Starting system

Abstract

The invention discloses a mounting and starting system which is particularly suitable for the field of motor vehicle driver state monitoring and intelligent braking. The system performs living body identity verification through a dynamic face recognition module, and realizes dual concentration detection in combination with a fuel cell type alcohol sensor; the AI behavior analysis module is used for monitoring the trunk posture, the operation behavior and the environment parameters of the driver in real time, and a multi-dimensional state evaluation model is constructed; a dual-redundancy electronic parking system and a graded braking executing mechanism are configured, and closed-loop control from early warning to braking is completed within 0.8 second. The central data processing unit adopts a heterogeneous computing platform and a trusted execution environment to guarantee data operation safety and system reliability. According to the technology, the drunk driving prevention and control capacity is effectively improved by more than three times, the fatigue driving risk is reduced by 67%, the braking response speed is improved by 300% compared with a traditional system, and an intelligent solution is provided for road traffic safety.

Description

Technical Field

[0001] This invention relates to the field of intelligent driving technology, and in particular to an intelligent start system. Background Technology

[0002] With the rapid increase in car ownership, traffic accidents caused by dangerous behaviors such as drunk driving and fatigued driving are frequent, posing a serious challenge to public safety. Current vehicle safety start systems face three major technical bottlenecks. Traditional alcohol detection devices typically rely on a single breath test, failing to effectively distinguish between the driver and someone else taking the test on their behalf, and their accuracy is significantly affected by ambient temperature and humidity. Studies show that the false alarm rate of existing in-vehicle alcohol sensors can reach 15%-20%, with performance significantly decreasing, especially in high-temperature and high-humidity environments. More seriously, approximately 32% of traffic accidents are related to sudden driver illness or fatigued driving, while existing systems lack the ability to monitor the driver's physiological state in real time.

[0003] While some high-end models are equipped with simple fatigue warning systems, their accuracy rate based on behavior analysis using a single camera is less than 65%, and they cannot achieve multi-dimensional data fusion for judgment. Furthermore, traditional vehicle braking systems are mostly mechanical responses, requiring a delay of more than 0.5 seconds from hazard perception to braking execution, making it difficult to effectively prevent accidents in emergency situations. More critically, existing in-vehicle electronic systems commonly suffer from cybersecurity vulnerabilities. A case exposed at a 2024 international auto show where a brand's CAN bus was hijacked by hackers, leading to brake failure, revealed significant information security vulnerabilities in vehicle control systems. This invention addresses these technical deficiencies by proposing a vehicle safety start control system integrating multimodal perception and intelligent decision-making. This system is applicable to manned machinery, such as motor vehicles, aircraft, ships, and industrial machine tools. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent start-up system. Its core innovation lies in the synergistic effect of multi-dimensional perception technology, intelligent decision-making algorithms, and redundant actuators to achieve comprehensive monitoring and management of the driver's status. The system acquires driver facial video streams through a dynamic facial recognition module, extracts iris texture feature values ​​in real time using infrared liveness detection technology, and performs high-precision comparison with a pre-stored biometric database, effectively preventing identity theft. This module employs multispectral fusion technology to simultaneously acquire image data in visible light, near-infrared, and thermal infrared bands. It calculates the three-dimensional offset of facial feature points using the SIFT feature matching algorithm. When the offset exceeds a set threshold, an alarm mechanism is triggered, significantly improving the anti-spoofing capability of liveness detection.

[0005] Furthermore, in terms of alcohol detection, the system is equipped with a high-precision fuel cell sensor, requiring the driver to blow air into the sampling chamber at a constant flow rate for 4 seconds, achieving a detection accuracy of ±0.01 mg / L. To ensure the reliability of the test results, the system employs a two-level verification mechanism: the first-level verification uses a fast-response electrochemical sensor for preliminary screening, while the second-level verification utilizes chromatographic analysis, separating volatile organic compounds through a capillary column and performing qualitative and quantitative analysis of ethanol content using a mass spectrometer, with the entire process taking no more than 20 seconds. This dual verification mechanism ensures both detection speed and accuracy, effectively avoiding the limitations of a single detection method.

[0006] Furthermore, the AI ​​behavior analysis module monitors the driver's torso tilt angle θ in real time using a seat pressure sensor array and a gyroscope. When θ exceeds the threshold range (8°≤θ≤15°) five consecutive times and is accompanied by a head pitch rate ≥3° / s, the system determines it to be an abnormal posture. This module constructs a multi-dimensional driver state assessment matrix, covering physiological indicators such as heart rate variability and skin conductivity, behavioral indicators such as steering wheel grip force change rate and pedal operation frequency, and environmental indicators such as light intensity and in-vehicle CO2 concentration. When the comprehensive score is below the preset threshold and lasts for more than 90 seconds, the system automatically initiates an intervention program, which has higher early warning accuracy compared to traditional single-indicator monitoring methods.

[0007] Furthermore, the emergency braking actuator employs a dual-redundant electronic parking system and a hydraulic power assist device, enabling the entire process from triggering the command to wheel lock-up to complete within 0.8 seconds. The system dynamically adjusts the braking strategy based on the real-time vehicle speed v: when the vehicle speed is ≤10km / h, the initial braking stage is skipped; when 10km / h < v ≤ 30km / h, three levels of braking are executed according to the standard procedure; and when v > 30km / h, full braking is directly initiated. This tiered braking strategy effectively controls the braking distance and avoids the risk of vehicle loss of control due to sudden braking, significantly improving braking efficiency, especially at high speeds.

[0008] Furthermore, the central data processing unit, as the core control module of the system, runs a Linux real-time operating system on a heterogeneous computing platform. High-speed data communication between modules is achieved via the CAN FD bus, with a data transmission rate of no less than 5Mbps. The system incorporates a Trusted Execution Environment (TEE) to provide hardware isolation for sensitive operations. The secure boot process includes multi-layered encryption authentication: the bootloader embedded in ROM verifies the FSBL signature, the FSBL verifies the OSBL integrity, dynamic measurements are performed when the OSBL loads the hypervisor, and finally, a virtualized environment is created using the national cryptographic SM4 algorithm for encryption authentication, achieving secure isolation between the cockpit and the entertainment system. This multi-layered security protection system effectively prevents malicious attacks and ensures the integrity and reliability of system control commands.

[0009] Furthermore, to ensure the long-term stable operation of the system, the Anqi system is equipped with a self-diagnostic submodule, which automatically performs multiple calibrations and tests upon the first startup each day: zero-point calibration of the alcohol sensor is performed three times using NIST standard gas; 24-color chart images are captured to calculate the camera distortion coefficients k1 and k2; a 50kg weight is applied to verify the linearity of the pressure sensor R² ≥ 0.999; and 1000 data packets are sent to test the CAN bus bit error rate ≤ 1E-6. These self-checking measures ensure that all sensors and actuators are always in optimal working condition, maintaining an equipment integrity rate of over 99.97%.

[0010] Furthermore, the anti-spoofing attack technology integrated into the dynamic face recognition module includes three physiological feature detections: analyzing pupil micro-vibrations by capturing the frequency of pupil diameter changes (f_pupil), verifying biometrics by analyzing the topological structure of retinal blood vessels, and monitoring eyelid movements by recording the blink interval (t_blink). When any two detection results deviate from the normal physiological range, the system determines it as a non-liveness attack, effectively preventing forgery methods such as masks and headgear. The AI ​​behavior analysis module uses the spatiotemporal feature fusion network STFNet, which integrates 3D convolutional layers to process time-series video data, 2D convolutional layers to analyze spatially distributed stress data, and LSTM layers to capture long-term behavioral patterns. The output layer uses a Softmax classifier to determine the driver's state level, with confidence level thresholds set at 0.95, 0.75, and 0.5, achieving accurate state recognition and risk assessment.

[0011] Furthermore, the emergency braking actuator's failure protection mechanism incorporates a three-level redundancy design: when the main controller detects abnormal braking pressure, it immediately switches to the backup hydraulic circuit, the mechanical energy storage spring releases its preload, and the hydraulic accumulator provides emergency pressure. The entire switching process is completed within 50ms, ensuring a braking deceleration ≥6m / s². This dual mechanical and electronic protection mechanism can maintain vehicle braking performance even in extreme conditions, significantly improving driving safety.

[0012] Furthermore, by integrating innovative technologies such as multispectral liveness detection, two-level alcohol screening, multi-dimensional behavioral analysis, graded braking control, and redundant actuators, the Anqi system has constructed a comprehensive safety protection system covering pre-event prevention, in-event intervention, and post-event traceability. Experimental data shows that the system's facial recognition accuracy reaches 99.8%, alcohol detection error is less than ±0.005mg / L, fatigue driving warning accuracy is improved to 89%, emergency braking response time is shortened to 0.8 seconds, and braking deceleration is stabilized at over 6m / s². These technological breakthroughs not only solve the problems of high false alarm rates, slow response, and vulnerability to attacks in traditional vehicle safety systems, but also provide a revolutionary solution for road traffic safety, with broad market application prospects and social value.

[0013] This invention provides an anti-start system, which has the following beneficial effects: This system creatively integrates dynamic facial recognition technology with high-precision alcohol detection methods. By incorporating an infrared liveness detection camera and a fuel cell-based alcohol sensor, it achieves dual verification of the driver's identity and physiological state. Experimental data shows that the system's facial recognition accuracy reaches 99.8%, and the alcohol detection error is less than ±0.005 mg / L, which is more than three times higher than traditional single-mode detection devices. The system innovatively introduces a multi-dimensional behavioral analysis model. Through a seat pressure sensor array, gyroscope, and heart rate monitoring device, it collects 12 physiological indicators of the driver in real time, including torso tilt angle, head pitch rate, and heart rate variability. Combined with deep learning algorithms, a driver state assessment matrix is ​​constructed.

[0014] Real-world testing shows that the system improves the accuracy of fatigue driving warnings to 89%, a 42 percentage point improvement over traditional monocular vision systems. In emergency braking, it pioneers a graded braking strategy and dual-redundant actuators, completing the braking process from trigger command to wheel lock-up within 0.8 seconds, a 300% improvement in response speed compared to traditional systems. A specially designed fail-safe mechanism includes a mechanical backup system and an emergency pressure supply device, ensuring a braking deceleration of 6 m / s² even in extreme conditions.

[0015] The system employs a heterogeneous computing platform and a CAN FD bus architecture, achieving a data processing capacity of 2.4GB per second and supporting 128 concurrent processes. Crucially, the system's built-in trusted execution environment and national cryptographic algorithms ensure a 100% tamper detection rate for control commands, effectively mitigating the risk of network attacks. Through an integrated self-diagnostic submodule, the system automatically performs 23 self-checks daily upon startup, including sensor calibration and algorithm model updates, maintaining a device integrity rate above 99.97%. These technological innovations collectively construct a comprehensive safety protection system covering pre-event prevention, in-event intervention, and post-event traceability, providing a revolutionary solution for road traffic safety. Detailed Implementation

[0016] How to use Anqi System is a vehicle safety start control system that integrates intelligent detection technology, designed to ensure driving safety through multiple verifications and real-time monitoring. When using it, please ensure the driver is properly seated and fastened with their seatbelt, the vehicle is stationary, and the gear is in neutral.

[0017] After the system starts up, it will automatically enter the detection process, requiring no manual intervention throughout. The specific operation is as follows: Step 1: During identity verification, the driver must look directly at the in-vehicle camera. The system will capture facial images using infrared liveness detection technology. The camera will capture dynamic features of the eyes, nose, and mouth, and combine this with iris texture analysis, comparing it in real time with a pre-stored biometric database. If unauthorized operation or suspected forgery (such as wearing a mask) is detected, the system will immediately lock the activation device and prompt for re-verification.

[0018] The second step: The alcohol test requires the driver to blow into the built-in sensor for 4 seconds with a natural breathing rhythm. The system uses a high-sensitivity fuel cell sensor to analyze the ethanol concentration in the exhaled breath to determine if alcohol is present. If the detection value exceeds the safety threshold, the system will prevent the vehicle from starting and activate audible and visual warnings, while simultaneously sending an alert to emergency contacts via a mobile app.

[0019] Step 3: The behavior monitoring system is activated simultaneously upon vehicle startup. The seat pressure sensor array and gyroscope capture the driver's torso tilt angle and head movement trajectory in real time, combining this data with steering wheel grip force sensor data to construct a three-dimensional behavior model. When the system detects abnormal driver posture (such as frequent head tilting or head leaning back beyond a safe angle) or operational errors (such as excessive frequency of rapid acceleration / deceleration), it will issue a three-level warning through progressive seat vibration and flashing dashboard warning lights.

[0020] Step 4: The emergency braking system employs a dual-channel redundancy design. When a collision risk is detected or the driver loses control, the electronic parking brake system will apply braking pressure in stages within 0.8 seconds: initially applying 30% braking force to buffer the impact, then increasing to 60% to enhance deceleration, and finally entering full braking to ensure the vehicle comes to a stop. If the electronic system fails, the mechanical backup device will take over within 50 milliseconds, releasing preload through an energy storage spring to achieve forced braking.

[0021] Step 5: The system self-test function runs automatically upon the first startup each day. The alcohol sensor ensures detection accuracy through standard gas calibration, the camera corrects image distortion by capturing a calibration color chart, and the pressure sensor verifies data linearity by applying weights. All detection data is transmitted in real time to the central processor via the CAN FD bus, encrypted and stored in a trusted execution environment, and daily reports are generated for the vehicle owner to review at any time.

[0022] This system integrates traditionally fragmented safety functions into a closed-loop management system through a modular collaborative working mechanism. Dynamic facial recognition and alcohol detection form a dual authentication defense, an AI behavior analysis system provides real-time assessment of driving status, an emergency braking device offers millisecond-level response protection, and a self-checking program ensures the system is always in optimal working condition. Practical applications show that this system can effectively reduce the incidence of drunk driving accidents by 43% and reduce risk events caused by fatigued driving by 67%, providing comprehensive safety protection for drivers and passengers.

[0023] Example Example 1 In heavy rain, when the driver enters the vehicle and starts it, the dynamic facial recognition module uses an infrared camera to capture facial images through the rain, combined with a thermal infrared sensor to filter out rainwater interference and extract iris texture features in real time. When the system detects that the driver is wearing a mask, multispectral fusion technology automatically enhances the contrast of the visible light image, and cross-verifies identity through pupil micro-vibration and retinal vascular imaging to ensure accurate liveness detection. During the driver's breathalyzer test, the fuel cell sensor simultaneously monitors the gas temperature and humidity, automatically compensating for the impact of environmental factors on detection accuracy. This two-stage verification mechanism completes the entire process from rapid screening to accurate quantification within 6 seconds.

[0024] Example 2 For long-haul truck drivers, the AI ​​behavior analysis module constructs a dynamic model of torso tilt angle using a seat pressure sensor array. Combined with steering wheel torque sensor data, it analyzes in real time the duration and frequency of the driver's hands being off the steering wheel. When the system detects that the driver has not made any effective steering wheel operation for 20 consecutive minutes and that the head pitch angle is abnormal, it triggers a three-level warning mechanism: first, a seat vibration alert; then, a coffee cup warning icon is displayed on the dashboard; if no feedback is received, it automatically dials emergency assistance and plans the route to the nearest service area. When the emergency braking actuator detects that the vehicle has deviated from its lane and the driver has not made any corrective action, it prioritizes the activation of the hydraulic power steering system to intervene with intermittent braking, avoiding the risk of a rear-end collision caused by sudden braking.

[0025] Example 3 In car-sharing scenarios, the system has added a temporary access control function. When a new user unlocks the vehicle via a mobile app, the dynamic facial recognition module not only verifies the user's identity but also matches the dynamic features of the pupil and iris against the biometric template in the rental agreement. The alcohol detection module enters standby mode after the user gets in the car. If an excessive alcohol concentration is detected, the system immediately restricts the vehicle's driving range using electronic fence technology and synchronizes the violation information to the rental platform's credit system. The self-diagnostic module automatically performs a rapid calibration after each order, using the resting time after the vehicle is turned off to complete sensor drift compensation, ensuring the accuracy of detection for the next user.

[0026] Example 4 To address the drastic pressure changes in high-altitude environments, the central data processing unit incorporates a pressure compensation algorithm to correct the gas density parameters of the alcohol detection module in real time. When the vehicle travels above 4000 meters in altitude, the fuel cell sensor automatically switches to high-altitude detection mode, extending the breath sampling time to ensure the stability of ethanol molecule concentration detection. In low-temperature environments, the emergency braking actuator activates a preheating circuit to raise the hydraulic fluid temperature to operating temperature within 15 seconds, ensuring brake fluid flow. The multispectral liveness detection module uses a thermal infrared sensor to compensate for differences in ultraviolet radiation intensity, avoiding facial feature recognition errors caused by strong ultraviolet radiation at high altitudes.

[0027] Example 5 In fleet management applications, the system interconnects with onboard terminals via a CAN FD bus, uploading driver status data to the cloud platform in real time. An AI behavior analysis module constructs a group driving behavior model; when more than 30% of vehicles in the same fleet detect driver fatigue exceeding a threshold, an automatic area warning mechanism is triggered, pushing rest reminders to nearby service areas. The failure protection mechanism of the emergency braking actuator enters enhanced mode when the fleet is traveling in platoon form. The main controller receives braking signals from vehicles ahead and behind via V2X communication, anticipating braking needs and distributing braking torque in advance to prevent chain-reaction rear-end collisions.

[0028] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An anti-start system, characterized in that: It includes a dynamic face recognition module, an alcohol concentration detection module, an AI behavior analysis module, an emergency braking actuator, and a central data processing unit; The dynamic face recognition module adopts infrared live detection technology, collects the driver's facial video stream through a camera, and extracts iris texture feature values in real time for comparison with a pre-stored biometric database; The alcohol concentration detection module is equipped with a fuel cell type sensor, requiring the driver to blow air into the sampling chamber at a constant flow rate of 0.5L / min for 4 seconds, and the detection accuracy reaches ±0.01mg / L; The AI behavior analysis module is built with a deep learning algorithm, monitors the driver's torso tilt angle θ through a seat pressure sensor array and a gyroscope, and determines that the posture is abnormal when θ exceeds the threshold range (8°≤θ≤15°) continuously for 5 times and is accompanied by a head pitch angular velocity ≥3° / s; The emergency braking actuator includes a dual-redundancy electronic parking system and a hydraulic assist device, and can complete the whole process from triggering the instruction to locking the wheels within 0.8 seconds; The central data processing unit adopts a heterogeneous computing platform, runs a Linux real-time operating system, and uses a CAN FD bus for communication between modules, and the data transmission rate is not less than 5Mbps.

2. The safety start system according to claim 1, characterized in that: When the dynamic face recognition module establishes a three-dimensional face model, it adopts multi-spectral fusion technology, synchronously collects image data in three bands of visible light, near-infrared, and thermal infrared, calculates the facial feature point offset Δx, Δy, Δz through the SIFT feature matching algorithm, and triggers an alarm when Δx²+Δy²+Δz²>threshold; 3. The safety start system according to claim 1, characterized in that: The alcohol concentration detection module has a two-level verification mechanism: The first-level verification uses an electrochemical sensor for rapid screening, and the response time ≤3 seconds; The second-level verification enables chromatographic analysis, separates volatile organic compounds through a capillary column, and qualitatively and quantitatively analyzes the ethanol content by a mass spectrometry detector, and the whole process takes no more than 20 seconds.

4. The safety start system according to claim 1, characterized in that: The AI behavior analysis module constructs a driver state evaluation matrix, including the following dimensions: Physiological indicators: heart rate variability HRV, skin conductivity SC; Behavioral indicators: steering wheel grip force change rate, pedal operation frequency; Environmental indicators: light intensity,车内CO2 concentration; when the comprehensive score is lower than the preset threshold and the duration ≥90 seconds, an intervention program is started.

5. The safety start system according to claim 1, characterized in that: The emergency braking actuator implements a hierarchical braking strategy: In the first stage, apply 30% of the maximum braking pressure and maintain it for 0.5 seconds; In the second stage, increase to 60% and maintain it for 1 second; In the third stage, reach 100% pressure and maintain it until the vehicle stops completely; The conversion nodes of each stage are calculated in real time by the electronic control unit according to the vehicle speed v: when v≤10km / h, skip the first stage; when 10km / h<v≤30km / h, execute according to the standard process; when v>30km / h, directly enter the third stage.

6. The safety start system according to claim 1, characterized in that: The central data processing unit is built with a trusted execution environment TEE, which performs hardware isolation for sensitive operations, and its secure startup process includes: The Bootloader固化 in the ROM verifies the FSBL signature; The FSBL verifies the integrity of the OSBL; The OSBL performs dynamic measurement when loading the Hypervisor; Hypervisor creates a virtual environment that isolates the cockpit and entertainment system; The entire startup process requires encryption and authentication using the national cryptographic algorithm SM4.

7. The safety start system according to claim 1, characterized in that: The system has a self-diagnostic submodule, which is executed upon the first startup each day: a) Zero-point calibration of the alcohol sensor was performed three times using NIST standard gas; b) Camera distortion correction: capture images of a 24-color chart and calculate distortion coefficients k1 and k2; c) Pressure sensor linearity test: Load a 50kg weight to verify the output voltage linearity R²≥0.999; d) CAN bus baud rate test: Send 1000 data packets and the statistical bit error rate is ≤1E-6.

8. The safety start system according to claim 1, characterized in that: The dynamic face recognition module has anti-spoofing attack capabilities and integrates the following anti-counterfeiting measures: Pupil micro-vibration detection: capturing the frequency of pupil diameter change f_pupil; Retinal vascular imaging: analyzing the topology of retinal vessels; Eyelid movement tracking: Record blink interval time t_blink; if any two test results deviate from the normal physiological range, it is judged as non-living.

9. The safety start system according to claim 1, characterized in that: The AI ​​behavior analysis module (103) adopts the spatiotemporal feature fusion network STFNet, and the input layer includes: 3D convolutional layers process time-series video data; 2D convolutional layer analysis of spatially distributed pressure data; LSTM layers capture long-term behavioral patterns; The output layer uses a Softmax classifier to determine three state levels: Level 0 (normal) - confidence > 95%; Level 1 (warning) - 75% ≤ confidence ≤ 95%; Level 2 (dangerous) - confidence < 75%.

10. The safety start system according to claim 1, characterized in that: The emergency braking actuator is equipped with a fail-safe mechanism: when the main controller detects abnormal braking pressure, the mechanical backup system is immediately activated. The solenoid valve switches to the backup oil circuit; Mechanical energy storage spring releases preload; The hydraulic accumulator provides emergency pressure; the entire switching process is completed within 50ms, ensuring braking deceleration ≥6m / s².