premium adjustment strategy associated with the report data. The assessment database includes an association table, which records the mapping relationship between the report data, physical condition and premium adjustment strategy.

[0018] The premium adjustment module is used to obtain the current premium data of the target object and adjust the premium data according to the physical condition using the premium adjustment strategy.

[0019] Thirdly, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the above-described behavior and posture-based data processing method.

[0020] Fourthly, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described behavior and attitude-based data processing method.

[0021] The proposed data processing method based on behavior and posture involves acquiring a sequence of stick figure images of a target object. Each stick figure image in the sequence is obtained by processing the image to be analyzed into a stick figure representing a region corresponding to the human body. The image to be analyzed is an image captured by a camera that matches the target object. The sequence of stick figure images is then input into a trained behavior and posture recognition model for feature extraction, yielding various feature data. These feature data are then input into a classification model for behavior and posture label classification, resulting in target labels corresponding to each feature data. Data analysis is performed based on these target labels to obtain report data corresponding to the target object. Finally, a matching query is performed on the report data according to a preset evaluation database to obtain the physical condition and premium adjustment strategy associated with the report data. The assessment database includes a correlation table that records the mapping relationship between report data, physical condition, and premium adjustment strategy. Finally, the current premium data of the target individual is obtained, and the premium adjustment strategy is applied to adjust the premium data based on the physical condition. This allows for the acquisition of a stick figure image sequence with individual information removed from the target individual, thus protecting the privacy of elderly people living at home. Through this stick figure image sequence, a behavior and posture recognition model, and a classification model, various target labels corresponding to the stick figure image sequence are obtained to determine the behavior and posture of the elderly people living at home. Further data analysis of the target labels yields report data, thus analyzing the behavior and posture of elderly people living at home. This avoids manual observation and assessment, improving the accuracy of the report data obtained from the data analysis. This allows for accurate adjustment of the target individual's premium data based on the report data, improving the precision and efficiency of insurance premium assessment. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] in:

[0024] Figure 1 This is an application environment diagram of a behavior and pose-based data processing method in one embodiment.

[0025] Figure 2 This is a flowchart of a behavior- and pose-based data processing method in one embodiment;

[0026] Figure 3This is a structural block diagram of a behavior and pose-based data processing device in one embodiment;

[0027] Figure 4 This is a structural block diagram of a computer device in one embodiment;

[0028] Figure 5 This is a structural block diagram of a computer device in another embodiment. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] The behavior- and pose-based data processing method provided in this invention can be applied to, for example... Figure 1In the application environment, client 110 communicates with server 120 via a network. Server 120 can obtain a sequence of stick figure images of the target object through client 110. Each stick figure image in the sequence is obtained by processing the image to be analyzed with stick figures representing the corresponding human body image region. The image to be analyzed is an image captured by a camera device in the target area that matches the target object. Server 120 inputs the stick figure image sequence into a trained behavior and posture recognition model to extract behavior and posture features, obtaining various feature data. Then, server 120 inputs each feature data into a classification model to classify behavior and posture labels, obtaining various target labels corresponding to each feature data. Based on the target labels, data analysis is performed to obtain report data corresponding to the target object. Then, server 120 performs matching queries on the report data according to a preset evaluation database to obtain the physical condition and premium adjustment strategy associated with the report data. The assessment database includes a correlation table that records the mapping relationship between report data, physical condition, and premium adjustment strategy. Finally, the current premium data of the target object is obtained, and the premium adjustment strategy is applied. The server 120 then adjusts the premium data based on the physical condition. It can acquire a stick figure image sequence with individual information of the target object removed to protect the privacy of elderly people living at home. Through this stick figure image sequence, a behavior and posture recognition model, and a classification model, various target labels corresponding to the stick figure image sequence are obtained to determine the behavior and posture of the elderly people living at home. Data analysis of the target labels is then performed to obtain report data, thus analyzing the behavior and posture of elderly people living at home. This avoids manual observation and assessment, improving the accuracy of the report data obtained from the data analysis. This allows for accurate adjustment of the target object's premium data based on the report data, improving the accuracy and efficiency of insurance premium assessment. The client 110 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server 120 can be implemented using a standalone server or a server cluster consisting of multiple servers. The present invention will now be described in detail through specific embodiments.

[0031] Please see Figure 2 As shown, Figure 2 A flowchart illustrating a behavior- and pose-based data processing method according to an embodiment of the present invention includes the following steps:

[0032] S101: Obtain a stick figure image sequence of the target object, wherein each stick figure image in the stick figure image sequence is an image obtained by processing the image region corresponding to the human body in the image to be analyzed, and the image to be analyzed is an image obtained by the camera device capturing the target region and matching the target object;

[0033] The target audience can be elderly people living at home. Each stick figure image in the stick figure image sequence is obtained by processing the image region corresponding to the human body in the image to be analyzed. As an example, stick figure processing can use computer vision technology to detect 18 human body key points and skeleton maps in the image to be analyzed, and connect the 18 detected key points with straight lines to obtain the stick figure image. As an example, an edge computing embedded device can be used to perform local preprocessing on the image to be analyzed, thereby extracting the stick figure image containing the stick figure skeleton data of the target object, and permanently deleting the image to be analyzed. It should be noted that the individual information of the target object is removed from the stick figure image, thereby protecting the individual privacy of the target object. The image to be analyzed is an image captured by the target camera device that matches the target object. As an example, the face in the image to be analyzed is compared with the face of the target object, and a match is determined based on the comparison result. As another example, the comparison result can be the similarity between the face in the image to be analyzed and the face of the target object. If the similarity is greater than a threshold, a match is determined. As another example, the faces in the image to be analyzed and the faces of the target object are input into a trained face classification model for face feature comparison, and the feature comparison results output by the trained face classification model are obtained. The face classification model can be trained based on the Transformer model.

[0034] S102: Input the stick figure image sequence into the trained behavior and posture recognition model to extract behavior and posture features and obtain various feature data;

[0035] Among them, the behavior and pose recognition model is a model trained based on a convolutional neural network. Specifically, the behavior and pose recognition model is used to identify the behavior and pose information of the target object in the stick figure image sequence and extract higher-level feature data, which refers to the feature data of behavior and pose.

[0036] Optionally, after obtaining each feature data, each feature data is encrypted using a preset encryption algorithm, for example, by randomly generating a key using a preset mathematical model.

[0037] S103: Input each of the aforementioned feature data into a classification model to classify the behavior and posture labels, and obtain each target label corresponding to each of the aforementioned feature data;

[0038] The classification model is a model trained based on the Transformer model.

[0039] In this embodiment, each feature data is input into a classification model for behavior and pose label classification, and the target label corresponding to each feature data is obtained. One feature data corresponds to one target label, thereby realizing the recognition of feature data.

[0040] The range of values ​​for the target label includes, but is not limited to, standing, walking, sitting, and lying down.

[0041] S104: Perform data analysis based on the target label to obtain the report data corresponding to the target object;

[0042] In this embodiment, after accurately identifying the behavior and posture of the target object and obtaining the target labels corresponding to the behavior and posture of the target object, data analysis is performed on the target labels to obtain the report data corresponding to the target object. The report data may include the user's daily behavior and abnormal behavior.

[0043] As one example, target tags are used to determine if a target object exhibits abnormal behavior, and each abnormal behavior is reported as data. For instance, in a restroom area, if a target object is lying down, this is considered abnormal behavior. As another example, the daily behavior of the target object is summarized using the timestamps of the behavior and posture corresponding to each target tag.

[0044] S105: Based on a preset assessment database, the report data is matched and queried to obtain the physical condition and premium adjustment strategy associated with the report data. The assessment database includes an association table, which records the mapping relationship between the report data, physical condition and premium adjustment strategy.

[0045] In this embodiment, the assessment database is a database summarized from human experience. The report data is matched and queried using the association tables of the assessment database to obtain the corresponding physical condition and premium adjustment strategy. As an example, the physical condition can include poor physical condition, good physical condition, etc. As another example, the premium adjustment strategy can be adjusting the premium to decrease or increase, etc.

[0046] S106: Obtain the current premium data of the target object, and adjust the premium data according to the physical condition using the premium adjustment strategy.

[0047] In this embodiment, the current premium data of the target object is obtained, wherein the premium data includes at least the target object's current insurance premium and compensation amount. Then, a premium adjustment strategy is adopted to make more reasonable adjustments to the premium data based on the target object's physical condition, thereby rationalizing the insurance premium and compensation amount.

[0048] In one implementation, the premium adjustment strategy includes reducing the premium, increasing the premium, and keeping the premium unchanged. The physical condition includes poor physical condition and better physical condition. The poor physical condition includes multiple poor physical condition levels, and the higher the poor physical condition level, the worse the physical condition of the target individual. The better physical condition includes multiple better physical condition levels, and the higher the better physical condition level, the better the physical condition. When the target individual's poor physical condition level is higher, the increase in premium adjustment is greater. When the target individual's better physical condition level is higher, the decrease in premium adjustment is greater.

[0049] In one implementation, reducing premiums includes multiple reduction levels; the higher the reduction level, the greater the premium reduction. Each reduction level corresponds one-to-one with a level of poor physical condition. Increasing premiums includes multiple increase levels; the higher the increase level, the greater the premium increase. Each increase level corresponds one-to-one with a level of good physical condition.

[0050] This embodiment proposes a data processing method based on behavior and posture. It acquires a sequence of stick figure images of the target object. Each stick figure image in the sequence is obtained by processing the image to be analyzed into a stick figure representing the corresponding human body region. The image to be analyzed is an image captured by a camera device that matches the target object within the target region. The sequence of stick figure images is then input into a trained behavior and posture recognition model for feature extraction, yielding various feature data. These feature data are then input into a classification model for behavior and posture label classification, resulting in target labels corresponding to each feature data. Data analysis is performed based on these target labels to obtain report data corresponding to the target object. Finally, a matching query is performed on the report data according to a preset evaluation database to obtain the associated physical condition and premium adjustment strategy. The assessment database includes a correlation table that records the mapping relationship between report data, physical condition, and premium adjustment strategy. Finally, the current premium data of the target individual is obtained, and the premium adjustment strategy is applied to adjust the premium data based on the physical condition. This allows for the acquisition of a stick figure image sequence with individual information removed from the target individual, thus protecting the privacy of elderly people living at home. Through this stick figure image sequence, a behavior and posture recognition model, and a classification model, various target labels corresponding to the stick figure image sequence are obtained to determine the behavior and posture of the elderly people living at home. Further data analysis of the target labels yields report data, thus analyzing the behavior and posture of elderly people living at home. This avoids manual observation and assessment, improving the accuracy of the report data obtained from the data analysis. This allows for accurate adjustment of the target individual's premium data based on the report data, improving the accuracy and efficiency of insurance premium assessment.

[0051] In one embodiment, the classification model is a model trained based on a first coding unit, a second coding unit, a third coding unit, and a classification unit. The step of inputting each feature data into the classification model for behavior and pose label classification to obtain each target label corresponding to each feature data includes:

[0052] S201: Obtain a reference image set, wherein the reference image set includes a first image of each normal behavior and posture and a second image of each abnormal behavior and posture;

[0053] S202: Input each of the feature data into the first encoding unit to obtain a first feature vector; input the first image into the second encoding unit to obtain a second feature vector; input the second image into the third encoding unit to obtain a third feature vector;

[0054] S203: Input the first feature vector, the second feature vector, and the third feature vector into the classification unit, calculate the similarity between the first feature vector and each of the second feature vectors respectively, and use them as a first similarity set; and calculate the similarity between the first feature vector and each of the third feature vectors respectively, and use them as a second similarity set.

[0055] S204: Determine the highest similarity in the first similarity set and the second similarity set, and take the label of the behavior and posture corresponding to the highest similarity as the target label corresponding to the feature data.

[0056] It should be noted that the first, second, and third coding units can all be trained based on Transformer, while the classification unit can be trained based on convolutional neural network.

[0057] In this embodiment, firstly, a reference image set is acquired, which includes first images of various normal behaviors and postures, and second images of various abnormal behaviors and postures. As an example, the first image may be an image of normal behaviors such as standing, walking, sitting, or lying down, and the second image may be an image of abnormal behaviors such as falling, calling for help, or disease characteristics.

[0058] Next, each feature data is input into the first coding unit to obtain the first feature vector output by the first coding unit. The first image is input into the second coding unit to obtain the second feature vector output by the second coding unit, and the second image is input into the third coding unit to obtain the third feature vector output by the third coding unit.

[0059] Next, each first feature vector, each second feature vector, and each third feature vector are input into the classification unit so that the classification unit calculates the similarity between the first feature vector and each second feature vector as the first similarity set, and calculates the similarity between the first feature vector and each third feature vector as the second similarity set.

[0060] Finally, the highest similarity is determined in the first and second similarity sets, and the label of the behavior and posture corresponding to the highest similarity is used as the target label for the feature data. As an example, for each feature data, if there are at least two similarities greater than a preset value in the first and second similarity sets, the feature data is labeled as suspicious.

[0061] As an example, similarity can be calculated by using cosine similarity.

[0062] The behavior and pose-based data processing method proposed in this embodiment acquires a reference image set, which includes first images of various normal behaviors and poses and second images of various abnormal behaviors and poses. Then, each feature data is input into a first encoding unit to obtain a first feature vector. The first image is input into a second encoding unit to obtain a second feature vector, and the second image is input into a third encoding unit to obtain a third feature vector. Subsequently, the first, second, and third feature vectors are input into a classification unit, and the similarity between the first feature vector and each of the second feature vectors is calculated, forming a first similarity set. The similarity between the first feature vector and each of the third feature vectors is calculated to form a second similarity set. The highest similarity is then determined within both the first and second similarity sets. The label corresponding to the behavior and posture with the highest similarity is taken as the target label corresponding to the feature data. This allows for accurate classification of the feature data of the target object's behavior and posture using a classification model trained on the first, second, third, and classification units, thereby obtaining the target label for each feature data. This improves the accuracy of identifying the target object's behavior and posture, making subsequent data analysis and premium adjustments based on the target labels more reasonable, and enhancing the accuracy and efficiency of insurance premium assessment.

[0063] In one embodiment, the report data includes anomaly reports, and the step of performing data analysis based on the target tag to obtain the report data corresponding to the target object includes:

[0064] S301: Determine whether there are any abnormal target labels among the target labels;

[0065] S302: If it exists, take the abnormal target label as the abnormal label, and obtain the abnormal frequency based on the number of abnormal labels and the number of target labels;

[0066] S303: Perform data analysis based on each target label and the anomaly frequency to obtain the anomaly report corresponding to the target object.

[0067] In this embodiment, it is determined whether there are abnormal target labels among the target labels. Abnormal target labels are the labels corresponding to images of abnormal behaviors such as falling, calling for help, and disease characteristics.

[0068] If there are abnormal target tags among the target tags, the abnormal target tags are regarded as abnormal tags. The abnormal frequency is obtained based on the number of abnormal tags and the number of target tags. As an example, the abnormal frequency is obtained by dividing the number of abnormal tags by the number of target tags.

[0069] Finally, data analysis is performed based on each target label and the frequency of anomalies to obtain an anomaly report corresponding to the target object. As an example, the degree of anomaly for each anomaly label is rated based on its frequency, and the rating result for each anomaly label is included as part of the anomaly report. As another example, the duration of the behavior and posture corresponding to each target label is obtained, and it is determined whether the duration corresponding to the target label is normal. For example, if the target label is sleep, then the behavior and posture are sleep; if the sleep duration exceeds 8 hours, it is considered a sleep abnormality, and the sleep abnormality is included as part of the anomaly report.

[0070] The behavior and posture-based data processing method proposed in this embodiment determines whether there are abnormal target tags among the target tags. If so, the abnormal target tags are used as abnormal tags. The abnormal frequency is obtained based on the number of abnormal tags and the number of target tags. Finally, data analysis is performed based on the target tags and the abnormal frequency to obtain the abnormal report corresponding to the target object. This method can calculate the abnormal frequency when abnormal tags exist and perform data analysis based on the target tags and the abnormal frequency, making the data content of the obtained abnormal report richer and improving the accuracy of the abnormal report.

[0071] In one embodiment, the step of performing data analysis based on each of the target tags and the anomaly frequency to obtain the anomaly report corresponding to the target object includes:

[0072] S401: Obtain the health data of the target object, wherein the health data of the target object includes medical records, physiological indicators, and medical history;

[0073] S402: Input the health data, each of the target labels, and the abnormal frequency into the trained fusion model to perform data fusion to obtain target fusion data;

[0074] S403: Input the target fusion data into the trained data analysis model for data analysis to obtain the anomaly report corresponding to the target object.

[0075] In this embodiment, the health data of the target object is acquired. This health data includes medical records, physiological indicators, and medical history. Medical records may be historical medical records from hospital visits. Physiological indicators may include heart rate, blood pressure, blood sugar, etc., and medical history may include family medical history and the target object's own medical history. Optionally, the health data may also include lifestyle habits, such as whether the target object smokes or drinks alcohol.

[0076] Next, the trained fusion model fuses health data, target labels, and anomaly frequencies to obtain target fused data. As an example, the fusion model includes a feature extraction unit and a fusion unit. Health data, target labels, and anomaly frequencies are input into the feature extraction unit for feature extraction, resulting in feature vector representations. These feature vector representations are then input into the fusion model for data fusion. As another example, the fusion model performs a weighted matrix summation of the feature vector representations. As yet another example, the fusion model concatenates the feature vector representations.

[0077] Finally, the target fused data is input into the trained data analysis model for data analysis, thereby generating an anomaly report and obtaining the anomaly report corresponding to the target object. The data analysis model is a language model, which can be trained based on the BERT (Bidirectional Encoder Representation from Transformers) model.

[0078] The behavior and posture-based data processing method proposed in this embodiment acquires the health data of the target object, including medical records, physiological indicators, and medical history. Then, the health data, various target labels, and the abnormal frequency are input into a trained fusion model for data fusion to obtain target fused data. Subsequently, the target fused data is input into a trained data analysis model for data analysis to obtain the abnormal report corresponding to the target object. This method can fuse health data, various target labels, and the abnormal frequency for data analysis, making the data content of the obtained abnormal report richer and improving its accuracy.

[0079] In one embodiment, the step of performing data analysis based on the target label to obtain report data corresponding to the target object further includes:

[0080] S501: Obtain the historical tag set of the target object;

[0081] S502: Input the historical label set and each of the target labels into the trained trend model to predict the trend and obtain the prediction result, wherein the trend model is a model trained based on the long short-term memory model;

[0082] S503: Based on the prediction results, determine the report data corresponding to the target object.

[0083] The historical tag set can include the tag set of the target object over the past week as the first tag set, the tag set of the target object over the past month as the second tag set, and the tag set of the target object over the past year as the third tag set, which is the Long Short-Term Memory model.

[0084] By using a trained trend model, trend prediction is performed based on the historical label set and each target label to obtain a prediction result. The trend model is a model trained using a long short-term memory (LSM) model. In one implementation, the trend model includes a first LSM model, a second LSM model, a third LSM model, an attention mechanism model, and a fully connected layer model. A first label set is input into the first LSM model to obtain a first output result; a second label set is input into the second LSM model to obtain a second output result; a third label set is input into the third LSM model to obtain a third output result; the first, second, and third output results are input into the attention mechanism model to obtain a fourth output result; and the fourth output result is input into the fully connected layer model to obtain the prediction result.

[0085] Finally, based on the prediction results, the corresponding report data for the target object is determined. The prediction results may include the probability of the target object exhibiting various abnormal behaviors in the future. As an example, the probability of the target object exhibiting various abnormal behaviors in the future is included as part of the report data.

[0086] The behavior and posture-based data processing method proposed in this embodiment obtains the historical label set of the target object, and then inputs the historical label set and each target label into a trained trend model for trend prediction to obtain the prediction result. The trend model is a model trained based on a long short-term memory model. Finally, based on the prediction result, the report data corresponding to the target object is determined. This method can use the historical label set and each target label to predict trends, thereby enriching the data content of the anomaly report and improving the accuracy of the anomaly report.

[0087] In one embodiment, the step of generating the stick figure image includes:

[0088] S601: Based on the preset human key point recognition model, feature extraction is performed on the image to be analyzed to obtain information on each key point and a timestamp, and stick figure processing is performed based on the information on each human key point and the timestamp to generate the stick figure image.

[0089] In this implementation, the timestamp is the acquisition time of the image containing human key point information. The human key point recognition model is trained based on a convolutional neural network. The human key point recognition model can be used to extract the human key point information corresponding to the target object from the image to be analyzed, so as to determine the pose of the target object, and to extract the timestamp in the image to determine the acquisition time corresponding to the pose of the target object. The key point locations in the key point information may include, but are not limited to: the positions of the monitored individual's eyes, nose, ears, neck, shoulders, elbows, wrists, knees, hip joints, ankle joints, hips, thumbs, little fingers, and heels.

[0090] Finally, stick figure processing is performed based on the key point information of each human body and the timestamp to generate the stick figure image corresponding to the target camera device. As an example, the key point information of the human body is concatenated in chronological order from morning to night according to the timestamps to obtain the stick figure image.

[0091] The behavior and posture-based data processing method proposed in this embodiment extracts features from the image to be analyzed according to a preset human key point recognition model, obtaining key point information and timestamps. Then, stick figure processing is performed based on the key point information and timestamps to generate the stick figure image. This method can protect the privacy of the target object in the image to be analyzed by privacy processing, and removes interference information (such as audio information, facial feature information of the first user, etc.) from the image to be analyzed. This makes the stick figure image fast, efficient, and requires less bandwidth and storage space for transmission over the network.

[0092] In one embodiment, the step of obtaining the current premium data of the target object and adjusting the premium data according to the physical condition using the premium adjustment strategy includes:

[0093] S701: Generate a user adjustment strategy based on the adjusted premium data and physical condition;

[0094] S702: The premium data, the physical condition status, and the user adjustment strategy are sent to the user terminal, wherein the user terminal is used to display the premium data, the physical condition status, and the user adjustment strategy on the display screen.

[0095] In this embodiment, a user adjustment strategy is generated based on the adjusted premium data and the individual's physical condition. This strategy includes personalized risk assessments and recommendations. For example, it assesses the risk level of a certain disease or accident and provides corresponding preventative measures and health management suggestions. These suggestions may include regular physical examinations, dietary improvements, increased exercise, and adherence to prescribed medications to help the target individual reduce risk and improve their health.

[0096] Finally, the premium data, physical condition, and user adjustment strategy are sent to the user terminal, which displays the data on the screen.

[0097] The behavior and posture-based data processing method proposed in this embodiment generates a user adjustment strategy based on the premium data and physical condition. Finally, the premium data, physical condition, and user adjustment strategy are sent to the user terminal. The user terminal displays the premium data, physical condition, and user adjustment strategy on a screen. This method can generate user adjustment strategies for target objects and display them on the user terminal, thereby improving the user experience.

[0098] Please see Figure 3 As shown, in one embodiment, a data processing apparatus based on behavior and posture is provided, the apparatus comprising:

[0099] The acquisition module 10 is used to acquire a stick figure image sequence of the target object, wherein each stick figure image in the stick figure image sequence is an image obtained by processing the image region corresponding to the human body in the image to be analyzed, and the image to be analyzed is an image obtained by the camera device capturing the target area and matching the target object;

[0100] Feature extraction module 20 is used to input the stick figure image sequence into a trained behavior and posture recognition model to extract behavior and posture features and obtain various feature data;

[0101] The classification module 30 is used to input each of the feature data into the classification model to classify the behavior and posture labels, and obtain each target label corresponding to each of the feature data;

[0102] Analysis module 40 is used to perform data analysis based on the target label to obtain report data corresponding to the target object;

[0103] The query module 50 is used to perform a matching query on the report data according to a preset assessment database to obtain the physical condition and premium adjustment strategy associated with the report data. The assessment database includes an association table, which records the mapping relationship between the report data, physical condition and premium adjustment strategy.

[0104] The premium adjustment module 60 is used to obtain the current premium data of the target object and adjust the premium data according to the physical condition using the premium adjustment strategy.

[0105] This embodiment proposes a data processing method based on behavior and posture. It acquires a sequence of stick figure images of the target object. Each stick figure image in the sequence is obtained by processing the image to be analyzed into a stick figure representing the corresponding human body region. The image to be analyzed is an image captured by a camera device that matches the target object within the target region. The sequence of stick figure images is then input into a trained behavior and posture recognition model for feature extraction, yielding various feature data. These feature data are then input into a classification model for behavior and posture label classification, resulting in target labels corresponding to each feature data. Data analysis is performed based on these target labels to obtain report data corresponding to the target object. Finally, a matching query is performed on the report data according to a preset evaluation database to obtain the associated physical condition and premium adjustment strategy. The assessment database includes a correlation table that records the mapping relationship between report data, physical condition, and premium adjustment strategy. Finally, the current premium data of the target individual is obtained, and the premium adjustment strategy is applied to adjust the premium data based on the physical condition. This allows for the acquisition of a stick figure image sequence with individual information removed from the target individual, thus protecting the privacy of elderly people living at home. Through this stick figure image sequence, a behavior and posture recognition model, and a classification model, various target labels corresponding to the stick figure image sequence are obtained to determine the behavior and posture of the elderly people living at home. Further data analysis of the target labels yields report data, thus analyzing the behavior and posture of elderly people living at home. This avoids manual observation and assessment, improving the accuracy of the report data obtained from the data analysis. This allows for accurate adjustment of the target individual's premium data based on the report data, improving the accuracy and efficiency of insurance premium assessment.

[0106] In one embodiment, the classification module 30 is further configured to acquire a reference image set, wherein the reference image set includes first images of various normal behaviors and postures and second images of various abnormal behaviors and postures; input each feature data into the first encoding unit to obtain a first feature vector, input the first image into the second encoding unit to obtain a second feature vector, input the second image into the third encoding unit to obtain a third feature vector; input the first feature vector, the second feature vector, and the third feature vector into the classification unit, calculate the similarity between the first feature vector and each of the second feature vectors respectively, as a first similarity set, and calculate the similarity between the first feature vector and each of the third feature vectors respectively, as a second similarity set; determine the highest similarity in the first similarity set and the second similarity set, and use the label of the behavior and posture corresponding to the highest similarity as the target label corresponding to the feature data.

[0107] In one embodiment, the analysis module 40 is further configured to determine whether there are any abnormal target tags among the target tags; if so, the abnormal target tags are used as abnormal tags, and the abnormal frequency is obtained based on the number of abnormal tags and the number of target tags; data analysis is performed based on each target tag and the abnormal frequency to obtain the abnormal report corresponding to the target object.

[0108] In one embodiment, the analysis module 40 is further configured to acquire the health data of the target object, wherein the health data of the target object includes medical records, physiological indicators, and medical history; input the health data, each of the target labels, and the abnormal frequency into a trained fusion model for data fusion to obtain target fusion data; input the target fusion data into a trained data analysis model for data analysis to obtain the abnormal report corresponding to the target object.

[0109] In one embodiment, the analysis module 40 is further configured to obtain the historical tag set of the target object; input the historical tag set and each of the target tags into a trained trend model for trend prediction to obtain a prediction result, wherein the trend model is a model trained based on a long short-term memory model; and determine the report data corresponding to the target object based on the prediction result.

[0110] In one embodiment, the behavior and posture-based data processing device is further configured to extract features from the image to be analyzed according to a preset human key point recognition model, obtain information on each key point and a timestamp, and perform stick figure processing based on the information on each human key point and the timestamp to generate the stick figure image.

[0111] In one embodiment, the behavior and posture-based data processing device is further configured to generate a user adjustment strategy based on the adjusted premium data and physical condition; and send the premium data, physical condition, and user adjustment strategy to a user terminal, wherein the user terminal is configured to display the premium data, physical condition, and user adjustment strategy on a display screen.

[0112] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with external clients via a network connection. When executed by the processor, the computer program implements server-side functions or steps of a behavior- and attitude-based data processing method.

[0113] In one embodiment, a computer device is provided, which may be a client, and its internal structure diagram may be as follows: Figure 5 As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with an external server via a network connection. When executed by the processor, the computer program implements client-side functions or steps of a behavior- and attitude-based data processing method.

[0114] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, performs the following steps:

[0115] A sequence of stick figure images of a target object is obtained, wherein each stick figure image in the sequence is an image obtained by processing the image region corresponding to the human body in the image to be analyzed, and the image to be analyzed is an image obtained by a camera device capturing the target region and matching the target object;

[0116] The stick figure image sequence is input into a trained behavior and posture recognition model to extract behavior and posture features, thereby obtaining various feature data.

[0117] Each of the aforementioned feature data is input into a classification model to classify the behavior and posture labels, thereby obtaining each target label corresponding to each of the aforementioned feature data.

[0118] Data analysis is performed based on the target tags to obtain the report data corresponding to the target object;

[0119] Based on a preset assessment database, the report data is matched and queried to obtain the physical condition and premium adjustment strategy associated with the report data. The assessment database includes an association table, which records the mapping relationship between the report data, physical condition, and premium adjustment strategy.

[0120] Obtain the current premium data of the target object, and adjust the premium data according to the physical condition using the premium adjustment strategy.

[0121] This embodiment proposes a data processing method based on behavior and posture. It acquires a sequence of stick figure images of the target object. Each stick figure image in the sequence is obtained by processing the image to be analyzed into a stick figure representing the corresponding human body region. The image to be analyzed is an image captured by a camera device that matches the target object within the target region. The sequence of stick figure images is then input into a trained behavior and posture recognition model for feature extraction, yielding various feature data. These feature data are then input into a classification model for behavior and posture label classification, resulting in target labels corresponding to each feature data. Data analysis is performed based on these target labels to obtain report data corresponding to the target object. Finally, a matching query is performed on the report data according to a preset evaluation database to obtain the associated physical condition and premium adjustment strategy. The assessment database includes a correlation table that records the mapping relationship between report data, physical condition, and premium adjustment strategy. Finally, the current premium data of the target individual is obtained, and the premium adjustment strategy is applied to adjust the premium data based on the physical condition. This allows for the acquisition of a stick figure image sequence with individual information removed from the target individual, thus protecting the privacy of elderly people living at home. Through this stick figure image sequence, a behavior and posture recognition model, and a classification model, various target labels corresponding to the stick figure image sequence are obtained to determine the behavior and posture of the elderly people living at home. Further data analysis of the target labels yields report data, thus analyzing the behavior and posture of elderly people living at home. This avoids manual observation and assessment, improving the accuracy of the report data obtained from the data analysis. This allows for accurate adjustment of the target individual's premium data based on the report data, improving the accuracy and efficiency of insurance premium assessment.

[0122] In one embodiment, a computer-readable storage medium is provided that stores a computer program, which, when executed by a processor, performs the following steps:

[0123] A sequence of stick figure images of a target object is obtained, wherein each stick figure image in the sequence is an image obtained by processing the image region corresponding to the human body in the image to be analyzed, and the image to be analyzed is an image obtained by a camera device capturing the target region and matching the target object;

[0124] The stick figure image sequence is input into a trained behavior and posture recognition model to extract behavior and posture features, thereby obtaining various feature data.

[0125] Each of the aforementioned feature data is input into a classification model to classify the behavior and posture labels, thereby obtaining each target label corresponding to each of the aforementioned feature data.

[0126] Data analysis is performed based on the target tags to obtain the report data corresponding to the target object;

[0127] Based on a preset assessment database, the report data is matched and queried to obtain the physical condition and premium adjustment strategy associated with the report data. The assessment database includes an association table, which records the mapping relationship between the report data, physical condition, and premium adjustment strategy.

[0128] Obtain the current premium data of the target object, and adjust the premium data according to the physical condition using the premium adjustment strategy.

[0129] This embodiment proposes a data processing method based on behavior and posture. It acquires a sequence of stick figure images of the target object. Each stick figure image in the sequence is obtained by processing the image to be analyzed into a stick figure representing the corresponding human body region. The image to be analyzed is an image captured by a camera device that matches the target object within the target region. The sequence of stick figure images is then input into a trained behavior and posture recognition model for feature extraction, yielding various feature data. These feature data are then input into a classification model for behavior and posture label classification, resulting in target labels corresponding to each feature data. Data analysis is performed based on these target labels to obtain report data corresponding to the target object. Finally, a matching query is performed on the report data according to a preset evaluation database to obtain the associated physical condition and premium adjustment strategy. The assessment database includes a correlation table that records the mapping relationship between report data, physical condition, and premium adjustment strategy. Finally, the current premium data of the target individual is obtained, and the premium adjustment strategy is applied to adjust the premium data based on the physical condition. This allows for the acquisition of a stick figure image sequence with individual information removed from the target individual, thus protecting the privacy of elderly people living at home. Through this stick figure image sequence, a behavior and posture recognition model, and a classification model, various target labels corresponding to the stick figure image sequence are obtained to determine the behavior and posture of the elderly people living at home. Further data analysis of the target labels yields report data, thus analyzing the behavior and posture of elderly people living at home. This avoids manual observation and assessment, improving the accuracy of the report data obtained from the data analysis. This allows for accurate adjustment of the target individual's premium data based on the report data, improving the accuracy and efficiency of insurance premium assessment.

[0130] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions on the server side and client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0131] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0132] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0133] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A data processing method based on behavior and posture, characterized in that, The method includes: A sequence of stick figure images of a target object is obtained, wherein each stick figure image in the sequence is an image obtained by processing the image region corresponding to the human body in the image to be analyzed, and the image to be analyzed is an image obtained by a camera device capturing the target region and matching the target object; The stick figure image sequence is input into a trained behavior and posture recognition model to extract behavior and posture features, thereby obtaining various feature data. Each of the aforementioned feature data is input into a classification model to classify the behavior and posture labels, thereby obtaining each target label corresponding to each of the aforementioned feature data. Data analysis is performed based on the target tags to obtain the report data corresponding to the target object; Based on a preset assessment database, the report data is matched and queried to obtain the physical condition and premium adjustment strategy associated with the report data. The assessment database includes an association table, which records the mapping relationship between the report data, physical condition, and premium adjustment strategy. Obtain the current premium data of the target object, and adjust the premium data according to the physical condition using the premium adjustment strategy; The report data includes anomaly reports. The step of performing data analysis based on the target tag to obtain the report data corresponding to the target object includes: Determine whether there are any abnormal target labels among the target labels; If an anomaly exists, the target label is taken as an anomaly label, and the anomaly frequency is obtained based on the number of anomaly labels and the number of target labels; Based on the target labels and the anomaly frequency, data analysis is performed to obtain the anomaly report corresponding to the target object; The step of analyzing data based on each target label and the anomaly frequency to obtain the anomaly report corresponding to the target object includes: Obtain the health data of the target object, wherein the health data of the target object includes medical records, physiological indicators, and medical history. The medical records are historical medical records of visits to the hospital. The physiological indicators are heart rate, blood pressure, and blood sugar. The medical history includes family medical history and the target object's own medical history. The health data, each of the target labels, and the abnormal frequency are input into a trained fusion model for data fusion to obtain target fusion data. The target fused data is input into a trained data analysis model for data analysis to obtain the anomaly report corresponding to the target object.

2. The data processing method based on behavior and posture according to claim 1, characterized in that, The classification model is a model trained based on a first coding unit, a second coding unit, a third coding unit, and a classification unit. The step of inputting each feature data into the classification model for behavior and pose label classification, and obtaining each target label corresponding to each feature data, includes: Obtain a baseline image set, wherein the baseline image set includes a first image of each normal behavior and posture and a second image of each abnormal behavior and posture; Each feature data is input into the first encoding unit to obtain a first feature vector; the first image is input into the second encoding unit to obtain a second feature vector; and the second image is input into the third encoding unit to obtain a third feature vector. The first feature vector, the second feature vector, and the third feature vector are input into the classification unit. The similarity between the first feature vector and each of the second feature vectors is calculated to form a first similarity set. The similarity between the first feature vector and each of the third feature vectors is also calculated to form a second similarity set. The highest similarity is determined in the first similarity set and the second similarity set, and the label of the behavior and posture corresponding to the highest similarity is taken as the target label corresponding to the feature data.

3. The data processing method based on behavior and posture according to claim 1, characterized in that, The step of performing data analysis based on the target label to obtain the report data corresponding to the target object also includes: Obtain the historical tag set of the target object; The historical label set and each of the target labels are input into the trained trend model to predict the trend and obtain the prediction result. The trend model is a model trained based on the long short-term memory model. Based on the prediction results, the report data corresponding to the target object is determined.

4. The data processing method based on behavior and posture according to claim 1, characterized in that, The steps for generating the stick figure image include: Based on a preset human key point recognition model, feature extraction is performed on the image to be analyzed to obtain information on each key point and a timestamp. Then, stick figure processing is performed based on the information on each human key point and the timestamp to generate the stick figure image.

5. The data processing method based on behavior and posture according to claim 1, characterized in that, The step of obtaining the current premium data of the target object, and then adjusting the premium data according to the physical condition using the premium adjustment strategy, includes: Based on the adjusted premium data and the patient's physical condition, a user adjustment strategy is generated; The premium data, the physical condition, and the user adjustment strategy are sent to the user terminal, wherein the user terminal is used to display the premium data, the physical condition, and the user adjustment strategy on a display screen.

6. A data processing device based on behavior and posture, characterized in that, The behavior and posture-based data processing device includes: The acquisition module is used to acquire a sequence of stick figure images of the target object, wherein each stick figure image in the sequence is an image obtained by processing the image to be analyzed with stick figures corresponding to the human body image region, and the image to be analyzed is an image obtained by a camera device capturing the target area and matching the target object; The feature extraction module is used to input the stick figure image sequence into the trained behavior and posture recognition model to extract the features of behavior and posture, and obtain various feature data. The classification module is used to input each of the feature data into the classification model to classify the behavior and posture labels, and obtain each target label corresponding to each of the feature data; The analysis module is used to perform data analysis based on the target label to obtain report data corresponding to the target object; The query module is used to match and query the report data according to a preset assessment database to obtain the physical condition and premium adjustment strategy associated with the report data. The assessment database includes an association table, which records the mapping relationship between the report data, physical condition and premium adjustment strategy. The premium adjustment module is used to obtain the current premium data of the target object, and adjust the premium data according to the physical condition using the premium adjustment strategy. The report data includes anomaly reports. The step of performing data analysis based on the target tag to obtain the report data corresponding to the target object includes: Determine whether there are any abnormal target labels among the target labels; If an anomaly exists, the target label is taken as an anomaly label, and the anomaly frequency is obtained based on the number of anomaly labels and the number of target labels; Based on the target labels and the anomaly frequency, data analysis is performed to obtain the anomaly report corresponding to the target object; The step of analyzing data based on each target label and the anomaly frequency to obtain the anomaly report corresponding to the target object includes: Obtain the health data of the target object, wherein the health data of the target object includes medical records, physiological indicators, and medical history. The medical records are historical medical records of visits to the hospital. The physiological indicators are heart rate, blood pressure, and blood sugar. The medical history includes family medical history and the target object's own medical history. The health data, each of the target labels, and the abnormal frequency are input into a trained fusion model for data fusion to obtain target fusion data. The target fused data is input into a trained data analysis model for data analysis to obtain the anomaly report corresponding to the target object.

7. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the behavior and posture-based data processing method as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the behavior and posture-based data processing method as described in any one of claims 1 to 5.

Images