Driving behavior risk detection method, system and device and storage medium

Abstract

The invention discloses a driving behavior risk detection method, system and device and a storage medium, and relates to the field of driving safety. The method comprises the following steps: collecting vehicle internal data, external environment data, driver biological characteristic data, vehicle-mounted video data and vehicle-mounted audio data; extracting a first feature from the vehicle internal data, extracting a second feature from the external environment data, extracting a third feature from the driver biological feature data, extracting a fourth feature from the vehicle-mounted video data, and extracting a fifth feature from the vehicle-mounted audio data; fusing the first feature, the second feature, the third feature, the fourth feature and the fifth feature to obtain a feature vector, and inputting the feature vector into a preset machine learning model to obtain a risk detection score; and when the risk detection score exceeds a threshold value, displaying risk information on a vehicle-mounted display screen and giving a voice prompt. By implementing the technical scheme provided by the invention, the driving risk can be comprehensively evaluated by integrating multi-source data, the early warning is timely performed, and the driving safety is improved.

Description

Technical Field

[0001] The present application relates to the technical field of driving safety, and specifically to a method, system, device and storage medium for detecting driving behavior risks. Background Art

[0002] With the continuous development of intelligent transportation systems, how to effectively monitor and evaluate the risks of driving behavior has become a key link in ensuring road traffic safety.

[0003] In order to deal with various risks that may arise during driving, most driving behavior monitoring systems on the market currently rely on a single data source (such as vehicle speed, braking frequency, etc.) to determine whether there are risks in driving behavior. However, this method often ignores the impact of the external driving environment, resulting in incomplete risk assessment and difficulty in accurately predicting the probability of potential accidents.

[0004] Therefore, there is an urgent need for a technical solution that can integrate multiple data sources and realize multi-dimensional risk assessment to improve the effectiveness of driving behavior risk detection. Summary of the invention

[0005] The present application provides a method, system, device and storage medium for detecting driving behavior risks, which can comprehensively and carefully evaluate the risks of driving behavior. This multi-dimensional data collection and analysis improves the accuracy of risk detection.

[0006] In a first aspect of the present application, a method for detecting driving behavior risk is provided, which is applied to a driving behavior risk detection platform, and the method comprises: Collecting vehicle internal data, external environment data, driver biometric data, in-vehicle video data, and in-vehicle audio data, wherein the vehicle internal data includes a first speed of the vehicle, target acceleration, current continuous driving time, and frequency of activating the accelerator and brake, the external environment data includes weather conditions, road type, a second speed of the nearest target vehicle around the vehicle, and a change in distance from the vehicle, the driver biometric data includes facial expressions, the in-vehicle video data includes eye attention tracks, gestures, and eye closure time, and the in-vehicle audio data includes the tone and intonation of the driver's speech; Extracting a first feature from the vehicle interior data, extracting a second feature from the external environment data, extracting a third feature from the driver biometric data, extracting a fourth feature from the vehicle video data, and extracting a fifth feature from the vehicle audio data; The first feature, the second feature, the third feature, the fourth feature and the fifth feature are integrated to obtain a feature vector, and the feature vector is input into a preset machine learning model to obtain a risk detection score; When the risk detection score exceeds a threshold, risk information is displayed on the vehicle display screen and a voice prompt is given.

[0007] Optionally, extracting a first feature from the vehicle interior data includes: Calculating an average speed of the first speed within a preset time period, calculating a standard deviation of the target acceleration within the preset time period, determining a first extreme value of the target acceleration within the preset time period, and determining a second extreme value of the first speed within the preset time period; When the first extreme value is greater than or equal to a first threshold, constructing a first feature according to the first extreme value, the second extreme value, the current continuous driving time, and the frequency of activating the accelerator and the brake; When the first extreme value is less than the first threshold, determining whether the standard deviation is greater than or equal to a second threshold; When the standard deviation is greater than or equal to the second threshold, constructing a first feature according to the standard deviation, the second extreme value, the current continuous driving time, and the frequency of activating the accelerator and the brake; When the standard deviation is less than the second threshold, a first feature is constructed according to the standard deviation, the average speed, the current continuous driving time, and the frequency of activating the accelerator and the brake.

[0008] Optionally, extracting the second feature from the external environment data includes: Determine whether a distance change rate is greater than or equal to a third threshold value according to the distance change, and when the distance change rate is greater than or equal to the third threshold value, determine whether the target vehicle is located in the target direction of the vehicle; When the target vehicle is located in a target direction of the vehicle, calculating a relative speed between the first speed and the second speed; A second feature is constructed according to the feature value of the weather condition, the feature value of the road type, and the relative speed.

[0009] Optionally, extracting a third feature from the driver's biometric data includes: In the captured image or video frame, using a facial detection algorithm to locate the driver's facial area; In the facial area, key feature points of the face are marked using a feature point detection algorithm, wherein the key feature points include the corners of the eyes, the corners of the mouth, and the eyebrows; Facial expression features are extracted based on the key feature points, the facial expression features including the degree of opening and closing of the eyes, the degree of curvature of the eyebrows, and the degree of uplift or droop of the corners of the mouth, and a third feature is constructed based on the facial expression features.

[0010] Optionally, extracting the fourth feature from the vehicle-mounted video data includes: Acquiring a sight line movement trajectory of the driver, and determining a distraction degree characteristic according to the sight line movement trajectory; Detecting and tracking the driver's hand position in the image frame, classifying the hand posture, identifying the driver's gesture action, and extracting gesture features from the gesture action, wherein the gesture features include gesture type, duration, and motion trajectory; Detecting the eye-closing duration of the driver, and acquiring the head posture and mouth movements of the driver, and determining fatigue driving characteristics through the eye-closing duration, the head posture and the mouth movements; A fourth feature is constructed based on the attention distraction degree feature, the gesture feature, and the fatigue driving feature.

[0011] Optionally, extracting the fifth feature from the vehicle-mounted audio data includes: Calculating volume differences and pitch differences between adjacent audio frames, determining volume features according to the volume differences, and determining pitch features according to the pitch differences, wherein the volume features include maximum volume, minimum volume, average volume, and standard deviation of volume changes, and the pitch features include average pitch and standard deviation of pitch changes; Converting the in-vehicle audio data into text, calculating the number of words in the text, determining a speech rate feature according to the number of words, and extracting content from the text, determining an emotional feature according to the content; A fifth feature is constructed based on the volume feature, the tone feature, the speech speed feature and the emotion feature.

[0012] Optionally, the fusing the first feature, the second feature, the third feature, the fourth feature and the fifth feature to obtain a feature vector includes: performing standardization processing on the first feature, the second feature, the third feature, the fourth feature, and the fifth feature; Perform weighted summation on the standardized first feature, the second feature, the third feature, the fourth feature, and the fifth feature to form a feature vector.

[0013] In a second aspect of the present application, a driving behavior risk detection system is provided, comprising a collection module, a feature module, a fusion module and a display module, wherein: a collection module configured to collect vehicle internal data, external environment data, driver biometric data, vehicle video data, and vehicle audio data, wherein the vehicle internal data includes a first speed of the vehicle, target acceleration, current continuous driving time, and frequency of activating the accelerator and brake, the external environment data includes weather conditions, road type, a second speed of the nearest target vehicle around the vehicle, and a change in distance from the vehicle, the driver biometric data includes facial expressions, the vehicle video data includes eye attention tracks, gestures, and eye closure time, and the vehicle audio data includes the tone and intonation of the driver's speech; a feature module configured to extract a first feature from the vehicle interior data, a second feature from the external environment data, a third feature from the driver biometric data, a fourth feature from the vehicle video data, and a fifth feature from the vehicle audio data; a fusion module, configured to fuse the first feature, the second feature, the third feature, the fourth feature and the fifth feature to obtain a feature vector, and input the feature vector into a preset machine learning model to obtain a risk detection score; The display module is configured to display risk information on the vehicle display screen and give a voice prompt when the risk detection score exceeds a threshold.

[0014] In the third aspect of the present application, an electronic device is provided, including a processor, a memory, a user interface and a network interface, the memory is used to store instructions, the user interface and the network interface are both used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes any one of the methods described above.

[0015] In a fourth aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores instructions, and when the instructions are executed, any of the methods described above is executed.

[0016] In summary, one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. By collecting multiple types of data (vehicle internal data, external environment data, driver biometric data, vehicle video data, and vehicle audio data), it is possible to comprehensively and carefully assess the risks of driving behavior. This multi-dimensional data collection and analysis makes risk detection more accurate and reliable; 2. By collecting and analyzing data in real time, potential risk behaviors can be discovered in a timely manner during driving, such as driver fatigue, inattention, speeding, etc., thereby effectively preventing traffic accidents; 3. Using machine learning models to process feature vectors can automatically identify risk behavior patterns and give corresponding risk detection scores. This intelligent processing method not only improves the efficiency and accuracy of risk detection, but also reduces the subjectivity and uncertainty of human judgment; 4. When risky behavior is detected, the risk information can be displayed on the in-vehicle display screen and voice prompts can be given to remind the driver to pay attention to safety. This intuitive feedback method helps the driver to correct bad driving habits in time and improve driving safety; 5. By detecting and warning of driving behavior risks in advance, this method helps drivers take preventive measures before potential dangers occur, thereby avoiding or reducing the occurrence of traffic accidents and protecting the lives and property of drivers and passengers. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flowchart of a method for detecting driving behavior risk disclosed in an embodiment of the present application; Figure 2 It is a module schematic diagram of a driving behavior risk detection system disclosed in an embodiment of the present application; Figure 3 It is a structural schematic diagram of an electronic device disclosed in an embodiment of the present application.

[0018] Explanation of the reference numerals: 201, acquisition module; 202, feature module; 203, fusion module; 204, display module; 301, processor; 302, communication bus; 303, user interface; 304, network interface; 305, memory. DETAILED DESCRIPTION

[0019] In order to enable technicians in this field to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments.

[0020] In the description of the embodiments of the present application, words such as "for example" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "for example" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "for example" or "for example" is intended to present related concepts in a specific way.

[0021] In the description of the embodiments of the present application, the meaning of the term "multiple" refers to two or more. For example, multiple systems refer to two or more systems, and multiple screen terminals refer to two or more screen terminals. In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. The terms "include", "comprise", "have" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

[0022] This embodiment discloses a driving behavior risk detection method, which is applied to a driving behavior risk detection platform. Figure 1 is a flow chart of the method for detecting driving behavior risk disclosed in the embodiment of the present application, such as Figure 1 As shown, the method comprises the following steps: S101, collecting vehicle internal data, external environment data, driver biometric data, in-vehicle video data, and in-vehicle audio data, wherein the vehicle internal data includes a first speed of the vehicle, target acceleration, current continuous driving time, and frequency of activating the accelerator and brake, the external environment data includes weather conditions, road type, a second speed of the nearest target vehicle around the vehicle, and a change in distance from the vehicle, the driver biometric data includes facial expressions, the in-vehicle video data includes eye attention tracks, gestures, and eye closure time, and the in-vehicle audio data includes the tone and intonation of the driver's speech; S102, extracting a first feature from the vehicle internal data, extracting a second feature from the external environment data, extracting a third feature from the driver's biometric data, extracting a fourth feature from the vehicle-mounted video data, and extracting a fifth feature from the vehicle-mounted audio data; S103, fusing the first feature, the second feature, the third feature, the fourth feature and the fifth feature to obtain a feature vector, and inputting the feature vector into a preset machine learning model to obtain a risk detection score; S104: When the risk detection score exceeds a threshold, risk information is displayed on the vehicle display screen and a voice prompt is given.

[0023] The first speed is used to record the current speed of the vehicle in real time, which is an important parameter for evaluating whether the driving behavior is safe. The target acceleration is used to record the current acceleration value of the vehicle, reflecting whether the current driving is smooth. The current continuous driving time is used to record the continuous driving time of the driver from the start of driving to the current moment, which is used to evaluate whether the driver is fatigued. The number and frequency of the driver's operation of the accelerator and brake are recorded to reflect the stability and responsiveness of the driver's control of the vehicle. The current weather information is obtained through on-board sensors or external data sources, such as sunny days, rainy days, snowy days, etc. Weather conditions have a significant impact on driving behavior. Identify the type of road on which the vehicle is located, such as highways, urban roads, rural roads, etc. Different types of roads have different requirements for driving behavior. Real-time acquisition of the speed of the target vehicle closest to the vehicle is used to evaluate relative speed and potential collision risks. Record the changes in the distance between the vehicle and surrounding vehicles over time to reflect the density of traffic and the congestion of the driving environment. The driver's facial expressions, such as smiling, frowning, yawning, etc., are captured through the on-board camera to evaluate the driver's emotional state and fatigue level. The driver's eye movement trajectory is tracked through video analysis technology to evaluate the driver's attention distribution. Identify the driver's gestures while driving, such as making phone calls, drinking water, eating, etc., which may distract the driver. Record the length of time the driver closes his eyes to assess whether the driver is in a state of fatigue. Analyze the driver's speaking tone and intonation through voice recognition technology, such as nervousness, anger, calmness, etc., to assess the driver's emotional state.

[0024] After collecting the above data, it is necessary to extract key features that can reflect the risk of driving behavior from each data source. These features may include statistics (such as mean, standard deviation, maximum, minimum, etc.), category labels (such as classification of weather conditions), time series data (such as speed, acceleration change trend), etc. First feature: features extracted from vehicle internal data, such as speed fluctuation range, acceleration change rate, continuous driving time threshold, etc. Second feature: features extracted from external environment data, such as the degree of influence of weather conditions on driving, the safety factor of road type, statistics of relative speed and distance changes, etc. Third feature: features extracted from driver biometric data, such as classification labels of facial expressions, blinking frequency, duration of facial expressions, etc. Fourth feature: features extracted from in-vehicle video data, such as the stability of eye attention trajectory, the category and frequency of gestures, the threshold of eye closure duration, etc. Fifth feature: features extracted from in-vehicle audio data, such as classification labels of tone and intonation, degree of emotional fluctuation, etc.

[0025] The above extracted features are fused to form a multi-dimensional feature vector. This process may involve steps such as feature standardization, feature selection, and feature transformation to ensure that different features have the same weight and comparability in the feature vector. The fused feature vector is input into a preset machine learning model (such as a neural network, support vector machine, decision tree, etc.). The preset machine learning model has been trained and verified with a large amount of historical data and can accurately identify risk patterns in driving behavior. The model will output a risk detection score to assess the risk level of the current driving behavior. When the risk detection score exceeds the preset threshold, the current driving behavior is considered to have a high risk. At this time, the system will display risk information (such as "Please pay attention to driving safety", "Avoid fatigue driving", etc.) on the on-board display screen and give voice prompts to remind the driver to pay attention to safety. These prompts are intended to help drivers promptly discover and correct bad driving habits and reduce the risk of traffic accidents.

[0026] By collecting vehicle internal data, external environment data, driver biometric data, in-vehicle video data, and in-vehicle audio data, driving behavior can be analyzed comprehensively and multi-dimensionally, thereby more accurately assessing risks. Real-time collection and processing of this data enables the system to promptly detect potential risk behaviors during driving, such as speeding, fatigue driving, distracted driving, etc., and give timely risk detection scores. When the risk score exceeds the preset threshold, the system can immediately display the risk information on the in-vehicle display screen and remind the driver through voice prompts, thereby effectively preventing traffic accidents. The fused feature vectors are processed using a machine learning model to achieve intelligent detection of driving behavior risks. This method not only improves the efficiency and accuracy of risk detection, but also reduces the need for manual intervention, making the driving process more automated and intelligent. By timely detecting and warning of driving behavior risks, the method helps drivers take preventive measures before potential dangers occur, thereby avoiding or reducing the occurrence of traffic accidents. This not only protects the lives of drivers and passengers, but also reduces the risk of vehicle damage and property loss.

[0027] Optionally, extracting a first feature from the vehicle interior data includes: Calculating an average speed of the first speed within a preset time period, calculating a standard deviation of the target acceleration within the preset time period, determining a first extreme value of the target acceleration within the preset time period, and determining a second extreme value of the first speed within the preset time period; When the first extreme value is greater than or equal to a first threshold, constructing a first feature according to the first extreme value, the second extreme value, the current continuous driving time, and the frequency of activating the accelerator and the brake; When the first extreme value is less than the first threshold, determining whether the standard deviation is greater than or equal to a second threshold; When the standard deviation is greater than or equal to the second threshold, constructing a first feature according to the standard deviation, the second extreme value, the current continuous driving time, and the frequency of activating the accelerator and the brake; When the standard deviation is less than the second threshold, a first feature is constructed according to the standard deviation, the average speed, the current continuous driving time, and the frequency of activating the accelerator and the brake.

[0028] Calculate the average speed of the vehicle at the first speed within a preset time period (such as the past one minute, five minutes, etc.). This parameter can reflect the average driving state of the vehicle during this period. Calculate the standard deviation of the target acceleration within the preset time period. The standard deviation is a statistic that measures the size of data fluctuations. Here it is used to reflect the fluctuation of acceleration, that is, the frequency and magnitude of the acceleration or deceleration of the vehicle during this period. First extreme value: Determine the maximum or minimum value of the target acceleration within the preset time period (depending on the specific situation, here it is assumed to be the maximum value). This parameter can reflect the maximum acceleration change experienced by the vehicle during this period. Second extreme value: Determine the maximum or minimum value of the first speed within the preset time period (also depending on the specific situation, here it is assumed to be the maximum value). This parameter can reflect the maximum driving speed of the vehicle during this period. If the first extreme value of the target acceleration exceeds a preset threshold (indicating that the vehicle has experienced a large acceleration change), the first feature is constructed based on the first extreme value, the second extreme value (maximum speed), the current continuous driving time, and the frequency of activating the accelerator and brake. These parameters together reflect the driving state of the vehicle when experiencing a large acceleration change, the driver's fatigue level, and the frequency of operation. If the first extreme value of the target acceleration does not exceed the first threshold, but the standard deviation exceeds the second preset threshold (indicating that the acceleration fluctuates greatly), the first feature is constructed based on the standard deviation, the second extreme value (maximum speed), the current continuous driving time, and the frequency of activating the accelerator and brake. These parameters reflect the driving state and frequency of operation of the vehicle when the acceleration fluctuates greatly. If the first extreme value and the standard deviation of the target acceleration do not exceed their respective thresholds (indicating that the vehicle is driving relatively smoothly), the first feature is constructed based on the standard deviation (which is small at this time but can still reflect certain fluctuations), the average speed, the current continuous driving time, and the frequency of activating the accelerator and brake. These parameters reflect the driving speed of the vehicle in a stable driving state, the driver's fatigue level, and the frequency of operation. Through the above condition judgment and feature construction process, the corresponding feature vectors can be extracted according to different driving behavior characteristics (such as large acceleration changes, large acceleration fluctuations, stable driving, etc.). These feature vectors will be subsequently input into the machine learning model for the calculation of risk detection scores. Since multiple driving behavior characteristics are taken into account, the driver's driving behavior and potential risks can be more comprehensively reflected, thereby improving the accuracy and reliability of risk detection.

[0029] By calculating the average speed of the first speed within a preset time period, the standard deviation and extreme value of the target acceleration within a preset time period, the dynamic changes in the vehicle driving process, such as the stability of the speed and the fluctuation of the acceleration, can be captured more finely. These statistics, as features, can more accurately reflect certain characteristics of driving behavior. According to the comparison between the extreme value (first extreme value) of the target acceleration and the preset threshold, it can be preliminarily judged whether there are possible risk behaviors such as sharp acceleration or deceleration. When the first extreme value exceeds the threshold, the first feature is constructed in combination with other relevant features (such as the second extreme value, the current continuous driving time, the frequency of activating the accelerator and brake), which helps to more accurately identify high-risk driving behaviors. According to the comparison between the standard deviation of the target acceleration and another preset threshold, the stability of the driving behavior can be further subdivided. When the standard deviation is large, it indicates that the acceleration fluctuates greatly, which may mean that the driving behavior is not stable enough; when the standard deviation is small, it indicates that the acceleration is relatively stable. According to these two situations, constructing different feature combinations respectively helps to describe the driving behavior more comprehensively. Through conditional judgment, different feature combinations can be constructed according to different situations to adapt to different driving scenarios. For example, when driving on a highway, the fluctuation of acceleration may be relatively small, while in a congested section of a city, the fluctuation of acceleration may be large. By constructing different feature combinations, driving behaviors in different scenarios can be more accurately reflected.

[0030] Optionally, extracting the second feature from the external environment data includes: Determine whether a distance change rate is greater than or equal to a third threshold value according to the distance change, and when the distance change rate is greater than or equal to the third threshold value, determine whether the target vehicle is located in the target direction of the vehicle; When the target vehicle is located in a target direction of the vehicle, calculating a relative speed between the first speed and the second speed; A second feature is constructed according to the feature value of the weather condition, the feature value of the road type, and the relative speed.

[0031] The first vehicle currently closest to the vehicle is determined, and the rate of change of the distance between the vehicle and the first vehicle is calculated. This rate of change reflects the relative movement speed between the two, that is, whether they are approaching or moving away. The system determines whether the rate of change of the distance is greater than or equal to a preset third threshold. The third threshold is used to determine whether there is a potential risk of collision. If the rate of change of the distance is large, it means that the relative movement speed between the two vehicles is very fast, and there may be a risk of collision. When the rate of change of the distance is greater than or equal to the third threshold, the system further determines whether the target vehicle is located in the target direction of the vehicle. Here, the "target direction" may refer to the direction in which the vehicle is currently traveling, or the direction that the driver intends to change (such as the direction indicated by the turn signal). If the target vehicle is located in the target direction of the vehicle, the relative movement between the two may be more dangerous because they are moving in the same direction. When it is determined that the target vehicle is located in the target direction of the vehicle, the system calculates the relative speed between the vehicle and the target vehicle. This relative speed is obtained by comparing the speeds of the two (the first speed and the second speed). The relative speed is an important feature because it directly reflects the approach speed between the two, so that the potential risk of collision can be evaluated. The system constructs the second feature based on the characteristic value of the weather condition, the characteristic value of the road type, and the calculated relative speed. Weather conditions (such as rainy days, snowy days, sunny days, etc.) and road types (such as highways, urban roads, rural roads, etc.) will affect the driving stability and safety of vehicles. Therefore, incorporating these factors as features into the risk detection model can improve the accuracy and adaptability of the model.

[0032] By judging whether the distance change rate is greater than or equal to the third threshold, potential changes in the external environment can be preliminarily identified, such as acceleration or deceleration of the vehicle ahead, road congestion, etc. This helps the vehicle to perceive changes in the surrounding environment more timely, thereby making more accurate driving decisions. When the distance change rate meets the conditions, it is further judged whether the target vehicle is in the target direction of the vehicle. This step helps the vehicle to more accurately identify potential risk vehicles related to its driving direction, thereby avoiding unnecessary interference and misjudgment. By calculating the relative speed of the first speed and the second speed, the speed difference between the vehicle and the target vehicle can be more intuitively reflected. This feature is of great significance for evaluating the relative position and potential collision risk between the two vehicles. The second feature is constructed according to the characteristic value of the weather condition, the characteristic value of the road type, and the relative speed. This method can comprehensively consider the impact of the external environment on driving behavior. For example, when driving on rainy or snowy days, the slippery road may cause the vehicle to brake more; while when driving on a highway, the vehicle speed is usually faster, and the potential risk is also increased accordingly. By introducing the characteristic values ​​of these environmental factors, the risk level of driving behavior can be more accurately assessed. The second feature constructed by the above steps can more comprehensively reflect the impact of the external environment on driving behavior and the relative relationship between the vehicle and the target vehicle. This helps the vehicle to judge potential risks more accurately, so that preventive measures can be taken in advance, such as slowing down and avoiding, to improve driving safety.

[0033] Optionally, extracting a third feature from the driver's biometric data includes: In the captured image or video frame, using a facial detection algorithm to locate the driver's facial area; In the facial area, key feature points of the face are marked using a feature point detection algorithm, wherein the key feature points include the corners of the eyes, the corners of the mouth, and the eyebrows; Facial expression features are extracted based on the key feature points, the facial expression features including the degree of opening and closing of the eyes, the degree of curvature of the eyebrows, and the degree of uplift or droop of the corners of the mouth, and a third feature is constructed based on the facial expression features.

[0034] Use facial detection algorithms, such as Haar features, HOG features combined with SVM classifiers, or deep learning-based facial detection methods (such as using convolutional neural networks (CNNs)) to accurately locate the driver's facial area in the captured image or video frame. These algorithms can recognize faces in complex backgrounds and give the location and size of the face. Get a rectangular box containing the driver's face, which accurately locates the location and range of the face. In the facial area, mark the key feature points of the face, which can reflect the shape and details of the face. Use feature point detection algorithms, such as ASM (active shape model), AAM (active appearance model), or deep learning-based methods (such as 68-point facial feature detection in the Dlib library). These algorithms can identify and mark the locations of key feature points such as the corners of the eyes, corners of the mouth, and eyebrows in the facial area. Get a series of coordinates of facial feature points, which accurately describe the shape and details of the face. According to the location of the facial feature points, extract facial expression features, which can reflect the driver's emotions and attention status. By calculating the distance or area change of the feature points in the eye area, it can be determined whether the eyes are open or closed, and the degree of opening and closing. By calculating the curvature or inclination angle of the feature points in the eyebrow area, it is possible to determine whether the eyebrows are up, down, or straight, and the degree of curvature. By calculating the position change of the feature points in the corner of the mouth area, it is possible to determine whether the corner of the mouth is up (indicating happiness or smiling) or down (indicating sadness or frustration), and the degree of up or down. A series of facial expression feature values ​​are obtained, which quantitatively describe the dynamic changes of the eyes, eyebrows, and corners of the mouth, thereby reflecting the driver's facial expression and emotional state. The extracted facial expression feature values ​​are integrated into a sub-feature vector as the third feature for subsequent driving behavior risk detection. Specifically, the feature values ​​such as the degree of opening and closing of the eyes, the degree of curvature of the eyebrows, and the degree of up or down of the corners of the mouth can be normalized (such as mapping the values ​​to the range of 0-1), and then combined into a sub-feature vector in a certain order.

[0035] By using the facial detection algorithm and the feature point detection algorithm, the driver's facial area can be accurately located and the key feature points of the face can be marked. This helps to capture the changes in the driver's facial expressions more accurately, thereby more accurately identifying the driver's fatigue, distraction, anger and other states. Not only are the key feature points of the face located, but also the facial expression features such as the degree of eye opening and closing, the degree of eyebrow curvature, and the degree of uplift or droop of the corners of the mouth are further extracted. These features can more finely reflect the driver's emotional state and attention level, and help to more comprehensively assess driving risks. The facial detection algorithm and the feature point detection algorithm usually have high real-time and robustness, and can work stably under different lighting conditions, head postures and occlusions. This means that the method can extract the driver's facial expression features in real time in various complex driving environments, providing timely and reliable information for driving behavior risk detection. Facial expression is an important reflection of the driver's psychological state and emotional state. By extracting and analyzing facial expression features, the method can provide an important basis for driving behavior risk detection. For example, when the driver shows fatigue or distraction, the system can issue a warning in time to remind the driver to pay attention to safety.

[0036] Optionally, extracting the fourth feature from the vehicle-mounted video data includes: Acquiring a sight line movement trajectory of the driver, and determining a distraction degree characteristic according to the sight line movement trajectory; Detecting and tracking the driver's hand position in the image frame, classifying the hand posture, identifying the driver's gesture action, and extracting gesture features from the gesture action, wherein the gesture features include gesture type, duration, and motion trajectory; Detecting the eye-closing duration of the driver, and acquiring the head posture and mouth movements of the driver, and determining fatigue driving characteristics through the eye-closing duration, the head posture and the mouth movements; A fourth feature is constructed based on the attention distraction degree feature, the gesture feature, and the fatigue driving feature.

[0037] The driver's facial image is captured by the on-board camera, and the driver's line of sight movement trajectory is tracked through image processing technology (such as pupil tracking, head posture estimation, etc.). The line of sight movement trajectory is analyzed to determine whether the driver's line of sight frequently deviates from the road ahead or stays in non-driving related areas (such as rearview mirrors, mobile phones, etc.) for a long time. According to the frequency and duration of line of sight deviation, the driver's distraction degree is quantified and used as part of the fourth feature. In the image frame, the driver's hand position is detected by image processing or deep learning technology, and it is continuously tracked. According to the hand position and movement trajectory, the hand posture is classified, such as holding the steering wheel, making a phone call, eating, etc. Further identify specific gestures, such as the type of gesture (such as pointing, waving, etc.), duration and movement trajectory. The identified gestures are converted into feature vectors, including gesture type, duration and movement trajectory, etc., as components of the fourth feature. The driver's eye closure state is detected by image processing technology, and the eye closure duration is calculated. At the same time, the driver's head posture (such as nodding, shaking head, etc.) and mouth movements (such as yawning, talking, etc.) are monitored. Combine the duration of eye closure, head posture and mouth movements to comprehensively assess the driver's fatigue level. For example, features such as long eye closure, frequent nodding or yawning may indicate that the driver is in a state of fatigue. The above-mentioned distraction level features, gesture features, fatigue driving features, etc. are integrated together to form the fourth feature. This feature vector can fully reflect the driver's behavioral state during driving, and provide an important basis for subsequent driving behavior analysis, risk assessment or early warning.

[0038] By obtaining the driver's line of sight movement trajectory, it is possible to determine whether the driver's attention is distracted, which is of great significance for timely detecting driver distracted driving and improving driving safety. Detecting and tracking the driver's hand position and posture, and identifying gestures can further understand the driver's operating habits and behavioral intentions during driving, which helps to detect abnormal driving behavior in a timely manner. Detecting the driver's eye closure duration, head posture and mouth movements, combined with this information, can accurately determine whether the driver is in a fatigue driving state. This plays a key role in preventing traffic accidents caused by fatigue driving. The extracted attention distraction level features, gesture features, and fatigue driving features provide rich data support for driving behavior analysis. These features can be used for training and optimization of driving behavior models to improve the intelligence level of intelligent driving systems. By real-time monitoring and analysis of the driver's line of sight, hand posture, gesture movements, and fatigue level, the intelligent driving assistance system can more intelligently judge the driver's state, issue warnings or take intervention measures in a timely manner, thereby improving driving safety and comfort.

[0039] Optionally, extracting the fifth feature from the vehicle-mounted audio data includes: Calculating volume differences and pitch differences between adjacent audio frames, determining volume features according to the volume differences, and determining pitch features according to the pitch differences, wherein the volume features include maximum volume, minimum volume, average volume, and standard deviation of volume changes, and the pitch features include average pitch and standard deviation of pitch changes; Converting the in-vehicle audio data into text, calculating the number of words in the text, determining a speech rate feature according to the number of words, and extracting content from the text, determining an emotional feature according to the content; A fifth feature is constructed based on the volume feature, the tone feature, the speech speed feature and the emotion feature.

[0040] The in-vehicle audio data is divided into multiple continuous audio frames, and the noise including music in the audio frames is removed. Then, the volume difference between adjacent audio frames is calculated, which can be achieved by comparing the audio energy or amplitude of each frame. The volume difference reflects the volume fluctuation of the driver's voice. Based on the volume difference, the maximum volume (i.e. the maximum volume value in the audio frame), the minimum volume (i.e. the minimum volume value in the audio frame), the average volume (the average of the volume of all audio frames), and the standard deviation of the volume change (reflecting the degree of volume fluctuation) can be calculated. Pitch is another important feature of speech, which reflects the fundamental frequency of speech. By calculating the pitch difference between adjacent audio frames, the pitch change of the driver's voice can be understood. Based on the pitch difference, the average pitch (the average of the pitch of all audio frames) and the standard deviation of the pitch change (reflecting the degree of pitch fluctuation) can be calculated. Using speech recognition technology, the in-vehicle audio data is converted into text. This step is the basis for subsequent analysis. The number of words in the text is calculated, and combined with the total duration of the audio data, the speaking rate (i.e. the number of words uttered per unit time) can be calculated. The speaking rate reflects how fast the driver speaks. Extract the content from the text and use natural language processing technology to perform sentiment analysis. Sentiment analysis can identify the type of emotions expressed in the text, such as positive, negative, neutral, etc. Emotional features reflect the emotional state of the driver during driving. The volume features (maximum volume, minimum volume, average volume, standard deviation of volume change), tone features (average tone, standard deviation of tone change), speech speed features and emotion features calculated above are integrated to construct the fifth feature.

[0041] By calculating the volume difference and pitch difference between adjacent audio frames, volume features and pitch features can be extracted, which can reflect the driver's emotional changes and attention status during driving. For example, a sudden increase in volume may indicate that the driver is emotionally excited or nervous, while a change in pitch may reflect the driver's fatigue or emotional ups and downs. After converting the in-vehicle audio data into text, the number of words in the text can be calculated to determine the speech rate feature. Speech rate is an important indicator reflecting the driver's speaking speed and thinking activity, and is of great significance for judging whether the driver is tired, distracted or nervous. By extracting the content in the text, the driver's emotional characteristics can be further analyzed. Emotion is one of the important influencing factors of driving behavior. By identifying the driver's emotional state, potential driving risks can be discovered in time. For example, negative emotions such as anger and anxiety may cause the driver to make dangerous driving behaviors. Combining information from multiple dimensions such as volume features, pitch features, speech rate features and emotional features, a more comprehensive driving behavior feature can be constructed. These features provide rich data support for intelligent driving assistance systems, which helps the system to more accurately judge the driver's status, issue warnings or take intervention measures in time, thereby improving driving safety.

[0042] Optionally, the fusing the first feature, the second feature, the third feature, the fourth feature and the fifth feature to obtain a feature vector includes: performing standardization processing on the first feature, the second feature, the third feature, the fourth feature, and the fifth feature; Perform weighted summation on the standardized first feature, the second feature, the third feature, the fourth feature, and the fifth feature to form a feature vector.

[0043] Standardization is an important step in data preprocessing. It helps to eliminate the dimensional differences between different features and make the features comparable in value. Standardization is particularly important in feature fusion because different features may come from different data sources or have different measurement units. Direct fusion may lead to inaccurate or difficult to interpret results. Standardization methods usually include: Z-score standardization: subtract the mean value of each feature from its value and then divide it by the standard deviation. This method makes the processed feature value have a mean of 0 and a standard deviation of 1. Min-Max standardization: scale the value of each feature to a specified range (usually between 0 and 1). This method is achieved by calculating the minimum and maximum values ​​of the feature value. Before the first feature, the second feature, the third feature, the fourth feature, and the fifth feature are fused, these features need to be standardized separately. This ensures that they have the same weight and comparability during the fusion process. Weighted summation is a common method for fusing multiple features into a feature vector. In this process, each feature is assigned a weight that reflects the importance or influence of the feature in the whole. The determination of weights is usually based on the importance, relevance, or empirical knowledge of the feature. In practical applications, weights can be determined by machine learning algorithms (such as principal component analysis, linear discriminant analysis, etc.) or expert scoring. The standardized value of each feature is multiplied by its corresponding weight, and then all weighted values ​​are added to obtain each component of the feature vector. All weighted sums are used as components of the feature vector to form a multidimensional vector.

[0044] Standardization is an important step in data preprocessing. It can unify the data range of different features to the same scale, thereby eliminating the dimensional differences between features. Through standardization, each feature is comparable in value, which is convenient for subsequent feature fusion and model training. Weighted summation is a common method to fuse multiple features into a single feature vector. In the weighted summation process, the weight of each feature reflects its importance in the overall feature vector. Through reasonable weight distribution, it can ensure that the feature vector can fully reflect the information of each feature, thereby improving the representation ability of the feature vector. Feature fusion is an important technology in the field of machine learning. It can combine multiple single features into more representative composite features. By fusing multiple features, a more complex and accurate model can be constructed, thereby enhancing the generalization ability of the model. This means that the model can better adapt to different data distributions and scene changes, and improve the accuracy and stability of prediction. Standardization and weighted summation are both relatively simple and efficient calculation processes. Through these two steps, multiple features can be quickly fused into feature vectors, providing strong support for subsequent model training and prediction. At the same time, since the normalization process eliminates the dimensional differences between features, it can avoid problems such as numerical overflow or underflow during the calculation process, thereby improving the accuracy and stability of the calculation.

[0045] This embodiment also discloses a driving behavior risk detection system. Figure 2 is a module diagram of a driving behavior risk detection system disclosed in an embodiment of the present application, such as Figure 2 As shown, the system includes a collection module 201, a feature module 202, a fusion module 203 and a display module 204, wherein: The acquisition module 201 is configured to collect vehicle internal data, external environment data, driver biometric data, vehicle video data, and vehicle audio data, wherein the vehicle internal data includes a first speed of the vehicle, a target acceleration, a current continuous driving time, and a frequency of activating the accelerator and the brake, the external environment data includes weather conditions, road types, a second speed of the nearest target vehicle around the vehicle, and a change in the distance from the vehicle, the driver biometric data includes facial expressions, the vehicle video data includes eye attention tracks, hand gestures, and eye closure time, and the vehicle audio data includes the tone and intonation of the driver's speech; A feature module 202 configured to extract a first feature from the vehicle interior data, a second feature from the external environment data, a third feature from the driver biometric data, a fourth feature from the vehicle video data, and a fifth feature from the vehicle audio data; A fusion module 203 is configured to fuse the first feature, the second feature, the third feature, the fourth feature and the fifth feature to obtain a feature vector, and input the feature vector into a preset machine learning model to obtain a risk detection score; The display module 204 is configured to display risk information on the vehicle display screen and give a voice prompt when the risk detection score exceeds a threshold.

[0046] Optionally, the feature module 202 is configured to: Calculating an average speed of the first speed within a preset time period, calculating a standard deviation of the target acceleration within the preset time period, determining a first extreme value of the target acceleration within the preset time period, and determining a second extreme value of the first speed within the preset time period; When the first extreme value is greater than or equal to a first threshold, constructing a first feature according to the first extreme value, the second extreme value, the current continuous driving time, and the frequency of activating the accelerator and the brake; When the first extreme value is less than the first threshold, determining whether the standard deviation is greater than or equal to a second threshold; When the standard deviation is greater than or equal to the second threshold, constructing a first feature according to the standard deviation, the second extreme value, the current continuous driving time, and the frequency of activating the accelerator and the brake; When the standard deviation is less than the second threshold, a first feature is constructed according to the standard deviation, the average speed, the current continuous driving time, and the frequency of activating the accelerator and the brake.

[0047] Optionally, the feature module 202 is configured to: Determine whether a distance change rate is greater than or equal to a third threshold value according to the distance change, and when the distance change rate is greater than or equal to the third threshold value, determine whether the target vehicle is located in the target direction of the vehicle; When the target vehicle is located in a target direction of the vehicle, calculating a relative speed between the first speed and the second speed; A second feature is constructed according to the feature value of the weather condition, the feature value of the road type, and the relative speed.

[0048] Optionally, the feature module 202 is configured to: In the captured image or video frame, using a facial detection algorithm to locate the driver's facial area; In the facial area, key feature points of the face are marked using a feature point detection algorithm, wherein the key feature points include the corners of the eyes, the corners of the mouth, and the eyebrows; Facial expression features are extracted based on the key feature points, the facial expression features including the degree of opening and closing of the eyes, the degree of curvature of the eyebrows, and the degree of uplift or droop of the corners of the mouth, and a third feature is constructed based on the facial expression features.

[0049] Optionally, the feature module 202 is configured to: Acquiring a sight line movement trajectory of the driver, and determining a distraction degree characteristic according to the sight line movement trajectory; Detecting and tracking the driver's hand position in the image frame, classifying the hand posture, identifying the driver's gesture action, and extracting gesture features from the gesture action, wherein the gesture features include gesture type, duration, and motion trajectory; Detecting the eye-closing duration of the driver, and acquiring the head posture and mouth movements of the driver, and determining fatigue driving characteristics through the eye-closing duration, the head posture and the mouth movements; A fourth feature is constructed based on the attention distraction degree feature, the gesture feature, and the fatigue driving feature.

[0050] Optionally, the feature module 202 is configured to: Calculating volume differences and pitch differences between adjacent audio frames, determining volume features according to the volume differences, and determining pitch features according to the pitch differences, wherein the volume features include maximum volume, minimum volume, average volume, and standard deviation of volume changes, and the pitch features include average pitch and standard deviation of pitch changes; Converting the in-vehicle audio data into text, calculating the number of words in the text, determining a speech rate feature according to the number of words, and extracting content from the text, determining an emotional feature according to the content; A fifth feature is constructed based on the volume feature, the tone feature, the speech speed feature and the emotion feature.

[0051] Optionally, the fusion module 203 is configured to: performing standardization processing on the first feature, the second feature, the third feature, the fourth feature, and the fifth feature; Perform weighted summation on the standardized first feature, the second feature, the third feature, the fourth feature, and the fifth feature to form a feature vector.

[0052] It should be noted that: when the device provided in the above embodiment realizes its function, only the division of the above functional modules is used as an example. In actual application, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the device and method embodiments provided in the above embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0053] This embodiment also discloses an electronic device, referring to Figure 3 The electronic device may include: at least one processor 301 , at least one communication bus 302 , a user interface 303 , a network interface 304 , and at least one memory 305 .

[0054] The communication bus 302 is used to realize the connection and communication between these components.

[0055] The user interface 303 may include a display screen (Display) and a camera (Camera). Optionally, the user interface 303 may also include a standard wired interface and a wireless interface.

[0056] The network interface 304 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0057] Among them, the processor 301 may include one or more processing cores. The processor 301 uses various interfaces and lines to connect various parts in the entire server, and executes various functions of the server and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory 305, and calling data stored in the memory 305. Optionally, the processor 301 can be implemented in at least one hardware form of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), and programmable logic array (Programmable Logic Array, PLA). The processor 301 can integrate one or a combination of a central processing unit (Central Processing Unit, CPU), a graphics processing unit (Graphics Processing Unit, GPU) and a modem. Among them, the CPU mainly processes the operating system, user interface and application programs; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor 301, and it can be implemented separately through a chip.

[0058] Among them, the memory 305 may include a random access memory (Random Access Memory, RAM) and may also include a read-only memory (Read-Only Memory). Optionally, the memory 305 includes a non-transitory computer-readable storage medium. The memory 305 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 305 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store data involved in the above-mentioned various method embodiments, etc. The memory 305 may also be optionally at least one storage device located away from the aforementioned processor 301. As Figure 3 As shown, the memory 305 as a computer storage medium may include an operating system, a network communication module, a user interface module, and an application program for a method for detecting driving behavior risks.

[0059] exist Figure 3 In the electronic device shown, the user interface 303 is mainly used to provide an input interface for the user and obtain data input by the user; and the processor 301 can be used to call the application program for the detection method of driving behavior risk stored in the memory 305. When executed by one or more processors 301, the electronic device executes one or more methods in the above-mentioned embodiments.

[0060] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the order of the actions described, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required for the present application.

[0061] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0062] In the several embodiments provided in this application, it should be understood that the disclosed devices can be implemented in other ways. For example, the device embodiments described above are only schematic, such as the division of units, which is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some service interfaces, and the indirect coupling or communication connection of devices or units can be electrical or other forms.

[0063] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0064] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0065] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a memory 305 and includes several instructions for a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the various embodiments of the present application. The aforementioned memory 305 includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a magnetic disk or an optical disk.

[0066] The above is only an exemplary embodiment of the present disclosure and cannot be used to limit the scope of the present disclosure. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. After considering the disclosure of the specification, those skilled in the art will easily think of other embodiments of the present disclosure. This application is intended to cover any modification, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the technical field that are not recorded in the present disclosure. The description and examples are only regarded as exemplary, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. A method for detecting driving behavior risk, characterized in that: Applied to a driving behavior risk detection platform, the method includes: Collecting vehicle internal data, external environment data, driver biometric data, in-vehicle video data, and in-vehicle audio data, wherein the vehicle internal data includes a first speed of the vehicle, target acceleration, current continuous driving time, and frequency of activating the accelerator and brake, the external environment data includes weather conditions, road type, a second speed of the nearest target vehicle around the vehicle, and a change in distance from the vehicle, the driver biometric data includes facial expressions, the in-vehicle video data includes eye attention tracks, gestures, and eye closure time, and the in-vehicle audio data includes the tone and intonation of the driver's speech; Extracting a first feature from the vehicle interior data, extracting a second feature from the external environment data, extracting a third feature from the driver biometric data, extracting a fourth feature from the vehicle video data, and extracting a fifth feature from the vehicle audio data; The first feature, the second feature, the third feature, the fourth feature and the fifth feature are integrated to obtain a feature vector, and the feature vector is input into a preset machine learning model to obtain a risk detection score; When the risk detection score exceeds a threshold, risk information is displayed on the vehicle display screen and a voice prompt is given.

2. The method for detecting driving behavior risk according to claim 1, characterized in that: The extracting a first feature from the vehicle interior data comprises: Calculating an average speed of the first speed within a preset time period, calculating a standard deviation of the target acceleration within the preset time period, determining a first extreme value of the target acceleration within the preset time period, and determining a second extreme value of the first speed within the preset time period; When the first extreme value is greater than or equal to a first threshold, constructing a first feature according to the first extreme value, the second extreme value, the current continuous driving time, and the frequency of activating the accelerator and the brake; When the first extreme value is less than the first threshold, determining whether the standard deviation is greater than or equal to a second threshold; When the standard deviation is greater than or equal to the second threshold, constructing a first feature according to the standard deviation, the second extreme value, the current continuous driving time, and the frequency of activating the accelerator and the brake; When the standard deviation is less than the second threshold, a first feature is constructed according to the standard deviation, the average speed, the current continuous driving time, and the frequency of activating the accelerator and the brake.

3. The method for detecting driving behavior risk according to claim 1, characterized in that: The extracting the second feature from the external environment data comprises: Determining whether a distance change rate is greater than or equal to a third threshold value according to the distance change, and determining whether the target vehicle is located in a target direction of the vehicle when the distance change rate is greater than or equal to the third threshold value; When the target vehicle is located in a target direction of the vehicle, calculating a relative speed between the first speed and the second speed; A second feature is constructed according to the feature value of the weather condition, the feature value of the road type, and the relative speed.

4. The method for detecting driving behavior risk according to claim 1, characterized in that: The extracting a third feature from the driver's biometric data comprises: In the captured image or video frame, using a facial detection algorithm to locate the driver's facial area; In the facial area, key feature points of the face are marked using a feature point detection algorithm, wherein the key feature points include the corners of the eyes, the corners of the mouth, and the eyebrows; Facial expression features are extracted based on the key feature points, the facial expression features including the degree of opening and closing of the eyes, the degree of curvature of the eyebrows, and the degree of uplift or droop of the corners of the mouth, and a third feature is constructed based on the facial expression features.

5. The method for detecting driving behavior risk according to claim 1, characterized in that: The extracting the fourth feature from the vehicle-mounted video data comprises: Acquiring a sight line movement trajectory of the driver, and determining a distraction degree characteristic according to the sight line movement trajectory; Detecting and tracking the driver's hand position in the image frame, classifying the hand posture, identifying the driver's gesture action, and extracting gesture features from the gesture action, wherein the gesture features include gesture type, duration, and motion trajectory; Detecting the eye-closing duration of the driver, and acquiring the head posture and mouth movements of the driver, and determining fatigue driving characteristics through the eye-closing duration, the head posture and the mouth movements; A fourth feature is constructed based on the attention distraction degree feature, the gesture feature, and the fatigue driving feature.

6. The method for detecting driving behavior risk according to claim 1, characterized in that: The extracting the fifth feature from the vehicle audio data comprises: Calculating volume differences and pitch differences between adjacent audio frames, determining volume features according to the volume differences, and determining pitch features according to the pitch differences, wherein the volume features include maximum volume, minimum volume, average volume, and standard deviation of volume changes, and the pitch features include average pitch and standard deviation of pitch changes; Converting the in-vehicle audio data into text, calculating the number of words in the text, determining a speech rate feature according to the number of words, and extracting content from the text, determining an emotional feature according to the content; A fifth feature is constructed based on the volume feature, the tone feature, the speech speed feature and the emotion feature.

7. The method for detecting driving behavior risk according to claim 1, characterized in that: The fusing the first feature, the second feature, the third feature, the fourth feature and the fifth feature to obtain a feature vector comprises: performing standardization processing on the first feature, the second feature, the third feature, the fourth feature, and the fifth feature; Perform weighted summation on the standardized first feature, the second feature, the third feature, the fourth feature, and the fifth feature to form a feature vector.

8. A driving behavior risk detection system, characterized in that: It includes acquisition module, feature module, fusion module and display module, among which: a collection module configured to collect vehicle internal data, external environment data, driver biometric data, vehicle video data, and vehicle audio data, wherein the vehicle internal data includes a first speed of the vehicle, target acceleration, current continuous driving time, and frequency of activating the accelerator and brake, the external environment data includes weather conditions, road type, a second speed of the nearest target vehicle around the vehicle, and a change in distance from the vehicle, the driver biometric data includes facial expressions, the vehicle video data includes eye attention tracks, hand gestures, and eye closing time, and the vehicle audio data includes the tone and intonation of the driver's speech; a feature module configured to extract a first feature from the vehicle interior data, a second feature from the external environment data, a third feature from the driver biometric data, a fourth feature from the vehicle video data, and a fifth feature from the vehicle audio data; a fusion module, configured to fuse the first feature, the second feature, the third feature, the fourth feature and the fifth feature to obtain a feature vector, and input the feature vector into a preset machine learning model to obtain a risk detection score; The display module is configured to display risk information on the vehicle display screen and give a voice prompt when the risk detection score exceeds a threshold.

9. An electronic device, characterized in that: It includes a processor, a memory, a user interface and a network interface, the memory is used to store instructions, the user interface and the network interface are both used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes 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 instructions, and when the instructions are executed, the method according to any one of claims 1 to 7 is executed.

Images