Method and system for preventing drunk driving in urban key area

Abstract

The invention discloses a method and system for preventing drunk driving in an urban key area, and relates to the field of urban traffic safety prevention and control, and the method comprises the following steps: obtaining entrance and exit video monitoring equipment information of the key area based on a designated urban drunk driving key area, so as to monitor the time when a vehicle enters and exits the key area, calculating the stay duration of the vehicle in the key area; image acquisition and behavior characteristic analysis are carried out on pedestrians in the key area to identify suspected drinkers, and early warning information is sent to corresponding vehicle owners according to the stay duration and the suspected drinker information; and performing non-contact alcohol detection on the driver of the driving-out vehicle at the parking lot exit of the key area, and executing corresponding intervention measures according to the detection result. And the pre-recognition, intelligent early warning and active intervention of the drunk driving behavior are realized.

Description

Technical Field

[0001] This invention relates to the field of urban traffic safety control, and in particular to a method and system for preventing drunk driving in key urban areas. Background Technology

[0002] Drunk driving seriously endangers road traffic safety. The existing drunk driving control system mainly relies on "post-incident checks, spot checks, and random checks," which clearly has the following limitations: First, existing methods are mostly post-event investigations, which can only identify and intercept vehicles after they have entered the road or when driving has already occurred. They lack the ability to proactively intervene before the vehicle enters the vehicle or leaves the parking lot, resulting in delayed control.

[0003] Secondly, the coverage of police deployment at fixed points is limited, making it impossible to achieve continuous and high-density control over key urban areas (such as areas with concentrated dining and entertainment), resulting in low law enforcement efficiency and high consumption of police resources.

[0004] Furthermore, existing electronic fence systems are limited in function, typically only recording vehicle trajectories and unable to proactively identify intoxication status by incorporating driver behavioral characteristics. Parking lots, serving as the "last mile" from which vehicles depart, generally lack alcohol testing capabilities, making it impossible to effectively screen vehicles about to exit, resulting in significant regulatory blind spots.

[0005] While some technical solutions attempt to introduce in-vehicle alcohol detection devices, these require hardware modifications to vehicles, resulting in high implementation costs, difficulties in widespread adoption, and a high false alarm rate. For example, Chinese invention patent CN104318711A discloses a driver's drunk driving warning system and its warning method, including an in-vehicle alcohol monitoring module and an information processing module installed in the vehicle. The in-vehicle alcohol monitoring module is connected to the information processing module, which communicates with a vehicle-to-everything (V2X) service platform and with other nearby vehicles via a vehicle-to-vehicle communication module. However, this system fails to form an effective closed loop with the public security traffic management platform, thus failing to achieve full-process control from risk identification to intervention and handling.

[0006] It is evident that designing a method and system for preventing drunk driving that forms a closed loop of "detection-alert-interception" based on key urban areas is of paramount importance. Summary of the Invention

[0007] In a first aspect, to address the aforementioned technical problems, the present invention provides a method for preventing drunk driving in key urban areas, comprising the following steps: Based on the designated key areas for drunk driving in the city, information from video surveillance equipment at the entrances and exits of the key areas is obtained, the time when vehicles enter and leave the key areas is monitored, and the duration of vehicle stay in the key areas is calculated. Image acquisition and behavioral feature analysis are performed on pedestrians within the key area to identify suspected intoxicated individuals. Based on the duration of their stay and information about the suspected intoxicated individuals, a warning message is sent to the corresponding vehicle owner. The behavioral feature analysis includes at least facial feature analysis and gait analysis. The method for identifying suspected intoxicated individuals includes: a) Concatenate the feature vectors obtained from the facial feature analysis and the feature vectors obtained from the gait feature analysis into a single composite feature vector; b) Train separate classifiers for facial features and gait features, and each classifier outputs the probability of drinking alcohol. and ; c) Combine the two independent drinking probabilities into a final drinking probability through weighted average decision fusion. It satisfies the formula:

[0008] In the formula, Indicates the probability of ultimately drinking alcohol; , Indicates different weighting coefficients; This represents the probability of drinking alcohol output by the facial feature classifier. This represents the probability of drinking alcohol output by the gait feature classifier; d) When the final probability of drinking exceeds a preset threshold, the person is identified as a suspected drinker and their information is recorded in the suspected drinker table; Non-contact alcohol testing is conducted on drivers of vehicles exiting the parking lot in the key area, and corresponding intervention measures are implemented based on the test results.

[0009] Furthermore, the facial feature analysis includes face color feature analysis and eye feature analysis.

[0010] Furthermore, the method for facial feature analysis includes: The acquired facial feature images were converted to HSV or YCrCb color spaces, and the average intensity value and distribution of the Cr component were analyzed.

[0011] Furthermore, the method for analyzing eye features includes: Based on the acquired facial feature images, the pupil size, the relative position of the iris and eyelid, the eye movement rate, and the blinking frequency are detected.

[0012] Furthermore, the gait analysis method includes: Binary contour images of the human body are extracted from video frames using background subtraction or semantic segmentation techniques. Based on deep learning methods, spatiotemporal features related to drinking status are extracted from the binary contour map or RGB video. By measuring the stability index of movement at key points in the human skeleton, it can be determined whether gait has become abnormal due to alcohol consumption.

[0013] Furthermore, the step of sending a warning message to the corresponding vehicle owner based on the duration of the stay and the information of the suspected intoxicated person specifically includes: If the vehicle remains stationary for longer than the preset time, a preliminary warning text message will be sent to the vehicle owner. The vehicle information of a suspected intoxicated person is compared with the vehicle information in the vehicle parking time meter. If the vehicle registered under the name of the suspected intoxicated person is in the current area, a medium-level risk intervention is triggered.

[0014] Furthermore, the non-contact alcohol detection includes infrared thermal imaging and remote alcohol sensing, and uses a video recognition module to perform facial recognition of the driver's seat and driver to match them with a list of suspected drinkers, wherein: If facial recognition matches a person suspected of having consumed alcohol, or if the infrared thermal imaging and alcohol remote sensing detect an anomaly, a breathalyzer retest is triggered. If the breathalyzer test result indicates that the person has consumed alcohol, the test result will be uploaded to the public security network system, and the gate will be prohibited from allowing passage and a voice alarm will be triggered. If the breathalyzer test result is that no alcohol was consumed, the gate will be raised to allow passage; If the infrared thermal imaging, the alcohol remote sensing detection, and the video recognition module do not determine that the person is suspected of drinking alcohol, the gate will allow passage directly.

[0015] Furthermore, the breathalyzer test is performed by an alcohol breathalyzer detection device. After receiving the breathalyzer test command, the device automatically extends to the vicinity of the driver's cab window, where the driver performs the breathalyzer test.

[0016] Furthermore, the method also includes: When the vehicle leaves the designated key area, the driver's facial image is captured and their identity is verified. If the driver is identified as a suspected intoxicated person, his vehicle is marked as a high-risk target for investigation and the information is sent to the traffic police terminal system.

[0017] Preferably, the key areas include at least one of the following: a concentrated area of ​​restaurants, an area surrounding entertainment venues, and a bar district.

[0018] A second aspect of the present invention provides a system for preventing drunk driving in key urban areas using the method described above, comprising: The area management module is used to delineate key areas and associate video surveillance equipment at import and export points; The dwell timer module is used to record the time when a vehicle enters and exits and to calculate the dwell time. The facial behavior recognition module is configured to identify individuals suspected of drinking based on video images; The early warning and intervention module is used to send corresponding early warning information to the vehicle owner based on the dwell time measured by the dwell timer module and the recognition result of the facial behavior recognition module. The exit detection module, integrated into the parking lot exit gate, is used to conduct alcohol tests on drivers and control their passage.

[0019] Preferably, the outlet detection module integrates an infrared thermal imaging unit, an alcohol gas remote sensing unit, a video recognition unit, a breath re-inspection interface, and an LED screen display unit, wherein: The video recognition unit is used to identify the driver's seat and driver, and to perform facial recognition to match the list of suspected intoxicated persons; The infrared thermal imaging unit and the alcohol gas remote sensing unit are used to detect abnormal facial temperature and alcohol content in the driver's exhaled breath. The LED screen display unit is used to display the detection status and warning information; If the infrared thermal imaging unit and the alcohol gas remote sensing unit detect an anomaly, or if the video recognition unit identifies the driver as a suspected drinker, the breathalyzer retest interface is triggered to perform an alcohol breathalyzer test.

[0020] Compared with the prior art, the embodiments of the present invention have the following beneficial effects: This invention constructs a multi-level prevention and control system consisting of "key area monitoring - facial behavior recognition - parking lot exit interception," which enables pre-emptive identification, intelligent early warning, and proactive intervention of drunk driving behavior. This improves the accuracy and real-time nature of drunk driving prevention and control, effectively reduces police manpower consumption and law enforcement costs, and forms a complete management closed loop from risk discovery to handling. It also realizes a fundamental shift from "post-incident investigation" to "pre-incident prevention" of drunk driving behavior. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart disclosed in the present invention; Figure 2 This is an architectural diagram of the multi-level prevention and control system disclosed in this invention; Figure 3 This is a flowchart of the alcohol detection process at a parking lot exit disclosed in this invention. Detailed Implementation

[0023] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] This invention aims to provide a method for preventing drunk driving in key urban areas. This invention utilizes existing road and intersection video surveillance, checkpoint and street monitoring, high-point cameras and other equipment to collect video images and videos. It only requires modification of the closed parking lot exit gate system and connection to the public security command platform. The implementation cost and difficulty are relatively small, so as to form a closed-loop governance of "discovery-intervention-interception" three-line defense.

[0025] Please see Figure 1-2 The method will be explained in detail below.

[0026] Step 1: Based on the designated key areas for drunk driving in the city, obtain information from the video surveillance equipment at the entrances and exits of the key areas to monitor the time when vehicles enter and leave the key areas, and calculate the duration of vehicle stay in the key areas.

[0027] It should be noted that the key areas are urban road areas designated by traffic management departments that are prone to drunk driving and have a high concentration of restaurants and entertainment venues, including at least one of the following: concentrated restaurant areas, areas surrounding entertainment venues, and bar districts.

[0028] Traffic management authorities identify equipment associated with imports in key areas from the equipment list associated with those areas, and mark it as imported equipment in the key area equipment list, thus forming a closed area with the key area export equipment list. Similarly, traffic management authorities identify equipment associated with exports in key areas from the equipment list associated with those areas, and mark it as export equipment in the key area equipment list, thus forming a closed area with the key area import equipment list.

[0029] Vehicles entering the key area through entrances are identified by smart checkpoint cameras installed at the entrances to the key area, which record the license plate number and entry time. The data is then written to the "Vehicle Dwelling Timer". Vehicles exit the key area via video recognition equipment installed at the exit, and the exit time is recorded. The vehicle is marked as "left". Vehicle dwell time. It updates every minute, counting vehicles that have not left the area (those not marked as "left"), and the current time. Vehicle dwell time .

[0030] In this embodiment, when the vehicle stays for a certain period of time... If the time exceeds 30 minutes, a low-level risk intervention will be triggered, and the platform will immediately send a text message to the car owner's registered mobile phone number to remind the car owner.

[0031] Step 2: Collect images and analyze the behavioral characteristics of pedestrians in key areas to identify suspected intoxicated individuals, and send warning messages to the corresponding vehicle owners based on the duration of their stay and information about suspected intoxicated individuals.

[0032] Specifically, video recognition equipment is connected to key areas to capture faces and recognize behaviors of people on the street. Edge computing terminals then perform noise reduction, image acquisition, and behavioral feature analysis. The behavioral feature analysis includes at least facial feature analysis and gait analysis. Facial feature analysis further includes face color feature analysis and eye feature analysis.

[0033] In a further proposed solution, the edge computing terminal data perception and preprocessing methods include: inputting raw video and audio streams captured by the camera; using object detection algorithms (such as YOLO and SSD) to locate "people" in video frames; and using tracking algorithms (such as DeepSORT) to track the same person in consecutive frames, providing a stable target region for subsequent analysis. Specifically, denoising algorithms, such as Non-Local Means Denoising (NMWD) or deep learning-based denoising models (such as DnCNN), are applied to the video frames. These algorithms effectively smooth noise while preserving edges and details. For the audio, speech analysis (such as articulation) is used, employing spectral subtraction or deep learning-based speech enhancement algorithms for denoising.

[0034] In a further approach, facial landmarks (such as eyes, nose, and corners of the mouth) are precisely located in the preprocessed image, and an affine transformation is performed to align the face to a standard pose. A deep convolutional neural network (such as FaceNet or ArcFace) is then used to map the aligned face image to a high-dimensional feature vector (embedding vector), which is used to uniquely identify a person.

[0035] To further identify the risk of drunk driving, this invention uses a face-behavior recognition model, combining facial features, gait stability (OpenPose skeletal keypoint detection), and time features (0-4 AM is a high-risk period) for comprehensive evaluation. Facial feature analysis methods include: based on acquired facial feature images and the characteristics that alcohol consumption leads to sluggish pupillary response, drooping eyelids, and dull gaze, detecting pupil size, the relative position of the iris and eyelid (PERCLOS - Percentage of Eyelid Closure over the Pupil over Time, an indicator of drowsiness), eye movement rate, and blink frequency.

[0036] The facial feature analysis algorithm satisfies the following formula:

[0037] In the formula, the characteristic function The definition of pupillary reflex abnormality Eyelid movement abnormality Eye movement coordination abnormality and abnormal blinking pattern Weighted fusion. In a specific example, ; ; ; .

[0038] Pupil reflex abnormality Satisfying the expression:

[0039] In the formula, Indicates the current pupil area. Indicates the individual's baseline pupillary area. Indicates the current pupillary light reflex delay time. This indicates the baseline reflection delay time.

[0040] Eyelid movement abnormality Satisfying the expression:

[0041] In the formula, This indicates the current eyelid opening (distance from the upper eyelid to the lower eyelid). Indicates the baseline eyelid opening. This indicates the current percentage of eyelid ptosis. This indicates the baseline percentage of eyelid ptosis.

[0042] Eye movement coordination disorder Satisfying the expression:

[0043] In the formula, This represents the smooth tracking gain (eye velocity / target velocity, awake ≈ 1). It represents the standard deviation of eye speed (a measure of stability). This represents the average eye speed. Indicates the current scan accuracy. This indicates the accuracy of the baseline scan.

[0044] Blinking pattern abnormality Satisfying the expression:

[0045] In the formula, This indicates the current blinking frequency (times / minute). Indicates the baseline blink frequency. This indicates the current average blink duration. Indicates the baseline blink duration. This indicates the current blinking speed (opening and closing rate). This indicates the baseline blink rate.

[0046] Gait analysis methods include determining whether alcohol consumption has caused gait abnormalities by using a stability index of movement at key points in the human skeleton. The stability index is calculated according to the following expression:

[0047] In the formula, Indicates the first Coordinates of key points (such as ankle, knee, hip, etc.); This represents the variance of key points over time (a measure of jitter). The standard deviation of the velocity at key points (a measure of motion smoothness); The standard deviation of acceleration at key points (measures the abruptness of motion); This represents the total number of keypoints (e.g., 25 keypoints using OpenPose). =25).

[0048] A further proposed solution would combine the results of facial feature analysis and gait feature analysis to make a more accurate judgment. First, feature vectors obtained from facial feature analysis (such as facial redness and PERCLOS) and feature vectors obtained from gait feature analysis (such as gait cycle and stability index) are concatenated into a comprehensive feature vector; then, independent classifiers are trained for facial features and gait features respectively, and each classifier outputs the probability of drinking alcohol.

[0049] Then, the two independent drinking probabilities are combined into a final drinking probability through weighted average decision fusion, which satisfies the formula:

[0050] In the formula, Indicates the probability of ultimately drinking alcohol; , Indicates different weighting coefficients; This represents the probability of drinking alcohol output by the facial feature classifier. This represents the probability of drinking alcohol output by the gait feature classifier.

[0051] , The rules for setting the weighting coefficients are shown in Table 1 below: Table 1

[0052] Overall score If the value is greater than 0.7, the person is identified as a suspected drinker. Basic information about the person is then retrieved using facial recognition technology (including their ID card information, mobile phone number, driver's license information, and associated vehicle information), and the results are pushed to the system and stored in the "Suspected Drinker Table".

[0053] Finally, the vehicle information of the "suspected intoxicated person" in the identification results is compared with the vehicle information in the vehicle parking time meter. If the vehicle under the name of the "suspected intoxicated person" is in the current area, a medium-level risk intervention is triggered, and the platform immediately sends a text message to the registered mobile phone number of the vehicle owner to warn the vehicle owner.

[0054] Step 3: Conduct non-contact alcohol testing on drivers of vehicles exiting parking lots in key areas, and implement corresponding intervention measures based on the test results.

[0055] In a further proposed solution, the parking lot exit gates integrate infrared thermal imaging, video recognition modules, LED screen display devices (or voice broadcast prompt devices), and alcohol breathalyzer detection devices. Simultaneously, the gate system is connected to the public security network system.

[0056] In this solution, the gate detection system uses infrared thermal imaging combined with video recognition to verify whether someone has consumed alcohol. Specifically: Driver's seat presence score Satisfying the expression:

[0057] In the formula, Indicates the confidence level of face detection (0-1); In the 3D head pose, if the yaw angle is ∈ [-30°, 30°] and located in the driver's side region, the value increases by 1; otherwise, the value decreases. This indicates that a human thermal outline exists in the driver's seat area of ​​the infrared thermal image; =0.4, =0.3, =0.3.

[0058] Alcohol thermal response index Satisfying the expression:

[0059] In the formula, = Unit: °C; its denominator is the experimentally calibrated upper limit of typical positive temperature rise for alcohol (normalized), truncated to [0,1].

[0060] Micro-motion instability index Satisfying the expression:

[0061] In the formula, The standard deviation of the yaw angle is expressed in degrees (°). This indicates the number of blinks per minute; This represents the energy of the nose tip displacement FFT in the 0.5-3Hz frequency band; all terms are truncated to [0,1].

[0062] In a further proposed solution, ambient light and temperature data would be obtained. To calculate the credibility of a video:

[0063] Calculate infrared confidence level Its normalized weights:

[0064] The normalization weighting coefficients are set as shown in Table 2: Table 2

[0065] The final fusion score is:

[0066] If S≥0.75, the output will be: Suspected drunk driving, triggering the gate interception mechanism; If 0.6 ≤ S < 0.75, the output will be: There is a risk, prompting a breath test. If S < 0.6, the output is: Normal, and the passage is allowed.

[0067] like Figure 3 As shown, the specific procedure for the driver to open the window for inspection when the vehicle arrives at the exit gate includes: (1) The driver's seat and driver are identified by the video recognition module to determine whether the person being tested is the driver when the subsequent detection behavior occurs; at the same time, face recognition is performed to obtain driver information and match it with the list of suspected drinkers; if the driver information is in the list of suspected drinkers, a breathalyzer test is required. The gate device automatically extends the alcohol breathalyzer test device for the driver to blow into the breath, and the LED prompt screen displays the corresponding prompt, such as "Hello driver, the system has determined that you are suspected of drinking alcohol. Please conduct a breathalyzer test."

[0068] (2) The system uses an infrared thermal imaging and video recognition fusion algorithm to determine whether the driver has been drinking. The detection results are transmitted to the gate control system in real time. The system combines video recognition and facial recognition data for multi-source fusion judgment. When the combined score of the infrared thermal imaging and video recognition fusion algorithm is greater than 0.6, the system automatically triggers the early warning mechanism, retrieves the list of suspected drinkers for secondary comparison, locks the target after confirmation, and starts the breathalyzer test process. At the same time, the LED screen displays the early warning information.

[0069] (3) When the above test results meet the breath test retest standard, the breath test device will automatically extend to the vicinity of the driver's cab window and the driver will conduct the breath test; if the test result is that the driver has been drinking, the test result will be uploaded to the public security network system and the LED screen will issue an alarm message, such as "You have been drinking and driving and cannot leave the parking lot. The information has been synchronized with the public security network system." The gate will not allow passage; if the test result is that the driver has not been drinking, the result will not be pushed to the public security network system and the gate will be raised to allow passage.

[0070] (4) If the video recognition module, infrared thermal imaging, and alcohol remote sensing detection do not determine that the person is suspected of drinking alcohol, the gate will allow the person to pass directly without the need for a breathalyzer test.

[0071] In addition, when a vehicle leaves a key area, the system captures the driver's facial image and performs identity verification. First, vehicles that remain for more than 30 minutes are marked as general screening targets, and the platform pushes them to the police linkage terminal held by traffic police to provide information support for traffic management departments to set up checkpoints. Second, if the driver's information is recorded in the "Suspected Intoxicated Persons List," the vehicle is marked as a high-risk screening target, and the vehicle information, capture location, driving direction, and captured image are sent to the system. The system then directs and dispatches nearby police forces to set up control based on the capture location information.

[0072] This invention also protects a system for preventing drunk driving in key urban areas using the above-described method. The system mainly includes an area management module, a stop-time tracking module, a facial recognition module, an early warning and intervention module, and an exit detection module. The area management module is used to delineate key areas and associate them with video surveillance equipment at entrances and exits. The stop-time tracking module records vehicle entry and exit times and calculates the duration of stay. The facial recognition module is configured to identify suspected intoxicated individuals based on video images. The early warning and intervention module sends corresponding early warning information to vehicle owners based on the stay duration recorded by the stop-time tracking module and the recognition results from the facial recognition module. The exit detection module is integrated into the parking lot exit gate for conducting alcohol tests on drivers and controlling their passage.

[0073] The exit detection module integrates a video recognition unit, an infrared thermal imaging unit, an alcohol gas remote sensing unit, a breathalyzer retest interface, and an LED screen or voice prompt unit. The video recognition unit identifies the driver and the driver's seat, and performs facial recognition to match the suspected intoxication list; the infrared thermal imaging unit and the alcohol gas remote sensing unit detect abnormal facial temperature and alcohol content in the driver's exhaled breath; the LED screen or voice prompt unit displays the detection status and warning information; if the video recognition unit identifies the driver as a suspected intoxicated person or the infrared thermal imaging unit and the alcohol gas remote sensing unit detect an anomaly, the breathalyzer retest interface is triggered to perform an alcohol breathalyzer test.

[0074] It should be noted that the alcohol detection module at the exit gate can be replaced with a combination of "laser alcohol remote sensing + millimeter wave personnel detection".

[0075] In addition, high-point PTZ cameras can be replaced by drones for patrol, covering temporary large-scale event areas, and can be connected to surveillance videos in catering and entertainment venues.

[0076] This invention integrates "time spent at regional checkpoints" with "face-behavior recognition" to create accurate profiles of vehicles at risk of drunk driving. It also proposes for the first time to link "parking lot exit gates + remote sensing alcohol detection" with checkpoints in key urban areas to form a third physical interception line. At the same time, multi-source data (checkpoints, street-side high-point video surveillance, parking lots) are fused at the edge to reduce the central computing pressure and improve real-time performance.

[0077] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for preventing drunk driving in key urban areas, characterized in that, Includes the following steps: Based on the designated key areas for drunk driving in the city, information from video surveillance equipment at the entrances and exits of the key areas is obtained, the time when vehicles enter and leave the key areas is monitored, and the duration of vehicle stay in the key areas is calculated. Image acquisition and behavioral feature analysis are performed on pedestrians within the key area to identify suspected intoxicated individuals. Based on the duration of their stay and information about the suspected intoxicated individuals, a warning message is sent to the corresponding vehicle owner. The behavioral feature analysis includes at least facial feature analysis and gait analysis. The method for identifying suspected intoxicated individuals includes: a) Concatenate the feature vectors obtained from the facial feature analysis and the feature vectors obtained from the gait feature analysis into a single composite feature vector; b) Train separate classifiers for facial features and gait features, and each classifier outputs the probability of drinking alcohol. and ; c) Combine the two independent drinking probabilities into a final drinking probability through weighted average decision fusion. It satisfies the formula: In the formula, Indicates the probability of ultimately drinking alcohol; , Indicates different weighting coefficients; This represents the probability of drinking alcohol output by the facial feature classifier. This represents the probability of drinking alcohol output by the gait feature classifier; d) When the final probability of drinking exceeds a preset threshold, the person is identified as a suspected drinker and their information is recorded in the suspected drinker table; Non-contact alcohol testing is conducted on drivers of vehicles exiting the parking lot in the key area, and corresponding intervention measures are implemented based on the test results.

2. The method for preventing drunk driving in key urban areas according to claim 1, characterized in that, The facial feature analysis includes complexion feature analysis and eye feature analysis.

3. The method for preventing drunk driving in key urban areas according to claim 2, characterized in that, The methods for facial feature analysis include: The acquired facial feature images were converted to HSV or YCrCb color spaces, and the average intensity value and distribution of the Cr component were analyzed.

4. The method for preventing drunk driving in key urban areas according to claim 2, characterized in that, The methods for analyzing eye features include: Based on the acquired facial feature images, the pupil size, the relative position of the iris and eyelid, the eye movement rate, and the blinking frequency are detected.

5. The method for preventing drunk driving in key urban areas according to claim 2, characterized in that, The gait analysis method includes: Binary contour images of the human body are extracted from video frames using background subtraction or semantic segmentation techniques. Based on deep learning methods, spatiotemporal features related to drinking status are extracted from the binary contour map or RGB video. By measuring the stability index of movement at key points in the human skeleton, it can be determined whether gait has become abnormal due to alcohol consumption.

6. The method for preventing drunk driving in key urban areas according to claim 1, characterized in that, The step of sending a warning message to the corresponding vehicle owner based on the duration of stay and the information of the suspected intoxicated person specifically includes: If the vehicle remains stationary for longer than the preset time, a preliminary warning text message will be sent to the vehicle owner. The vehicle information of a suspected intoxicated person is compared with the vehicle information in the vehicle parking time meter. If the vehicle registered under the name of the suspected intoxicated person is in the current area, a medium-level risk intervention is triggered.

7. The method for preventing drunk driving in key urban areas according to claim 1, characterized in that, The non-contact alcohol detection includes infrared thermal imaging and remote alcohol sensing, and uses a video recognition module to identify the driver's seat and driver's face to match them with a list of suspected drinkers. If facial recognition matches a person suspected of having consumed alcohol, or if the infrared thermal imaging and alcohol remote sensing detect an anomaly, a breathalyzer retest is triggered. If the breathalyzer test result indicates that the person has consumed alcohol, the test result will be uploaded to the public security network system, and the gate will be prohibited from allowing passage and a voice alarm will be triggered. If the breathalyzer test result is that no alcohol was consumed, the gate will be raised to allow passage; If the infrared thermal imaging, the alcohol remote sensing detection, and the video recognition module do not determine that the person is suspected of drinking alcohol, the gate will allow passage directly.

8. The method for preventing drunk driving in key urban areas according to claim 7, characterized in that, The breathalyzer test is performed by an alcohol breathalyzer detection device. After receiving the breathalyzer test command, the device automatically extends to the vicinity of the driver's cab window, where the driver performs the breathalyzer test.

9. The method for preventing drunk driving in key urban areas according to claim 1, characterized in that, The method further includes: When the vehicle leaves the designated key area, the driver's facial image is captured and their identity is verified. If the driver is identified as a suspected intoxicated person, his vehicle is marked as a high-risk target for investigation and the information is sent to the traffic police terminal system.

10. The method for preventing drunk driving in key urban areas according to claim 1, characterized in that, The key areas include at least one of the following: concentrated dining areas, areas surrounding entertainment venues, and bar districts.

11. A system for preventing drunk driving in key urban areas using the method described in any one of claims 1-10, characterized in that, include: The area management module is used to delineate key areas and associate video surveillance equipment at import and export points; The dwell timer module is used to record the time when a vehicle enters and exits and to calculate the dwell time. The facial behavior recognition module is configured to identify individuals suspected of drinking based on video images; The early warning and intervention module is used to send corresponding early warning information to the vehicle owner based on the dwell time measured by the dwell timer module and the recognition result of the facial behavior recognition module. The exit detection module, integrated into the parking lot exit gate, is used to conduct alcohol tests on drivers and control their passage.

12. The system for preventing drunk driving in key urban areas according to claim 11, characterized in that, The outlet detection module integrates an infrared thermal imaging unit, an alcohol gas remote sensing unit, a video recognition unit, a breath re-inspection interface, and an LED screen display unit, wherein: The video recognition unit is used to identify the driver's seat and driver, and to perform facial recognition to match the list of suspected intoxicated persons; The infrared thermal imaging unit and the alcohol gas remote sensing unit are used to detect abnormal facial temperature and alcohol content in the driver's exhaled breath. The LED screen display unit is used to display the detection status and warning information; If the video recognition unit identifies the driver as a suspected intoxicated person or the infrared thermal imaging unit and the alcohol gas remote sensing unit detect an anomaly, the breathalyzer retest interface is triggered to perform an alcohol breathalyzer test.

Images