Riding helmet automatic rescue method and system based on multi-source data fusion

Abstract

The invention discloses a riding helmet automatic rescue system and method based on multi-source data fusion, and the system comprises a sensor module which is used for collecting automobile state data and riding environment data, and the output end of the sensor module is connected to a data processing unit; the data processing unit is used for receiving the automobile state data and the riding environment data and performing risk assessment according to the received data to obtain an assessed risk level signal, and the output end of the data processing unit is connected to the rescue starting device; the rescue triggering device outputs a corresponding rescue signal according to the received risk level signal, and the output end of the rescue triggering device is connected to the communication module; and after receiving the rescue signal, the communication module sends the rescue signal to an external rescue system and executes a rescue response. The method has the advantages that real-time monitoring, accurate judgment and graded response of riding risks are achieved by means of auxiliary driving and lane early warning technologies through multi-source data fusion and an AI algorithm, and riding safety and rescue efficiency are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent wearable devices, traffic safety technology and artificial intelligence, in particular to a riding helmet system integrating sensor network, AI algorithm, ADAS and emergency rescue linkage, which is used for real-time monitoring of riding risks and triggering automatic rescue, and is suitable for riding scenes such as bicycles, electric bicycles and shared bicycles. BACKGROUND

[0002] Riding safety is an important issue in the field of traffic travel and sports health. With the acceleration of urbanization and the popularization of green travel concept, riding as an environmentally friendly and healthy mode of travel is favored by more and more people. However, riders are easily disturbed by vehicles, pedestrians, road conditions and other factors on the road, and there is a high risk of safety. Traditional riding safety equipment (such as helmets, reflective clothing, etc.) mainly provides passive protection and cannot actively intervene or timely send a distress signal when an accident occurs, resulting in delayed rescue and increased risk of injury to riders. Therefore, developing an intelligent riding safety system that can actively warn and automatically rescue has become an important research direction in the field of riding safety.

[0003] Currently, there are some intelligent riding helmets and related safety technologies on the market. For example, some riding helmets are equipped with acceleration sensors and gyroscopes to detect collisions and trigger alarms; some helmets are equipped with Bluetooth modules that can connect with mobile phones to realize navigation, calling and other functions. In addition, some high-end riding helmets integrate cameras to record the riding process. In terms of rescue, existing technologies mainly trigger rescue processes through manual triggering or third-party discovery of accidents, such as sending distress signals through mobile phone APPs or using smart bracelets to detect the state of riders. These technologies have improved riding safety to some extent, but still have problems such as single function and insufficient intelligence.

[0004] Although existing technologies have improved riding safety to some extent, there are still the following shortcomings: first, the risk judgment of existing intelligent helmets mainly relies on a single sensor (such as an acceleration sensor), which is prone to false positives or false negatives due to misjudgment, and cannot accurately reflect the real riding risk. Second, existing technologies lack comprehensive perception of the surrounding environment and cannot conduct comprehensive risk analysis in combination with traffic conditions (such as vehicles, pedestrians, lane lines, etc.). Finally, existing rescue mechanisms usually need to be manually triggered or rely on external devices, and cannot quickly and automatically start the rescue process after an accident occurs, resulting in low rescue efficiency. And different rescue measures need to be taken for different levels of accidents. Therefore, there is an urgent need for a riding safety solution that can combine multiple data sources to achieve accurate risk judgment and automatic rescue. SUMMARY

[0005] The present application aims to overcome the deficiencies of the prior art, and provides a riding helmet automatic risk early warning and rescue system and method, which realizes real-time monitoring, accurate judgment and hierarchical response of riding risk through multi-source data fusion and AI algorithm, and improves riding safety and rescue efficiency by means of auxiliary driving and lane warning technology.

[0006] In order to achieve the above-mentioned purpose, the technical scheme adopted by the present application is as follows: a riding helmet automatic rescue system based on multi-source data fusion, comprising a sensor module, a data processing unit, a communication module and a rescue triggering device;

[0007] The sensor module is used to collect automobile state data and riding environment data, and its output end is connected to the data processing unit.

[0008] The data processing unit is used to receive automobile state data and riding environment data and perform risk assessment according to the received data to obtain an evaluated risk level signal, and its output end is connected to the rescue triggering device.

[0009] The rescue triggering device outputs a corresponding rescue signal according to the received risk level signal, and its output end is connected to the communication module; the communication module sends a rescue signal to an external rescue system after receiving the rescue signal and executes a rescue response.

[0010] The rescue system further comprises

[0011] The user interaction module is connected with the rescue triggering device, receives the rescue signal, and sends a warning and alarm signal to the rider or the riding vehicle to execute a low-level rescue response.

[0012] The sensor module comprises any one or a combination of a three-axis acceleration sensor, a gyroscope, a camera, a heart rate sensor and / or a barometric pressure sensor; wherein,

[0013] The three-axis acceleration sensor detects the acceleration change of the rider to identify abnormal actions such as falling down and collision;

[0014] The gyroscope monitors the riding posture to determine whether a rollover or loss of control occurs;

[0015] The camera captures surrounding environment and lane information to identify obstacles, vehicles and pedestrians;

[0016] The heart rate sensor monitors the physiological state of the rider to determine whether an accident is caused by a sudden illness;

[0017] The barometric pressure sensor detects altitude changes to assist in determining whether a fall occurs.

[0018] The data processing unit is built-in with an AI algorithm, and the AI algorithm performs comprehensive risk assessment by fusing sensor data to obtain a risk level.

[0019] The communication module is used to send a rescue request signal to the outside world rescue system, which includes at least one of 4G / 5G module, Bluetooth module and V2X module;

[0020] Among them, 4G / 5G: for communication with rescue platform and emergency contact person;

[0021] Bluetooth: connected with the rider's mobile phone or vehicle ADAS system;

[0022] V2X (vehicle-to-everything communication): interact with nearby vehicles and traffic management center.

[0023] The rescue triggering device includes a one-key SOS button and / or an automatic fall detection alarm module;

[0024] The rescue triggering device sends a rescue signal after receiving the risk level signal sent by the data processing unit; or the rider manually triggers the one-key SOS button to send an emergency rescue signal; or the automatic fall detection alarm module detects and identifies the severity of the fall and automatically changes to trigger and send a rescue signal.

[0025] The user interaction module includes bone conduction earphones and LED tail lights;

[0026] Among them: bone conduction earphones are used to provide voice warning, alarm and navigation reminder signals, which are integrated in the riding helmet;

[0027] LED tail lights are used to display the riding state, which are integrated on the outside rear of the head, to provide riding state information to the rear.

[0028] A riding helmet automatic rescue method based on multi-source data fusion, through the real-time collection of physiological data, motion data and surrounding environment data of the rider by the sensor module built-in the riding helmet, using the risk assessment model constructed, combined with the collected data for dynamic risk assessment, when the risk level reaches the preset threshold, the system automatically triggers the rescue mechanism and responds to the rescue according to the risk level classification.

[0029] According to the collision risk, it is divided into first-level response and second-level response;

[0030] Among them, the first-level response of rescue includes that the helmet vibrates slightly at a preset frequency through the built-in vibration module, and the voice prompt module sends a gentle prompt sound; a query window pops up on the display screen on the helmet, if the rider does not operate within 15 seconds, automatically sends a message containing the rider's location and brief accident description to the preset emergency contact person;

[0031] The secondary response of the rescue includes that a communication module of the helmet immediately sends detailed help-seeking information including an accurate position, an accident type, and an injury degree preliminarily judged according to sensor data to an emergency contact person, and simultaneously sends a rescue request to a nearby hospital and a traffic police department, and attaches accident scene surrounding environment information; the helmet starts a high-brightness flashing warning light and a high-decibel alarm to attract the attention of surrounding pedestrians; if the helmet is connected with a vital sign monitoring device, heart rate and blood pressure data are acquired through the monitoring device and continuously transmitted to rescue personnel in real time.

[0032] A manual trigger button is arranged on the helmet, when the rider encounters danger but the system does not respond in time, or the rider subjectively judges that rescue is needed, long-pressing the button can manually trigger rescue; after triggering, according to current sensor data and positioning information, a rescue request is sent to the emergency contact person and the rescue agency according to the secondary response standard, and the warning light and the alarm are started.

[0033] The advantages of the present application are that: by means of assisted driving and lane warning technology, through multi-source data fusion and AI algorithm, real-time monitoring, accurate judgment and grading response of riding risk are realized, and the riding safety and rescue efficiency are improved. The following technical advantages are possessed:

[0034] 1: Multi-dimensional data fusion: combined with vehicle-mounted assisted driving and helmet sensors, the accident determination accuracy is improved.

[0035] 2: Grading rescue mechanism: dynamically adjusting the response strategy according to the severity of the accident, optimizing the allocation of rescue resources.

[0036] 3: Human-vehicle collaborative protection: realizing the whole-link automation from risk warning to accident rescue. BRIEF DESCRIPTION OF DRAWINGS

[0037] The contents expressed by each figure in the specification of the present application and the marks in the figures are briefly described as follows:

[0038] Figure 1 It is a structural block diagram of the rescue system of the present application;

[0039] Figure 2 It is a work flow chart of the rescue system of the present application. DETAILED DESCRIPTION

[0040] The specific embodiments of the present application are further explained in detail by describing the optimal embodiment with reference to the drawings.

[0041] The present application aims to provide a riding helmet automatic risk warning and rescue system and method, by means of assisted driving and lane warning technology, through multi-source data fusion and AI algorithm, real-time monitoring, accurate judgment and grading response of riding risk are realized, and the riding safety and rescue efficiency are improved.

[0042] Automatic rescue method of cycling helmet, comprising:

[0043] 1. Collecting cycling physiological information, cyclist body posture information and cycling collision information through a sensor module, wherein the cycling physiological information includes heart rate; the body posture information includes normal information and abnormal information; and the cycling collision information includes cycling acceleration change and cycling surrounding environment information;

[0044] 2. Determining falling information, collision information and body posture abnormal time of the cyclist through a data processing unit, wherein the falling information is obtained by analyzing the cycling physiological information, the cycling collision information is obtained by analyzing the cycling acceleration change and the cycling surrounding environment information, and the body posture abnormal time is obtained by analyzing the time interval of normal and abnormal cycling posture of the cyclist;

[0045] 3. Determining the cycling accident level according to the falling information, the body posture abnormal time and the cycling collision information through the data processing unit, wherein when the cyclist falls slightly and the body posture abnormal time is less than or equal to 2s, and there is no collision during cycling, the cycling accident level is a minor accident; when the cyclist falls and the body posture abnormal time is greater than 2s, and there is a severe collision during cycling, the cycling accident level is a serious accident

[0046] 5. Triggering a multi-level accident rescue mechanism according to the multi-level cycling accident level through a rescue triggering device, including an artificial triggering method and a cycling system self-triggering method;

[0047] 6. Executing rescue response through a rescue response module, and executing multi-level rescue response according to the multi-level accident level, wherein a first-level rescue response is executed for a minor accident, and a second-level rescue response is executed for a serious accident.

[0048] The cycling system comprises:

[0049] 1. Sensor module: collecting cycling state data and cycling environment data;

[0050] 2. Data processing unit: connected with the sensor module, receiving cycling state data and cycling environment data, processing data and performing risk assessment;

[0051] 3. Rescue triggering device: connected with the data processing unit, receiving the risk level signal sent by the data processing unit and sending a rescue signal

[0052] 4. Rescue response module

[0053] a) User interaction module: connected with the rescue triggering device, receiving the rescue signal, and sending early warning and alarm signals to the cyclist or the cycling vehicle, and executing a first-level rescue response.

[0054] b) Communication module: connected with the rescue trigger device, receives the rescue signal, sends the rescue signal to the external rescue system, and executes the secondary rescue response.

[0055] Real-time collection of physiological data, motion data and surrounding environment data of the rider through the built-in sensor modules (such as acceleration sensor, gyroscope, heart rate sensor, GPS module, camera, etc.) in the cycling helmet, combined with external data sources such as assisted driving system, lane warning system, etc. The risk assessment model constructed by machine learning algorithm is used for dynamic risk judgment and estimation. When the risk level reaches the preset threshold, the system automatically triggers a multi-dimensional rescue mechanism, including sending a distress message to emergency contacts and rescue agencies, starting the sound and light alarm device, and conducting voice communication with rescue personnel. In addition, the system also has functions such as real-time feedback, user interaction, data storage and analysis, helping the rider to adjust behavior in time and optimize the cycling experience. Through the intelligent and proactive safety protection and rescue mechanism, the invention significantly improves the cycling safety and rescue efficiency, and is suitable for various cycling scenarios, with wide application prospects.

[0056] The technical solution is specifically divided into four parts: system composition, risk judgment method, automatic rescue grading response mechanism and privacy and compliance design.

[0057] I. System composition

[0058] 1. Sensor module: collect cycling state data and cycling environment data;

[0059] Three-axis acceleration sensor: detects the acceleration change of the rider, identifies abnormal actions such as falling down and collision.

[0060] Gyroscope: monitors the riding posture and judges whether it has rolled over or lost control.

[0061] Camera: captures surrounding environment and lane information, identifies obstacles, vehicles and pedestrians.

[0062] Heart rate sensor: monitors the physiological state of the rider and judges whether the accident is caused by sudden illness.

[0063] Air pressure sensor: detects altitude changes to assist in determining whether a fall has occurred (such as a bridge or cliff).

[0064] 2. Data processing unit: connected with the sensor module, receives the cycling state data and cycling environment data, processes the data, and performs risk assessment;

[0065] Built-in AI algorithm, integrates sensor data and ADAS information, and performs comprehensive risk assessment; supports edge computing to reduce data transmission delay.

[0066] 3. Communication module: connected with rescue trigger device, receives rescue signal, sends rescue signal to external rescue system, executes secondary / tertiary rescue response

[0067] 4G / 5G: for communication with rescue platform, emergency contact.

[0068] Bluetooth: connected with cyclist's phone or vehicle ADAS system.

[0069] V2X (vehicle-to-everything communication): interacts with nearby vehicles and traffic management center.

[0070] 4. Rescue trigger device: connected with data processing unit, receives risk level signal from data processing unit, sends rescue signal, integrated in helmet,

[0071] One-key SOS button: cyclist manually triggers emergency rescue.

[0072] Automatic fall detection alarm module: determines fall severity based on AI algorithm, automatically triggers rescue.

[0073] 5. User interaction module: connected with rescue trigger device, receives rescue signal, and sends warning / alarm signal to cyclist or cycling vehicle, executes primary rescue response.

[0074] Bone conduction earphone: provides voice warning and navigation prompt, integrated in helmet.

[0075] LED tail light: displays cycling status (e.g. normal, warning, emergency), integrated in rear of helmet.

[0076] II. Risk judgment method

[0077] Cycling helmet integrates acceleration sensor, gyroscope sensor, high-definition camera, and millimeter wave radar. Acceleration sensor monitors cyclist's linear acceleration in real time, can sensitively capture subtle changes such as acceleration, deceleration, and vibration; gyroscope sensor accurately tracks cyclist's posture changes such as turning and inclination angle; high-definition camera collects image information in front and around the cyclist, providing intuitive data for environment analysis; millimeter wave radar detects the distance, speed, and angle between the cyclist and surrounding objects. Deep learning algorithm is used to analyze camera images, identify lane lines, traffic signs, vehicles, pedestrians, etc. Combined with millimeter wave radar data, a risk assessment model is constructed to calculate the relative motion parameters of the cyclist and surrounding objects, and compare them with dynamic safety threshold. For example, when a front vehicle is detected to be suddenly braking, the distance between the cyclist and the front vehicle rapidly shortens below the safety threshold, and the speed difference exceeds the set value, it is determined that there is a risk; if the acceleration and gyroscope sensor detects that the cyclist's posture is abnormal, such as inclination angle exceeding 30° for a short time and lasting more than 0.5 seconds, it is also confirmed that there is a risk.

[0078] III. Automatic Rescue Graded Response Mechanism

[0079] Primary Response (Minor Accident): When the system determines a minor accident, such as a cyclist's minor fall, a brief abnormal body posture that quickly recovers, and no obvious collision signs, the primary response is initiated. The helmet vibrates slightly at a frequency of 2-3 times per second through the built-in vibration module, while the voice prompt module emits a gentle prompt sound, such as "You seem to have a minor condition, please confirm if you need help." An inquiry window pops up on the display screen, and if the cyclist does not operate within 15 seconds, the system automatically sends a text message to the preset emergency contact containing the cyclist's location (obtained through high-precision GPS positioning) and a brief accident description (such as suspected minor fall).

[0080] Secondary Response (Serious Accident): If a serious accident is detected, such as a severe collision, a long-term abnormal posture (more than 2 seconds), it is determined as a serious accident, and the secondary response is initiated. The helmet's communication module (supporting 4G / 5G network) immediately sends detailed rescue information to the emergency contact, including accurate location, accident type (collision, serious fall, etc.), and preliminary judgment of injury level (such as possible risk of fracture, head impact, etc.) based on sensor data. At the same time, it sends a rescue request to nearby hospitals and traffic police departments, with the surrounding environment information of the accident site (obtained through camera image analysis). The helmet turns on the high-brightness flashing warning light (brightness 500-800 lumens, flashing frequency 5-8 times per second) and high-decibel alarm (volume 110-130 decibels) to attract the attention of surrounding pedestrians. If the helmet is connected to a life sign monitoring device (such as a smart bracelet), it continuously transmits real-time data such as heart rate and blood pressure to rescue personnel.

[0081] Manual Trigger Mechanism: A physical button is provided on the side of the helmet or an easily accessible location. When the cyclist encounters danger but the system does not respond in time, or the cyclist subjectively judges the need for rescue, a long press of the button for 3 seconds can manually trigger the rescue. After triggering, the system sends a rescue request to the emergency contact and rescue agencies according to the current sensor data and positioning information, according to the secondary response standard, and starts the warning light and alarm.

[0082] IV. Privacy and Compliance Design

[0083] Data anonymization processing: The surrounding environment video captured by the camera only stores the fragments before and after the accident, and the facial information is automatically blurred.

[0084] Users can customize emergency contacts and data sharing permissions, in line with the requirements of the "Personal Information Protection Law".

[0085] Data acquisition principle: The acceleration sensor in this cycling helmet is based on the piezoelectric effect. When subjected to acceleration, the internal piezoelectric material will generate a change in electric charge. By measuring the acceleration of the cyclist in all directions, the accuracy can reach ±0.01g, and it can capture the instantaneous acceleration change when the cyclist applies the brake.

[0086] Gyroscope sensor uses the principle of Coriolis force. The internal vibration element will produce a change in vibration frequency or phase when rotating due to Coriolis force, thereby accurately measuring the rotation angle and angular velocity of the cyclist's head. The measurement range can reach ±2000° / s, which can clearly distinguish the cyclist's turning, tilting and other posture changes.

[0087] High-definition camera uses CMOS image sensor to convert optical signal to electrical signal, and then through analog-to-digital conversion and image processing algorithm, output digital image for analysis. It has 120° ultra-wide angle shooting capability and can fully cover the forward field of view.

[0088] Millimeter wave radar works in the frequency band of 76GHz-81GHz. The transmitted millimeter wave signal is reflected back after encountering surrounding objects. By analyzing the frequency difference (based on the Doppler effect) and time difference between the transmitted signal and the reflected signal, the distance, relative speed and angle of the object and the cyclist are accurately calculated. The distance accuracy can reach ±3cm, and the speed accuracy can reach ±0.2m / s.

[0089] Risk judgment principle: The images collected by the camera are analyzed based on the deep learning algorithm of convolutional neural network (CNN). CNN extracts image features such as lane line shape, traffic sign pattern, vehicle and pedestrian contour through multiple convolution layers, and then reduces the dimension of features through pooling layer. Finally, through the full connection layer, it is classified and recognized. Under the training of a large number of labeled data, the recognition accuracy of various targets can reach more than 95%. Combined with the distance and speed data obtained by the millimeter wave radar, the preset risk assessment algorithm is used to compare the relative motion parameters of the cyclist and the surrounding objects with the dynamic safety threshold. The safety threshold will be adjusted in real time according to the factors such as cycling speed, road type (such as urban street, highway, rural road), etc. For example, in urban streets, due to complex road conditions and more vehicles and pedestrians, the safety distance threshold will be relatively small. On the highway, considering the fast cycling speed, the safety distance threshold will be increased accordingly. When the relative speed exceeds a certain value and the distance rapidly shortens to below the safety threshold within a short time, it is determined that there is a collision risk. At the same time, the data of acceleration sensor and gyroscope sensor are used to analyze the posture stability of the cyclist. When it is detected that the cyclist's posture changes abnormally, such as the inclination angle exceeding 45° within a short time and the duration exceeding 0.8 seconds, combined with other sensor data to confirm whether an accident has occurred.

[0090] Automatic rescue tiered response principle: Once the risk assessment module identifies an accident, it initiates different levels of rescue response based on the severity. Level 1 response, for minor incidents, primarily uses the helmet's built-in vibration module, voice prompt module, and display to inquire and prompt the rider, leveraging their tactile, auditory, and visual senses to empower them to independently choose whether to request rescue. If the rider does not respond, a brief message is automatically sent to emergency contacts. Level 2 response, for serious incidents, utilizes the communication module, using 4G / 5G networks, and according to pre-set communication protocols, to send a distress signal containing key information such as the precise location, detailed accident type, and preliminary injury details to emergency contacts and rescue agencies. Simultaneously, the helmet's warning lights and siren activate, using light and sound signals to attract the attention of nearby personnel and increase the chances of rescue. A manual trigger mechanism provides riders with a way to independently control rescue efforts. When the rider deems rescue necessary, they can directly trigger the system to initiate the rescue process according to the level 2 response standard by pressing a physical button.

[0091] like Figure 1 、 2 The specific embodiments of the present invention are further described in detail.

[0092] Hardware configuration:

[0093] Sensors: Bosch's BMI270 accelerometer and gyroscope fusion sensor features low power consumption, high precision, and high stability, enabling precise detection of various motion changes during riding. The high-definition camera, powered by the Sony IMX686, boasts 64 megapixels and an f / 1.8 aperture, capturing clear, high-resolution images and acquiring high-quality image data for analysis even in complex lighting conditions. The millimeter-wave radar, powered by Infineon's BGT60TR13C, operates in a stable frequency band and accurately detects objects within a 30-meter radius, achieving distance resolutions of ±2 cm and velocity resolutions of ±0.1 m / s.

[0094] Processing and Communication Module: The core processor is Rockchip's RK3588S, based on the ARM architecture and featuring quad-core Cortex-A76 and quad-core Cortex-A55 processors. Coupled with a powerful GPU, it boasts exceptional computing power, enabling rapid execution of deep learning algorithms and complex risk assessment programs. The communication module utilizes Quectel's RM500Q-GN 5G module, supporting both 5G NSA / SA dual-modes, with download speeds up to 2.5Gbps and upload speeds up to 1.25Gbps, ensuring fast and stable transmission of emergency information. A Bluetooth 5.2 module is also integrated for short-range communication with mobile phones and other devices.

[0095] Other components: equipped with a lithium polymer battery with a capacity of 4000mAh, providing stable power support for the entire system, with a battery life of 8h-10h. A physical button with obvious touch and easy operation is set on the side of the helmet as a manual trigger button, the button surface has anti-slip design, which is convenient for riders to operate quickly and accurately in emergency situations. Inside the helmet, there is also a high-brightness LED warning light with a brightness of 800lm, and a high-decibel alarm with a volume of 130dB, ensuring that it can effectively attract the attention of surrounding people in emergency situations.

[0096] Software system:

[0097] Data acquisition and preprocessing: a special driver is written to realize real-time acquisition of sensor data. The collected data is first denoised, for example, Kalman filter algorithm is used for acceleration sensor and gyroscope sensor data to remove noise interference and improve data accuracy and stability; the camera image data is preprocessed such as denoising and contrast enhancement to improve the accuracy and efficiency of subsequent image recognition.

[0098] Data acquisition and preprocessing driver content as follows:

[0099] Step one, acceleration and gyroscope sensor data acquisition and preprocessing

[0100] Driver initialization: Bosch BMI270 acceleration and gyroscope fusion sensor is selected, the driver initializes at system startup to establish a stable communication connection with the sensor and sets the sensor's working mode to high-precision detection mode to meet the precise detection needs of various motion state changes during riding.

[0101] Data acquisition: according to the set sampling frequency (such as 100Hz), the acceleration and gyroscope data are read from the sensor at regular intervals to obtain the acceleration of the rider in each direction and the rotation angle and angular velocity information of the head.

[0102] Kalman filter denoising: to improve the accuracy and stability of the data, Kalman filter algorithm is used to process the acceleration and gyroscope data. Kalman filter estimates the state at the last time and the measurement at the current time, and through the prediction and update steps, the sensor data is optimized to effectively remove noise interference, making the data more accurately reflect the actual motion state of the rider.

[0103] Step two, high-definition camera data acquisition and preprocessing

[0104] Camera Driver Initialization: For the Sony IMX686 high-definition camera, the driver first initializes and configures the camera's parameters, such as resolution (set to 6400 pixels corresponding resolution), aperture (f / 1.8), exposure time, etc., to adapt to different light conditions and shooting needs.

[0105] Image Acquisition: The driver controls the camera to acquire images at a frame rate of 30 frames per second, ensuring real-time capture of the environment in front of and around the rider.

[0106] Image Preprocessing: The collected image data is denoised using median filtering or Gaussian filtering algorithms to remove noise points and improve image clarity. At the same time, contrast enhancement is performed by adjusting the image's gray scale range to make details more obvious, facilitating subsequent image recognition and analysis.

[0107] Step Three, Millimeter Wave Radar Data Acquisition and Preprocessing

[0108] Radar Driver Initialization: For the BGT60TR13C millimeter wave radar of Infineon, the driver initializes the radar at startup, setting the radar's working frequency band (76GHz~81GHz), transmission power, and other parameters to ensure the radar works normally and accurately detects surrounding objects.

[0109] Data Acquisition: According to the radar's working cycle, the driver reads the detected object's distance, relative speed, and angle information from the radar at regular intervals.

[0110] Data Calibration and Smoothing: To correct errors in the millimeter wave radar's measurement process, the driver calibrates the collected data based on the radar's characteristic parameters and actual measurement conditions. It also uses methods such as sliding average filtering to smooth the data, reducing fluctuations and improving data stability and reliability.

[0111] Step Four, Data Transmission and Storage

[0112] After preprocessing, the sensor data is packaged by the driver according to a certain format and transmitted to the data processing unit (such as using RK3588S of Rockchip as the core processor) through the system's data bus. In the data processing unit, the data is stored as needed for subsequent analysis and review, and real-time data is provided to the risk judgment and decision-making module for real-time riding risk assessment and accident judgment.

[0113] Through the detailed design and implementation of the sensor data collection and preprocessing drivers, accurate and real-time high-quality sensor data can be obtained, providing a solid data foundation for the normal operation and function implementation of the multi-source data fusion-based cycling helmet automatic rescue system.

[0114] Risk judgment and decision-making: A convolutional neural network model is built based on the PyTorch deep learning framework for image recognition. The model is trained on a large number of actual cycling scene images and can accurately identify various traffic elements. The risk assessment algorithm calculates the relative motion parameters of the cyclist and the surrounding objects in real time based on sensor data and compares them with the pre-set dynamic safety threshold. The safety threshold is adjusted in real time according to the cycling speed, road conditions, etc. For example, in different road conditions, different safety distance and speed threshold ranges are determined through analysis of a large amount of historical accident data. When the risk assessment result reaches the corresponding accident level, the corresponding rescue response process is started.

[0115] Automatic rescue grading response: The first-level response function is realized by writing vibration control programs, speech synthesis programs, and display screen drivers. When the first-level response is triggered, the vibration module works according to the pre-set vibration mode (such as 2 vibrations per second), the speech prompt module synthesizes the corresponding voice information (such as "You seem to have a minor condition, please confirm whether you need help, if you do not need help, please ignore, 15 seconds later will automatically contact the emergency contact person") and plays it through the built-in speaker, and a query window pops up on the display screen, which contains "Need help" and "No need for help" options for easy operation by the cyclist. The second-level response function is realized through the 5G communication module and the pre-set rescue communication protocol. When the second-level response is triggered, the communication module sends a message containing the latitude and longitude position (accurate to six decimal places), the accident type (collision, serious fall, etc.), and the preliminary injury situation (if connected to a life sign monitoring device, data such as heart rate, blood pressure, and oxygen saturation can be obtained) to the emergency contact person in priority order; then sends a detailed rescue request to nearby hospitals, traffic police departments, and other rescue agencies, including high-definition images of the accident scene (taken by the camera and compressed before sending), surrounding environment information, and the cyclist's health profile information (if connected to a health management platform). The manual trigger mechanism is realized by writing button detection programs and rescue trigger programs. When the manual trigger button is pressed for 3 seconds, the second-level response rescue process is immediately started, ensuring that the cyclist can obtain timely rescue in an emergency.

[0116] Actual application scenario demonstration:

[0117] Assuming a cyclist is riding at 30km / h on an urban road, a car in front suddenly changes lanes. The millimeter wave radar quickly detects the speed change and rapid distance reduction of the vehicle in front, while the camera identifies that the turn signal of the vehicle is not on and the vehicle body cuts into the lane where the cyclist is located. After the sensor data is transmitted to the RK3588S processor, the risk assessment algorithm calculates that the approach speed of the cyclist and the vehicle in front exceeds the safety threshold, and the distance will be reduced to a dangerous distance within 2 seconds, determining that there is a collision risk. Since the collision is relatively minor, the cyclist only loses balance for a short time but quickly recovers, and the system determines that it is a minor accident, triggering a level one response. The helmet triggers a vibration, a voice prompt, and a display screen asking the cyclist if he needs help. If the cyclist does not operate within 15 seconds, the system automatically sends a message to the emergency contact containing the location and suspected minor collision. If the cyclist falls during the avoidance process and does not get up for a long time (more than 3 seconds), the acceleration sensor and gyroscope sensor detect that the cyclist's posture is abnormal for a long time, and the system determines that it is a serious accident, triggering a level two response. The 5G communication module immediately sends a distress message to the emergency contact, informing the accident location, the type of serious fall accident, and the possible existence of a fall injury; at the same time, it sends a rescue request to the nearby hospital and traffic police department, attaching the image of the accident scene and the basic health information of the cyclist. The LED warning light of the helmet starts to flash rapidly, and the alarm emits a high-decibel alarm sound to attract the attention of surrounding pedestrians. If the cyclist feels unwell or encounters other emergencies during the ride, he can directly long-press the manual trigger button for 3 seconds, and the system will immediately send a rescue request to the emergency contact and rescue agencies according to the level two response standard, and start the warning light and alarm, to save valuable rescue time for the cyclist.

[0118] Obviously, the specific implementation of the present application is not limited by the above-mentioned manner, as long as various non-essential improvements are made using the method concept and technical solutions of the present application, they are within the protection scope of the present application.

Claims

1. Cycling helmet automatic rescue system based on multi-source data fusion, characterized by: It includes sensor module, data processing unit, communication module and rescue trigger device; The sensor module is used to collect vehicle status data and riding environment data, and its output end is connected to the data processing unit; The data processing unit is used to receive vehicle status data and riding environment data and perform risk assessment based on the received data to obtain an assessed risk level signal, and its output end is connected to the rescue triggering device; The rescue trigger device outputs a corresponding rescue signal according to the received risk level signal, and its output end is connected to the communication module; After receiving the rescue signal, the communication module sends a rescue signal to an external rescue system and executes a rescue response.

2. The automatic rescue system for cycling helmets based on multi-source data fusion according to claim 1, characterized in that: The rescue system also includes User interaction module: connects to the rescue trigger device, receives rescue signals, and sends early warning and alarm signals to the cyclist or the riding vehicle to perform low-level rescue response.

3. The automatic rescue system for cycling helmets based on multi-source data fusion according to claim 1 or 2, characterized in that: The sensor module includes any one or a combination of a three-axis acceleration sensor, a gyroscope, a camera, a heart rate sensor and / or an air pressure sensor; wherein, Three-axis acceleration sensor: detects changes in the rider's acceleration to identify abnormal movements such as falls and collisions; Gyroscope: monitors riding posture to determine whether rollover or loss of control occurs; Camera: Captures surrounding environment and lane information to identify obstacles, vehicles, and pedestrians; Heart rate sensor: monitors the rider's physiological state to determine whether an accident is caused by a sudden illness; Air pressure sensor: Detects altitude changes to help determine whether a fall has occurred.

4. The automatic rescue system for cycling helmets based on multi-source data fusion according to claim 1 or 2, characterized in that: The data processing unit has a built-in AI algorithm, and the AI ​​algorithm performs comprehensive risk assessment by fusing sensor data to obtain a risk level.

5. The automatic rescue system for cycling helmets based on multi-source data fusion according to claim 1 or 2, characterized in that: The communication module is used to send a rescue request signal to an external rescue system, and includes at least one of a 4G / 5G module, a Bluetooth module, and a V2X module; 4G / 5G: used for communication with rescue platforms and emergency contacts; Bluetooth: connects to the rider’s mobile phone or vehicle ADAS system; V2X (Vehicle-to-Everything) communication: interacts with nearby vehicles and traffic management centers.

6. The automatic rescue system for cycling helmets based on multi-source data fusion according to claim 1 or 2, characterized in that: The rescue trigger device includes a one-touch SOS button and or an automatic fall detection alarm module; The rescue trigger device sends a rescue signal after receiving the risk level signal from the data processing unit; or sends an emergency rescue signal after the rider manually triggers the one-touch SOS button; or automatically changes the trigger to send a rescue signal after detecting and identifying the severity of the fall through the automatic fall detection alarm module.

7. The automatic rescue system for cycling helmets based on multi-source data fusion according to claim 1 or 2, characterized in that: The user interaction module includes bone conduction headphones and LED taillights; Among them: bone conduction headphones are used to provide voice warning, alarm and navigation reminder signals, which are integrated into the cycling helmet; The LED taillight is used to display the riding status. It is integrated at the rear outside of the head and tail to provide riding status information to the rear.

8. A method for automatic rescue of cycling helmets based on multi-source data fusion, characterized by: The built-in sensor module in the cycling helmet collects the rider's physiological data, motion data and surrounding environment data in real time, and performs dynamic risk assessment based on the collected data. When the risk level reaches the preset threshold, the system automatically triggers the rescue mechanism and responds to rescue according to the risk level.

9. The automatic rescue method for cycling helmets based on multi-source data fusion according to claim 8, characterized in that: Divided into rescue primary response and secondary response according to collision risk; The first-level rescue response includes the helmet vibrating slightly at a preset frequency through the built-in vibration module, while the voice prompt module emits a gentle prompt tone. A query window pops up on the helmet's display screen. If the rider does not respond within 15 seconds, a text message containing the rider's location and a brief description of the accident is automatically sent to the preset emergency contact. The secondary rescue response includes: the helmet's communication module immediately sends a detailed distress message to the emergency contact, including the precise location, type of accident, and the extent of the injury initially determined based on sensor data; At the same time, a rescue request is sent to nearby hospitals and traffic police departments, with information about the surrounding environment of the accident scene attached; the helmet turns on high-brightness flashing warning lights and high-decibel sirens to attract the attention of pedestrians around; if the helmet is connected to a vital signs monitoring device, it obtains heart rate and blood pressure data through the monitoring device and continuously transmits heart rate, blood pressure and other data to rescue personnel in real time.

10. The automatic rescue method for cycling helmets based on multi-source data fusion according to claim 9, characterized in that: A manual trigger button is set on the helmet. When the rider encounters danger but the system does not respond in time, or the rider subjectively determines that rescue is needed, long pressing the button can manually trigger rescue; Once triggered, based on the current sensor data and positioning information, a rescue request is sent to emergency contacts and rescue agencies in accordance with the secondary response standard, and the warning lights and sirens are activated at the same time.

Images