Vehicle-mounted alcohol detection and intelligent control device based on multi-sensor fusion

Abstract

The invention relates to the technical field of vehicle safety control, and discloses a vehicle-mounted alcohol detection and intelligent control device based on multi-sensor fusion, which comprises a headrest body, a multi-mode sensing system, a data processing unit and a grading response control unit, wherein the multi-mode sensing system is used for sensing the alcohol state; the data processing unit is used for carrying out fusion processing on data acquired by the multi-modal sensing system and outputting an alcohol concentration detection result and confidence; the hierarchical response control unit is used for executing hierarchical control operation. Through three-mode deep fusion of the fuel cell sensor, the infrared spectrum module and the millimeter wave radar, a triple evidence chain of expiration, blood and physical signs is constructed, and three-dimensional and high-specificity perception of the alcohol state is realized; an improved D-S evidence theory algorithm introducing dynamic weight and time filtering is adopted, sensor local failure or data conflict caused by environmental interference is automatically recognized and restrained, and the detection error rate is greatly increased.

Description

Technical Field

[0001] This invention relates to the field of vehicle safety control technology, specifically to an in-vehicle alcohol detection and intelligent control device based on multi-sensor fusion. Background Technology

[0002] With increasingly stringent road traffic safety regulations and heightened public awareness of safety, in-vehicle alcohol detection technology, as a proactive safety measure to prevent drunk driving accidents, has received widespread attention and development in recent years. However, existing technologies still have significant shortcomings in terms of reliability, practicality, and safety, hindering their large-scale commercial application. The current mainstream technological approaches mainly suffer from the following limitations:

[0003] 1. Passive and easily circumvented detection methods: Most solutions, such as breathalyzers based on steering wheels or seat belts, require the driver to actively blow air to activate the detection. This not only disrupts driving continuity and results in a poor user experience, but also poses a serious risk of cheating (such as using pre-prepared clean air or having a passenger blow air on behalf of the driver), making it impossible to achieve effective passive supervision.

[0004] 2. Detection accuracy is greatly affected by environmental interference: The reliability of a single sensor is insufficient: If only semiconductor (such as MQ-3) or electrochemical fuel cell sensors are used, their response characteristics are easily affected by drastic changes in temperature and humidity in the vehicle environment (-40℃~85℃), resulting in baseline drift and decreased sensitivity. Furthermore, they are prone to false alarms under the interference of volatile organic compounds such as perfumes and cleaning agents.

[0005] Insufficient depth of dual-modal fusion: Existing few fusion solutions combine gas sensors and cameras, but visual detection fails severely in backlight, low light, or when the driver is wearing sunglasses. Essentially, it is still motion-assisted recording and cannot fundamentally improve the specificity of alcohol feature recognition.

[0006] 3. The control strategy is singular and rigid, posing safety hazards: Most devices adopt a "one-size-fits-all" control logic of "locking the vehicle when the detection exceeds the standard", such as forcibly prohibiting driving by cutting off the ignition circuit or prohibiting the engine from starting; although it prevents drunk driving, in sudden emergency situations (such as needing to drive to the hospital for emergency treatment), completely depriving the vehicle of control may cause secondary risks and lacks the flexibility to deal with complex real-world scenarios. Summary of the Invention

[0007] The purpose of this invention is to provide an in-vehicle alcohol detection and intelligent control device based on multi-sensor fusion, so as to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides the following technical solution:

[0009] A vehicle-mounted alcohol detection and intelligent control device based on multi-sensor fusion includes a headrest body, a multimodal sensing system, a data processing unit, and a hierarchical response control unit; wherein,

[0010] The multimodal sensing system is integrated into the headrest body and includes: a fuel cell alcohol sensor for detecting breath alcohol concentration; an infrared spectroscopy module for indirectly measuring blood alcohol concentration; and a millimeter-wave radar for monitoring respiratory rate and vital signs.

[0011] The data processing unit is connected to the multimodal sensing system and is used to perform fusion processing on the data collected by the multimodal sensing system using a fusion algorithm based on the improved DS evidence theory, and output the alcohol concentration detection results and confidence level.

[0012] The graded response control unit is connected to the data processing unit and is used to perform graded control operations based on the alcohol concentration detection results.

[0013] As a further embodiment of the present invention: the data processing unit adopts an improved... The fusion algorithm of evidence theory uses the following method to fuse data:

[0014] S21. Preprocess the multimodal sensing data and convert the preprocessed data into their respective basic probability assignment functions;

[0015] S22. Calculate the conflict coefficient between the evidence from each sensor.

[0016] S23. Assign a dynamic weighting factor to each sensor based on the real-time confidence level of each sensor;

[0017] S24. Based on dynamic weighting factors, an improved method is adopted. The combination rules synthesize the weighted evidence to obtain the initial fusion result;

[0018] S25. Filter the time series of the initial fusion results and output the final alcohol concentration detection results and confidence level.

[0019] As a further aspect of the present invention: in step S21, the specific method for preprocessing the multimodal sensing data and converting it into a basic probability assignment function is as follows:

[0020] S211. Perform temperature and baseline drift compensation on the output data of the fuel cell alcohol sensor; perform ambient light intensity compensation on the output data of the infrared spectroscopy module; perform motion artifact filtering on the output data of the millimeter-wave radar;

[0021] S212. Convert the preprocessed sensor data into basic probability assignment functions on the recognition framework according to the preset correspondence.

[0022] As a further aspect of the present invention: in step S22, the method for calculating the conflict coefficient between the evidence from each sensor is as follows:

[0023] S221. Calculate the relationship between the basic probability assignment functions of each pair of sensors. Distance, to quantify the degree of conflict between the two;

[0024] S222. Calculate a comprehensive conflict coefficient for each sensor, where the comprehensive conflict coefficient is the ratio between the basic probability assignment function of that sensor and the basic probability assignment functions of all other sensors. The average distance.

[0025] As a further aspect of the present invention: in step S23, the method for allocating the dynamic weighting factor of the sensor is as follows:

[0026] S231. Based on the real-time operating status parameters of each sensor, evaluate its corresponding real-time confidence level; wherein, the real-time confidence level of the fuel cell alcohol sensor is determined according to the accuracy drift coefficient caused by the deviation between its operating temperature and the calibration temperature.

[0027] The real-time confidence level of the infrared spectroscopy module is determined based on the ambient light intensity detected by the ambient light sensor, and the confidence level value is negatively correlated with the light intensity.

[0028] The real-time confidence level of millimeter-wave radar is determined based on the signal-to-noise ratio of its echo signal;

[0029] S232. Normalize the real-time confidence scores of all sensors to obtain the dynamic weighting factor for each sensor.

[0030] As a further aspect of the present invention: in step S24, the method for calculating the initial fusion result is as follows:

[0031] S241. Using the dynamic weight factor of each sensor as a discount factor, the probability quality assigned to each proposition in its basic probability assignment function is weighted, and the remaining probability quality is assigned to the entire set of the recognition frame; thus realizing the discount operation on the basic probability assignment function of each sensor.

[0032] S242, Use The combination rule synthesizes the basic probability assignment functions of all discounted sensors to obtain the initial fusion result.

[0033] As a further aspect of the present invention: in step S24, the specific method for outputting the alcohol concentration detection result and confidence level is as follows:

[0034] S251. The initial fusion results at multiple consecutive time points are input as a time series into a filter for processing.

[0035] S252. Output the filtered alcohol concentration state estimate for the current moment as the final alcohol concentration detection result.

[0036] S253. Calculate the confidence level of the final detection result based on the internal state or output error of the filter.

[0037] As a further aspect of the present invention, the strategy for the hierarchical response control unit to perform hierarchical control operations is as follows:

[0038] When the alcohol concentration is within the first threshold range, an early warning control is activated.

[0039] When the alcohol concentration is within the second threshold range, speed limit control is implemented, limiting the vehicle's maximum speed to a preset value;

[0040] Automatic parking control is activated when the alcohol concentration reaches or exceeds the third threshold.

[0041] A method for vehicle-mounted alcohol detection and intelligent control based on multi-sensor fusion includes the following steps:

[0042] S1. Multimodal sensing data acquisition: Through the fuel cell alcohol sensor, infrared spectroscopy module and millimeter-wave radar integrated in the headrest, breath alcohol concentration data, blood alcohol concentration indirect information data and respiratory rate and vital signs parameter data are collected respectively.

[0043] S2. Data fusion processing: Using an improved DS evidence theory fusion algorithm, the multi-source data collected in step S1 is preprocessed, conflict detected, dynamically weighted, and evidence synthesized, and the fused alcohol concentration detection results and confidence level are output.

[0044] S3. Graded response control: Based on the alcohol concentration detection results output in step S2, execute the corresponding graded control operation.

[0045] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0046] This invention constructs a triple evidence chain of breath, blood, and vital signs through the deep fusion of three modes of fuel cell sensors, infrared spectroscopy modules, and millimeter-wave radar, achieving a three-dimensional and highly specific perception of alcohol status.

[0047] An improved DS evidence theory algorithm with the introduction of dynamic weights and time filtering is adopted to automatically identify and suppress sensor local failures or data conflicts caused by environmental interference, which greatly improves the detection error rate.

[0048] It adopts a three-level control logic of warning, speed limit and automatic parking, and implements differentiated management dynamically based on the alcohol concentration threshold; while effectively preventing dangerous driving, it retains the necessary operational leeway and time window for special scenarios such as emergency avoidance, avoiding secondary safety problems caused by excessive control. Attached Figure Description

[0049] Figure 1 This is a schematic diagram of a vehicle-mounted alcohol detection and intelligent control device based on multi-sensor fusion.

[0050] Figure 2 This is a schematic diagram of the framework of an in-vehicle alcohol detection and intelligent control device based on multi-sensor fusion;

[0051] Figure 3 This is a timing diagram of hierarchical control in an in-vehicle alcohol detection and intelligent control device based on multi-sensor fusion;

[0052] Figure 4 This is a flowchart illustrating a method for vehicle-mounted alcohol detection and intelligent control based on multi-sensor fusion.

[0053] In the diagram: 101, fuel cell alcohol sensor; 102, infrared spectroscopy module; 103, millimeter-wave radar; 104, main control board. Detailed Implementation

[0054] Please see Figures 1-3 In this embodiment of the invention, an in-vehicle alcohol detection and intelligent control device based on multi-sensor fusion includes a headrest body, a multimodal sensing system, a data processing unit, and a hierarchical response control unit; wherein,

[0055] The outer shell of the headrest body is made of a material with a density of Slow rebound The material ensures both neck support comfort and provides an acoustically isolated environment for the sensor array; the internally embedded thickness... glass fiber reinforced The skeleton, with its mesh structure optimized through finite element simulation, significantly improves the stiffness of key components, ensuring the mounting flatness accuracy of the multimodal sensing system. .

[0056] A multimodal sensing system is integrated into the headrest body, including: a fuel cell alcohol sensor 101 for detecting breath alcohol concentration, which is sealed with a fluororubber ring. Shore hardness It is fixed below the ventilation hole at the top center of the headrest body; and adopts a labyrinth-style dustproof structure (compliant with...). Protection standards), aperture honeycomb-shaped air intake holes The spacing array arrangement ensures efficient gas flow while also enabling filtration. Particle size Particulate matter 40; measuring range is Sensitivity is Response time Second.

[0057] The infrared spectroscopy module 102, used for indirect measurement of blood alcohol concentration, employs... The inclined installation design incorporates a light source (operating wavelength range of...) and The detectors are symmetrically distributed on the left wing of the headrest body; the beam is focused onto the driver's neck skin area (spot diameter) using an aspherical focusing lens. This enables non-invasive alcohol concentration monitoring.

[0058] And a millimeter-wave radar 103 for monitoring respiratory rate and vital signs, which through A thick magnesium-aluminum alloy shield (with conductive anodized surface treatment) is located on the right wing of the headrest body, operating at a frequency of [frequency band missing]. The antenna beam angle is The ranging range is This allows for multi-dimensional monitoring of the driver's breathing gases, neck skin characteristics, and head position.

[0059] A main control board 104 is installed in the center of the back of the headrest, and the main control board 104 is connected to the sensor via a ribbon cable.

[0060] The data processing unit is connected to the multimodal sensing system and is used to fuse the data collected by the multimodal sensing system using a fusion algorithm based on improved DS evidence theory, and output the alcohol concentration detection result and confidence level; its specific implementation method is as follows:

[0061] S21. Preprocess the multimodal sensing data and convert the preprocessed data into their respective basic probability assignment functions; the specific implementation steps are as follows:

[0062] S211. Perform temperature and baseline drift compensation on the output data of the fuel cell alcohol sensor 101; perform ambient light intensity compensation on the output data of the infrared spectroscopy module 102; and perform motion artifact filtering on the output data of the millimeter-wave radar 103.

[0063] Taking the fuel cell alcohol sensor 101 as an example, assuming that the original alcohol concentration reading corresponding to the output current is 0.078 mg / L, the built-in temperature sensor detects that the chip temperature is 45℃, while the calibration temperature is 25℃;

[0064] According to the temperature-drift characteristic curve of this sensor model, it was found that at 45℃, its reading has a positive drift of +30%.

[0065] Then, the correction calculation is performed: the estimated true concentration. raw readings Drift rate ;

[0066] S212. Convert the preprocessed sensor data into data for the recognition framework according to a preset correspondence. The basic probability assignment function is based on the calibrated experimental data between the readings of each sensor and the alcohol concentration, and the mapping from the physical quantity reading to the basic probability assignment function is realized through the fuzzy membership function.

[0067] Recognition Framework It includes three propositions: "normal driving," "driving under the influence of alcohol," and "driving while intoxicated"; and is defined as follows: Normal driving, Drunk driving Drunk driving

[0068] Suppose the vehicle is driving at midday in summer, under strong sunlight; at a certain moment, the three sensors... After preliminary signal processing, the output yields the following preliminary data:

[0069] Fuel Cell Alcohol Sensor 101 Output current corresponds to alcohol concentration reading However, the sensor has a high internal temperature, which may cause slight drift.

[0070] Infrared spectroscopy module 102 Analyzing the reflectance spectrum of the driver's neck, the blood alcohol concentration reading was obtained. However, strong ambient light may interfere with the spectral signal.

[0071] Millimeter-wave radar 103 The equivalent concentration for determining the degree of alcohol effect was determined through respiratory rate and heart rate variability analysis. The radar signal quality is good.

[0072] First, the raw readings are filtered (to remove electrical noise) and normalized (mapped to a uniform scale); then, a basic probability assignment function is generated based on the pre-calibrated reading-confidence model for each sensor. The details are as follows:

[0073] of Because the reading is (exceeding the legal limit for drunk driving) The confidence levels for normal driving, drunk driving, and driving under the influence of alcohol are as follows: ;

[0074] of The reading is This is at an extremely low level, and is allocated as: m ;

[0075] of The reading was 0.10 mg / L, which is considered a high-risk level. The allocation was as follows: ;

[0076] It can be concluded that sensor data with different physical dimensions can be transformed into a basic probability assignment function that can be compared and calculated within the same probabilistic framework (identification framework Ω). );

[0077] S22. Calculate the conflict coefficient between evidence from various sensors (such as fuel cell alcohol sensor 101, infrared spectroscopy module 102, and millimeter-wave radar 103); the specific implementation method is as follows:

[0078] S221. Calculate the relationship between the basic probability assignment functions of each pair of sensors. Distance, to quantify the degree of conflict between the two; distance ;in, and The first and the The basic probability assignment function vector of each sensor For identification framework The number of single-element propositions, for example, 3; Indicates the first The sensor for the first The basic probability quality of each proposition assignment; for example... Corresponding to the propositions , , ;

[0079] S222. Calculate a comprehensive conflict coefficient for each sensor, where the comprehensive conflict coefficient is the ratio between the basic probability assignment function of that sensor and the basic probability assignment functions of all other sensors. The average distance;

[0080] Assuming the confidence level is calculated based on the steps above, (Right now: , , ; ; ;

[0081] Then, calculate distance:

[0082] ;

[0083] ;

[0084] ;

[0085] Assuming a conflict coefficient threshold of 0.6, the conflict analysis is shown in Table 1 below;

[0086] Table 1 Conflict Analysis Table

[0087] Evidence against distance Comparison with threshold Conflict determination 0.687 ＞0.6 High conflict 0.557 ＜0.6 Low conflict (not exceeding the threshold) 0.805 ＞0.6 High conflict

[0088] Next, calculate the overall conflict coefficient of each sensor:

[0089] Conflict coefficient ;

[0090] Conflict coefficient ;

[0091] Conflict coefficient ;

[0092] We can conclude that: The conflict coefficient of (infrared spectrum) was the highest (0.746), indicating that its evidence (strongly supporting "normal driving") seriously contradicts the evidence from the other two sensors (both supporting "drunk driving"); this is likely due to data distortion caused by strong ambient light interference.

[0093] (Fuel Cells) and The conflict coefficients of the two (millimeter-wave radar 103) are relatively low and close, indicating that the evidence from both is somewhat consistent (both point to the influence of alcohol), but the specific degree (driving under the influence) differs. The discrepancies in the definition of drunk driving are within a reasonable range.

[0094] This conflict coefficient This will be directly used for the next step of dynamic weight allocation: the weight of sensors with high conflict coefficients should be reduced.

[0095] S23. Based on the real-time confidence level of each sensor, assign a dynamic weighting factor to each sensor; the specific implementation steps are as follows:

[0096] S231. Based on the real-time operating status parameters of each sensor, evaluate its corresponding real-time confidence level; wherein, the real-time confidence level of the fuel cell alcohol sensor 101 is determined according to the accuracy drift coefficient caused by the deviation between its operating temperature and the calibration temperature.

[0097] The real-time confidence level of the infrared spectroscopy module 102 is determined based on the ambient light intensity detected by the ambient light sensor, and the confidence level value is negatively correlated with the light intensity.

[0098] The real-time confidence level of the millimeter-wave radar 103 is determined based on the signal-to-noise ratio of its echo signal.

[0099] S232. Normalize the real-time confidence scores of all sensors to obtain the dynamic weighting factor for each sensor; where, the dynamic weighting factor... Calculated using the following formula: ;in, For the first Real-time confidence level of each sensor, This represents the total number of sensors;

[0100] Assume that the fuel cell alcohol sensor 101 The internal temperature was found to be too high, and the calculated accuracy drift coefficient was 0.8.

[0101] Infrared spectroscopy module 102 Strong light was detected, and according to the preset model, its confidence level was reduced to 0.4.

[0102] Millimeter-wave radar 103 The radar signal signal-to-noise ratio (SNR) is excellent, with a confidence level of 0.9.

[0103] Normalization yields dynamic weights:

[0104] Original confidence level and: ;

[0105] Weight: ; ;

[0106] It can be concluded that the weight allocation accurately reflects the sensor reliability in the current scenario: the infrared spectral module 102 is susceptible to interference from strong light. The weighting was significantly reduced; the millimeter-wave radar 103 had stable performance. The fuel cell alcohol sensor 101 has the highest weighting; it is susceptible to drift. Appropriate weighting;

[0107] S24. Based on dynamic weighting factors, an improved method is adopted. The combination rule synthesizes the weighted evidence to obtain an initial fusion result; the specific implementation steps are as follows:

[0108] S241. Using the dynamic weighting factor of each sensor as a discount factor, weight the probability quality assigned to each proposition in its basic probability assignment function, and assign the remaining probability quality to the entire set of the recognition frame; implement the discount operation on the basic probability assignment function of each sensor, wherein the discounted basic probability assignment function satisfies the following condition:

[0109] ;

[0110] ;

[0111] in, For the first Dynamic weighting factors for each sensor Assign a function to the original basic probabilities. The function is assigned to the basic probability after discount; Ω is the complete set of recognition frames;

[0112] S242, Use The combination rule synthesizes the basic probability assignment functions of all discounted sensors to obtain the initial fusion result;

[0113] Assume that the basic probability assignment function for each sensor is... Weighting is performed, and the remaining probabilities are allocated to the uncertain terms. ;

[0114] for The new confidence level allocation after weighting normal driving, drunk driving, and driving under the influence of alcohol is as follows: , , Uncertainty Confidence level: ;

[0115] for , , , , ;

[0116] for , , , , ;

[0117] Reuse Rules are synthesized sequentially , , The specific calculations are as follows:

[0118] First, calculate the collision quality.

[0119] ;

[0120] but, and Synthesis, intermediate results obtained The details are as follows:

[0121]

[0122] ;

[0123]

[0124] ;

[0125]

[0126]

[0127] ;

[0128] Next, the conflict quality results are calculated.

[0129] ;

[0130] but, and Synthesize to obtain the final synthetic result. The details are as follows:

[0131]

[0132] ;

[0133]

[0134] ;

[0135]

[0136]

[0137] ;

[0138] It can be concluded that: drunk driving The probability is 0.25 for drunk driving. The probability is 0.30, and the sum of the two is 0.55, indicating that there is more than a 50% chance that the driver is under the influence of alcohol.

[0139] However, at the same time, uncertainties The probability is as high as 0.35, which reflects significant conflict among sensor evidence (especially...). (Compared to other sensors), the fusion system provides an honest measurement of this;

[0140] Compared to the unmodified direct synthesis, the improved method did not suffer from [the following issues]: The strong opposition (normal) completely negated the possibility of drunk driving, and did not because of... The strong support for (drunk driving) led to an arbitrary determination of drunk driving; the result is more reasonable and stable.

[0141] S25. Filter the time series of the initial fusion results to output the final alcohol concentration detection results and confidence levels; the specific implementation steps are as follows:

[0142] S251. The initial fusion results at multiple consecutive time points are input as a time series into a filter for processing; wherein, the filter is a Kalman filter, and the probability value or scalar estimate of alcohol concentration for the drunk driving proposition in the initial fusion results constitutes a time series.

[0143] For example, the probability estimation sequence for the proposition C regarding drunk driving. ;

[0144] Scalar estimates of alcohol concentration ;

[0145] S252. Output the filtered alcohol concentration state estimate for the current moment as the final alcohol concentration detection result.

[0146] S253. Calculate the confidence level of the final detection result based on the internal state or output error of the filter; wherein, the confidence level is calculated based on the estimation error covariance of the Kalman filter output;

[0147] Suppose the input sequence is: arrive Moment : It can be seen that, in and An unusual spike appeared at a certain moment. and It may be caused by momentary air disturbances or electromagnetic pulse interference.

[0148] The output sequence after Kalman filtering: Abnormal spikes are effectively smoothed; Final output: Current time. The filtered probability of drunk driving is The confidence level is calculated based on the filter residuals;

[0149] It can be concluded that time series filtering effectively suppresses random impulse interference in the single fusion result, and the output... Compared to the original value It better represents a stable trend; assuming the system's automatic parking threshold is set to... The current result This will prevent the extreme control from being triggered, avoiding malfunctions caused by momentary interference and improving system stability.

[0150] The graded response control unit is connected to the data processing unit and is used to execute graded control operations based on the alcohol concentration detection results; for example... Figure 3 As shown, the execution strategy for hierarchical control operations is as follows:

[0151] When the alcohol concentration is within the first threshold range, an early warning control is implemented; for example, the first threshold range is 0.02–0.05 mg / L.

[0152] When the blood alcohol concentration is within the second threshold range, speed limit control is implemented, limiting the vehicle's maximum speed to a preset value; for example, the second threshold range is 0.05~0.08mg / L.

[0153] When the alcohol concentration reaches or exceeds the third threshold, automatic parking control is activated, such as when the third threshold is ≥0.08mg / L.

[0154] Please see Figure 4 In this embodiment of the invention, a method for vehicle-mounted alcohol detection and intelligent control based on multi-sensor fusion includes the following steps:

[0155] S1. Multimodal sensing data acquisition: Through the fuel cell alcohol sensor 101, infrared spectroscopy module 102 and millimeter-wave radar 103 integrated in the headrest, breath alcohol concentration data, blood alcohol concentration indirect information data and respiratory rate and vital signs parameter data are collected respectively.

[0156] S2. Data fusion processing: Using an improved DS evidence theory fusion algorithm, the multi-source data collected in step S1 is preprocessed, conflict detected, dynamically weighted, and evidence synthesized, and the fused alcohol concentration detection results and confidence level are output.

[0157] S3. Graded Response Control: Based on the alcohol concentration detection result output in step S2, execute the corresponding graded control operation; the logic of the graded control operation is as follows:

[0158] If the alcohol concentration remains in the range of 0.02 to 0.05 mg / L for a first predetermined time, an early warning control will be triggered, including voice prompts and dashboard warnings.

[0159] If the blood alcohol concentration is in the range of 0.05 to 0.08 mg / L and continues for a second predetermined time, speed limit control is triggered, and the vehicle speed is limited by the vehicle controller (ECU), such as limiting the maximum speed to 20 km / h.

[0160] If the blood alcohol concentration is ≥0.08mg / L and remains so for a third predetermined time, automatic parking control will be triggered to slow down the vehicle and bring it to a safe stop.

[0161] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A vehicle-mounted alcohol detection and intelligent control device based on multi-sensor fusion, characterized in that, It includes a headrest body, a multimodal sensing system, a data processing unit, and a hierarchical response control unit; among which, The multimodal sensing system is integrated into the headrest body and includes: a fuel cell alcohol sensor for detecting breath alcohol concentration; an infrared spectroscopy module for indirectly measuring blood alcohol concentration; and a millimeter-wave radar for monitoring respiratory rate and vital signs. The data processing unit is connected to the multimodal sensing system and is used to perform fusion processing on the data collected by the multimodal sensing system using a fusion algorithm based on the improved DS evidence theory, and output the alcohol concentration detection results and confidence level. The graded response control unit is connected to the data processing unit and is used to perform graded control operations based on the alcohol concentration detection results.

2. The in-vehicle alcohol detection and intelligent control device based on multi-sensor fusion according to claim 1, characterized in that, The data processing unit adopts an improved The fusion algorithm of evidence theory uses the following method to fuse data: S21. Preprocess the multimodal sensing data and convert the preprocessed data into their respective basic probability assignment functions; S22. Calculate the conflict coefficient between the evidence from each sensor. S23. Assign a dynamic weighting factor to each sensor based on the real-time confidence level of each sensor; S24. Based on dynamic weighting factors, an improved method is adopted. The combination rules synthesize the weighted evidence to obtain the initial fusion result; S25. Filter the time series of the initial fusion results and output the final alcohol concentration detection results and confidence level.

3. The in-vehicle alcohol detection and intelligent control device based on multi-sensor fusion according to claim 2, characterized in that, In step S21, the specific method for preprocessing the multimodal sensing data and converting it into a basic probability assignment function is as follows: S211. Perform temperature and baseline drift compensation on the output data of the fuel cell alcohol sensor; perform ambient light intensity compensation on the output data of the infrared spectroscopy module; perform motion artifact filtering on the output data of the millimeter-wave radar; S212. Convert the preprocessed sensor data into basic probability assignment functions on the recognition framework according to the preset correspondence.

4. The in-vehicle alcohol detection and intelligent control device based on multi-sensor fusion according to claim 2, characterized in that, In step S22, the method for calculating the conflict coefficient between the sensor evidence is as follows: S221. Calculate the relationship between the basic probability assignment functions of each pair of sensors. Distance, to quantify the degree of conflict between the two; S222. Calculate a comprehensive conflict coefficient for each sensor, where the comprehensive conflict coefficient is the ratio between the basic probability assignment function of that sensor and the basic probability assignment functions of all other sensors. The average distance.

5. The in-vehicle alcohol detection and intelligent control device based on multi-sensor fusion according to claim 2, characterized in that, In step S23, the method for allocating the dynamic weighting factors of the sensor is as follows: S231. Based on the real-time operating status parameters of each sensor, evaluate its corresponding real-time confidence level; wherein, the real-time confidence level of the fuel cell alcohol sensor is determined according to the accuracy drift coefficient caused by the deviation between its operating temperature and the calibration temperature. The real-time confidence level of the infrared spectroscopy module is determined based on the ambient light intensity detected by the ambient light sensor, and the confidence level value is negatively correlated with the light intensity. The real-time confidence level of millimeter-wave radar is determined based on the signal-to-noise ratio of its echo signal; S232. Normalize the real-time confidence scores of all sensors to obtain the dynamic weighting factor for each sensor.

6. The in-vehicle alcohol detection and intelligent control device based on multi-sensor fusion according to claim 2, characterized in that, In step S24, the initial fusion result is calculated as follows: S241. Using the dynamic weighting factor of each sensor as a discount factor, the probability quality assigned to each proposition in its basic probability assignment function is weighted, and the remaining probability quality is assigned to the entire set of the recognition frame; thus realizing the discount operation on the basic probability assignment function of each sensor. S242, Use The combination rule synthesizes the basic probability assignment functions of all discounted sensors to obtain the initial fusion result.

7. The in-vehicle alcohol detection and intelligent control device based on multi-sensor fusion according to claim 2, characterized in that, In step S24, the specific method for outputting the alcohol concentration detection results and confidence level is as follows: S251. The initial fusion results at multiple consecutive time points are input as a time series into a filter for processing. S252. Output the filtered alcohol concentration state estimate for the current moment as the final alcohol concentration detection result. S253. Calculate the confidence level of the final detection result based on the internal state or output error of the filter.

8. The in-vehicle alcohol detection and intelligent control device based on multi-sensor fusion according to claim 1, characterized in that, The strategy for the hierarchical response control unit to perform hierarchical control operations is as follows: When the alcohol concentration is within the first threshold range, an early warning control is activated. When the alcohol concentration is within the second threshold range, speed limit control is implemented, limiting the vehicle's maximum speed to a preset value; Automatic parking control is activated when the alcohol concentration reaches or exceeds the third threshold.

9. A control method for implementing the multi-sensor fusion-based vehicle alcohol detection and intelligent control device as described in claims 1-8, characterized in that, Includes the following steps: S1. Multimodal sensing data acquisition: Through the fuel cell alcohol sensor, infrared spectroscopy module and millimeter-wave radar integrated in the headrest, breath alcohol concentration data, blood alcohol concentration indirect information data and respiratory rate and vital signs parameter data are collected respectively. S2. Data fusion processing: Using an improved DS evidence theory fusion algorithm, the multi-source data collected in step S1 is preprocessed, conflict detected, dynamically weighted, and evidence synthesized, and the fused alcohol concentration detection results and confidence level are output. S3. Graded response control: Based on the alcohol concentration detection results output in step S2, execute the corresponding graded control operation.

Images