Non-contact drunk driving detection method and system adopting large-linewidth tunable laser

Abstract

The invention discloses a non-contact drunk driving detection method and system adopting a large-linewidth tunable laser, and relates to the technical field of laser detection, and the method comprises the steps: collecting transmission light signals and corresponding electric signals under different linewidth conditions through experiments, synchronously recording experiment parameters, obtaining an original data set, carrying out the preprocessing of the original data set, and carrying out the detection of drunk driving. Generating a training set, a verification set and a test set, and training a BP (Back Propagation) neural network to obtain a pre-trained noise suppression model; the method comprises the following steps of: generating a wavelength-tunable modulated laser beam by adopting a large-linewidth tunable laser, directionally projecting the modulated laser beam into a vehicle cab, and enabling laser to penetrate through the cab space and carry alcohol gas concentration information to obtain a transmission light signal; focusing the transmission light signal to a photoelectric detector through an optical convergence device to obtain a noisy electric signal, inputting the noisy electric signal into a pre-trained noise suppression model, and outputting a de-noised electric signal; performing synchronous demodulation and harmonic analysis on the de-noised electric signal by using a phase-locked amplification detection technology, extracting a second harmonic signal amplitude associated with the modulation frequency, and determining a second harmonic signal characteristic amplitude through peak detection; according to the calibration relation between the second harmonic signal characteristic amplitude and the alcohol gas concentration, the concentration value of the alcohol gas in the cab is calculated in an inversion mode through a least square fitting algorithm, an alcohol concentration threshold value is set, the drunk driving condition of a driver is judged, a detection result is generated, and an alarm signal is output.

Description

Technical Field

[0001] The present invention relates to the technical field of laser detection, and in particular to a non-contact drunk driving detection method and system using a large linewidth tunable laser. Background Art

[0002] In recent years, non-contact detection technologies based on tunable diode laser absorption spectroscopy (TDLAS) have gradually emerged. They achieve remote detection by modulating the laser wavelength to scan the gas absorption peak. However, existing systems rely on ultra-narrow linewidth lasers (2 - 5 MHz), which are costly and susceptible to frequency drift caused by temperature and environmental perturbations, requiring additional temperature control devices, further driving up costs and reducing practicality.

[0003] Due to the high cost and low stability of ultra-narrow linewidth lasers in existing TDLAS solutions, it is difficult to achieve large-scale applications. On the one hand, narrow linewidth lasers have stringent requirements for the stability of optical devices and need frequent calibration and maintenance. On the other hand, environmental factors (such as temperature fluctuations and air turbulence) are likely to cause laser frequency drift, leading to linewidth broadening and a decline in detection accuracy. How to reduce hardware costs while overcoming the interference of linewidth broadening to signal quality has become the core challenge in enhancing the practicality of non-contact drunk driving detection technology. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a non-contact drunk driving detection method using a large linewidth tunable laser to solve the problems of high cost and environmental sensitivity in the prior art.

[0006] To solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a non-contact drunk driving detection method using a large linewidth tunable laser, which includes collecting transmitted light signals and corresponding electrical signals under different linewidth conditions through experiments, synchronously recording experimental parameters to obtain an original data set, and preprocessing the original data set to generate a training set, a validation set, and a test set, training a BP neural network to obtain a pre-trained noise suppression model;

[0008] Generating a wavelength-tunable modulated laser beam using a large linewidth tunable laser, directing the modulated laser beam into the vehicle cab, enabling the laser to penetrate the cab space and carry alcohol gas concentration information, and obtaining a transmitted light signal;

[0009] Focusing the transmitted light signal onto a photodetector through an optical focusing device to obtain a noisy electrical signal, and inputting the noisy electrical signal into the pre-trained noise suppression model to output a denoised electrical signal;

[0010] Using the lock-in amplification detection technology, synchronously demodulate and perform harmonic analysis on the denoised electrical signal, extract the amplitude of the second harmonic signal associated with the modulation frequency, and determine the characteristic amplitude of the second harmonic signal through peak detection;

[0011] According to the calibration relationship between the characteristic amplitude of the second harmonic signal and the alcohol gas concentration, use the least squares fitting algorithm to inversely calculate the concentration value of the alcohol gas in the cab, set the alcohol concentration threshold, judge the drunk driving situation of the driver, generate the detection result and output an alarm signal.

[0012] As a preferred scheme of the non-contact drunk driving detection method using a large linewidth tunable laser according to the present invention, wherein: the experimental parameters include laser parameters, ambient temperature, and calibrated alcohol concentration;

[0013] The preprocessing includes performing initial noise screening, normalization processing, and proportional division;

[0014] The initial noise screening is to identify and filter out abnormal data points and high-frequency noise interference in the original data set through statistical analysis methods, and retain the effective signal characteristics;

[0015] The normalization processing is to map the effective signal characteristics to a preset numerical range;

[0016] The proportional division is to divide the normalized data set into a training set, a validation set, and a test set according to a standardized ratio.

[0017] As a preferred scheme of the non-contact drunk driving detection method using a large linewidth tunable laser according to the present invention, wherein: training the BP neural network to obtain a pre-trained noise suppression model, and the specific steps are as follows,

[0018] Based on the dimension of the training set, randomly initialize the weight parameters of the BP neural network, and construct the network topology structure of the input layer, hidden layer, and output layer;

[0019] Input the noisy electrical signal in the training set into the BP neural network, calculate the predicted value of the output layer through forward propagation, and compare it with the true value of the denoised electrical signal in the validation set, and calculate the prediction error based on a preset loss function;

[0020] Through the backpropagation algorithm, use the prediction error to adjust the weight parameters of the neural network layer by layer, and optimize the mapping relationship of noise suppression;

[0021] Iteratively execute the forward propagation and backpropagation processes, and evaluate the model performance in real time in combination with the validation set. When the prediction error converges to a preset threshold, terminate the training;

[0022] Use the test set to verify the generalization performance of the trained noise suppression model;

[0023] Save the neural network weight parameters that meet the generalization requirements to form a pre-trained noise suppression model.

[0024] As a preferred solution of the non-contact drunk driving detection method using a large linewidth tunable laser according to the present invention, wherein: a wavelength tunable modulated laser beam is generated by a large linewidth tunable laser, and the modulated laser beam is directionally projected into the vehicle cab, so that the laser penetrates the cab space and carries alcohol gas concentration information to obtain a transmitted light signal. The specific steps are as follows.

[0025] Based on the modulation signal generating unit, a superimposed modulation signal of a low-frequency sawtooth wave signal and a high-frequency sine wave signal is generated, and is loaded onto the large linewidth tunable laser through the laser driving control unit to drive the laser to output a wavelength tunable modulated laser beam.

[0026] The modulated laser beam is directionally projected into the cab through the vehicle side window. Utilizing the absorption characteristics of alcohol gas for a specific wavelength laser, the laser penetrates the cab space and undergoes an absorption effect with alcohol gas molecules to form a transmitted light signal carrying concentration information.

[0027] By adjusting the injection current of the laser, the laser wavelength is periodically scanned to cover the characteristic absorption spectral line of alcohol gas.

[0028] The transmitted light signal penetrating the cab is received by an optical focusing device.

[0029] As a preferred solution of the non-contact drunk driving detection method using a large linewidth tunable laser according to the present invention, wherein: the transmitted light signal is focused onto a photodetector through an optical focusing device to obtain a noisy electrical signal, and is input into a pre-trained noise suppression model to output a denoised electrical signal. The specific steps are as follows.

[0030] The transmitted light signal penetrating the cab is received by an optical focusing device, and the transmitted light signal is focused onto the photosensitive area of the photodetector by an optical lens group to complete the conversion of the optical signal to an electrical signal, generating a noisy electrical signal including the laser linewidth noise and environmental interference.

[0031] The noisy electrical signal is subjected to preliminary filtering processing to suppress the high-frequency noise components and retain the low-frequency signal characteristics related to the absorption of alcohol gas.

[0032] The filtered noisy electrical signal is input into a pre-trained noise suppression model, and through the non-linear mapping relationship of the neural network module, the noise component and the effective signal component are separated to generate a pure denoised electrical signal.

[0033] As a preferred embodiment of the non-contact drunk driving detection method using a large linewidth tunable laser according to the present invention, the following steps are included: using lock-in amplification detection technology, synchronously demodulating and performing harmonic analysis on the denoised electrical signal, extracting the amplitude of the second harmonic signal associated with the modulation frequency, and determining the characteristic amplitude of the second harmonic signal through peak detection. The specific steps are as follows:

[0034] Based on the reference signal generated by the modulation signal generation unit, synchronously demodulate the denoised electrical signal through a bi-phase lock-in amplifier to separate the fundamental frequency signal and the second harmonic signal synchronized with the modulation frequency.

[0035] Use a band-pass filter to screen the frequency band of the demodulated signal, retain the target band signal containing the second harmonic component, and filter out the out-of-band noise interference.

[0036] Calculate the amplitude of the second harmonic signal in the target frequency band through a harmonic amplitude extraction module, determine the characteristic amplitude at the center position of the absorption spectrum line based on the peak detection algorithm, and compare the characteristic amplitude with the preset calibration curve of the alcohol gas concentration.

[0037] As a preferred embodiment of the non-contact drunk driving detection method using a large linewidth tunable laser according to the present invention, the following steps are included: according to the calibration relationship between the characteristic amplitude of the second harmonic signal and the alcohol gas concentration, use the least squares fitting algorithm to inversely calculate the concentration value of the alcohol gas in the driver's cab, set an alcohol concentration threshold, determine the drunk driving situation of the driver, generate a detection result and output an alarm signal. The specific steps are as follows:

[0038] Using pre-experiment calibration data, establish a non-linear mapping relationship between the characteristic amplitude of the second harmonic signal and the alcohol gas concentration through the least squares fitting algorithm to generate a concentration calibration curve, and the expression is:

[0039] C = a 1 X 2 + a 2 X + a 3 ;

[0040] Where C is the alcohol gas concentration value, a 1 is the quadratic term fitting coefficient, a 2 is the linear term fitting coefficient, a 3 is the constant term fitting coefficient, and X is the second harmonic amplitude.

[0041] Input the characteristic amplitude of the second harmonic signal detected in real time into the concentration calibration curve, and inversely calculate the real-time concentration value of the alcohol gas in the driver's cab through the fitting equation.

[0042] Set an alcohol concentration determination threshold according to the legal drunk driving standard, compare the real-time concentration value with the alcohol concentration determination threshold, and if it exceeds the alcohol concentration determination threshold, it is determined as a drunk driving behavior.

[0043] Generate a detection report containing alcohol concentration values and determination results.

[0044] In a second aspect, the present invention provides a non-contact drunk driving detection system using a large linewidth tunable laser, including a pre-training module, a modulated laser module, a denoising module, an amplitude acquisition module, and a detection output module;

[0045] The pre-training module is used to collect transmitted light signals and corresponding electrical signals under different linewidth conditions through experiments, synchronously record experimental parameters, obtain an original data set, preprocess the original data set, generate a training set, a validation set, and a test set, train a BP neural network, and obtain a pre-trained noise suppression model;

[0046] The modulated laser module is used to generate a wavelength-tunable modulated laser beam using a large linewidth tunable laser, direct the modulated laser beam into the vehicle cab, enable the laser to penetrate the cab space and carry alcohol gas concentration information, and obtain a transmitted light signal;

[0047] The denoising module is used to focus the transmitted light signal onto a photodetector through an optical focusing device to obtain a noisy electrical signal, and input it into the pre-trained noise suppression model to output a denoised electrical signal;

[0048] The amplitude acquisition module is used to perform synchronous demodulation and harmonic analysis on the denoised electrical signal using lock-in amplifier detection technology, extract the amplitude of the second harmonic signal associated with the modulation frequency, and determine the characteristic amplitude of the second harmonic signal through peak detection;

[0049] The detection output module is used to invert and calculate the concentration value of alcohol gas in the cab according to the calibration relationship between the characteristic amplitude of the second harmonic signal and the alcohol gas concentration, set an alcohol concentration threshold, judge the drunk driving situation of the driver, generate a detection result, and output an alarm signal.

[0050] In a third aspect, the present invention provides a computer device, including a memory and a processor, where the memory stores a computer program, and: when the computer program is executed by the processor, any step of the non-contact drunk driving detection method using a large linewidth tunable laser as described in the first aspect of the present invention is implemented.

[0051] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and: when the computer program is executed by the processor, any step of the non-contact drunk driving detection method using a large linewidth tunable laser as described in the first aspect of the present invention is implemented.

[0052] The beneficial effects of the present invention are as follows: Through the collaborative application of a large linewidth tunable laser and BP neural network denoising technology, the present invention significantly reduces the system hardware cost, effectively suppresses the influence of linewidth broadening and environmental interference on the detection accuracy, and achieves high-precision alcohol concentration detection equivalent to that of an ultra-narrow linewidth laser. Through the non-linear noise suppression ability of the neural network, the system can operate stably without a complex temperature control device, adapt to complex environmental conditions, and support vehicle dynamic passing detection, significantly improving law enforcement efficiency. In addition, the non-contact detection method avoids traffic interference and privacy disputes of traditional methods, providing an innovative solution for drunk driving screening with low cost, high reliability, and easy large-scale deployment. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can obtain other drawings based on these drawings without creative efforts.

[0054] Figure 1 It is a hardware module diagram of the non-contact drunk driving detection method using a large linewidth tunable laser in Embodiment 1.

[0055] Figure 2 It is a model acquisition diagram of the non-contact drunk driving detection method using a large linewidth tunable laser in Embodiment 1.

[0056] Figure 3 It is a diagram of the second harmonic amplitude signal before denoising of the non-contact drunk driving detection method using a large linewidth tunable laser in Embodiment 1.

[0057] Figure 4 It is a diagram of the second harmonic amplitude signal after denoising of the non-contact drunk driving detection method using a large linewidth tunable laser in Embodiment 1.

[0058] Figure 5 It is a diagram of the retrieved alcohol gas concentration without denoising of the non-contact drunk driving detection method using a large linewidth tunable laser in Embodiment 1.

[0059] Figure 6 It is a diagram of the retrieved alcohol gas concentration after denoising of the non-contact drunk driving detection method using a large linewidth tunable laser in Embodiment 1. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0060] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the following detailed description of the specific embodiments of the present invention will be made in conjunction with the accompanying drawings of the specification.

[0061] In the following description, numerous specific details are set forth to provide a thorough understanding of the present invention. However, the present invention may be practiced in other ways than those specifically described herein. Those skilled in the art can make similar generalizations without departing from the spirit of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0062] Secondly, as used herein, an "embodiment" or "an embodiment" refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The appearances of "in an embodiment" in different places in this specification do not all refer to the same embodiment, nor are they separate or alternative embodiments that exclude each other.

[0063] Example 1, referring to Figure 1 and Figure 2 , is the first embodiment of the present invention. This embodiment provides a non-contact drunk driving detection method using a large linewidth tunable laser, including the following steps:

[0064] S1. Collect transmitted light signals and corresponding electrical signals under different linewidth conditions through experiments, synchronously record experimental parameters to obtain an original data set, and preprocess the original data set to generate a training set, a validation set, and a test set. Train a BP neural network to obtain a pre-trained noise suppression model, including the following steps:

[0065] The experimental parameters include laser parameters, environmental temperature, and calibrated alcohol concentration. Specifically, the linewidth range of the laser parameters is 3 MHz to 2 GHz (corresponding to a wavelength change of 0.11 pm to 0.074 nm), the central wavelength is 3345 nm (the alcohol characteristic absorption peak), and the output power is 10 to 100 mW. The environmental temperature is -20°C to 50°C (covering extreme environmental working conditions). The calibrated alcohol concentration is 20 ppm to 200 ppm (corresponding to the range from drunk driving to drunk driving in violation of the law).

[0066] The initial noise screening is to identify and filter out abnormal data points and high-frequency noise interference in the original data set through statistical analysis methods, and retain the effective signal characteristics. Specifically, abnormal points in the original data are identified and filtered out through the 3σ criterion (triple standard deviation) (such as amplitudes exceeding the mean ± 3σ range); a low-pass digital filter with a cut-off frequency of 50 kHz is used to filter out high-frequency noise (such as the thermal noise of the photodetector), and the low-frequency signal related to alcohol absorption (0 to 20 kHz) is retained.

[0067] The normalization process is to map the effective signal characteristics to a preset numerical range. Specifically, the signal amplitude is mapped to the [0, 1] interval using min-max normalization (Min-Max Scaling). The environmental temperature parameter is standardized using Z-score (mean = 0, standard deviation = 1).

[0068] The normalized dataset is divided into a training set, a validation set, and a test set according to a standardized ratio. Specifically, it is divided in the ratio of 70% training set, 15% validation set, and 15% test set; the training set covers the entire line width (3 MHz to 2 GHz) and temperature range (-20 °C to 50 °C) to ensure the generalization of the model.

[0069] Based on the dimensions of the training set, randomly initialize the weight parameters of the BP neural network and construct the network topology of the input layer, hidden layer, and output layer. Specifically, the network structure: input layer (1000 nodes, corresponding to the number of electrical signal sampling points), 3 hidden layers (each layer has 512 nodes, ReLU activation function), output layer (1000 nodes, linear activation); the loss function uses mean squared error (MSE); the optimizer uses the Adam algorithm (learning rate 0.001, batch size 256); the convergence condition is that the change rate of the validation set loss function for 10 consecutive iterations < 0.1% or the maximum training cycle is 500; save the weights when the relative error of the test set ≤ 5.5% (for example, the error is 1.75% when the line width is 300 MHz).

[0070] Input the noisy electrical signals in the training set into the BP neural network, calculate the predicted values of the output layer through forward propagation, and compare them with the true values of the denoised electrical signals in the validation set. Calculate the prediction error based on the preset loss function. Specifically, the loss function uses mean squared error (MSE). The optimizer uses the Adam algorithm (learning rate 0.001, batch size 256).

[0071] Through the backpropagation algorithm, use the prediction error to adjust the weight parameters of the neural network layer by layer and optimize the mapping relationship of noise suppression. Specifically, the convergence condition is that the change rate of the validation set loss function for 10 consecutive iterations < 0.1% or the maximum training cycle is 500.

[0072] Iteratively execute the forward propagation and backpropagation processes, and evaluate the model performance in real-time in combination with the validation set. Terminate the training when the prediction error converges to the preset threshold. Specifically, save the weights when the relative error of the test set ≤ 5.5% (for example, the error is 1.75% when the line width is 300 MHz).

[0073] It should be noted that through the full-parameter coverage experiment, ensure the representativeness and generalization of the model training dataset, and avoid the "cold start" problem in actual deployment. Through the combination of physical noise characteristics and statistical rules, achieve the preliminary purification of signal features and reduce the complexity of subsequent neural network training. Through strict verification processes and error control, improve the detection accuracy of large line width lasers to a level comparable to that of traditional ultra-narrow line width systems, and verify the feasibility of the technical solution.

[0074] S2. Use a tunable wide-linewidth laser to generate a wavelength-tunable modulated laser beam, and direct the modulated laser beam into the vehicle cab so that the laser penetrates the cab space and carries alcohol gas concentration information to obtain a transmitted light signal. The steps are as follows:

[0075] Based on the modulation signal generation unit, generate a superimposed modulation signal of a low-frequency sawtooth wave signal and a high-frequency sine wave signal, and load it onto the tunable wide-linewidth laser through the laser driver control unit to drive the laser to output a wavelength-tunable modulated laser beam. Specifically, the modulation signal generation unit generates a superimposed signal of a 10 Hz low-frequency sawtooth wave (wavelength scanning) and a 10 kHz high-frequency sine wave (frequency modulation), and loads the superimposed signal onto the tunable wide-linewidth laser through the laser driver control unit to drive the laser to output a modulated laser beam with continuously tunable wavelength. The injection current adjustment range of the laser: 50 - 200 mA, corresponding to a wavelength tuning range of 3344.9 nm - 3345.1 nm (covering the alcohol characteristic absorption peak ±0.1 nm).

[0076] Direct the modulated laser beam into the cab through the vehicle side window. Utilize the absorption characteristics of alcohol gas for a specific wavelength laser, so that the laser penetrates the cab space and undergoes an absorption interaction with alcohol gas molecules to form a transmitted light signal carrying concentration information. Specifically, the modulated laser beam is directed through a collimating lens group (focal length 50 mm, spot diameter 5 mm) onto the vehicle side window (glass thickness ≤6 mm), penetrates the cab space (optical path is about 0.5 m), and exits from the other side window.

[0077] By adjusting the injection current of the laser, make the laser wavelength periodically scan to cover the characteristic absorption spectral line of alcohol gas. Specifically, by adjusting the laser injection current (step 0.1 mA), make the laser wavelength periodically scan in the range of 3344.95 nm - 3345.05 nm at a frequency of 10 Hz (covering the half-width of the alcohol absorption spectral line of 0.1 nm).

[0078] Receive the transmitted light signal penetrating the cab through an optical focusing device. Specifically, use a HgCdTe cooled photodetector (response wavelength 2 - 5 μm, detectivity 1×10 10 / cm·Hz 1 / / 2 / W), and focus the transmitted light signal onto the photosensitive surface of the detector (diameter 1 mm) through an optical focusing device (focal length 100 mm, NA = 0.3).

[0079] It should be noted that the sawtooth wave is used for periodic scanning of the alcohol absorption peak (wavelength tuning), the sine wave is used to generate the modulation sideband, and the second harmonic signal is extracted by combining the phase-locked amplification technology to suppress the low-frequency noise. Selecting a wavelength of 3345 nm can avoid interference from other gases to the greatest extent, and the mid-infrared band is suitable for penetrating the window glass (the transmittance of ordinary glass in the 3-5 μm band is >80%).

[0080] S3. Focus the transmitted light signal onto a photodetector through an optical focusing device to obtain a noisy electrical signal, and input it into a pre-trained noise suppression model to output a denoised electrical signal, including the following steps.

[0081] Receive the transmitted light signal penetrating the cab through an optical focusing device, and use an optical lens group to focus the transmitted light signal onto the photosensitive area of the photodetector to complete the conversion of the optical signal to an electrical signal, generating a noisy electrical signal containing laser linewidth noise and environmental interference. Specifically, the current signal output by the photodetector is converted into a voltage signal through a transimpedance amplifier (gain 10 4 V / A, bandwidth 0-100 kHz) to generate a noisy electrical signal (noise sources include laser relaxation oscillation noise and detector shot noise).

[0082] Perform preliminary filtering on the noisy electrical signal to suppress the high-frequency noise component and retain the low-frequency signal characteristics related to alcohol gas absorption. Specifically, a 6th-order Butterworth low-pass filter (cutoff frequency 50 kHz, stopband attenuation ≥40 dB / dec) is used to suppress the high-frequency noise (such as laser relaxation oscillation noise >100 kHz) and retain the low-frequency signal related to alcohol absorption (0-20 kHz).

[0083] Input the filtered noisy electrical signal into a pre-trained noise suppression model, and separate the noise component and the effective signal component through the non-linear mapping relationship of the neural network module to generate a pure denoised electrical signal. Specifically, the filtered electrical signal (sampling rate 200 kHz, number of points 1000 / ms) is input into a pre-trained BP neural network (input layer with 1000 nodes, hidden layer with 512×3, output layer with 1000 nodes), and the non-linear separation of noise and signal is achieved through the ReLU activation function to output a denoised electrical signal.

[0084] It should be noted that the gain and bandwidth of the transimpedance amplifier need to match the output characteristics of the detector to avoid signal saturation or high-frequency distortion. The Butterworth filter has a flat passband and a linear phase response, which is suitable for retaining the signal waveform characteristics. The BP neural network learns the noise distribution characteristics through end-to-end training, without the need to manually design filter parameters, and can adapt to complex noise environments.

[0085] S4. Using lock-in amplification detection technology, synchronously demodulate and perform harmonic analysis on the denoised electrical signal, extract the amplitude of the second harmonic signal associated with the modulation frequency, and determine the characteristic amplitude of the second harmonic signal through peak detection, including the following steps:

[0086] Based on the reference signal generated by the modulation signal generation unit, synchronously demodulate the denoised electrical signal through a dual-phase lock-in amplifier to separate the fundamental frequency signal and the second harmonic signal synchronized with the modulation frequency. Specifically, the dual-phase lock-in amplifier (reference signal: 10 kHz sine wave) performs quadrature demodulation on the denoised electrical signal and outputs the in-phase component and the quadrature component.

[0087] Use a band-pass filter to screen the frequency band of the demodulated signal, retain the target band signal containing the second harmonic component, and filter out out-of-band noise interference. Specifically, use a band-pass filter (passband 19 - 21 kHz, bandwidth 2 kHz) to extract the second harmonic signal (20 kHz) and suppress the fundamental frequency (10 kHz) and other harmonic interferences.

[0088] Calculate the amplitude of the second harmonic signal in the target frequency band through the harmonic amplitude extraction module, determine the characteristic amplitude at the center position of the absorption line based on the peak detection algorithm, and compare the characteristic amplitude with the preset calibration curve of the alcohol gas concentration. Specifically, determine the characteristic amplitude at the center position of the absorption line through the Gaussian fitting algorithm.

[0089] It should be noted that dual-phase demodulation can eliminate the influence of phase drift and improve the stability of harmonic amplitude detection. The amplitude of the second harmonic signal has an approximately linear relationship with the alcohol concentration and is less affected by laser intensity fluctuations. Gaussian fitting can eliminate baseline drift and accurately extract the amplitude of the absorption peak.

[0090] S5. According to the calibration relationship between the characteristic amplitude of the second harmonic signal and the alcohol gas concentration, use the least squares fitting algorithm to inversely calculate the concentration value of the alcohol gas in the cab, set the alcohol concentration threshold, judge the drunk driving situation of the driver, generate the detection result and output an alarm signal, including the following steps:

[0091] Using the pre-experiment calibration data (the second harmonic amplitudes corresponding to alcohol concentrations of 20 / 60 / 100 / 140 / 200 ppm are 0.3624 / 1.0722 / 1.7597 / 2.4257 / 3.3908 (unit: V).), establish a non-linear mapping relationship between the characteristic amplitude of the second harmonic signal and the alcohol gas concentration through the least squares fitting algorithm, generate the concentration calibration curve, and the expression is:

[0092] C = a 1 X 2 + a 2 X + a 3 ;

[0093] Wherein, C is the alcohol gas concentration value, a 1 is the quadratic term fitting coefficient, a 2 is the linear term fitting coefficient, a 3 is the constant term fitting coefficient, and X is the second harmonic amplitude;

[0094] Input the characteristic amplitude of the second harmonic signal detected in real time into the concentration calibration curve, and inversely calculate the real-time concentration value of the alcohol gas in the cab through the fitting equation. Specifically, input the characteristic amplitude of the second harmonic signal detected in real time into the concentration calibration curve, and inversely calculate the real-time concentration value of the alcohol gas in the cab through the fitting equation.

[0095] Set the alcohol concentration determination threshold according to the legal drunk driving standard, compare the real-time concentration value with the alcohol concentration determination threshold, and if it exceeds the alcohol concentration determination threshold, it is determined as a drunk driving behavior. Specifically, the alcohol concentration in exhaled breath needs to be converted to blood alcohol concentration through Henry's constant (blood / gas partition ratio 2100:1). According to the GB / T 21254-2017 standard, set the alcohol concentration determination threshold for drunk driving as C≥20 ppm (corresponding to blood alcohol concentration BAC≥0.02%); for drunk driving: C≥80 ppm (BAC≥0.08%).

[0096] Generate a detection report including the alcohol concentration value and the determination result. Specifically, use the determination result to output a detection report in the format of alcohol concentration, blood alcohol concentration, determination result, and timestamp through the RS-485 or wireless module (Wi-Fi / 4G).

[0097] It should be noted that the quadratic polynomial can effectively fit the non-linear relationship between alcohol concentration and harmonic amplitude (absorption saturation effect). Data encryption transmission (AES-256) ensures the legality of the law enforcement process and privacy protection.

[0098] This embodiment also provides a non-contact drunk driving detection system using a large linewidth tunable laser, including: a pre-training module, a modulated laser module, a denoising module, an amplitude acquisition module, and a detection output module;

[0099] The pre-training module is used to collect the transmitted light signals and corresponding electrical signals under different linewidth conditions through experiments, synchronously record the experimental parameters, obtain the original data set, preprocess the original data set, generate the training set, validation set, and test set, train the BP neural network, and obtain a pre-trained noise suppression model;

[0100] The modulated laser module is used to generate a wavelength-tunable modulated laser beam using a large linewidth tunable laser, direct the modulated laser beam into the vehicle cab, make the laser penetrate the cab space and carry the alcohol gas concentration information, and obtain the transmitted light signal;

[0101] A denoising module, which is used to focus the transmitted optical signal onto a photodetector through an optical focusing device to obtain a noisy electrical signal, and input it into a pre-trained noise suppression model to output a denoised electrical signal;

[0102] An amplitude acquisition module, which is used to perform synchronous demodulation and harmonic analysis on the denoised electrical signal by using lock-in amplification detection technology, extract the amplitude of the second harmonic signal associated with the modulation frequency, and determine the characteristic amplitude of the second harmonic signal through peak detection;

[0103] A detection output module, which is used to inversely calculate the concentration value of alcohol gas in the cab by using the least square fitting algorithm according to the calibration relationship between the characteristic amplitude of the second harmonic signal and the alcohol gas concentration, set an alcohol concentration threshold, judge the drunk driving situation of the driver, generate a detection result and output an alarm signal.

[0104] This embodiment also provides a computer device, which is applicable to the situation of the non-contact drunk driving detection method using a large linewidth tunable laser, and includes: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the non-contact drunk driving detection method using a large linewidth tunable laser as proposed in the above embodiment.

[0105] In summary, through the collaborative application of the large linewidth tunable laser and the BP neural network denoising technology, the present invention significantly reduces the system hardware cost, and at the same time effectively suppresses the influence of linewidth broadening and environmental interference on the detection accuracy, realizing high-precision alcohol concentration detection equivalent to that of an ultra-narrow linewidth laser. Through the non-linear noise suppression ability of the neural network, the system can operate stably without a complex temperature control device, adapt to complex environmental conditions, and support vehicle dynamic passing detection, significantly improving the law enforcement efficiency. In addition, the non-contact detection method avoids the traffic interference and privacy disputes of traditional methods, providing an innovative solution for drunk driving screening with low cost, high reliability and easy large-scale deployment.

[0106] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A non-contact drunk driving detection method using a large linewidth tunable laser, characterized by: include, Through experiments, the transmitted light signals and corresponding electrical signals under different line width conditions are collected, and the experimental parameters are recorded synchronously to obtain the original data set, and the original data set is preprocessed to generate the training set, verification set and test set, train the BP neural network, and obtain the pre-trained noise suppression model; A large linewidth tunable laser is used to generate a wavelength tunable modulated laser beam, which is directed into the vehicle cab, so that the laser penetrates the cab space and carries the alcohol gas concentration information to obtain a transmitted light signal; The transmitted light signal is focused onto the photodetector through an optical focusing device to obtain a noisy electrical signal, which is then input into a pre-trained noise suppression model to output a denoised electrical signal; The phase-locked amplification detection technology is used to synchronously demodulate and harmonic analyze the de-noised electrical signal, extract the amplitude of the second harmonic signal associated with the modulation frequency, and determine the characteristic amplitude of the second harmonic signal through peak detection; According to the calibration relationship between the characteristic amplitude of the second harmonic signal and the alcohol gas concentration, the least square fitting algorithm is used to inversely calculate the concentration of alcohol gas in the cab, and the alcohol concentration threshold is set to judge the driver's drunk driving situation, generate detection results and output warning signals.

2. The non-contact drunk driving detection method using a large linewidth tunable laser as claimed in claim 1, characterized in that: The experimental parameters include laser parameters, ambient temperature and calibrated alcohol concentration; The preprocessing includes preliminary noise screening, normalization processing and proportional division; The noise initial screening is to identify and filter out abnormal data points and high-frequency noise interference in the original data set through statistical analysis methods, and retain effective signal characteristics; The normalization process is to map the effective signal characteristics to a preset value range; The proportional division is to divide the normalized data set into a training set, a validation set and a test set according to a standardized ratio.

3. The non-contact drunk driving detection method using a large linewidth tunable laser as claimed in claim 2, characterized in that: The BP neural network is trained to obtain a pre-trained noise suppression model. The specific steps are as follows: Based on the dimension of the training set, the weight parameters of the BP neural network are randomly initialized to construct the network topology of the input layer, hidden layer and output layer; The noisy electrical signal in the training set is input into the BP neural network, and the output layer prediction value is calculated through forward propagation. It is compared with the true value of the denoised electrical signal in the validation set, and the prediction error is calculated based on the preset loss function. Through the back propagation algorithm, the prediction error is used to adjust the weight parameters of the neural network layer by layer to optimize the mapping relationship of noise suppression; Iterate the forward propagation and back propagation processes, evaluate the model performance in real time based on the validation set, and terminate the training when the prediction error converges to the preset threshold. Use the test set to verify the generalization performance of the trained noise suppression model; The neural network weight parameters that meet the generalization requirements are saved to form a pre-trained noise suppression model.

4. The non-contact drunk driving detection method using a large linewidth tunable laser as claimed in claim 3, characterized in that: A large linewidth tunable laser is used to generate a wavelength tunable modulated laser beam, which is directed into the vehicle cab so that the laser penetrates the cab space and carries the alcohol gas concentration information to obtain the transmitted light signal. The specific steps are as follows: A superposition modulation signal of a low-frequency sawtooth wave signal and a high-frequency sine wave signal is generated based on a modulation signal generating unit, and loaded to a large linewidth tunable laser through a laser driving control unit, so as to drive the laser to output a modulated laser beam with tunable wavelength; The modulated laser beam is directed into the cab through the vehicle side window. The absorption characteristics of alcohol gas to lasers of a specific wavelength are used to make the laser penetrate the cab space and absorb the alcohol gas molecules to form a transmitted light signal carrying concentration information. By adjusting the injection current of the laser, the laser wavelength is periodically scanned to cover the characteristic absorption spectrum of alcohol gas; The transmitted light signal penetrating the cab is received by the optical focusing device.

5. The non-contact drunk driving detection method using a large linewidth tunable laser as claimed in claim 4, characterized in that: The transmitted light signal is focused onto the photodetector through an optical converging device to obtain a noisy electrical signal, which is then input into a pre-trained noise suppression model to output a denoised electrical signal. The specific steps are as follows: The optical focusing device receives the transmitted light signal penetrating the cab, and uses the optical lens group to focus the transmitted light signal onto the photosensitive area of ​​the photodetector, completing the conversion of the optical signal into an electrical signal, and generating a noisy electrical signal containing the laser line width noise and environmental interference; Perform preliminary filtering on the noisy electrical signal to suppress high-frequency noise components and retain low-frequency signal characteristics related to alcohol gas absorption; The filtered noisy electrical signal is input into a pre-trained noise suppression model, and the noise component and the effective signal component are separated through the nonlinear mapping relationship of the neural network module to generate a denoised pure electrical signal.

6. The non-contact drunk driving detection method using a large linewidth tunable laser as claimed in claim 5, characterized in that: The phase-locked amplification detection technology is used to synchronously demodulate and harmonic analyze the de-noised electrical signal, extract the amplitude of the second harmonic signal associated with the modulation frequency, and determine the characteristic amplitude of the second harmonic signal through peak detection. The specific steps are as follows: Based on the reference signal generated by the modulation signal generation unit, the de-noised electrical signal is synchronously demodulated by a dual-phase lock-in amplifier to separate the baseband signal and the second harmonic signal synchronized with the modulation frequency; The demodulated signal is filtered using a bandpass filter to retain the target frequency band signal containing the second harmonic component and filter out the out-of-band noise interference; The amplitude of the second harmonic signal in the target frequency band is calculated through the harmonic amplitude extraction module, and the characteristic amplitude at the center of the absorption spectrum is determined based on the peak detection algorithm, and the characteristic amplitude is compared with the preset alcohol gas concentration calibration curve.

7. The non-contact drunk driving detection method using a large linewidth tunable laser as claimed in claim 6, characterized in that: According to the calibration relationship between the characteristic amplitude of the second harmonic signal and the alcohol gas concentration, the least square fitting algorithm is used to inversely calculate the concentration of alcohol gas in the cab, and the alcohol concentration threshold is set to judge the driver's drunk driving situation, generate the detection results and output the warning signal. The specific steps are as follows: Using the pre-experimental calibration data, the nonlinear mapping relationship between the characteristic amplitude of the second harmonic signal and the alcohol gas concentration is established through the least squares fitting algorithm to generate a concentration calibration curve, which is expressed as follows: C=a1X 2 +a2X+a3; Among them, C is the alcohol gas concentration value, a1 is the quadratic fitting coefficient, a2 is the linear fitting coefficient, a3 is the constant fitting coefficient, and X is the second harmonic amplitude; The characteristic amplitude of the second harmonic signal detected in real time is input into the concentration calibration curve, and the real-time concentration value of the alcohol gas in the cab is calculated by inversion through the fitting equation; Set an alcohol concentration threshold according to the legal drunk driving standard, compare the real-time concentration value with the alcohol concentration threshold, and determine it as drunk driving if it exceeds the alcohol concentration threshold; Generate a test report including the alcohol concentration value and the judgment result.

8. A non-contact drunk driving detection system using a large linewidth tunable laser, based on the non-contact drunk driving detection method using a large linewidth tunable laser according to any one of claims 1 to 7, characterized in that: Including, pre-training module, laser modulation module, denoising module, amplitude acquisition module and detection output module; The pre-training module is used to collect the transmitted light signal and the corresponding electrical signal under different line width conditions through experiments, synchronously record the experimental parameters, obtain the original data set, and pre-process the original data set to generate a training set, a verification set and a test set, train the BP neural network, and obtain a pre-trained noise suppression model; The modulated laser module is used to generate a wavelength-tunable modulated laser beam using a large linewidth tunable laser, and project the modulated laser beam into the vehicle cab in a direction so that the laser penetrates the cab space and carries the alcohol gas concentration information to obtain a transmitted light signal; The denoising module is used to focus the transmitted light signal to the photodetector through an optical converging device to obtain a noisy electrical signal, input a pre-trained noise suppression model, and output a denoised electrical signal; The amplitude acquisition module is used to perform synchronous demodulation and harmonic analysis on the de-noised electrical signal using a phase-locked amplification detection technology, extract the amplitude of the second harmonic signal associated with the modulation frequency, and determine the characteristic amplitude of the second harmonic signal through peak detection; The detection output module is used to inversely calculate the concentration value of alcohol gas in the cab based on the calibration relationship between the characteristic amplitude of the second harmonic signal and the alcohol gas concentration using a least squares fitting algorithm, and to set an alcohol concentration threshold to determine the driver's drunk driving situation, generate a detection result and output a warning signal.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the non-contact drunk driving detection method using a large linewidth tunable laser as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the non-contact drunk driving detection method using a large linewidth tunable laser as described in any one of claims 1 to 7 are implemented.

Images