Intelligent verification system for alcohol content of exhaled gas based on image recognition

Abstract

The invention provides an exhaled gas alcohol content intelligent verification system based on image recognition, relates to the technical field of intelligent detection, and aims to solve the problems that a traditional sensor is prone to environmental interference, needs frequent calibration and is insufficient in dynamic response. Comprising a gas-sensitive developing unit, a multispectral dynamic imaging unit, a data processing unit and a closed-loop control unit, the gas-sensitive color development unit adopts a reversible gas-sensitive film and generates a color change signal through an alcohol specific color development reaction; the multispectral dynamic imaging unit is combined with an ambient light suppression and dynamic region positioning technology to capture spectral reflectivity time sequence data in a color development process; the data processing unit maps the chromaticity characteristic into an alcohol concentration value based on a Gaussian process regression model, and triggers near-infrared auxiliary verification through confidence evaluation; and the closed-loop control unit maintains a constant detection environment through the temperature control and airflow stabilization module. According to the scheme, non-intrusive high-precision detection of the alcohol concentration in a complex environment can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent detection, and particularly relates to an intelligent verification system for the alcohol content in exhaled breath based on image recognition. Background Art

[0002] Traditional exhaled breath alcohol detection technologies mainly rely on electrochemical sensors or semiconductor gas-sensitive elements, and their detection accuracy is easily interfered by factors such as environmental temperature and humidity, airflow fluctuations, and sensor aging. For example, electrochemical sensors need to replace the electrolyte regularly and are prone to baseline drift in high-humidity environments; although semiconductor sensors have lower costs, they have poor selectivity and are easily affected by other volatile organic compounds (such as acetone), resulting in a high false alarm rate. In addition, existing technologies mostly adopt single-point static detection, which cannot capture the dynamic change process of alcohol concentration in real time and lack a self-calibration mechanism, relying on manual periodic calibration, with high maintenance costs. In complex environments (such as extreme temperatures and light changes), the stability of traditional sensors significantly decreases, making it difficult to meet the requirements of high-reliability detection in fields such as law enforcement and medical treatment. Summary of the Invention

[0003] The present invention provides an intelligent verification system for the alcohol content in exhaled breath based on image recognition to solve the problems in the prior art that the detection accuracy is easily interfered by the environment, requires frequent manual calibration, and has insufficient dynamic response.

[0004] The present invention provides an intelligent verification system for the alcohol content in exhaled breath based on image recognition, including: a gas-sensitive colorimetric unit, a reversible gas-sensitive thin film configured to have a specific colorimetric reaction with alcohol, and its color change has a preset functional relationship with the alcohol concentration; A multi-spectral dynamic imaging unit, configured with a multi-spectral light source and an image sensor, for capturing the spectral reflectance time-series data of the colorimetric thin film during the dynamic reaction process; A data processing unit, mapping the chromaticity features in the spectral reflectance time-series data to alcohol concentration values based on a machine learning model and generating a confidence evaluation result; A closed-loop control unit, integrating a temperature control device and an airflow control device, and maintaining a constant temperature environment and a stable exhaled airflow rate in the gas chamber through a feedback control algorithm.

[0005] Furthermore, the multi-spectral dynamic imaging unit includes: An ambient light suppression module, extracting the ambient light interference signal from the continuously collected spectral images through a time-series difference algorithm; A dynamic region positioning module, based on the edge gradient features of the colorimetric thin film, real-time segmenting the effective reaction region, where the effective reaction region is a continuous region in the colorimetric thin film with uniform color change and meeting the preset morphological conditions, and excluding the non-uniform regions caused by airflow fluctuations; The spectral data generation module aggregates the multi-band reflectance data within the effective reaction region according to the time series to generate a dynamic spectral feature matrix.

[0006] Furthermore, the data processing unit includes: The chromaticity feature extraction module converts the dynamic spectral feature matrix to a preset color space and extracts the combined feature vector composed of saturation and hue. The concentration mapping module performs non-linear fitting on the combined feature vector through a Gaussian process regression model and outputs the alcohol concentration prediction value and the confidence interval. The cross-validation module, when the width of the confidence interval exceeds the threshold, triggers the acquisition of auxiliary data from the near-infrared transmission spectral channel, and recomputes the concentration value after weighted fusion with the chromogenic features.

[0007] Furthermore, the training method of the Gaussian process regression model includes: collecting historical chromogenic reaction data, including the spectral reflectance time series data and the corresponding standard concentration true values at different alcohol concentrations; fitting the non-linear relationship between the chromaticity features and the concentration values through a kernel function optimization algorithm, and dynamically adjusting the model hyperparameters based on the Bayesian criterion; performing robustness verification on the trained model, and if the error rate on the interference test set exceeds the preset upper limit, re-perform feature selection and kernel function matching.

[0008] Furthermore, the closed-loop control unit includes: The temperature control module monitors the chamber temperature in real time and compares it with the preset target value. If a temperature deviation is detected, dynamic compensation is performed through a semiconductor temperature control device. The air flow stability module adjusts the output power of the micro air pump according to the fluctuation data fed back by the expiratory flow rate sensor using a proportional-integral-derivative algorithm to maintain the flow rate within the preset range. The anomaly handling module, when the temperature or flow rate continuously exceeds the limit for a set duration, triggers system self-locking and generates a hardware fault alarm signal.

[0009] Furthermore, the system includes an adaptive calibration unit, and its execution logic includes: activating the built-in reference sample before each detection and collecting the chromogenic feature data of the reference sample; calculating the feature deviation amount between the real-time chromogenic feature and the reference sample. If the deviation amount exceeds the tolerance threshold, start the model parameter correction program; inject the deviation compensation amount into the machine learning model through an incremental learning algorithm to update the concentration mapping relationship until the feature deviation amount returns to within the tolerance range.

[0010] Further, the reset method of the gas-sensitive color display unit includes: after the detection is completed, start the thin-film heating device, apply a preset temperature field to the color display thin film to trigger the reverse reaction to restore the initial color; verify the color state after reset through an optical sensor, if the color difference from the initial state exceeds the allowable range, mark the thin film as invalid and prohibit subsequent detections; count the cumulative usage times of the thin film, and when the number reaches the preset life threshold, automatically generate a replacement reminder signal.

[0011] Further, the system includes a user interaction unit, and the user interaction unit includes: dynamically drawing a curve of the change in alcohol concentration, and superimposing and displaying the confidence interval boundary and the legal threshold reference line; when the detected value exceeds the legal threshold, automatically generate a verifiable digital certificate including the timestamp, concentration value, and calibration status; upload the digital certificate to the supervision platform through an encrypted channel, and receive the compliance certification result feedback from the platform.

[0012] Further, the execution logic of the incremental learning algorithm is: calculate the model weight correction gradient according to the feature deviation amount, and limit the gradient update amplitude to prevent overfitting; retain the statistical distribution characteristics of the historical training data, and balance the weight ratio of new and old data through a sliding window mechanism; perform local verification on the corrected model after calibration, if the verification error does not decrease, roll back to the previous stable version.

[0013] This solution realizes the dynamic color display reaction of alcohol concentration through the reversible thin film of the gas-sensitive color display unit, combines multi-spectral dynamic imaging and the Gaussian process regression model, and significantly improves the detection accuracy and anti-interference ability; the closed-loop control unit maintains a constant detection environment through the temperature control and air flow stabilization module, solving the problem of temperature and humidity sensitivity of traditional sensors; the adaptive calibration unit dynamically corrects the model parameters using the incremental learning algorithm, extends the calibration period and reduces the maintenance cost; the dynamic area positioning module excludes the air flow fluctuation interference area, ensuring high detection stability in complex environments, and overall solves the defects of the existing technology that the accuracy is easily affected by the environment, requires frequent manual calibration, and has insufficient dynamic response. Brief Description of the Drawings

[0014] Figure 1 It is an architecture diagram of an intelligent verification system for the alcohol content in exhaled breath based on image recognition according to the present invention. Detailed Embodiments

[0015] The present invention relates to an intelligent verification system for the alcohol content in exhaled breath based on image recognition, aiming to realize the accurate detection of the alcohol content in exhaled breath by combining image recognition technology with multi-feature analysis.

[0016] The above technical solutions will be described in detail below in conjunction with the accompanying drawings of the specification and specific embodiments to better understand the above technical solutions. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments for explaining the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention. In addition, it should be noted that, for the sake of convenience of description, only the parts related to the present invention rather than all are shown in the drawings.

[0017] Embodiment 1 As Figure 1 shown in the architecture diagram of an intelligent verification system for the alcohol content in exhaled gas based on image recognition, it specifically includes: A gas-sensitive color-changing unit, which is equipped with a reversible gas-sensitive thin film that undergoes a specific color-changing reaction with alcohol, and the color change has a preset functional relationship with the alcohol concentration; the gas-sensitive thin film is loaded with a specific redox catalyst (such as CrO2 / SiO2) through nanomaterials. When alcohol molecules come into contact with the surface of the thin film, a redox reaction is triggered, causing the color of the thin film to gradually change from the initial color (such as orange-red) to the target color (such as green); the dynamic process of the color change has a non-linear mapping relationship with the alcohol concentration, and a "color-concentration" conversion model is established through pre-experiment calibration. For example, when the alcohol concentration increases from 0.01 mg / L to 0.1 mg / L, the hue (H value) of the thin film linearly changes from 120° to 60°, and the saturation (S value) increases from 30% to 80%; the thin film has reversibility and can be reset to the initial color state by heating after the detection is completed to ensure repeated use.

[0018] A multi-spectral dynamic imaging unit, which is equipped with a multi-spectral light source and an image sensor for capturing the spectral reflectance time-series data of the color-changing thin film during the dynamic reaction process; specifically, a narrow-band multi-spectral light source (such as 520 nm ± 5 nm, 600 nm ± 5 nm) is used to cover the sensitive band of the color-changing reaction and avoid environmental light interference; a high-frame-rate CMOS sensor (such as 30 frames per second) is configured to collect the dynamic spectral image sequence of the color-changing thin film in real time; based on an edge detection algorithm (such as the Canny operator), continuous regions with uniform color changes in the thin film (defined as the effective reaction region) are segmented to exclude edge blurring or non-uniform regions caused by air flow disturbance; through the time-series differential algorithm, the static environmental light component is separated from the continuous image sequence, and the dynamic color-changing signal is retained. For example, the difference operation is performed on two adjacent frames of images to eliminate the fixed background noise.

[0019] A data processing unit that maps the chromaticity features in the spectral reflectance time series data to alcohol concentration values based on a machine learning model and generates a confidence evaluation result. Specifically, it converts the spectral image of the effective reaction region to the HSV color space and extracts the joint feature vector composed of saturation (S) and hue (H). For example, the H value is sensitive to changes under low-concentration alcohol, and the S value changes significantly under high-concentration alcohol. The Matérn kernel function is used to fit the non-linear relationship between chromaticity features and alcohol concentration. The training data includes historical chromogenic reaction data (such as 100 groups of H / S values at different concentrations) and gas chromatography (GC) standard concentration values. When the width of the confidence interval output by the model exceeds a threshold (such as ±5%), it triggers the near-infrared transmission spectrum (1450nm) auxiliary detection channel to improve the accuracy by weighted fusion of bimodal data.

[0020] A closed-loop control unit that integrates a temperature regulation device and an air flow control device to maintain a constant temperature environment and a stable exhalation flow rate in the gas chamber through a feedback control algorithm. Specifically, it monitors the gas chamber temperature in real time. If it deviates from the preset range (such as 25±0.5°C), it dynamically compensates by heating or cooling with a Peltier element. It combines historical temperature fluctuation data to predict trends and dynamically adjusts the integral term weight of the PID algorithm to prevent overshoot. It monitors the exhalation flow rate through a MEMS flow sensor. If the detected flow rate fluctuation exceeds a threshold (such as ±5%), it adjusts the power output of the micro air pump. It uses a feedforward-feedback composite control algorithm to quickly respond to changes in exhalation pressure.

[0021] Furthermore, the multispectral dynamic imaging unit includes: An ambient light suppression module that extracts the ambient light interference signal from the continuously acquired spectral images through a temporal difference algorithm; A dynamic region localization module that, based on the edge gradient features of the chromogenic film, real-time segments the effective reaction region, where the effective reaction region is a continuous region in the chromogenic film with uniform color change and meeting the preset morphological conditions, and excludes the non-uniform regions caused by air flow fluctuations; A spectral data generation module that aggregates the multi-band reflectance data within the effective reaction region according to the time series to generate a dynamic spectral feature matrix.

[0022] Specifically, in the continuously acquired spectral image sequence (such as 30 frames per second), pixel-level difference operations are performed on adjacent two frames of images to extract the dynamically changing chromogenic signal. High-frequency noise is eliminated through sliding window mean filtering, and low-frequency chromogenic reaction components are retained. Ambient light interference is eliminated.

[0023] The Canny edge detection algorithm is used to identify the boundary of the color-developing film and generate a binary mask. Based on the morphological closing operation, the internal holes of the film are filled, and a continuous region with uniform color change (defined as the effective reaction region) is extracted; the chromaticity mean value of the pixels within the effective reaction region is calculated as the basic data for subsequent analysis. The edge-blurred region caused by air flow disturbance is excluded.

[0024] The reflectance data of multiple bands (such as 520nm, 600nm) within the effective reaction region are aligned in time series. A dynamic spectral feature matrix is constructed, where the rows represent time points and the columns represent the band reflectance intensities; standardized input data are generated to adapt to the processing requirements of the machine learning model.

[0025] Furthermore, the data processing unit includes: A chromaticity feature extraction module that converts the dynamic spectral feature matrix to a preset color space and extracts the joint feature vector composed of saturation and hue; A concentration mapping module that performs non-linear fitting on the joint feature vector through a Gaussian process regression model and outputs the predicted value and confidence interval of the alcohol concentration; A cross-validation module that, when the width of the confidence interval exceeds the threshold, triggers the acquisition of auxiliary data from the near-infrared transmission spectral channel, and recomputes the concentration value after weighted fusion with the color-developing features.

[0026] Specifically, the dynamic spectral feature matrix is converted from the RGB space to the HSV space, and the joint feature vector of saturation (S) and hue (H) is extracted; the H and S values are normalized to eliminate the influence of illumination intensity differences; the feature dimension is compressed to 2D.

[0027] A Gaussian process regression (GPR) model is used to fit the non-linear relationship between the chromaticity features and the alcohol concentration with the Matérn5 / 2 kernel function; the kernel function hyperparameters (such as length scale, variance) are automatically adjusted through Bayesian optimization; so that the prediction error of the model within the concentration range of 0.01 - 2.0mg / L becomes smaller.

[0028] When the width of the confidence interval output by the GPR model exceeds the threshold (such as ±5%), the near-infrared transmission spectrum (1450nm) auxiliary detection is triggered; the color-developing data and the transmission data are weighted and fused (the weights are dynamically assigned according to the confidence), and the alcohol concentration is recomputed.

[0029] Furthermore, the training method of the Gaussian process regression model includes: collecting historical color reaction data, including spectral reflectance time series data under different alcohol concentrations and the corresponding standard concentration true values; fitting the nonlinear relationship between chromaticity characteristics and concentration values ​​through a kernel function optimization algorithm, and dynamically adjusting the model hyperparameters based on the Bayesian criterion; performing robustness verification on the trained model, and if the error rate on the interference test set exceeds a preset upper limit, re-executing feature selection and kernel function matching.

[0030] Specifically, historical color reaction data is collected, covering different temperature, humidity and alcohol concentration conditions; abnormal data (such as sensor failure frames) is eliminated, and missing values ​​are interpolated and filled.

[0031] The grid search method is used to evaluate the fitting effects of different kernel functions (such as RBF and Matérn) and select the optimal kernel type; based on the marginal likelihood maximization criterion, the kernel function hyperparameters are dynamically adjusted.

[0032] Verify the generalization ability of the model on the interference test set (including simulated ambient light mutation and airflow disturbance data); if the error rate exceeds the threshold (such as 5%), re-execute feature selection or introduce regularization terms.

[0033] Furthermore, the closed-loop control unit includes: The temperature control module monitors the air chamber temperature in real time and compares it with the preset target value. If a temperature deviation is detected, dynamic compensation is performed through a semiconductor temperature control device; The airflow stabilization module uses a proportional-integral-differential algorithm to adjust the output power of the micro air pump according to the fluctuation data fed back by the exhalation flow rate sensor to maintain the flow rate within a preset range; The exception handling module triggers the system to self-lock and generates a hardware fault alarm signal when the temperature or flow rate exceeds the limit continuously for more than the set time.

[0034] Specifically, for example, the air chamber temperature is monitored in real time by a DS18B20 temperature sensor with a sampling frequency of 10 Hz; if the temperature deviates from a preset range (such as 25 ± 0.5 ° C), the Peltier element is triggered for dynamic compensation (the heating / cooling power is PID-adjusted according to the deviation amount).

[0035] The exhalation flow rate is fed back in real time through the MEMS flow sensor. If the flow rate exceeds the limit (such as 1.5L / min±5%), the air pump speed is adjusted through PWM. Feedforward control is used to predict the change of exhalation pressure and adjust the air pump output in advance. The flow rate stability is improved to avoid local overload of the membrane.

[0036] If the temperature or flow rate continuously exceeds the limit for a set duration (e.g., 10 seconds), the system will trigger self-locking and generate a fault code (e.g., E01: Temperature anomaly, E02: Flow rate anomaly); an alarm will be given synchronously through a buzzer and an LED indicator, and the fault log will be recorded.

[0037] Furthermore, the adaptive calibration unit: activates the built-in reference sample before each detection, collects the color development characteristic data of the reference sample; calculates the characteristic deviation between the real-time color development and the reference sample. If the deviation exceeds the tolerance threshold, the model parameter correction program will be started; the deviation compensation amount will be injected into the machine learning model through the incremental learning algorithm to update the concentration mapping relationship until the characteristic deviation returns within the tolerance range.

[0038] Specifically, before detection, start the built-in reference color comparison block (sealed with 0.1 mg / L alcohol gas) and collect its color development characteristic data ( , ); calculate the deviation between the real-time color development characteristic ( , ) and the reference value: , .

[0039] If > 2° or > 5%, start the incremental learning algorithm to update the GPR model parameters.

[0040] Convert the deviation into the correction gradient of the model weight, and limit the single update amplitude not to exceed 10% of the original weight.

[0041] Keep a sliding window of historical data (e.g., the data of the most recent 100 detections) to prevent model drift. Furthermore, the reset method of the gas-sensitive color development unit includes: after the detection is completed, start the thin-film heating device, apply a preset temperature field to the color development thin film to trigger the reverse reaction to restore the initial color; verify the color state after reset through an optical sensor. If the color difference from the initial state exceeds the allowable range, mark the thin film as failed and prohibit subsequent detections; count the cumulative usage times of the thin film. When the number reaches the preset life threshold, an automatic replacement reminder signal will be generated.

[0042] Specifically, after the detection is completed, start the thin-film heating device (50 ± 2 °C) and keep it for 30 seconds to trigger the reverse oxidation reaction of CrO2 and restore the initial orange-red color.

[0043] Measure the LAB color difference of the thin film after reset through a color sensor , if > 5 (compared with the initial state), it is determined that the reset fails.

[0044] Statistically count the cumulative usage times of the thin film. When the number of times exceeds the threshold (such as 500 times), prompt to replace the thin film through the interactive interface.

[0045] Furthermore, the user interaction unit includes: dynamically drawing the curve of the alcohol concentration change, and superimposing and displaying the confidence interval boundary and the legal threshold reference line; when the detected value exceeds the legal threshold, automatically generating a verifiable digital certificate including the timestamp, concentration value, and calibration status; uploading the digital certificate to the supervision platform through an encrypted channel, and receiving the compliance certification result feedback from the platform.

[0046] Specifically, in real-time draw the alcohol concentration-time curve, and superimpose and display the confidence interval (such as the shaded area represents 95% confidence); when the concentration approaches the legal threshold (0.15 mg / L), the curve color gradually changes from green to red.

[0047] Generate a detection report including the timestamp, device serial number, and SHA-256 signature, and upload it to the supervision platform through the HTTPS protocol; if the network is interrupted, locally store the encrypted report until the connection is restored.

[0048] Furthermore, the execution logic of the incremental learning algorithm is: calculate the model weight correction gradient according to the feature deviation amount, and limit the gradient update amplitude to prevent overfitting; retain the statistical distribution characteristics of the historical training data, and balance the weight ratio of the new and old data through the sliding window mechanism; perform local verification on the corrected model after calibration, and if the verification error does not decrease, roll back to the previous stable version.

[0049] Specifically, the optimization logic of the incremental learning algorithm: Calculate the model weight correction amount according to the feature deviation amount , where α is the learning rate (default 0.01), X is the input feature, is the expected output deviation; further, retain the most recent N sets of training data (N = 100), and the weight of the new data is 1.2 times that of the old data to prevent the dilution of historical data; After the update, use the validation set (20% historical data) to test the model error. If the error increase > 1%, roll back to the previous stable version.

[0050] This system realizes non-invasive and high-precision alcohol detection through the technical combination of gas-sensitive color reaction, multi-spectral dynamic imaging, closed-loop environmental control, and adaptive calibration, and solves the problems of large environmental interference and frequent calibration of traditional sensors.

[0051] It should be understood that the disclosed embodiments of the present invention and the above descriptions enable those skilled in the art to implement the present invention using the present invention. At the same time, the present invention is not limited to the above-mentioned part of the embodiments. It should be understood that those of ordinary skill in the art can still modify the technical solutions recorded in the foregoing embodiments or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. An intelligent detection system for alcohol content in exhaled gas based on image recognition, characterized in that: include: The gas-sensitive color-developing unit is equipped with a reversible gas-sensitive film that undergoes a specific color-developing reaction with alcohol, and its color change is in a preset functional relationship with the alcohol concentration; A multi-spectral dynamic imaging unit, equipped with a multi-spectral light source and an image sensor, is used to capture the spectral reflectance time series data of the color-developing film during the dynamic reaction process; A data processing unit maps the chromaticity features in the spectral reflectance time series data to alcohol concentration values ​​based on a machine learning model and generates a confidence assessment result; The closed-loop control unit integrates a temperature control device and an airflow control device to maintain a constant temperature environment and a stable exhalation flow rate in the air chamber through a feedback control algorithm.

2. The intelligent detection system for alcohol content in exhaled gas based on image recognition as claimed in claim 1, characterized in that: The multi-spectral dynamic imaging unit comprises: The ambient light suppression module extracts the ambient light interference signal from the continuously collected spectral images through the time difference algorithm; The dynamic area positioning module segments the effective reaction area in real time based on the edge gradient characteristics of the color film. The effective reaction area is a continuous area in the color film where the color changes evenly and meets the preset morphological conditions, and excludes the non-uniform area caused by airflow fluctuations. The spectral data generation module aggregates the multi-band reflectance data in the effective reaction area in time series to generate a dynamic spectral feature matrix.

3. The intelligent detection system for alcohol content in exhaled gas based on image recognition as claimed in claim 1, characterized in that: The data processing unit comprises: The chromaticity feature extraction module converts the dynamic spectrum feature matrix into a preset color space and extracts the joint feature vector composed of saturation and hue; A concentration mapping module performs nonlinear fitting on the joint feature vector through a Gaussian process regression model and outputs a predicted alcohol concentration value and a confidence interval; In the cross-validation module, when the confidence interval width exceeds the threshold, the near-infrared transmission spectrum channel is triggered to collect auxiliary data, and the concentration value is recalculated after weighted fusion with the color rendering feature.

4. The intelligent detection system for alcohol content in exhaled gas based on image recognition as claimed in claim 3, characterized in that: The training method of the Gaussian process regression model includes: collecting historical color reaction data, including spectral reflectance time series data under different alcohol concentrations and the corresponding standard concentration true values; fitting the nonlinear relationship between chromaticity characteristics and concentration values ​​through a kernel function optimization algorithm, and dynamically adjusting the model hyperparameters based on the Bayesian criterion; performing robustness verification on the trained model, and if the error rate on the interference test set exceeds a preset upper limit, re-executing feature selection and kernel function matching.

5. The intelligent detection system for alcohol content in exhaled gas based on image recognition as claimed in claim 1, characterized in that: The closed-loop control unit comprises: The temperature control module monitors the air chamber temperature in real time and compares it with the preset target value. If a temperature deviation is detected, dynamic compensation is performed through a semiconductor temperature control device; The airflow stabilization module uses a proportional-integral-differential algorithm to adjust the output power of the micro air pump according to the fluctuation data fed back by the exhalation flow rate sensor to maintain the flow rate within a preset range; The exception handling module triggers the system to self-lock and generate a hardware fault alarm signal when the temperature or flow rate exceeds the limit continuously for more than the set time.

6. The intelligent detection system for alcohol content in exhaled gas based on image recognition as claimed in claim 1, characterized in that: The system also includes an adaptive calibration unit, whose execution logic includes: activating a built-in reference sample before each detection and collecting color characteristic data of the reference sample; calculating the characteristic deviation between the real-time color characteristic and the reference sample, and if the deviation exceeds the tolerance threshold, starting the model parameter correction program; injecting the deviation compensation amount into the machine learning model through the incremental learning algorithm, and updating the concentration mapping relationship until the characteristic deviation is restored to the tolerance range.

7. The intelligent detection system for alcohol content in exhaled gas based on image recognition as claimed in claim 1, characterized in that: The resetting method of the gas-sensitive colorimetric unit includes: starting the film heating device after the detection is completed, applying a preset temperature field to the colorimetric film to trigger a reverse reaction to restore the initial color; verifying the color state after reset by an optical sensor, if the color difference with the initial state exceeds the allowable range, marking the film as invalid and prohibiting subsequent detection; counting the cumulative number of times the film is used, and automatically generating a replacement reminder signal when the number reaches a preset life threshold.

8. The intelligent detection system for alcohol content in exhaled gas based on image recognition as claimed in claim 1, characterized in that: The system further includes a user interaction unit, which includes: dynamically drawing an alcohol concentration change curve, and superimposing and displaying the confidence interval boundary and the legal threshold reference line; when the detection value exceeds the legal threshold, automatically generating a verifiable digital certificate containing a timestamp, concentration value and calibration status; uploading the digital certificate to the supervision platform through an encrypted channel, and receiving the compliance certification results fed back by the platform.

9. The intelligent detection system for alcohol content in exhaled gas based on image recognition as claimed in claim 6, characterized in that: The execution logic of the incremental learning algorithm is as follows: calculate the model weight correction gradient according to the feature deviation, and limit the gradient update amplitude to prevent overfitting; retain the statistical distribution characteristics of historical training data, and balance the weight ratio of new and old data through a sliding window mechanism; perform local verification on the corrected model after calibration, and roll back to the previous stable version if the verification error is not reduced.

Images