Ultra-long-range motion monitoring management system based on machine learning

Abstract

The invention discloses a machine learning-based ultra-long-range motion monitoring management system, which is characterized in that the rehabilitation condition of a patient is indicated on an acquisition module by adopting lamps with different colors, and a single relation between data information acquired by the acquisition module and patient identity information is established; face recognition is carried out on a patient wearing the sports bracelet to judge whether the patient is a patient needing to be monitored or not, the same color matched with a corresponding color lamp displayed on the sports bracelet is adopted to select the monitored patient in a frame mode, and the health condition of the patient is monitored, and the positioning condition and the movement condition of the patient are monitored; then, the identity information of the patient singly corresponding to the sports bracelet and the face recognition identity information collected by the camera are compared and confirmed, and the color lamp displayed on the sports bracelet collected by the camera, the color lamp data fed back to the background server by the sports bracelet and the color lamp displayed according to the rehabilitation condition of the patient are compared and judged; accuracy of collected data of the patient is achieved.

Description

Technical Field

[0001] The present invention relates to a motion monitoring and management system, in particular to an ultra-long-range motion monitoring and management system based on machine learning, and belongs to the technical field of motion monitoring and management systems. Background Art

[0002] In the existing technology, the rehabilitation status of patients is monitored by, for example, a sports bracelet and the information is fed back to the system or a mobile phone. This monitoring method is not suitable for comprehensive monitoring of patients in hospitals, and it is also prone to patient information mismatch. The patient loses the sports bracelet, or the patient no longer needs rehabilitation training monitoring and continuous monitoring due to the inability to update data and unified management in a timely manner. Therefore, an ultra-long-range motion monitoring and management system based on machine learning is designed to solve the above problems. Summary of the invention

[0003] The main purpose of the present invention is to provide an ultra-long-range motion monitoring and management system based on machine learning.

[0004] The purpose of the present invention can be achieved by adopting the following technical solutions:

[0005] The ultra-long-range motion monitoring and management system based on machine learning includes a collection module, which is used to collect the patient's rehabilitation motion information and feed it back to the background server;

[0006] The backend server is used to indicate the patient's recovery status on the acquisition module using lights of different colors, and to establish a unique relationship between the data information collected by the acquisition module and the patient's identity information;

[0007] The acquisition module includes a camera and a sports bracelet;

[0008] The camera is used to perform facial recognition on patients wearing sports bracelets to determine whether they are patients who need to be monitored, and to select the monitored patients by matching the corresponding color lights displayed on the sports bracelets with the same color frame;

[0009] Sports bracelets, used to monitor the patient's health, location and exercise status;

[0010] An information comparison module is used to compare and confirm the patient identity information corresponding to the sports bracelet with the face recognition identity information collected by the camera;

[0011] The color lights displayed on the sports bracelet collected by the camera, the color light data fed back to the background server by the sports bracelet, and the color lights that should be displayed for the patient's recovery status are compared and determined.

[0012] Preferably, the acquisition module further includes a fitness tracker, a motion sensor, a heart rate belt, a triple accelerometer, and a pressure sensitive mat / carpet;

[0013] Fitness trackers to measure patients’ steps, distance, calories burned, heart rate, and sleep quality data;

[0014] Heart rate belts, which use electrocardiogram or optical sensor technology to monitor the patient's heart rate in real time;

[0015] Three accelerometers, used to measure the patient's acceleration and movement posture;

[0016] Pressure sensitive mat / carpet for monitoring pressure distribution and balance during jumping and stepping movements.

[0017] Preferably, the camera, sports bracelet, fitness tracker, motion sensor, heart rate belt, three accelerometers, and pressure sensitive mat / carpet are all coupled with a data standardization module;

[0018] The data standardization module is used to standardize the raw data collected by each device, convert it into a unified data format, formulate a unified data field specification by using a common data protocol or format, and then transmit it to the background server.

[0019] Preferably, the acquisition module is coupled to the power amplifier module, and the power amplifier module is coupled to the long-distance communication module, and the data is transmitted to the data standardization module via the long-distance communication module.

[0020] Preferably, the following monitoring and management method is also included:

[0021] Step 1: The camera obtains the location information and motion data transmitted by the acquisition module through the background server;

[0022] Step 2: Use the camera to perform facial recognition on the patient using the acquisition module to determine whether the patient is the one who needs to be monitored;

[0023] Step 3: After the determination is completed, the patient is continuously monitored and the data is transmitted to the backend server;

[0024] Step 4: The backend server feeds back the transmitted patient data to the information comparison module for patient identification and comparison and feeds back to the camera;

[0025] Step 5: Use the camera to select and calibrate the patients who need to be monitored into three levels: blue, orange, and red.

[0026] Preferably, performing face recognition in step 2 to determine whether the patient is a patient who needs to be monitored specifically includes the following steps:

[0027] S11: locate the face in the image and use Haar feature cascade to locate the approximate position of the face;

[0028] S12: accurately locate the detected face so that the feature points in the face image are aligned and in the same position;

[0029] S13: Draw a bounding box or outline in the image to mark the position of the face, using a rectangular bounding box or a polygonal outline to include the entire face area;

[0030] S14: Apply a face recognition algorithm to identify the identity and feature vector of the face.

[0031] Preferably, the Haar feature cascade used in step S11 specifically includes the following steps:

[0032] S21: Features are defined based on local differences in grayscale pixels in the image;

[0033] S22: By sliding windows of different sizes and shapes on the image, the Haar feature values ​​in the mouth are calculated to construct a large number of Haar features;

[0034] S23: Use the AdaptiveBoosting algorithm to select the best Haar features and build a classifier;

[0035] S24: AdaBoost is used to assign a weight to each feature according to the importance of different Haar features, and a strong classifier is formed by combining multiple weak classifiers;

[0036] S25: Multiple strong classifiers are cascaded into a cascade classifier. The idea of ​​the cascade is to filter out negative samples through multiple stages of classifiers, thereby reducing the detection operations on areas where no faces exist in the image. Each stage of the classifier is composed of multiple strong classifiers. Only when they all pass, the area is considered to contain a face.

[0037] Preferably, the Haar feature values ​​are collected as eyes, nose, and mouth;

[0038] Calculate the difference between the average pixel value of the eyes, nose, and mouth and the average pixel value of the surrounding area:

[0039] Suppose you want to calculate the difference between the average pixel value of a certain area (such as area A) and the average pixel value of the surrounding area (such as area B);

[0040] The average pixel value of area A can be expressed as:

[0041]

[0042] Where P irepresents the value of the i-th pixel in area A;

[0043] N A Indicates the number of pixels in area A.

[0044] The pixel average value of area B is expressed as:

[0045]

[0046] Where P i represents the value of the i-th pixel in area B;

[0047] N B Indicates the number of pixels in region B.

[0048] So the difference formula is:

[0049] Preferably, the information comparison module includes the following operation steps when performing information comparison:

[0050] Compare the identity information obtained by face recognition with the data information of the acquisition module;

[0051] While collecting face recognition data, the color of the flashing light of the acquisition module is also obtained synchronously;

[0052] Compare the identity information obtained by face recognition, the data information of the acquisition module, the identity information corresponding to the acquisition module, and the color of the flashing light of the acquisition module;

[0053] After the comparison is completed, the boundary box or outline is monitored and analyzed using three levels of blue, orange and red, and the colors of the boundary box or outline of the three levels of blue, orange and red are consistent with the colors of the flashing lights on the acquisition module.

[0054] Preferably, before the collection module is used, the patient using it is required to register in the background server and bind his / her name, ID number, and mobile phone number;

[0055] Ensure data synchronization between the acquisition module and the backend server. When the acquisition module acquires data, the data should be transmitted to a specific backend server, which should be able to identify the acquisition module and the corresponding patient of the data source;

[0056] Establishing an association between the acquisition module and the patient's identity information, binding the acquired data with the patient's identity information so that the corresponding data can be subsequently queried and matched in a database or data storage system through the identity information;

[0057] The system and database installed in the backend server establish a single corresponding relationship between the collection module and the identity information;

[0058] The data information of the camera's face recognition also corresponds to the identity information. The camera's face recognition identity information is matched with the identity information of the acquisition module to determine whether they are consistent. If they are consistent, the light color is matched. If they are inconsistent, an alarm is triggered and the data is fed back to the backend server for data processing and manual verification.

[0059] If they are consistent, the lighting colors are matched as follows:

[0060] Color coding: Establish a set of color coding rules between the acquisition module and the camera annotation system, set specific color codes for different patient physical characteristics, such as green, orange, and red, to ensure the correspondence between the bright color of the acquisition module and the annotated color box;

[0061] Set up a real-time data transmission channel between the acquisition module and the camera annotation system. When the light color of the bracelet changes, the color data will be transmitted to the annotation system in real time.

[0062] In the camera annotation system, a color detection algorithm is used to detect specific color boxes in the video stream using common computer vision algorithms;

[0063] By comparing and matching the color data of the light on the acquisition module with the color boxes detected in the video stream, the color data and the boxes are matched by calculating the similarity between the colors or using a classification algorithm. When the match is successful, the box corresponding to the color of the light on the bracelet can be used to accurately identify the patient's status, accurately mark the patient, and accurately authenticate the patient's identity.

[0064] Beneficial technical effects of the present invention:

[0065] The ultra-long-range motion monitoring and management system based on machine learning provided by the present invention establishes a system for indicating the patient's rehabilitation status on the acquisition module using different color lights, establishes a unique relationship between the data information collected by the acquisition module and the patient's identity information, and uses face recognition to determine whether the patient wearing a sports bracelet is a patient who needs to be monitored, and uses the same color frame to select the monitored patient by matching with the corresponding color light displayed on the sports bracelet, and monitors the patient's health condition and positioning and exercise status, and then compares and confirms the single corresponding patient identity information of the sports bracelet with the face recognition identity information collected by the camera, and compares and determines the color light displayed on the sports bracelet collected by the camera, the color light data fed back to the background server by the sports bracelet, and the color light that should be displayed for the patient's rehabilitation status, thereby achieving the accuracy of the patient, the accuracy of the collected data, and the color display and camera color frame selection methods, which can efficiently and simply monitor the patient in real time, compare the problem of data mismatch, and promptly discover and solve the problem. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 It is an overall system diagram of a device according to a preferred embodiment of the ultra-long-range motion monitoring and management system based on machine learning of the present invention;

[0067] Figure 2 A diagram of a data transmission system of an acquisition module according to a preferred embodiment of an ultra-long-range motion monitoring and management system based on machine learning of the present invention;

[0068] Figure 3 This is a flowchart of a preferred embodiment of the ultra-long-range motion monitoring and management system based on machine learning according to the present invention. DETAILED DESCRIPTION

[0069] In order to make the technical solution of the present invention more clear and specific to those skilled in the art, the present invention is further described in detail below in conjunction with embodiments and drawings, but the implementation manner of the present invention is not limited thereto.

[0070] Embodiment 1

[0071] In the hospital, some heart disease patients and patients who need rehabilitation training need to be supervised whether they are undergoing rehabilitation training and their heart conditions;

[0072] The present invention installs cameras in multiple sports environments in the hospital, uses the cameras to monitor the sports conditions of patients who need rehabilitation training, records the patients' sports movements, and helps evaluate the accuracy and fluency of postures and movements through video playback.

[0073] When determining patient information, facial recognition technology and data information transmitted by smart bracelets are used to determine which patients need to be monitored for rehabilitation training and heart disease patients;

[0074] It includes a collection module, which is used to collect the patient's rehabilitation movement information and feed it back to the background server;

[0075] The backend server is used to indicate the patient's recovery status on the acquisition module using lights of different colors, and to establish a unique relationship between the data information collected by the acquisition module and the patient's identity information;

[0076] The acquisition module includes a camera and a sports bracelet;

[0077] The camera is used to perform facial recognition on patients wearing sports bracelets to determine whether they are patients who need to be monitored, and to select the monitored patients by matching the corresponding color lights displayed on the sports bracelets with the same color frame;

[0078] Sports bracelets, used to monitor the patient's health, location and exercise status;

[0079] An information comparison module is used to compare and confirm the patient identity information corresponding to the sports bracelet with the face recognition identity information collected by the camera;

[0080] The color lights displayed on the sports bracelet collected by the camera, the color light data fed back to the background server by the sports bracelet, and the color lights that should be displayed for the patient's recovery status are compared and determined.

[0081] In this embodiment, the acquisition module also includes a fitness tracker, a motion sensor, a heart rate belt, a three-accelerometer, and a pressure-sensitive mat / carpet;

[0082] Fitness trackers to measure patients’ steps, distance, calories burned, heart rate, and sleep quality data;

[0083] Heart rate belts, which use electrocardiogram or optical sensor technology to monitor the patient's heart rate in real time;

[0084] Three accelerometers, used to measure the patient's acceleration and movement posture;

[0085] Pressure sensitive mat / carpet for monitoring pressure distribution and balance during jumping and stepping movements.

[0086] Preferably, the camera, sports bracelet, fitness tracker, motion sensor, heart rate belt, three accelerometers, and pressure sensitive mat / carpet are all coupled with a data standardization module;

[0087] The data standardization module is used to standardize the raw data collected by each device, convert it into a unified data format, formulate a unified data field specification by using a common data protocol or format, and then transmit it to the background server.

[0088] In this embodiment, the acquisition module is coupled to the power amplifier module, and the power amplifier module is coupled to the long-distance communication module, and the data is transmitted to the data standardization module through the long-distance communication module.

[0089] In this embodiment, the following monitoring and management method is also included:

[0090] Step 1: The camera obtains the location information and motion data transmitted by the acquisition module through the background server;

[0091] Step 2: Use the camera to perform facial recognition on the patient using the acquisition module to determine whether the patient is the one who needs to be monitored;

[0092] Step 3: After the determination is completed, the patient is continuously monitored and the data is transmitted to the backend server;

[0093] Step 4: The backend server feeds back the transmitted patient data to the information comparison module for patient identification and comparison and feeds back to the camera;

[0094] Step 5: Use the camera to select and calibrate the patients who need to be monitored into three levels: blue, orange, and red.

[0095] Embodiment 2

[0096] The specific technical solutions are as follows:

[0097] The camera obtains the location information and motion data transmitted by the sports bracelet through the background server;

[0098] The camera performs facial recognition on patients wearing sports bracelets to determine whether they are patients who need to be monitored;

[0099] After the determination is made, the patient is continuously monitored and the data is transmitted to the backend server;

[0100] The backend server identifies and analyzes the transmitted patient data and feeds it back to the motion camera;

[0101] Patients who need to be monitored are divided into three levels: blue, orange and red through motion cameras for monitoring and analysis.

[0102] The above method is used to confirm which patients need to be monitored through double determination, thereby improving the accuracy of monitoring.

[0103] Furthermore, in the face recognition process, not only the face of the patient to be monitored needs to be recognized, but also the smart bracelet worn by the patient needs to be recognized, and the color of the flashing light on the smart bracelet needs to be recognized to achieve further recognition accuracy.

[0104] The specific technical solutions include the following:

[0105] Locate the face in the image and use Haar feature cascade to locate the approximate position of the face;

[0106] Accurately locate the detected face so that the feature points (eyes, nose, mouth) in the face image are aligned and in the same position;

[0107] Draw a bounding box or outline in the image to mark the location of the face, using a rectangular bounding box or a polygonal outline to contain the entire face area;

[0108] Apply face recognition algorithms to identify the identity and feature vectors of faces.

[0109] Compare the identity information obtained through facial recognition with the data information of the sports bracelet;

[0110] While collecting facial recognition data, the color of the flashing light of the sports bracelet is also collected synchronously;

[0111] Compare the identity information obtained by face recognition, the data information of the sports bracelet and the color of the flashing light of the sports bracelet;

[0112] After the comparison is completed, the boundary box or outline is monitored and analyzed using three levels of blue, orange and red, and the colors of the boundary box or outline at the three levels of blue, orange and red are consistent with the colors of the flashing lights on the sports bracelet.

[0113] The Haar feature cascade used in the present invention specifically includes the following technical solutions:

[0114] Perform feature extraction: Features are defined based on local differences in grayscale pixels in an image.

[0115] By sliding windows of different sizes and shapes on the image, the Haar feature values ​​in the mouth, such as contrast, edges, etc., are calculated. By adjusting the position, size and feature extraction method of the window, a large number of Haar features can be constructed.

[0116] Boost: Use the Adaptive Boosting (AdaBoost) algorithm to select the best Haar features and build a classifier.

[0117] AdaBoost is a machine learning algorithm that can assign a weight to each feature according to the importance of different Haar features, and then combine multiple weak classifiers into a strong classifier.

[0118] Cascade classifier combines multiple strong classifiers into a cascade classifier. The idea of ​​cascade is to filter out negative samples through multiple stages of classifiers, thereby reducing the detection operations on areas where there are no faces in the image. Each stage of the classifier is composed of multiple strong classifiers. Only when they all pass, the area is considered to contain a face.

[0119] Sliding window detection applies a cascade classifier to a sliding window in the image to determine whether there is a face in the window. Accurate face location can be achieved by sliding windows of different sizes and gradually reducing the size of the window.

[0120] Haar feature values ​​are collected for eyes, nose, and mouth;

[0121] Specifically, the difference between the average pixel value of the eyes, nose, and mouth and the average pixel value of the surrounding area is calculated:

[0122] Suppose you want to calculate the difference between the average pixel value of a certain area (such as area A) and the average pixel value of the surrounding area (such as area B);

[0123] The average pixel value of area A can be expressed as:

[0124]

[0125] Where P i represents the value of the i-th pixel in area A;

[0126] N A Indicates the number of pixels in area A.

[0127] The pixel average value of area B is expressed as:

[0128]

[0129] Where P i represents the value of the i-th pixel in area B;

[0130] N B Indicates the number of pixels in region B.

[0131] So the difference formula is:

[0132] Before applying the face recognition algorithm to identify the identity of a face, a database is first established and compared as follows:

[0133] Collect a dataset of face images with identity signatures. The dataset should include multiple images of multiple people, and each image should correspond to the identity signature of one person.

[0134] Use the face detection algorithm to locate and extract the face area in the image, and the algorithm still uses the Haar feature cascade;

[0135] Extract features from detected face images. Commonly used face feature extraction algorithms include principal component analysis (PCA), linear discriminant analysis (LDA), and local binary pattern (LBP). These algorithms convert face images into high-dimensional feature vector representations.

[0136] Use feature vectors for face matching or classification, use distance measurement algorithms (such as Euclidean distance, cosine similarity) to calculate the similarity score between two feature vectors, and then determine the degree of match between the face to be identified and the known identity by comparing the similarity score with a preset threshold;

[0137] Based on the result of feature matching or classification, the identity of the face to be identified is determined. Based on the similarity score and threshold setting, binary classification (whether it is a known identity) or multi-classification (determine which known identity it belongs to) can be performed.

[0138] Optimize face recognition algorithms by adjusting feature extraction methods, distance measurement algorithms, and threshold settings to improve recognition accuracy. In addition, feedback training (such as increasing sample data) can be used to improve the performance of the algorithm in actual applications.

[0139] Therefore, facial recognition of patients is achieved and then the matching between the face and the identity is compared;

[0140] Before using the smart bracelet, patients are required to register in the system and bind their identity, name, ID number, and mobile phone number;

[0141] Ensure data synchronization between smart bracelets and related systems. When the bracelet collects data, the data should be transmitted to a specific system or server. The system should be able to identify the bracelet from which the data came and the corresponding patient.

[0142] Establish an association between the smart bracelet and the patient's identity information, bind the collected data to the patient's identity information so that the corresponding data can be queried and matched in the database or data storage system through the identity information later;

[0143] Therefore, the system and database installed in the background server establish a single corresponding relationship between the smart bracelet and the identity information;

[0144] The data information of the camera's face recognition also has a single corresponding relationship with the identity information. Therefore, an identity information matching module is established to match the camera's face recognition identity information with the smart bracelet's identity information to determine whether they are consistent. If they are consistent, the light color is matched. If they are inconsistent, an alarm is triggered and feedback is sent to the background server for data processing and manual verification.

[0145] If they are consistent, the lighting colors are matched as follows:

[0146] Color coding is performed and a set of color coding rules is established between the smart bracelet and the camera annotation system. Specific color codes such as green, orange, and red are set for different patient physical characteristics to ensure the correspondence between the bright color of the bracelet and the labeled color box.

[0147] Real-time data transmission: set up a real-time data transmission channel between the bracelet and the camera annotation system. When the color of the bracelet light changes, the color data will be transmitted to the annotation system in real time.

[0148] Color detection algorithm,In the camera annotation system, a color detection algorithm is used to detect specific color boxes in the video stream using common computer vision algorithms, such as color space conversion, color filtering, and image segmentation techniques.

[0149] Color matching, by comparing and matching the color data of the bracelet light with the color boxes detected in the video stream, the color data and the boxes can be matched by calculating the similarity between the colors or using a classification algorithm (such as a support vector machine). When the match is successful, the box corresponding to the color of the bracelet light can be used to accurately identify the patient's status, accurately mark the patient, and accurately authenticate the patient's identity.

[0150] However, when monitoring patients in hospitals, due to the constantly changing distance and obstacles, the present invention adopts a long-range communication module, LoRa (low-power wide area network) or Sigfox (low-power wide area self-organizing network) module, and adopts low-power, long-distance transmission wireless technology to achieve ultra-long-range communication.

[0151] In addition, the smart bracelet is equipped with a power amplifier module to enhance the transmission power of the signal so that the signal can be transmitted over a longer distance.

[0152] Secondly, in order to ensure that the patient's condition can be monitored in real time and avoid data loss during transmission, the following solution is adopted in the present invention:

[0153] Set monitoring goals and indicators, make it clear whether the monitoring goal is signal quality or data transmission, and determine key indicators such as signal strength, data loss rate, and delay so that the monitoring system can provide useful data and alarms.

[0154] The monitoring equipment should be reasonably deployed according to the scale layout of the actual communication network. Monitoring equipment can be set up around key communication nodes, key equipment or at the edge of the coverage area to ensure that the monitoring equipment can fully cover the target area.

[0155] Set monitoring parameters and thresholds, and set monitoring parameter thresholds according to actual conditions. For example, set the minimum threshold for signal strength, and if it is lower than the threshold, an alarm will be triggered.

[0156] Real-time monitoring and data collection: The monitoring system should be able to monitor signal quality and data transmission in real time, collect data and record key monitoring indicators for subsequent analysis and judgment.

[0157] Establish an alarm mechanism to notify relevant personnel or automatically trigger corresponding actions. For example, when the signal quality is lower than the threshold or the transmission is abnormal, send an alarm message in time.

[0158] Solve communication problems, respond to alarms in a timely manner, and quickly locate and resolve communication problems. You can work with network engineers or technical support personnel to analyze the root cause of the problem and take appropriate measures to repair it.

[0159] Regular evaluation and optimization: Regularly evaluate the effectiveness of the monitoring system, optimize and improve it based on feedback and data analysis, and ensure that the monitoring system continues to effectively protect communication stability.

[0160] Secondly, the present invention uses different data collection devices to collect data from patients in different scenarios:

[0161] Specifically, it includes the collection of data from fitness trackers, motion sensors, heart rate belts, three accelerometers, and pressure-sensitive mats / carpets, and the standardization of the collected data. The raw data collected by each device is standardized and converted into a unified data format. This can be achieved by using a common data protocol or format, such as JSON or XML. A unified data field specification can be formulated to ensure that the data collected by different devices are consistent in format, and then all are transmitted to the background server. Just like before using a smart bracelet, the patient using it is required to register in the system and bind his or her name, ID number, and mobile phone number.

[0162] Ensure that data synchronization can be achieved between the acquisition equipment and related systems. When the acquisition device collects data, the data should be transmitted to a specific system or server. The system should be able to identify the acquisition device and the corresponding patient of the data source. At the same time, green, yellow and red lights are set on these devices to display the patient's status. When in use, cameras are installed near the patient to achieve three-in-one data matching and scientific management functions.

[0163] The above description is only a further embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical solutions and concepts of the present invention within the scope disclosed by the present invention, which belong to the protection scope of the present invention.

Claims

1. An ultra-long-range motion monitoring and management system based on machine learning, including a collection module, which is used to collect the patient's rehabilitation motion information and feed it back to the background server; Features: The backend server is used to indicate the patient's recovery status on the acquisition module using lights of different colors, and to establish a single relationship between the data information collected by the acquisition module and the patient's identity information; The acquisition module includes a camera and a sports bracelet; The camera is used to perform facial recognition on patients wearing sports bracelets to determine whether they are patients who need to be monitored, and to select the monitored patients by matching the corresponding color lights displayed on the sports bracelets with the same color frame; Sports bracelets, used to monitor the patient's health, location and exercise status; An information comparison module is used to compare and confirm the patient identity information corresponding to the sports bracelet with the face recognition identity information collected by the camera; The color lights displayed on the sports bracelet collected by the camera, the color light data fed back to the background server by the sports bracelet, and the color lights that should be displayed for the patient's recovery status are compared and determined.

2. The ultra-long-range motion monitoring and management system based on machine learning according to claim 1, characterized in that: The acquisition module also includes a fitness tracker, motion sensor, heart rate belt, three accelerometers, and pressure-sensitive mat / carpet; Fitness trackers to measure patients’ steps, distance, calories burned, heart rate, and sleep quality data; Heart rate belts, which use electrocardiogram or optical sensor technology to monitor the patient's heart rate in real time; Three accelerometers, used to measure the patient's acceleration and movement posture; Pressure sensitive mat / carpet for monitoring pressure distribution and balance during jumping and stepping movements.

3. The ultra-long-range motion monitoring and management system based on machine learning according to claim 2 is characterized in that: Cameras, sports bracelets, fitness trackers, motion sensors, heart rate belts, three accelerometers, and pressure-sensitive mats / carpets are all coupled with data standardization modules; The data standardization module is used to standardize the raw data collected by each device, convert it into a unified data format, formulate a unified data field specification by using a common data protocol or format, and then transmit it to the background server.

4. The ultra-long-range motion monitoring and management system based on machine learning according to claim 3 is characterized in that: The acquisition module is coupled to the power amplifier module, and the power amplifier module is coupled to the long-distance communication module, and the data is transmitted to the data standardization module through the long-distance communication module.

5. The ultra-long-range motion monitoring and management system based on machine learning according to claim 4 is characterized in that: It also includes the following monitoring and management methods: Step 1: The camera obtains the location information and motion data transmitted by the acquisition module through the background server; Step 2: Use the camera to perform facial recognition on the patient using the acquisition module to determine whether the patient is the one who needs to be monitored; Step 3: After the determination is completed, the patient is continuously monitored and the data is transmitted to the backend server; Step 4: The backend server feeds back the transmitted patient data to the information comparison module for patient identification and comparison and feeds back to the camera; Step 5: Use the camera to select and calibrate the patients who need to be monitored into three levels: blue, orange, and red.

6. The ultra-long-distance motion monitoring and management system based on machine learning according to claim 5, characterized in that: In step 2, face recognition is performed to determine whether the patient is a patient that needs to be monitored, which specifically includes the following steps: S11: locate the face in the image and use Haar feature cascade to locate the approximate position of the face; S12: accurately locate the detected face so that the feature points in the face image are aligned and in the same position; S13: Draw a bounding box or outline in the image to mark the position of the face, using a rectangular bounding box or a polygonal outline to include the entire face area; S14: Apply a face recognition algorithm to identify the identity and feature vector of the face.

7. The ultra-long-distance motion monitoring and management system based on machine learning according to claim 6, characterized in that: The Haar feature cascade used in step S11 specifically includes the following steps: S21: Features are defined based on local differences in grayscale pixels in the image; S22: By sliding windows of different sizes and shapes on the image, the Haar feature values ​​in the mouth are calculated to construct a large number of Haar features; S23: Use Adaptive Boosting algorithm to select the best Haar features and build a classifier; S24: AdaBoost is used to assign a weight to each feature according to the importance of different Haar features, and a strong classifier is formed by combining multiple weak classifiers; S25: Multiple strong classifiers are cascaded into a cascade classifier. The idea of ​​the cascade is to filter out negative samples through multiple stages of classifiers, thereby reducing the detection operations on areas where no faces exist in the image. Each stage of the classifier is composed of multiple strong classifiers. Only when they all pass, the area is considered to contain a face.

8. The ultra-long-distance motion monitoring and management system based on machine learning according to claim 7, characterized in that: Haar feature values ​​are collected for eyes, nose, and mouth; Calculate the difference between the average pixel value of the eyes, nose, and mouth and the average pixel value of the surrounding area: Suppose you want to calculate the difference between the average pixel value of a certain area (such as area A) and the average pixel value of the surrounding area (such as area B); The average pixel value of area A can be expressed as: Where P i represents the value of the i-th pixel in area A; N A Indicates the number of pixels in area A. The pixel average value of area B is expressed as: Where P i represents the value of the i-th pixel in area B; N B Indicates the number of pixels in region B. So the difference formula is:

9. The ultra-long-distance motion monitoring and management system based on machine learning according to claim 8, characterized in that: The information comparison module includes the following steps when performing information comparison: Compare the identity information obtained by face recognition with the data information of the acquisition module; While collecting face recognition data, the color of the flashing light of the acquisition module is also obtained synchronously; Compare the identity information obtained by face recognition, the data information of the acquisition module, the identity information corresponding to the acquisition module, and the color of the flashing light of the acquisition module; After the comparison is completed, the boundary box or outline is monitored and analyzed using three levels of blue, orange and red, and the colors of the boundary box or outline of the three levels of blue, orange and red are consistent with the colors of the flashing lights on the acquisition module.

10. The ultra-long-distance motion monitoring and management system based on machine learning according to claim 9, characterized in that: Before using the collection module, patients are required to register and bind their ID, name, ID number, and mobile phone number in the backend server; Ensure data synchronization between the acquisition module and the backend server. When the acquisition module acquires data, the data should be transmitted to a specific backend server, which should be able to identify the acquisition module and the corresponding patient of the data source; Establishing an association between the acquisition module and the patient's identity information, binding the acquired data with the patient's identity information so that the corresponding data can be subsequently queried and matched in a database or data storage system through the identity information; The system and database installed in the backend server establish a single corresponding relationship between the collection module and the identity information; The data information of the camera's face recognition also corresponds to the identity information. The camera's face recognition identity information is matched with the identity information of the acquisition module to determine whether they are consistent. If they are consistent, the light color is matched. If they are inconsistent, an alarm is triggered and the data is fed back to the backend server for data processing and manual verification. If they are consistent, the lighting colors are matched as follows: Color coding: Establish a set of color coding rules between the acquisition module and the camera annotation system, set specific color codes for different patient physical characteristics, such as green, orange, and red, to ensure the correspondence between the bright color of the acquisition module and the annotated color box; Set up a real-time data transmission channel between the acquisition module and the camera annotation system. When the light color of the bracelet changes, the changed color data is transmitted to the annotation system in real time. In the camera annotation system, a color detection algorithm is used to detect specific color boxes in the video stream. Use a common computer vision algorithm; By comparing and matching the color data of the light on the acquisition module with the color boxes detected in the video stream, and by calculating the similarity between colors or using a classification algorithm to match the color data and the boxes, when the match is successful, the acquisition module lights up the box corresponding to the color, so as to accurately identify the patient's status, accurately mark it, and accurately authenticate the patient's identity.

Images