Detection method and device, vehicle and storage medium

Abstract

The invention relates to a detection method and device, a vehicle and a storage medium, the detection method is used for the vehicle, the vehicle comprises a first detection module, an air blowing type alcohol detector and a camera, and the method comprises the steps that air in the vehicle is collected through the first detection module to detect whether the air in the vehicle contains alcohol or not; when the air in the vehicle contains alcohol, first video data of a first driver of the vehicle are collected through a camera, and the first video data comprise action features and expression features of the first driver; sending the action features and the expression features to a first recognition model to determine the drunk probability of the first driver; when the drinking probability is larger than a first preset value, the alcohol concentration of gas exhaled by the first driver is detected through a blowing type alcohol detector; when the alcohol concentration is larger than or equal to the second preset value, the vehicle is controlled to be locked, and an alarm is given out, the probability of drunk driving behaviors is reduced, and the inspection efficiency of the drunk driving behaviors is improved.

Description

Technical Field

[0001] This invention relates to the field of display technology, and more specifically to a detection method, apparatus, vehicle, and storage medium. Background Technology

[0002] Existing technologies and traditional methods for detecting drunk driving have drawbacks. They rely heavily on roadside enforcement and spot checks, which are passive management methods that are difficult to detect drunk driving behavior comprehensively and in a timely manner. They also have blind spots and time lags, making it impossible to detect all vehicles in real time, resulting in low efficiency in detecting drunk driving behavior. Summary of the Invention

[0003] The purpose of this invention is to provide a detection method, device, vehicle, and storage medium that improves the efficiency of detecting drunk driving behavior.

[0004] To achieve the objectives of this invention, the following technical solution is provided:

[0005] In a first aspect, the present invention provides a detection method for a vehicle, the vehicle including a first detection module, a breathalyzer, and a camera. The method includes: collecting air inside the vehicle through the first detection module to detect whether the air inside the vehicle contains alcohol; when the air inside the vehicle contains alcohol, collecting first video data of a first driver of the vehicle through the camera, the first video data including the first driver's action features and facial expression features; sending the action features and the facial expression features to a first recognition model to determine the probability that the first driver has consumed alcohol; when the probability of consuming alcohol is greater than a first preset value, detecting the alcohol concentration of the first driver's exhaled breath through the breathalyzer; when the alcohol concentration is greater than or equal to a second preset value, controlling the vehicle to lock and issuing a warning.

[0006] In this embodiment, the first detection module collects air from inside the vehicle to detect whether the air contains alcohol. When the air contains alcohol, a camera collects first video data of the first driver, including the driver's motion and facial features. The motion and facial features are sent to a first recognition model to determine the probability that the driver has consumed alcohol. When the probability of consuming alcohol is greater than a first preset value, a breathalyzer is used to detect the alcohol concentration in the driver's exhaled breath. When the alcohol concentration is greater than or equal to a second preset value, the vehicle is locked and a warning is issued. This method allows for a tiered testing of whether a driver has consumed alcohol, improving the comprehensiveness of the detection, reducing the probability of drunk driving, and enhancing driving safety.

[0007] In one embodiment, sending the action features and the facial expression features to a first recognition model to determine the probability that the first driver has consumed alcohol includes: using the first recognition model to recognize the action features to determine a first probability that the first driver has consumed alcohol; using the first recognition model to recognize the facial expression features to determine a second probability that the first driver has consumed alcohol; and determining the probability of consuming alcohol based on the first probability and the second probability.

[0008] In this embodiment, the method for determining the probability that the first driver has consumed alcohol is further defined in detail. First, the action features and facial expression features are fed into the first recognition model, which identifies the head movement speed and upper limb movement speed in the action features, and the pupil focusing degree and facial color in the facial expression features. Through model processing, a first probability and a second probability are obtained. Then, the two are combined to finally determine the probability that the first driver has consumed alcohol. This step-by-step recognition and comprehensive judgment allows for a more detailed and comprehensive determination of the probability of consuming alcohol. Implementing this method can effectively improve the accuracy of drunk driving detection. By recognizing the action features and facial expression features separately and obtaining the corresponding probabilities, and then comprehensively determining them, the driver's state can be analyzed in a comprehensive and detailed manner, avoiding misjudgments due to the randomness or uncertainty of a single feature, and ensuring a more reliable detection of whether the driver has consumed alcohol.

[0009] In one embodiment, the step of using the first recognition model to identify the action features to determine a first probability that the first driver has consumed alcohol includes: the action features including a first head movement speed and a first upper limb movement speed of the first driver; determining a preset head movement speed corresponding to the first driver consuming alcohol; using the first recognition model to identify the matching degree between the first head movement speed and the preset head movement speed to determine a first sub-probability that the first driver has consumed alcohol; determining a preset upper limb movement speed corresponding to the first driver consuming alcohol; using the first recognition model to identify the matching degree between the first upper limb movement speed and the preset upper limb movement speed to determine a second sub-probability that the first driver has consumed alcohol; and determining the first probability based on the first sub-probability and the second sub-probability.

[0010] In this embodiment, preset head movement speed and preset upper limb movement speed are determined for the head movement speed and upper limb movement speed in the movement features. The preset speed is a standard reference value obtained by statistical analysis of a large amount of movement feature data of drivers who have been drinking. Then, the matching degree between the driver's actual head movement speed and the preset head movement speed, and the matching degree between the driver's actual limb movement speed and the preset upper limb movement speed are calculated using the first recognition model. The corresponding sub-probabilities are determined based on these two matching degrees. Finally, the first probability of the movement feature is obtained by combining these two sub-probabilities.

[0011] This helps to more accurately identify whether a driver is in an abnormal state of intoxication, avoids judgment bias caused by focusing only on a certain action feature, and enhances the reliability and effectiveness of the entire drunk driving detection system. This hierarchical method of identification and probability determination down to each action feature makes the first probability more scientific and reasonable.

[0012] In one embodiment, the step of using the first recognition model to identify the facial expression features to determine a second probability that the first driver has consumed alcohol includes: the facial expression features including a first pupil focus and a first facial color of the first driver; determining a preset pupil focus corresponding to the first driver consuming alcohol; using the first recognition model to identify the matching degree between the preset pupil focus and the first pupil focus to determine a third sub-probability that the first driver has consumed alcohol; determining a preset facial color corresponding to the first driver consuming alcohol; using the first recognition model to identify the matching degree between the preset facial color and the first facial color to determine a fourth sub-probability that the first driver has consumed alcohol; and determining the second probability based on the third sub-probability and the fourth sub-probability.

[0013] In this embodiment, facial expression features involve pupil focus and facial color. Similarly, preset pupil focus and preset facial color corresponding to alcohol consumption are first determined. These two preset values ​​are typical feature values ​​derived from statistical analysis of a large amount of facial expression data of drivers who have consumed alcohol. Then, a first recognition model is used to calculate the matching degree between the actual first pupil focus and the preset pupil focus, and the matching degree between the first facial color and the preset facial color. Based on these two matching degrees, the corresponding third and fourth sub-probabilities are determined. Finally, these two sub-probabilities are combined to obtain the second probability. This method can more sensitively capture subtle abnormalities in the facial expressions of drivers who have consumed alcohol, which helps to more accurately identify whether a driver has consumed alcohol, avoiding misjudgments or omissions caused by inaccurate facial expression feature judgments. The determination of the second probability more accurately reflects the driver's possible state of intoxication. This method can effectively improve the comprehensiveness, reliability, and accuracy of judging the probability of a driver consuming alcohol based on facial expression features.

[0014] In one embodiment, detecting the alcohol concentration of the first driver's exhaled breath using the breathalyzer includes: using the camera to determine whether the subject of the test is the first driver; when the subject of the test is the first driver, controlling the breathalyzer to contact the first driver's lips; controlling the breathalyzer to collect the exhaled breath of the first driver; and controlling the breathalyzer to obtain the alcohol concentration of the exhaled breath.

[0015] In this embodiment, a camera is used to determine whether the subject of the test is the primary driver, ensuring the targetedness and accuracy of the test and preventing others from taking the test on their behalf. Only when it is confirmed that the subject is the primary driver will the breathalyzer contact the driver's lips, guiding the driver to blow into the breath correctly, and then collecting the exhaled breath and obtaining the alcohol concentration data. This series of steps ensures the standardization, effectiveness, and accuracy of the breathalyzer test process, enabling precise determination of the alcohol concentration in the driver's exhaled breath.

[0016] In one embodiment, controlling the vehicle to lock and issuing a warning includes: the vehicle includes a steering device, a power device, and an audible device; the steering device is used to steer the vehicle; the power device is used to drive the vehicle; controlling the steering device and the power device to lock; and controlling the audible device to emit a warning sound.

[0017] This embodiment further clarifies the specific operational methods for controlling vehicle locking and issuing warnings. The vehicle's steering and power systems are key components for vehicle operation; locking them can immediately stop the vehicle from moving, preventing the continuation of drunk driving and its potential harm. Simultaneously, controlling the sound device to emit a warning sound can promptly alert the driver and surrounding personnel that the vehicle is in an abnormal state, drawing attention and serving as a warning and deterrent. This comprehensive control measure, addressing both vehicle operation and on-site warning aspects, effectively counters drunk driving situations, ensures traffic safety, and facilitates timely detection and handling of drunk driving by traffic enforcement personnel.

[0018] In one embodiment, the method further includes: when the vehicle is locked, detecting the alcohol concentration of the second driver's exhaled breath using the breathalyzer; when the alcohol concentration is zero, controlling the vehicle to unlock and stopping the warning.

[0019] In this implementation, when a vehicle is locked due to a detected driver having consumed alcohol, considering the possibility that other drivers who have not consumed alcohol may need to use the vehicle, a step is added to test the alcohol concentration of a second driver using a breathalyzer. If the test result shows an alcohol concentration of zero, indicating that the second driver has not consumed alcohol, then the vehicle is unlocked, the warning is stopped, and the vehicle's normal driving status is restored. This process demonstrates that while strictly regulating drunk driving, the normal use needs of the vehicle are also taken into account, avoiding unnecessary trouble for users due to misjudgment or special circumstances causing the vehicle to be unusable for a long time. This improves the system's flexibility and user-friendliness.

[0020] Secondly, this application also provides a detection apparatus, including a method for performing the method provided in the first aspect or any embodiment of the first aspect.

[0021] Thirdly, this application also provides a vehicle including a processor, a memory, and at least one program, wherein the at least one program is stored in the memory and configured to be executed by the processor, the program including instructions for performing the method provided in the first aspect or any embodiment of the first aspect.

[0022] Fourthly, this application also provides a computer-readable storage medium storing a computer program that causes a computer to execute the method provided in the first aspect or any embodiment of the first aspect. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0024] Figure 1 This is a schematic diagram illustrating an application scenario of a detection method according to one embodiment.

[0025] Figure 2 This is a flowchart illustrating a detection method according to one embodiment;

[0026] Figure 3 This is a schematic diagram of the structure of a detection device according to one embodiment;

[0027] Figure 4 This is a schematic diagram of the structure of a vehicle component device according to one embodiment.

[0028] Explanation of reference numerals in the attached figures:

[0029] 101-User, 102-Vehicle, 103-Server, 300-Detection device, 301-Acquisition module, 302-Processing module, 401-Processor, 402-Memory, 403-At least one program. Detailed Implementation

[0030] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. 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.

[0031] It should be noted that when a component is said to be "fixed" to another component, it can be directly on the other component or it can be in a middle component. When a component is said to be "connected" to another component, it can be directly connected to the other component or it may be in a middle component.

[0032] Unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used in this invention includes any and all combinations of one or more of the associated listed items.

[0033] The following detailed description of some embodiments of the present invention is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0034] Please see Figure 1 , Figure 1 This is a schematic diagram illustrating an application scenario of a detection method according to one embodiment. For example... Figure 1 As shown in the diagram, the application scenario includes a user 101, a vehicle 102, and a server 103. Optionally, the vehicle 102 includes at least one first detection module, at least one breathalyzer, and at least one camera. The vehicle 102 can be a gasoline-powered vehicle, an electric vehicle, or a hybrid vehicle; this application does not limit the type and structure of the vehicle 102.

[0035] Optionally, server 103 can be a standalone server 103, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms. Server 103 can also be implemented through a cluster of server 103 composed of multiple sub-servers.

[0036] It should be noted that, Figure 1 The number and form of each device in the system shown, as well as the number of users 101 and servers 103, are for illustrative purposes only and do not constitute a limitation on the embodiments of this application. One server 103 is connected to multiple vehicles 102.

[0037] User 101 can operate vehicle 102 to control vehicle 102 to perform corresponding actions. Server 103 may include multiple personal computers (PCs), cloud virtual hosts, cloud databases, and cloud storage, etc.

[0038] The following describes the detection method provided in the embodiments of this application. The detection method can be executed by a detection device, which can be implemented by software and / or hardware, and can generally be integrated into the vehicle 102 or the server 103.

[0039] Please refer to Figure 2 , Figure 2 This is a flowchart illustrating a detection method according to one embodiment. Taking the application of this method to the processing of a detection method as an example, the detection device may include a server or electronic device. The method includes the following steps S201-S205, wherein,

[0040] S201: The first detection module collects air from inside the vehicle to detect whether the air inside the vehicle contains alcohol.

[0041] S202: When the air inside the vehicle contains alcohol, the first video data of the first driver of the vehicle is collected by a camera. The first video data includes the first driver's action characteristics and facial expression characteristics.

[0042] S203: Send the action features and facial expression features to the first recognition model to determine the probability that the first driver has been drinking.

[0043] The action and facial expression features are sent to the first recognition model to determine the probability that the first driver has been drinking, including:

[0044] The first recognition model is used to identify action features to determine the first probability that the first driver has been drinking.

[0045] Using a first recognition model to identify action features to determine a first probability that the first driver has consumed alcohol, the method includes: identifying action features including the first head movement speed and the first upper limb movement speed of the first driver; determining a preset head movement speed corresponding to the first driver consuming alcohol; using the first recognition model to identify the matching degree between the first head movement speed and the preset head movement speed to determine a first sub-probability that the first driver has consumed alcohol; determining a preset upper limb movement speed corresponding to the first driver consuming alcohol; using the first recognition model to identify the matching degree between the first upper limb movement speed and the preset upper limb movement speed to determine a second sub-probability that the first driver has consumed alcohol; and determining the first probability based on the first sub-probability and the second sub-probability.

[0046] Optionally, the upper limb includes the shoulder, upper arm, forearm, and hand, which are not limited in this application.

[0047] In one embodiment, optionally, the first probability can be a weighted sum of the first sub-probability and the second sub-probability, or it can be the maximum or minimum value between the first sub-probability and the second sub-probability, etc., and is not limited thereto. The preset weights of the first sub-probability and the second sub-probability are preset according to requirements, and are not limited in this application.

[0048] In one embodiment, optionally, the first recognition model is derived from artificial intelligence. Optionally, the action features include, but are not limited to, head movements (such as nodding, shaking the head, looking down at a phone), hand gestures (such as making hand gestures, steering wheel operation amplitude and rhythm), and body movements (such as leaning forward, leaning back). The first recognition model achieves real-time recognition, capable of rapidly processing video or image data streams and promptly feeding back the driver's action feature analysis results during vehicle operation, with latency controlled within a preset millisecond to meet the needs of real-time warnings and driver assistance. The first recognition model exhibits robustness to different lighting conditions, individual driver differences (such as body type, clothing), and changes in the vehicle's internal environment, ensuring that the recognition accuracy remains stable above a preset probability in various real-world driving scenarios.

[0049] Optionally, the action feature recognition using the first recognition model includes the following recognition steps (optionally, this step can also be used to recognize facial expression features using the first recognition model; this recognition step can be applied to all situations requiring recognition in this application, depending on the circumstances). First, multiple high-definition cameras are installed inside the vehicle, positioned to target key perspectives such as the driver's front, side, and upper body side views, ensuring comprehensive capture of the driver's action features. The cameras should have a high frame rate and appropriate resolution (e.g., 1080p or higher) to provide sufficiently clear and smooth image data for analysis. Experimental scenarios for collecting the first video data are designed, covering different time periods (daytime, nighttime, dusk, etc.), different weather conditions (sunny, rainy, cloudy, etc.), different road types (urban roads, highways, rural roads, etc.), and different driving tasks (normal driving, turning, parking, etc.) to ensure the diversity and representativeness of the collected data. Simultaneously, drivers of different ages, genders, and body types are invited to participate in the experiment to fully consider individual differences. During the data collection process, key action events are marked with timestamps and type labels, such as "the driver started looking down at his phone at second X and lasted for X seconds," for subsequent data annotation and model training.

[0050] The second step involves assembling a professional annotation team to perform detailed motion feature annotations on the collected initial video data. Annotations include key information such as the type of driver's movements (e.g., specific types of head movements), start and end times, movement amplitude, and movement frequency. Strict annotation standards and quality control procedures are established, and annotation results are regularly sampled, checked, and evaluated to ensure the accuracy and consistency of the annotated data. The annotation accuracy should reach or exceed the initial accuracy (preset according to requirements) to meet the needs of model training.

[0051] The third step involves selecting a suitable primary recognition model architecture for processing video and image sequence data, based on project requirements and the specific needs for motion feature recognition. For example, considering that motion feature recognition requires combining temporal and spatial information, a hybrid model architecture combining convolutional neural networks (CNNs) with recurrent neural networks (RNNs) or long short-term memory (LSTM) networks can be chosen, such as a CNN-LSTM model. The CNN portion extracts spatial features from the image (e.g., the position and posture of the driver's body parts), while the LSTM portion captures the changing patterns and dependencies of the motion over time. The performance of different model architectures on similar tasks is compared and evaluated (e.g., accuracy, recall, F1 score, etc.) as well as model complexity (e.g., number of parameters, training and inference time). The model that meets the recognition performance requirements while having relatively low computational resource consumption and fast inference speed is selected as the foundational architecture for the primary recognition model.

[0052] The fourth step involves preprocessing the acquired first video data to improve the model's training efficiency and recognition performance. Preprocessing steps include: Grayscale conversion: If the model has low sensitivity to color information, color images can be converted to grayscale to reduce data dimensionality and computational load. Size normalization: All images are uniformly adjusted to the fixed size required for model input (e.g., 224×224 or 128×128 pixels) to ensure image data consistency and facilitate model processing. Noise removal: Filtering algorithms (such as Gaussian filtering, median filtering, etc.) are used to remove noise points in the images, improving image quality and reducing noise interference with model training. Data augmentation: Through random rotation, translation, scaling, flipping, and other image transformation operations, the size of the training dataset is expanded to enhance the model's adaptability to different viewpoints, poses, and environmental changes, thereby improving the model's generalization performance.

[0053] The fifth step is to divide the labeled dataset into a certain proportion (e.g., 70% as the training set, 15% as the validation set, and 15% as the test set). The training set is used for model parameter learning and optimization, the validation set is used to periodically evaluate the model's performance during training to prevent overfitting and adjust the model's hyperparameters based on the validation results, and the test set is used for final performance evaluation after the model training is completed to ensure that the model has reliable generalization ability on unseen data.

[0054] The sixth step is to select a suitable optimization algorithm for model training and set key training parameters such as the learning rate (the initial learning rate can be set as the first learning rate and adjusted using a learning rate decay strategy based on changes in validation loss during training), batch size (e.g., 32 or 64), and number of training epochs (determined based on dataset size and model convergence). Define the model's loss function; for multi-class action feature recognition tasks, the cross-entropy loss function can be used. Simultaneously, evaluation metrics such as accuracy, recall, and F1 score should be used to assess the model's training process.

[0055] Step 7: The first recognition model is trained in batches using the training set data. In each batch, preprocessed image data is input into the model, and forward propagation is used to calculate the model's output and loss value. Then, the model parameters are updated using the backpropagation algorithm to gradually optimize the model's performance. During training, the model is periodically validated using a validation set, and changes in validation loss and evaluation metrics are recorded. If the validation loss no longer decreases within several consecutive training epochs, or if overfitting occurs (the training set loss continues to decrease while the validation set loss begins to increase), training is stopped, and the current optimal model parameters are saved. Based on the validation results, the model's hyperparameters (such as learning rate, number of network layers, number of neurons, etc.) are adjusted and optimized, and the training process is repeated until the model achieves satisfactory performance on the validation set.

[0056] Step 8: After model training, a comprehensive performance evaluation is conducted using an independent test set. Evaluation metrics include accuracy, recall, precision, and confusion matrix. A detailed analysis of the model's recognition performance for different types of driver action features is performed to identify its strengths, weaknesses, and action categories prone to misclassification. Receiver action feature curves and precision-recall curves are plotted to visually demonstrate the model's performance at different thresholds, further determining the optimal decision threshold to balance accuracy and recall and meet the needs of practical applications.

[0057] Step nine involves taking corresponding optimization measures based on the model evaluation results, targeting action categories with poor recognition performance or existing problems. For example, for certain action categories with low recognition accuracy, the reasons can be analyzed, such as insufficient data, inaccurate labeling, or insufficient model representation of these features. If it's a data issue, supplementary training data can be added and retraining can be performed. If it's a model issue, consider adding a specific extraction layer for these action features in the model architecture or adjusting the model's parameter configuration to enhance the model's recognition ability for these features. If the model has high real-time requirements in practical applications, but the current model's inference speed is slow, model compression techniques (such as pruning, quantization, and knowledge distillation) can be used to optimize the model. While maintaining the model's recognition performance as much as possible, the model's size and computational load can be reduced, improving the model's inference efficiency to meet the real-time processing requirements of the in-vehicle environment.

[0058] The tenth step is to integrate the trained first recognition model into the vehicle's electronic control system or dedicated driver assistance system. This involves developing corresponding software interfaces and data transmission modules to enable real-time transmission of image data captured by the camera to the model for processing, and feeding back the driver action feature recognition results output by the model to other relevant vehicle systems (such as fatigue warning systems and autonomous driving control systems). Ensuring the stability and reliability of the system during vehicle operation requires adapting and optimizing hardware devices (such as cameras and computing platforms), and properly handling resource allocation and task scheduling during model operation to prevent recognition function failure due to system resource contention or hardware malfunctions.

[0059] In this embodiment, the method for determining the probability that the first driver has consumed alcohol is further defined in detail. First, the action features and facial expression features are fed into the first recognition model, which identifies the head movement speed and upper limb movement speed in the action features, and the pupil focusing degree and facial color in the facial expression features. Through model processing, a first probability and a second probability are obtained. Then, the two are combined to finally determine the probability that the first driver has consumed alcohol. This step-by-step recognition and comprehensive judgment allows for a more detailed and comprehensive determination of the probability of consuming alcohol. Implementing this method can effectively improve the accuracy of drunk driving detection. By recognizing the action features and facial expression features separately and obtaining the corresponding probabilities, and then comprehensively determining them, the driver's state can be analyzed in a comprehensive and detailed manner, avoiding misjudgments due to the randomness or uncertainty of a single feature, and ensuring a more reliable detection of whether the driver has consumed alcohol.

[0060] In this embodiment, preset head movement speed and preset upper limb movement speed are determined for the head movement speed and upper limb movement speed in the movement features. The preset speed is a standard reference value obtained by statistical analysis of a large amount of movement feature data of drivers who have been drinking. Then, the matching degree between the driver's actual head movement speed and the preset head movement speed, and the matching degree between the driver's actual limb movement speed and the preset upper limb movement speed are calculated using the first recognition model. The corresponding sub-probabilities are determined based on these two matching degrees. Finally, the first probability of the movement feature is obtained by combining these two sub-probabilities.

[0061] This helps to more accurately identify whether a driver is in an abnormal state of intoxication, avoids judgment bias caused by focusing only on a certain action feature, and enhances the reliability and effectiveness of the entire drunk driving detection system. This hierarchical method of identification and probability determination down to each action feature makes the first probability more scientific and reasonable.

[0062] In one embodiment, optionally, in addition to the first head movement speed and the first upper limb movement speed, the first video data also includes the driver's body tilt angle and seat adjustment frequency. For example, changes in the driver's body tilt angle are detected by a pressure sensor installed under the seat, and the seat adjustment frequency is recorded by the control signal of the seat adjustment motor. These refined motion features can more comprehensively reflect the driver's physiological and behavioral state, providing richer information for alcohol consumption identification. To accurately collect these refined motion features, corresponding sensors need to be deployed in the vehicle, and the sensor data needs to be collected and processed in real time. Simultaneously, feature extraction and normalization processing are performed on the collected data before it is input into the first recognition model for analysis.

[0063] In one embodiment, optionally, at least one of several matching degree calculation methods, such as dynamic time warping, correlation coefficient calculation, and Euclidean distance, is used to evaluate head movement velocity and upper limb movement velocity. For example, the results of different methods are considered comprehensively to improve the accuracy and reliability of the matching degree evaluation. For instance, for time-series data of head movement velocity, the dynamic time warping algorithm can be used to calculate its similarity to a preset head movement velocity template; simultaneously, the correlation coefficient between the two is calculated to assess their linear correlation. Finally, the results of these two methods are fused to obtain a comprehensive matching degree score. To achieve this extension, in-depth research and implementation of different matching degree calculation methods are needed, and a suitable fusion strategy needs to be determined to ensure the accuracy and reliability of the evaluation results.

[0064] Optionally, a dynamic time warping algorithm is used to identify the similarity between the first head movement velocity and the preset head movement velocity (optionally, this step can also be used to identify the matching degree between the preset pupil focus and the first pupil focus using the first recognition model; this step can be applied to all cases in this application where the matching degree needs to be calculated, depending on the circumstances), and this similarity is regarded as the matching degree. The first step is to obtain data on the first head movement velocity and the preset head movement velocity. This data is typically a series of velocity values ​​that change over time. Assume the first head movement velocity sequence is S = [s1, s2, s3, ..., sn], and the preset head movement velocity sequence is P = [p1, p2, p3, ..., pm], where sn and pm represent the head velocity values ​​at different time points.

[0065] The second step is to normalize the data. Since different head movement speed data may come from different measurement devices or units, normalization can eliminate these differences. A common normalization method is to scale the data to the [0,1] interval. The formula is: for each speed value x, the normalized value in the sequence is (x-min) / (max-min), where min is the minimum value in the sequence and max is the maximum value in the sequence.

[0066] The third step is to construct a distance matrix. Create an n x m distance matrix D, where each element D[i][j] represents the distance between the i-th element in the first head motion velocity sequence and the j-th element in the preset head motion velocity sequence. Here, Euclidean distance is used for calculation, i.e., D[i][j] = (ab). 2 , where a is the normalized first head motion velocity value, and b is the normalized preset head motion velocity value.

[0067] The fourth step is to find the optimal path. Starting from the top left element D[0][0] of the matrix and ending at the bottom right element D[n-1][m-1], find a path that minimizes the sum of all elements on the path. The sum of the elements on this path can be represented as the first distance between the two head motion velocity sequences.

[0068] The fifth step is to find the optimal path using dynamic programming. For each element D[i][j], its predecessor can be the element to its left D[i-1][j], the element above it D[i][j-1], or the element at the top left D[i-1][j-1]. The state transition equation for dynamic programming is D[i][j] = D[i][j] + min(D[i-1][j], D[i][j-1], D[i-1][j-1]). By calculating the cumulative distance of each element step by step in this way, the entire distance matrix is ​​finally obtained.

[0069] The sixth step is to obtain the similarity score by subtracting the first distance from 1 and dividing by a normalization factor. For example, the similarity score is 1 - (DTW distance) / (max(DTW possible distance)), where DTW distance is the first distance and max(DTW possible distance) can be determined based on the range of the data. The value of this similarity score is between [0,1]. The larger the value, the more similar the two head movement speed sequences are, and the higher the matching degree.

[0070] In one embodiment, alternatively, besides simple arithmetic averaging or weighted averaging, more complex fusion strategies can be employed to synthesize the first and second sub-probabilities, such as evidence-theory-based fusion or Bayesian-theory-based fusion, to fully consider the uncertainty and conflicting information between the two sub-probabilities, thereby obtaining a more accurate first probability. For example, using the Dempster-Shafer (DS) synthesis rule in evidence theory, the first and second sub-probabilities are treated as evidence from two different sources. By calculating their belief function, likelihood function, and rejection function, a comprehensive trust assignment function is synthesized, thus obtaining the first probability. This method can effectively handle conflicts and uncertainties between sub-probabilities, improving the credibility of the fusion result.

[0071] In one embodiment, optionally, a rich feature library of intoxication status is established, collecting a large number of samples of head movement speed and upper limb movement speed of different drivers at different levels of intoxication, as well as other relevant features (such as facial expressions and voice tone). Through in-depth analysis and mining of this feature library, more representative and discriminative feature templates are extracted for matching with real-time collected action features. For example, a large amount of sample data is collected on drivers' intoxication behavior, and this data is labeled and classified. Simultaneously, data augmentation techniques can be used to expand and transform the sample data to improve the diversity and representativeness of the feature library. In practical applications, the feature library can be updated and optimized regularly to adapt to changes in driver behavior patterns and the addition of new data.

[0072] In one embodiment, optionally, a personalized alcohol consumption status feature template is established for each driver. Considering individual differences among drivers in physiological structure, driving habits, and reactions to alcohol consumption, feature templates are created for each driver's alcohol-consuming and non-alcohol-consuming states by long-term recording of their driving behavior data. For example, the vehicle's intelligent connectivity system can be used to record driver behavior data at different times and in different scenarios, including head movement speed, upper limb movement speed, and facial expressions. By analyzing and modeling this data, a personalized feature template is generated for each driver. During alcohol consumption identification, the real-time collected action features are matched with the corresponding personalized feature template, thereby improving the accuracy and adaptability of the identification. Simultaneously, the personalized feature templates can be updated periodically to reflect changes in driver behavior patterns.

[0073] In one embodiment, optionally, the recognition parameters and thresholds for head movement speed and upper limb movement speed can be adaptively adjusted according to different driving scenarios (such as urban roads, highways, night driving, and inclement weather). For example, on highways, the driver's upper limb movement speed may be relatively stable, in which case the recognition threshold for upper limb movement speed can be appropriately increased; while on urban roads, due to frequent steering and gear shifting operations, the upper limb movement speed may vary significantly, in which case the recognition threshold can be appropriately decreased. Simultaneously, the recognition parameters can be further optimized by combining information such as vehicle speed and steering wheel rotation angle to improve recognition performance and accuracy in different driving scenarios.

[0074] In one embodiment, optionally, in addition to head movement speed and upper limb movement speed, other multi-source data (such as vehicle trajectory, braking force, and accelerator pedal depth) can be used to verify and supplement the alcohol consumption identification results. For example, when abnormalities are detected in the driver's head movement speed and upper limb movement speed, further analysis can be performed to determine if the vehicle's trajectory deviates significantly or becomes unstable, if the braking force suddenly increases or decreases, or if the accelerator pedal depth changes frequently. Through comprehensive analysis of multi-source data, the driver's alcohol consumption status can be assessed more comprehensively, improving identification performance and accuracy in different driving scenarios and reducing the false positive rate. To achieve this extension, corresponding sensors are deployed within the vehicle, and data fusion algorithms are developed to integrate and analyze data from different types of sensors.

[0075] In one embodiment, optionally, during the recognition process, the system provides real-time feedback to the driver on the recognition results of head movement speed and upper limb movement speed. The driver is allowed to correct the recognition results through specific operations (such as entering a password or fingerprint verification) after confirming they have not consumed alcohol. The entire recognition process is recorded, and the audio and video recordings, along with the recognition results, are stored on the local vehicle system and a cloud server. The system can use the driver's feedback information as new training samples to continuously optimize the recognition model and improve its accuracy and adaptability. Simultaneously, the system can record the driver's feedback operations for further analysis. For example, frequent corrections may indicate that the driver is taking chances or has actually consumed alcohol, allowing for appropriate intervention measures to be taken.

[0076] The first recognition model is used to identify facial features to determine the second probability that the first driver has been drinking.

[0077] Using a first recognition model to identify facial expression features to determine a second probability that the first driver has consumed alcohol, the method includes: identifying facial expression features including the first driver's first pupil focus and first facial color; determining a preset pupil focus corresponding to the first driver consuming alcohol; using the first recognition model to identify the matching degree between the preset pupil focus and the first pupil focus to determine a third sub-probability that the first driver has consumed alcohol; determining a preset facial color corresponding to the first driver consuming alcohol; using the first recognition model to identify the matching degree between the preset facial color and the first facial color to determine a fourth sub-probability that the first driver has consumed alcohol; and determining a second probability based on the third and fourth sub-probabilities.

[0078] In this embodiment, facial expression features involve pupil focus and facial color. Similarly, preset pupil focus and preset facial color corresponding to alcohol consumption are first determined. These two preset values ​​are typical feature values ​​derived from statistical analysis of a large amount of facial expression data of drivers who have consumed alcohol. Then, a first recognition model is used to calculate the matching degree between the actual first pupil focus and the preset pupil focus, and the matching degree between the first facial color and the preset facial color. Based on these two matching degrees, corresponding third and fourth sub-probabilities are determined. Finally, these two sub-probabilities are combined to obtain the second probability. This method can more sensitively capture subtle abnormalities in the facial expressions of drivers who have consumed alcohol, which helps to more accurately identify whether a driver has consumed alcohol, avoiding misjudgments or omissions caused by inaccurate facial expression feature judgments. The second probability more accurately reflects the driver's state of intoxication. This method can effectively improve the comprehensiveness, reliability, and accuracy of judging the probability of a driver consuming alcohol based on facial expression features.

[0079] In one embodiment, optionally, the second probability can be a weighted sum of the third and fourth sub-probabilities, or the maximum or minimum value between the third and fourth sub-probabilities, etc., without limitation. The preset weights of the third and fourth sub-probabilities are preset according to requirements, and are not limited in this application.

[0080] In one embodiment, optionally, the types and parameters of facial expression features can be further refined. Besides pupil focus and facial color, facial expression features can also include changes in the corners of the mouth, eyebrow movements, and the degree of eye opening and closing. For example, a high-resolution camera can capture subtle changes in the driver's facial expressions, and image processing algorithms can be used to quantify and analyze the degree of upward or downward movement of the corners of the mouth, the degree of furrowing or relaxing of the eyebrows, and the frequency of eye opening and closing. These refined facial expression features can more comprehensively reflect the driver's emotions and physiological state, providing richer information for alcohol consumption identification. To accurately acquire these refined facial expression features, high-resolution and high-frame-rate cameras are required, and the acquired image data needs to be processed and analyzed in real time.

[0081] In one embodiment, optionally, multiple matching degree calculation methods are employed to evaluate pupil focus and facial color, such as similarity calculation based on color histograms, matching based on texture features, and feature similarity measurement based on deep learning. The results of different methods are considered comprehensively to improve the accuracy and reliability of the matching degree evaluation. For example, for facial color features, a color histogram can be extracted, and its Bach distance with a preset facial color template can be calculated; for pupil focus features, texture features of the eye region can be extracted, and convolutional neural networks can be used for feature matching and similarity measurement. Finally, the results of these two methods are fused to obtain a comprehensive matching degree score.

[0082] In one embodiment, optionally, a rich feature database of intoxication status is established, collecting a large number of pupil focusing and facial color samples from different drivers at different levels of intoxication, as well as other relevant features (such as head movements and voice tone). Through in-depth analysis and mining of this feature database, more representative and discriminative feature templates are extracted for matching with real-time collected facial expression features. For example, experimental studies on driver intoxication behavior can be conducted in collaboration with medical and research institutions to collect a large amount of sample data, which is then labeled and classified. Simultaneously, data augmentation techniques can be used to expand and transform the sample data to improve the diversity and representativeness of the feature database. In practical applications, the feature database can be updated and optimized regularly to adapt to changes in driver behavior patterns and the addition of new data.

[0083] In one embodiment, optionally, the recognition results of pupil focusing and facial color can be interacted with the vehicle's safety system. For example, when it is detected that the driver may have been under the influence of alcohol, it can work in conjunction with the vehicle's lane departure warning system to improve the sensitivity of the warning; and in conjunction with the automatic emergency braking system to shorten the braking reaction time. Furthermore, the recognition results can be combined with the vehicle's fatigue driving detection system. When both signs of driver intoxication and fatigue are detected simultaneously, higher-level safety intervention measures can be initiated, such as automatically reducing vehicle speed and vibrating the seat as a reminder. Through close integration with other safety systems, a comprehensive vehicle safety protection system is formed, effectively reducing the risk of traffic accidents caused by drunk driving.

[0084] Optionally, a seat sensor is installed on the driver's seat. The seat sensor detects the number of times the driver moves his back and the movement of his back. The ratio of the number of times the driver moves his back to the normal number of times he moves his back is used as the probability that he has been drinking alcohol. Alternatively, the ratio of the movement of the driver's back to the normal number of times he moves his back is used as the probability that he has been drinking alcohol. The above-mentioned installation of a seat sensor can be used as a reference for the detection of the probability of being drinking alcohol, or this measure can be omitted. This application is not limited to this.

[0085] Optionally, a first sensor is installed on the driver's seat. The first sensor detects the number of times the driver's buttocks move and the movement of the buttocks. The ratio of the number of times the driver's buttocks move to the normal buttock movement is used as the probability of having consumed alcohol. Alternatively, the ratio of the driver's buttock movement to the normal buttock movement is used as the probability of having consumed alcohol. The above-mentioned installation of the first sensor can be used as a reference for the detection of the probability of having consumed alcohol, or this measure can be omitted. This application is not limited to this.

[0086] In one embodiment, optionally, the recognition parameters and thresholds for pupil focusing and facial color can be adaptively adjusted according to different driving scenarios (such as urban roads, highways, night driving, and inclement weather). For example, during night driving, due to low light, facial color features may be difficult to capture accurately. In this case, the weight of facial color features can be appropriately reduced, while the weight of pupil focusing features can be increased. In inclement weather conditions, such as heavy fog or heavy rain, the driver's facial expressions may be affected. In this case, the corresponding recognition parameters can be adjusted to improve the accuracy and robustness of recognition. Simultaneously, the recognition parameters can be further optimized by combining information such as vehicle speed and ambient light intensity to adapt to the recognition needs of different driving scenarios.

[0087] In one embodiment, optionally, in addition to pupil focusing and facial color, other multi-source data (such as vehicle trajectory, braking force, accelerator pedal depth, etc.) can be used to verify and supplement the alcohol consumption identification results. For example, when abnormal facial expressions are detected, further analysis can be conducted to determine if the vehicle's trajectory deviates significantly or becomes unstable, if braking force suddenly increases or decreases, or if accelerator pedal depth changes frequently. Through comprehensive analysis of multi-source data, the driver's alcohol consumption status can be assessed more comprehensively, reducing the false positive rate.

[0088] In one embodiment, the probability of having consumed alcohol is determined based on a first probability and a second probability.

[0089] The probability of having consumed alcohol can be a weighted sum of the first and second probabilities, or the maximum or minimum value between the first and second probabilities, etc., and is not limited here. The preset weights of the first and second probabilities are pre-set according to requirements, and are not limited in this application.

[0090] Optionally, driver intoxication can be determined by identifying the following behaviors: staggering, incoherent speech, shouting, erratic pupil movement, abnormal ear twitching, and poor complexion. Alternatively, it can be determined by the rhythm of the steering wheel (when the detection method is used for automobiles) or pedals (when the detection method is used for e-bikes). For example, an upper limb sensor can be installed on the steering wheel to detect abnormal upper limb twitching. Similarly, a foot sensor can be installed on the pedals to detect abnormal foot twitching.

[0091] In one embodiment, alternatively, besides using a single first recognition model, multiple model fusion can be employed. This involves combining different types of models (such as rule-based models, statistical models, and deep learning models) and integrating their outputs through methods like voting, weighted averaging, or stacking to improve the accuracy and robustness of the probability assessment of alcohol consumption. For example, a rule-based model can be used to extract obvious signs of alcohol consumption from driver behavior features (such as frequent head shaking and upper limb tremors), while a deep learning model can be used to uncover subtle changes in facial expressions. The results of both can then be fused using a weighted average to obtain the final probability of alcohol consumption. To achieve model fusion, the outputs of different models need to be standardized to ensure comparability. Simultaneously, appropriate fusion weights need to be determined, which can be trained and optimized using experimental data to achieve the best fusion effect.

[0092] In one embodiment, optionally, the weights of action features and facial expression features in the assessment process can be dynamically adjusted, considering their different degrees of influence on the probability of alcohol consumption. For example, facial expression features may better reflect a driver's intoxication status when driving at high speed or in an emergency, thus increasing their weight; while action features may be more crucial when the vehicle is stationary or moving at low speed, thus increasing their weight. A weight adjustment model based on driver status and vehicle driving environment can be established to calculate the weight coefficients of action features and facial expression features in real time. This model can use driving state parameters such as vehicle speed, acceleration, and steering wheel angle, as well as environmental parameters such as in-vehicle light and noise, as input features. Through online learning and adaptive adjustment, it dynamically determines the weights of both features to improve the relevance and accuracy of the probability of alcohol consumption assessment.

[0093] In one embodiment, alternatively, besides simply averaging the first and second probabilities or other simple combinations, more complex probability calculation methods, such as Bayesian networks and Markov chains, can be used to comprehensively consider the correlation and dependency between action features and facial expression features, thereby more accurately determining the probability of having consumed alcohol. For example, a probabilistic graphical model can be constructed using a Bayesian network, with action features and facial expression features as input nodes and the probability of having consumed alcohol as the output node. By learning the conditional probability distribution in the data, the probability of the driver having consumed alcohol can be inferred. This method can fully utilize prior knowledge and statistical patterns in the data to uncover potential connections between features and improve the credibility of the evaluation results.

[0094] In one embodiment, optionally, when identifying action and facial expression features, a feature selection algorithm can be used to select the most representative and discriminative subset from numerous features to reduce data dimensionality and improve the model's computational efficiency and generalization ability. For example, recursive feature elimination and principal component analysis can be used to filter action features (such as head movement speed, upper limb movement speed, body tilt angle, etc.) and facial expression features (such as pupil focus, facial color, changes in the corners of the mouth, etc.) to select the feature subset that contributes the most to the assessment of the probability of having consumed alcohol. In practical applications, the feature selection results can be evaluated and updated periodically to adapt to changes in driver behavior patterns and the addition of new data.

[0095] In one embodiment, optionally, the recognition strategies for action and facial features, as well as the evaluation criteria for the probability of alcohol consumption, can be adaptively adjusted according to different driving scenarios (such as urban roads, highways, and rural roads). For example, on highways, a driver's action features may be relatively stable, and facial features may receive more attention; while on urban roads, frequent steering and braking may lead to more prominent action features. A driving scenario recognition module can be established to determine the current driving scenario in real time using vehicle GPS data, road type, traffic flow, and other information, and adjust the parameters and evaluation criteria of the recognition model accordingly to improve detection performance in different scenarios.

[0096] In one embodiment, optionally, the driver's historical driving data (such as past drinking records, driving behavior patterns, etc.) can be incorporated into the assessment of the probability of having consumed alcohol to achieve personalized assessment results. For example, for drivers who frequently drive under the influence of alcohol, the assessment threshold can be appropriately lowered to increase detection sensitivity; while for drivers with a good historical record, the threshold can be appropriately raised to reduce false alarms. A driver profiling system can be established to model and analyze each driver's driving habits, drinking behavior, etc., transforming historical data into prior knowledge, which can then be combined with real-time detection data to generate personalized assessment results of the probability of having consumed alcohol.

[0097] In one embodiment, optionally, when a vehicle has multiple drivers, a separate alcohol consumption probability assessment model is established for each driver by identifying their individual identities. For example, the vehicle's smart key system, facial recognition technology, or fingerprint recognition technology can be used to distinguish different drivers, and a personalized alcohol consumption probability assessment can be performed based on each driver's specific behavioral characteristics and historical records. This improves the accuracy of the assessment and avoids misjudgments caused by individual driver differences. Simultaneously, an independent driving safety report can be generated for each driver, helping them understand their driving behavior and drinking habits, thus promoting improved driving safety.

[0098] In one embodiment, the probability assessment of alcohol consumption can optionally be interacted with other vehicle safety systems (such as lane departure warning systems and adaptive cruise control systems) to achieve coordinated safety warnings and control. For example, when the probability of alcohol consumption is high, the lane departure warning threshold can be lowered to issue an earlier warning; or in adaptive cruise control, the vehicle speed can be automatically reduced to increase the safe distance from the vehicle in front. Through close integration with other safety systems, a comprehensive vehicle safety protection system is formed, effectively reducing the risk of traffic accidents caused by drunk driving.

[0099] In one embodiment, optionally, a data visualization system is developed to display various data (such as action characteristics, facial expression characteristics, primary probability, secondary probability, etc.) during the alcohol consumption probability assessment process to the driver or safety supervisor in the form of intuitive charts, images, or videos. For example, the driver's action trajectory, facial expression changes, and the changing trend of the alcohol consumption probability can be displayed in real time through the vehicle's central control display screen or a mobile application. This helps the driver understand their status in a timely manner and take appropriate measures; at the same time, it also provides safety supervisors with detailed analytical data to facilitate accident investigations and safety assessments.

[0100] Optionally, a seatbelt sensor detection system can be developed to detect whether a driver has consumed alcohol. A seatbelt sensor is installed on the seatbelt. Based on the force and posture of the user pulling the seatbelt, the seatbelt sensor detects the force and posture of the user pulling the seatbelt. When the force or posture of the user pulling the seatbelt is abnormal, it is determined that the driver may have consumed alcohol.

[0101] S204: When the probability of having consumed alcohol is greater than the first preset value, the alcohol concentration of the first driver's breath is detected by a breathalyzer.

[0102] The method of detecting the alcohol concentration of the first driver's breath using a breathalyzer includes: using a camera to determine whether the subject of the test is the first driver; when the subject of the test is the first driver, controlling the breathalyzer to contact the first driver's lips; controlling the breathalyzer to collect the first driver's breath; and controlling the breathalyzer to obtain the alcohol concentration of the exhaled breath.

[0103] In this embodiment, a camera is used to determine whether the subject of the test is the primary driver, ensuring the targetedness and accuracy of the test and preventing others from taking the test on their behalf. Only when it is confirmed that the subject is the primary driver will the breathalyzer contact the driver's lips, guiding the driver to blow into the breath correctly, and then collecting the exhaled breath and obtaining the alcohol concentration data. This series of steps ensures the standardization, effectiveness, and accuracy of the breathalyzer test process, enabling precise determination of the alcohol concentration in the driver's exhaled breath.

[0104] In one embodiment, alternatively, in addition to using a camera for identity verification, other identity verification methods, such as fingerprint recognition, voiceprint recognition, and iris recognition, can be combined to improve the accuracy and reliability of the identity verification of the detected object. For example, a fingerprint sensor can be integrated into the contact part of the breathalyzer. When the breathalyzer touches the driver's lips, fingerprint information is simultaneously collected for identity verification. The fingerprint recognition result is fused with the facial recognition result collected by the camera. Only when both confirm that the detected object is the first driver will the breathalyzer be activated to detect the alcohol concentration. This can effectively prevent non-drivers from taking the test on behalf of the driver, ensuring the authenticity and validity of the test results.

[0105] Optionally, when using a breathalyzer, the driver must sit in the driver's seat to take the breathalyzer test. A dashcam or camera will record the process. If someone else blows the breathalyzer, the result is invalid. The video recording will be uploaded to the cloud for use in determining liability. Optionally, a exemption period can be set for drivers with no prior alcohol consumption record, during which they are exempt from alcohol testing. Optionally, facial recognition can be used to identify the driver, or real-time driver identification can be performed, with re-identification required if the driver is not the intended recipient.

[0106] In one embodiment, optionally, various contact control technologies can be employed to ensure accurate contact between the breathalyzer and the driver's lips. For example, the vehicle may include infrared and ultrasonic sensors to detect the distance between the breathalyzer and the driver's lips. When the distance reaches a suitable range, the gas collection function is automatically activated. Alternatively, a pressure sensor can be installed on the contact portion of the breathalyzer. When the pressure applied by the lips reaches a certain threshold, successful contact is confirmed, and gas collection begins. Furthermore, an adjustable breathalyzer holder can be designed to automatically adjust the angle and position of the breathalyzer based on the driver's facial features and position to ensure accurate contact with the lips. These contact control technologies can improve the ease of operation and accuracy of the breathalyzer.

[0107] In one embodiment, the gas collection process of the breathalyzer can be optimized to improve collection efficiency and accuracy. For example, a flow sensor can be used to monitor the gas flow rate in real time to ensure a sufficient volume of exhaled gas is collected; a gas filtration device can be used to remove impurities and moisture from the gas, improving detection accuracy; and multi-point sampling technology can be used to collect multiple gas samples at different time points to avoid detection errors caused by the driver's brief breath-holding or uneven breathing. Simultaneously, an intelligent gas collection algorithm can be developed to automatically adjust the collection time and volume based on real-time monitored gas flow rate and composition information, ensuring that representative gas samples are collected.

[0108] In one embodiment, optionally, multiple alcohol concentration detection methods can be fused, such as electrochemical detection, semiconductor oxide detection, and infrared spectroscopy. By comparing and verifying the detection results of different methods, the accuracy and reliability of alcohol concentration detection can be improved. For example, electrochemical alcohol sensors can provide highly sensitive and selective detection results, but may be affected by ambient temperature and humidity; infrared spectroscopy can detect alcohol concentration in gases non-contactly, and is fast and accurate, but the equipment cost is relatively high. Combining these two methods can complement each other's advantages and improve the reliability of the detection results. To achieve this expansion, it is necessary to conduct in-depth research on the principles, advantages, disadvantages, and applicable conditions of different detection methods, and design corresponding fusion algorithms and detection processes.

[0109] In one embodiment, optionally, in addition to obtaining the alcohol concentration value, the test results can be further analyzed and processed, such as assessing the driver's level of alcohol consumption and estimating blood alcohol content, and this information can be fed back to the vehicle control system and the driver. For example, based on the conversion relationship between alcohol concentration and blood alcohol content, the driver's blood alcohol content can be estimated and compared with the legal drunk driving standard to determine whether the driver is driving under the influence or intoxicated. Simultaneously, the test results and assessment information can be fed back to the driver in real time through the vehicle's central control display screen or a mobile application to remind them to pay attention to driving safety. Furthermore, the test results can be uploaded to a cloud server for traffic management departments to conduct data statistics and analysis, providing support for traffic enforcement and safety management.

[0110] In one embodiment, optionally, a temperature compensation algorithm is designed to compensate for sensor performance drift caused by changes in ambient temperature; a waterproof and dustproof design is adopted to ensure the normal operation of the breathalyzer under harsh weather conditions; and anti-electromagnetic interference measures are added to prevent interference from vehicle electronic devices on the breathalyzer's detection results. Through these environmentally adaptable designs, the reliability and accuracy of the breathalyzer in various practical application scenarios can be guaranteed, improving its adaptability and stability under different environmental conditions.

[0111] In one embodiment, optionally, a user guidance function can be added to help the driver use the breathalyzer correctly. For example, the vehicle's voice prompt system or display screen can guide the driver on how to properly hold the breathalyzer in their mouth and how to exhale evenly; indicator lights or vibration feedback devices can be installed on the breathalyzer to indicate the driver's current testing status (such as ready, collecting data, collecting data, etc.). In addition, an intelligent guidance algorithm can be designed to adjust the guidance strategy in real time based on the driver's operation behavior. For example, when insufficient exhalation is detected, the driver is promptly reminded to continue exhaling; when an operational error is detected, the testing process is automatically restarted and corresponding corrective guidance is provided.

[0112] In one embodiment, optionally, the detection data from a breathalyzer is recorded and stored to establish a driver alcohol detection database. The stored information may include the detection time, location, alcohol concentration value, and result. Analysis of this database can reveal drivers' drunk driving habits and trends, providing data support for personalized safety reminders and management. For example, if a driver is found to be frequently driving under the influence of alcohol during a specific time period or location, early warnings and interventions can be issued. Simultaneously, the stored data can serve as a basis for traffic enforcement, providing relevant departments with information for verification and processing when necessary.

[0113] In one embodiment, the breathalyzer can optionally work in conjunction with other vehicle systems (such as the vehicle's anti-theft system, remote video system, etc.). For example, when a driver's blood alcohol concentration exceeds the legal limit, in addition to locking the vehicle and issuing a warning, the vehicle's anti-theft system can be automatically triggered to prevent unauthorized personnel from moving the vehicle. Simultaneously, the test results and vehicle status information can be sent to a pre-designated guardian or management department via the remote video system so that appropriate measures can be taken in a timely manner. Furthermore, it can be linked to the vehicle's insurance system, using the alcohol test results as one of the bases for adjusting insurance rates, thereby economically deterring drivers from driving under the influence of alcohol.

[0114] In one embodiment, optionally, the breath-type alcohol analyzer is designed with self-calibration and maintenance functions to ensure its long-term stable operation. For example, the analyzer can be automatically calibrated periodically using a standard alcohol gas of known concentration to check and correct the sensitivity and accuracy of the sensor; a self-diagnostic program can be developed to monitor and diagnose faults in real time for various components of the analyzer (such as gas sensors, flow sensors, control circuits, etc.), promptly identifying and alerting potential problems. Furthermore, an easily replaceable sensor module can be designed, allowing users to conveniently replace the sensor themselves when its performance deteriorates or it is damaged, reducing maintenance costs and downtime.

[0115] S205: When the alcohol concentration is greater than or equal to the second preset value, the vehicle is locked and a warning is issued.

[0116] Controlling vehicle locking and issuing a warning includes: the vehicle includes a steering mechanism, a power unit, and an audible device, the steering mechanism being used to steer the vehicle, and the power unit being used to drive the vehicle; controlling the steering mechanism and the power unit to lock; and controlling the audible device to emit a warning sound.

[0117] This embodiment further clarifies the specific operation method of controlling vehicle locking and issuing warnings. Optionally, the vehicle's steering and power systems are key components for vehicle operation; locking them can immediately stop the vehicle from moving, preventing the continuation of drunk driving and the potential harm it may cause. Simultaneously, controlling the sound device to emit a warning sound can promptly alert the driver and surrounding personnel that the vehicle is in an abnormal state, drawing attention and serving as a warning and deterrent. This comprehensive control measure, addressing both vehicle operation and on-site warning aspects, effectively responds to drunk driving situations, ensures traffic safety, and also facilitates timely detection and handling of drunk driving by traffic enforcement personnel.

[0118] In one embodiment, alternatively, in addition to locking the steering and power systems, other critical vehicle systems, such as the ignition and transmission systems, can also be locked to ensure the vehicle cannot start. For example, the engine can be prevented from starting by cutting off the power supply to the ignition system; or the vehicle can be prevented from shifting gears by controlling the transmission's gear-locking mechanism. Simultaneously, a tiered locking strategy can be designed, employing different levels of locking measures based on the alcohol concentration. For example, when the alcohol concentration is slightly higher than a second preset value, only some functions are locked (such as limiting vehicle speed or prohibiting shifting to higher gears); when the alcohol concentration is significantly higher than the second preset value, all functions are locked. This tiered locking strategy can minimize unnecessary restrictions on the driver while ensuring driving safety.

[0119] In one embodiment, optionally, the warning methods can be enriched. Besides audible warnings, warnings can be provided through various means such as the vehicle's lighting system, display screen, and seat vibration. For example, the vehicle's external turn signals and hazard warning lights can flash at specific frequencies to convey warning information to surrounding vehicles and pedestrians; the vehicle's central control display screen can show red warning signs and text prompts to remind the driver that the vehicle has been locked; and the seat vibration device can provide tactile warnings to the driver, enhancing the perceived effect of the warning. Simultaneously, an intelligent warning strategy can be designed to select the most appropriate combination of warning methods based on the vehicle's environment and status. For example, in noisy environments, light and seat vibration warnings are prioritized; in quiet environments, audible warnings and display screen prompts are used.

[0120] In one embodiment, optionally, the content and details of the warning message can be increased. Besides a simple warning, the driver can be provided with specific reasons for the warning, suggested actions, and other information. For example, the vehicle's central control display screen may show a message such as "High blood alcohol concentration detected. The vehicle is locked. Do not attempt to start the vehicle. Contact a designated driver service or wait for roadside assistance." Simultaneously, contact information for nearby designated driver services or navigation routes can be provided to facilitate the driver's next steps. Furthermore, the warning message can be categorized into different levels and types, such as minor warnings, serious warnings, and emergency warnings, with corresponding warning message content and prompts tailored to different situations.

[0121] In one embodiment, optionally, the vehicle's communication module can send vehicle locking and warning information to a remote server or a pre-set guardian's mobile phone in real time to promptly notify relevant personnel of the vehicle's abnormal status. For example, when a vehicle is locked because the driver has been drinking, the system automatically sends a notification SMS to the owner's mobile phone, informing them of the vehicle's location, the reason for the locking, and other information. Simultaneously, the remote server can record and manage all vehicle locking events, providing data support to traffic management departments and assisting them in traffic enforcement and safety management. Furthermore, a mobile application can be developed to allow the owner or guardian to view the vehicle's status in real time and remotely control some of the vehicle's functions (such as unlocking the vehicle and turning off warnings) for convenient emergency handling.

[0122] In one embodiment, optionally, the intensity and manner of the warning can be adjusted by incorporating the vehicle's environmental perception system, such as surrounding vehicle detection and road condition recognition. For example, in congested or densely populated environments, the intensity of the warning can be increased (e.g., by increasing the volume of the warning sound or the frequency of light flashing) to ensure that the warning information is effectively conveyed to the driver and surrounding pedestrians; on open roads, the warning intensity can be appropriately reduced to avoid excessive interference with the surrounding environment. Through this extended environmental perception, the effectiveness and relevance of the warning can be improved, while reducing unnecessary interference with the surrounding environment.

[0123] In one embodiment, the vehicle locking and warning functions can optionally be coordinated with other safety systems (such as automatic parking systems, remote diagnostic systems, etc.). For example, when the vehicle is locked due to the driver having consumed alcohol, the automatic parking system is automatically activated to safely park the vehicle on the roadside; simultaneously, the remote diagnostic system can perform a comprehensive inspection of the vehicle to ensure that there are no other safety hazards. Furthermore, it can be integrated with the vehicle's emergency rescue system, automatically contacting emergency rescue services and providing the vehicle's location and driver information when a potentially dangerous situation is detected (such as impaired consciousness due to high blood alcohol concentration) for timely assistance.

[0124] In one embodiment, optionally, each warning event is recorded in detail, including the warning time, reason for the warning, vehicle location, driver information, etc., and these records are stored in the vehicle's local storage and a cloud server. Analysis of the warning records can provide insights into the vehicle's safety status and the driver's driving behavior patterns, providing data support for vehicle insurance, leasing, and maintenance. For example, insurance companies can assess a vehicle's risk level based on warning records and develop personalized insurance plans; leasing companies can use warning records to better manage leased vehicles. Simultaneously, warning records can serve as evidence for traffic enforcement, provided to relevant departments for verification and processing when necessary.

[0125] In one embodiment, users can optionally customize the warning method and content according to their needs and preferences. For example, users can select different warning sound types, light flashing modes, and display prompts in the vehicle's settings menu; they can also set the warning sensitivity and trigger conditions, such as issuing a warning only when the alcohol concentration exceeds a certain value, or disabling the warning function for a specific time period. This user-customizable extension can improve the personalization and user experience of the warning system, while meeting the needs of different users in different scenarios.

[0126] In one embodiment, optionally, in addition to issuing warnings within the vehicle, the warning information can also be disseminated to surrounding vehicles and traffic infrastructure via the vehicle's communication module, enabling vehicle-to-the-world communication and collaboration. For example, warning messages can be sent to surrounding vehicles via dedicated short-range communication or cellular vehicle-to-everything (V2X) technology to remind other drivers to take evasive action; simultaneously, vehicle status information can be sent to traffic infrastructure such as traffic lights to assist traffic management departments in traffic flow control and accident prevention. This extended warning dissemination can broaden the coverage and impact of warnings, improving overall road traffic safety.

[0127] In one embodiment, optionally, a warning effectiveness evaluation mechanism is established to assess the effectiveness and timeliness of the warning by monitoring the driver's reaction to the warning and changes in vehicle status. For example, the driver's actions after receiving the warning (such as whether they attempt to unlock the vehicle or contact a designated driver service) are recorded to analyze the warning's guiding effect on the driver's behavior; simultaneously, the vehicle's driving status after the warning (such as whether it slows down or stops) is monitored to assess the warning's impact on vehicle safety. Based on the evaluation results, the warning system can be optimized and improved to enhance its practicality and reliability.

[0128] The method also includes: when the vehicle is locked, detecting the alcohol concentration of the second driver's breath using a breathalyzer; when the alcohol concentration is zero, unlocking the vehicle and stopping the warning.

[0129] In this implementation, when a vehicle is locked due to a detected driver having consumed alcohol, considering the possibility that other drivers who have not consumed alcohol may need to use the vehicle, a step is added to test the alcohol concentration of a second driver using a breathalyzer. If the test result shows an alcohol concentration of zero, indicating that the second driver has not consumed alcohol, then the vehicle is unlocked, the warning is stopped, and the vehicle's normal driving status is restored. This process demonstrates that while strictly regulating drunk driving, the normal use needs of the vehicle are also taken into account, avoiding unnecessary trouble for users due to misjudgment or special circumstances causing the vehicle to be unusable for a long time. This improves the system's flexibility and user-friendliness.

[0130] In one embodiment, alternatively, besides detecting zero alcohol concentration in the second driver's exhaled breath, other unlocking conditions can be added, such as identity verification of the second driver and a safety assessment of the vehicle's surrounding environment. For example, the second driver can be required to verify their identity via fingerprint recognition, facial recognition, or password verification to prevent unauthorized personnel from unlocking the vehicle. Simultaneously, the vehicle's environmental perception system assesses the safety of the current environment, such as whether it is in a no-parking zone or whether there are any surrounding hazards. The vehicle is only allowed to unlock when all unlocking conditions are met. This multi-condition unlocking mechanism can improve the security and reliability of vehicle unlocking.

[0131] In one embodiment, optionally, after the vehicle is unlocked, various vehicle functions are restored gradually, rather than all at once, to ensure driving safety. For example, the vehicle's starting function is restored first, but the maximum speed and some advanced driver assistance functions are limited; after the vehicle has been driven for a period of time and no abnormalities are detected, other functions are gradually restored. Simultaneously, the driving behavior and status of the second driver can be monitored in real time. If abnormal driving signs are detected (such as frequent lane changes, sudden acceleration, etc.), the vehicle locking and warning procedures are restarted. This extended function restoration after unlocking provides necessary driving convenience for the second driver while ensuring driving safety.

[0132] In one embodiment, optionally, after the vehicle is unlocked, the second driver's blood alcohol concentration and driving behavior are continuously monitored to ensure that they remain sober and safe throughout the driving process. For example, the second driver's blood alcohol concentration can be retested periodically using a breathalyzer, with the testing frequency set according to the time interval or mileage. Simultaneously, the vehicle's driving behavior monitoring system analyzes the second driver's movement and facial expressions in real time; if any abnormalities are detected, the vehicle locking and warning procedures are immediately restarted. This continuous monitoring allows for the timely detection of potential safety hazards, preventing traffic accidents caused by the second driver being intoxicated or in poor condition.

[0133] In one embodiment, optionally, an unlocking authorization function is added, allowing the vehicle owner or other authorized personnel to remotely authorize vehicle unlocking. For example, when a vehicle is locked because the first driver has been intoxicated, the owner can receive an unlocking request via a mobile application and, after verifying the identity and status of the second driver, remotely authorize the vehicle to unlock. Simultaneously, the validity period and scope of the authorization can be set, such as the authorization being valid only within a specific time period, or only allowing the vehicle to travel to a designated location. This extended unlocking authorization enhances the flexibility and security of vehicle use, especially when the owner cannot handle the situation personally.

[0134] In one embodiment, optionally, after the vehicle is unlocked, its driving range can be restricted to ensure the second driver travels within a safe area. For example, using the vehicle's GPS positioning system and geofencing technology, a permitted driving area can be set, such as allowing the vehicle to travel only within designated urban areas, or restricting the vehicle from traveling on highways. When the vehicle approaches or exceeds the set driving range, a timely warning is issued, and corresponding measures are taken, such as reducing speed or forcibly stopping the vehicle. This extended driving range restriction effectively reduces the risk of accidents caused by the second driver's unfamiliarity with the road or driving into dangerous areas.

[0135] In one embodiment, optionally, after unlocking, the vehicle's driving mode is automatically adjusted based on the second driver's driving habits and the vehicle's status. For example, if the second driver is a novice driver, the vehicle can automatically switch to novice mode, providing more sensitive braking feedback, more stable power output, etc.; if the vehicle is in energy-saving mode, it will remain in energy-saving mode after unlocking to meet the user's driving preferences and the vehicle's operating needs. Simultaneously, in conjunction with the vehicle's intelligent connectivity system, the vehicle's seat position, rearview mirror angle, and other configurations can be automatically adjusted based on the second driver's historical driving records and preference settings, providing the driver with a more comfortable and convenient driving environment.

[0136] In one embodiment, optionally, after the vehicle is unlocked, safe driving education information, such as the dangers of driving under the influence of alcohol and safe driving techniques, is provided to the second driver via the vehicle's central control display screen or voice prompt system. For example, when the vehicle is locked because the first driver has been drinking, the system automatically plays a short video about the dangers of driving under the influence of alcohol when the second driver unlocks the vehicle, reminding them to pay attention to driving safety. Simultaneously, practical safe driving advice can be provided, such as maintaining a safe following distance and obeying traffic rules, to help the second driver improve their safety awareness and driving skills.

[0137] In one embodiment, optionally, emergency service support, such as roadside assistance and medical first aid, can be provided to the second driver after unlocking. For example, after the vehicle is unlocked, the system automatically sends the vehicle's location and driver information to a pre-set emergency service provider so that assistance can be provided promptly when needed. Simultaneously, emergency service contact information and usage guidelines can be displayed on the vehicle's central control screen to facilitate quick access to help for the second driver in emergencies. Furthermore, collaboration with emergency service providers can be established to provide users with more information, such as vehicle malfunction details and the driver's health status, to allow for advance preparation and more efficient rescue services.

[0138] In one embodiment, optionally, the driving behavior and vehicle status information of the second driver after unlocking the vehicle are recorded to establish a database of unlocked driving behavior. The stored information may include driving time, mileage, driving speed, and alcohol concentration test results. Analysis of this database can assess the second driver's driving habits and risk level, providing data support for vehicle insurance, leasing, and other businesses. For example, insurance companies can assess the second driver's risk level based on unlocked driving behavior records and develop personalized insurance plans; leasing companies can use the records to better manage leased vehicles and ensure safe vehicle use.

[0139] In this embodiment, the first detection module collects air from inside the vehicle to detect whether the air contains alcohol. When the air contains alcohol, a camera collects first video data of the first driver, including the driver's motion and facial features. The motion and facial features are sent to a first recognition model to determine the probability that the driver has consumed alcohol. When the probability of consuming alcohol is greater than a first preset value, a breathalyzer is used to detect the alcohol concentration in the driver's exhaled breath. When the alcohol concentration is greater than or equal to a second preset value, the vehicle is locked and a warning is issued. This method allows for a tiered testing of whether a driver has consumed alcohol, improving the comprehensiveness of the detection, reducing the probability of drunk driving, and enhancing driving safety.

[0140] Optionally, the driver's voice can be used to determine if they are driving under the influence of alcohol. For example, ask the driver to say "Open Sesame, my beloved car" three times. Alternatively, ask the driver to say a tongue twister to assess their level of consciousness and thought process. For example, ask someone from Guangdong to say "Every country has its own national anthem." Or give the driver a riddle to solve. The above methods of using the driver's voice to determine if they are driving under the influence can be used as supplementary methods for detecting whether a driver has been drinking, or this measure can be omitted; this application is not limited to these methods.

[0141] Optionally, the in-vehicle AI voice box continuously recognizes and analyzes the cabin dialogue, identifying statements related to alcohol such as "how many drinks I've had" or "I'm a little dizzy." If the confidence level is greater than 0.8, an enhanced alcohol detection process is initiated: the screen prompts "Please exhale into the breathalyzer," and the breathalyzer quickly measures the breath alcohol. The above content, combined with the cabin dialogue, can be used as a reference for determining whether the driver is driving under the influence of alcohol. Alternatively, this measure can be omitted, and this application is not limited to this.

[0142] It should be noted that, optionally, this detection method can also be applied to electric bicycles or electric tricycles, etc., and this application does not limit it.

[0143] Please see Figure 3 , Figure 3 This is a schematic diagram of a detection device according to one embodiment. Based on the above system architecture, the detection device 300 can be a server or a vehicle, or it can be a module within a server. The detection device 300 includes at least: a data acquisition module 301 and a processing module 302, wherein...

[0144] The acquisition module 301 acquires the air inside the vehicle through the first detection module; when the air inside the vehicle contains alcohol, the acquisition module 301 acquires the first video data of the first driver of the vehicle, the first video data including the first driver's action characteristics and facial expression characteristics.

[0145] The processing module 302 detects whether the air inside the vehicle contains alcohol; the processing module 302 sends the action features and facial expression features to the first recognition model to determine the probability that the first driver has consumed alcohol; when the probability of consuming alcohol is greater than a first preset value, the processing module 302 detects the alcohol concentration of the first driver's exhaled breath using a breathalyzer; when the alcohol concentration is greater than or equal to a second preset value, the processing module 302 controls the vehicle to lock and issues a warning.

[0146] In one possible example, the processing module 302 is used to identify action features using a first recognition model to determine a first probability that the first driver has been drinking; to identify facial expression features using the first recognition model to determine a second probability that the first driver has been drinking; and to determine the probability of having been drinking based on the first and second probabilities.

[0147] In one possible example, the action features include the first head movement speed and the first upper limb movement speed of the first driver. The processing module 302 is used to determine a preset head movement speed corresponding to the first driver having consumed alcohol; use a first recognition model to identify the matching degree between the first head movement speed and the preset head movement speed to determine a first sub-probability that the first driver has consumed alcohol; determine a preset upper limb movement speed corresponding to the first driver having consumed alcohol; use the first recognition model to identify the matching degree between the first upper limb movement speed and the preset upper limb movement speed to determine a second sub-probability that the first driver has consumed alcohol; and determine a first probability based on the first sub-probability and the second sub-probability.

[0148] In one possible example, the facial features include the first driver's first pupil focus and first facial color. The processing module 302 is used to determine a preset pupil focus corresponding to the first driver having consumed alcohol; use a first recognition model to identify the matching degree between the preset pupil focus and the first pupil focus to determine a third sub-probability that the first driver has consumed alcohol; determine a preset facial color corresponding to the first driver having consumed alcohol; use the first recognition model to identify the matching degree between the preset facial color and the first facial color to determine a fourth sub-probability that the first driver has consumed alcohol; and determine a second probability based on the third and fourth sub-probabilities.

[0149] In one possible example, the processing module 302 is used to use a camera to determine whether the detected object is the first driver; when the detected object is the first driver, it controls the breathalyzer to contact the first driver's lips; controls the breathalyzer to collect the breath exhaled by the first driver; and controls the breathalyzer to obtain the alcohol concentration of the exhaled breath.

[0150] In one possible example, the vehicle includes a steering mechanism, a power unit, and an audible device. The steering mechanism is used to steer the vehicle, the power unit is used to drive the vehicle, and the processing module 302 is used to control the locking of the steering mechanism and the power unit; and to control the audible device to emit a warning sound.

[0151] In one possible example, when the vehicle is locked, the processing module 302 is used to detect the alcohol concentration of the second driver's breath using a breathalyzer; when the alcohol concentration is zero, the vehicle is unlocked and the warning is stopped.

[0152] Please see Figure 4 , Figure 4 This is a schematic diagram of the structure of a vehicle component according to one embodiment. (For example...) Figure 4 As shown, the vehicle 102 includes a processor 401, a memory 402, and at least one program 403. The at least one program 403 is stored in the memory 402 and configured to be executed by the processor 401. The program 403 includes instructions for performing the following steps:

[0153] The first detection module collects air from inside the vehicle to detect whether the air inside the vehicle contains alcohol.

[0154] When the air inside the vehicle contains alcohol, the camera captures the first video data of the first driver of the vehicle. The first video data includes the first driver's action characteristics and facial expression characteristics.

[0155] The action and facial features are sent to the first recognition model to determine the probability that the first driver has been drinking;

[0156] When the probability of having consumed alcohol exceeds a first preset value, the alcohol concentration in the breath of the first driver is detected by a breathalyzer.

[0157] When the alcohol concentration is greater than or equal to the second preset value, the vehicle will be locked and a warning will be issued.

[0158] In one possible example, program 403 is specifically used to perform the following instructions:

[0159] The first recognition model is used to identify action features to determine the first probability that the first driver has been drinking;

[0160] The first recognition model is used to identify facial features to determine a second probability that the first driver has been drinking;

[0161] The probability of having consumed alcohol is determined based on the first probability and the second probability.

[0162] In one possible example, program 403 is specifically used to perform the following instructions:

[0163] The motion characteristics include the first head movement speed and the first upper limb movement speed of the first driver;

[0164] Determine the preset head movement speed corresponding to the first driver's alleged intoxication;

[0165] The first recognition model is used to identify the matching degree between the first head movement speed and the preset head movement speed in order to determine the first sub-probability that the first driver has been drinking alcohol.

[0166] Determine the preset upper limb movement speed corresponding to the first driver's alleged intoxication;

[0167] The first recognition model is used to identify the matching degree between the speed of the first upper limb movement and the preset speed of the upper limb movement in order to determine the second sub-probability that the first driver has been drinking alcohol.

[0168] The first probability is determined based on the first sub-probability and the second sub-probability.

[0169] In one possible example, program 403 is specifically used to perform the following instructions:

[0170] Facial features include the first driver's first pupil focus and first facial color;

[0171] Determine the preset pupil focus level corresponding to the driver's alleged intoxication;

[0172] The first recognition model is used to identify the matching degree between the preset pupil focus and the first pupil focus in order to determine the third probability that the first driver has been drinking alcohol;

[0173] Determine the preset facial color corresponding to whether the first driver has consumed alcohol;

[0174] The first recognition model is used to identify the matching degree between the preset facial color and the first facial color in order to determine the fourth sub-probability that the first driver has been drinking alcohol.

[0175] The second probability is determined based on the third and fourth sub-probabilities.

[0176] In one possible example, program 403 is specifically used to perform the following instructions:

[0177] Use a camera to determine if the detected object is the first driver;

[0178] When the subject of the test is the first driver, the breathalyzer is controlled to contact the first driver's lips;

[0179] The breathalyzer is used to collect the breath of the first driver.

[0180] Control the breathalyzer to obtain the alcohol concentration of exhaled breath.

[0181] In one possible example, program 403 is specifically used to perform the following instructions:

[0182] A vehicle includes a steering system, a power system, and a sound system. The steering system is used to steer the vehicle, and the power system is used to drive the vehicle.

[0183] Controls the locking of the steering and power units;

[0184] The sound control device emits an alarm sound.

[0185] In one possible example, program 403 is specifically used to perform the following instructions:

[0186] When the vehicle is locked, the alcohol concentration of the second driver's breath is detected by a breathalyzer.

[0187] When the alcohol concentration is zero, the vehicle will unlock and the warning will stop.

[0188] Those skilled in the art will understand that, for ease of explanation, Figure 4 Only one memory 402 and processor 401 are shown in the illustration. In a real terminal or server, multiple processors and memories may exist. Memory can also be called storage medium or storage device, etc., and this application embodiment does not limit this.

[0189] It should be understood that in this application, the processor can be a Central Processing Unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor can also be a general-purpose microprocessor, graphics processing unit (GPU), or one or more integrated circuits to execute relevant programs to achieve the functions required by the embodiments of this application.

[0190] Processor 401 can also be an integrated circuit chip with signal processing capabilities. In implementation, each step of this application can be completed through integrated logic circuits in the hardware of processor 401 or through software instructions. The processor 401 described above can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The steps of the methods disclosed in the embodiments of this application can be directly manifested as execution by a hardware decoding processor, or as execution by a combination of hardware and software modules in the decoding processor. The software modules can be located in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory 402. Processor 401 reads information from memory 402 and, in conjunction with its hardware, completes the functions required by the units included in the methods, apparatus, and storage media of the embodiments of this application.

[0191] It should also be understood that the memory mentioned in the embodiments of this application can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. Non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Synchronous DRAM (SDRAM), Double Data Rate Synchronous DRAM (DDR SDRAM), Enhanced Synchronous DRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), and Direct Rambus RAM (DR RAM). The memory can also be, but is not limited to, Compact Disc Read-Only Memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed discs, laser discs, optical discs, digital universal discs, Blu-ray discs, etc.), magnetic disk storage media, or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures that can be accessed by a computer. The memory can be independent and connected to the processor via a bus. The memory can also be integrated with the processor. The memory can store programs. When the program stored in the memory is executed by the processor, the processor is used to execute the various steps of the determination method in the above embodiments of this application.

[0192] It should be noted that when the processor is a general-purpose processor, DSP, ASIC, FPGA, or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component, the memory (storage module) is integrated into the processor. It should be noted that the memory described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0193] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0194] In implementation, each step of the above method can be completed by integrated logic circuits in the processor's hardware or by instructions in software. The steps of the method disclosed in the embodiments of this application can be directly implemented by a hardware processor, or by a combination of hardware and software modules within the processor. The software modules can reside in mature storage media in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. Since this storage medium is located in memory, the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method; to avoid repetition, these will not be described in detail here.

[0195] Those skilled in the art will recognize that the various illustrative logical blocks (ILBs) and steps described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

[0196] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented, in whole or in part, as a computer-programmed program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a processor, all or part of the flow or function according to the embodiments of this application is generated. The computer can be a general-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic) or wireless (e.g., infrared, wireless, microwave, etc.) means, or from one website, computer, server, or data center to a mobile phone processor via a wired means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive), etc.

[0197] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A detection method, characterized in that, The detection method is used in a vehicle, the vehicle including a first detection module, a breathalyzer, and a camera, the method comprising: The first detection module collects air from inside the vehicle to detect whether the air inside the vehicle contains alcohol. When the air inside the vehicle contains alcohol, the camera captures first video data of the first driver of the vehicle. The first video data includes the first driver's action features and facial expression features. The action features and facial expression features are sent to a first recognition model to determine the probability that the first driver has been drinking alcohol; When the probability of having consumed alcohol is greater than a first preset value, the alcohol concentration of the first driver's exhaled breath is detected by the breathalyzer. When the alcohol concentration is greater than or equal to a second preset value, the vehicle is locked and a warning is issued.

2. The detection method according to claim 1, characterized in that, The step of sending the action features and the facial expression features to the first recognition model to determine the probability that the first driver has consumed alcohol includes: The first recognition model is used to identify the action features to determine the first probability that the first driver has been drinking alcohol; The first recognition model is used to identify the facial features to determine a second probability that the first driver has been drinking alcohol; The probability of having consumed alcohol is determined based on the first probability and the second probability.

3. The detection method according to claim 2, characterized in that, The step of using the first recognition model to identify the action features to determine the first probability that the first driver has been drinking includes: The motion characteristics include the first head motion speed and the first upper limb motion speed of the first driver; Determine the preset head movement speed corresponding to the first driver having consumed alcohol; The first recognition model is used to identify the degree of matching between the first head movement speed and the preset head movement speed in order to determine the first sub-probability that the first driver has been drinking alcohol. Determine the preset upper limb movement speed corresponding to the first driver having consumed alcohol; The first recognition model is used to identify the matching degree between the first upper limb movement speed and the preset upper limb movement speed to determine a second sub-probability that the first driver has been drinking alcohol. The first probability is determined based on the first sub-probability and the second sub-probability.

4. The detection method according to claim 2, characterized in that, The step of using the first recognition model to recognize the facial expression features to determine a second probability that the first driver has been drinking includes: The facial features include the first driver's first pupil focus and first facial color; Determine the preset pupil focus level corresponding to the first driver's alleged alcohol consumption; The first recognition model is used to identify the matching degree between the preset pupil focus and the first pupil focus in order to determine the third sub-probability that the first driver has been drinking alcohol; Determine the preset facial color corresponding to the first driver's alleged alcohol consumption; The first recognition model is used to identify the matching degree between the preset facial color and the first facial color in order to determine the fourth sub-probability that the first driver has been drinking alcohol. The second probability is determined based on the third sub-probability and the fourth sub-probability.

5. The detection method according to claim 1, characterized in that, The step of detecting the alcohol concentration in the breath of the first driver using the breathalyzer includes: The camera is used to determine whether the detected object is the first driver; When the target of the test is the first driver, the breathalyzer is controlled to contact the first driver's lips; The breathalyzer is controlled to collect the breath of the first driver. The breathalyzer is controlled to obtain the alcohol concentration of the exhaled breath.

6. The detection method according to claim 1, characterized in that, The control of locking the vehicle and issuing a warning includes: The vehicle includes a steering device, a power device, and a sound device; the steering device is used to steer the vehicle, and the power device is used to drive the vehicle. Lock the steering device and the power unit; Control the sound device to emit a warning sound.

7. The detection method according to claim 1, characterized in that, The method further includes: When the vehicle is locked, the alcohol concentration of the second driver's breath is detected by the breathalyzer. When the alcohol concentration is zero, the vehicle is unlocked and the warning is stopped.

8. A detection device, characterized in that, Includes methods for performing any one of claims 1-7.

9. A vehicle, characterized in that, The method includes a processor, a memory, and at least one program, wherein the at least one program is stored in the memory and configured to be executed by the processor, the program including instructions for performing the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that causes a computer to execute in order to implement the method as described in any one of claims 1-7.

Images