A safe driving warning method based on driving recorder

Abstract

The present invention relates to the field of intelligent vehicle technology, and specifically, to a safe driving warning method based on a driving recorder. It comprises the following steps: the driving recorder has a built-in sound recognition processor, and the monitored sound source signal is compared with the target audio sample in the database. Once the target sound source is identified, the driving recorder performs real-time analysis on the image. After the target vehicle is identified, the optimal avoidance scheme is obtained based on the positioning of the real-time target sound source and the collected traffic data, and then the current state of the self-driving vehicle is checked. If the state of the self-driving vehicle is safe, the warning and avoidance instructions are immediately conveyed to the driver through voice, otherwise, the avoidance scheme is adjusted. This method enhances the adaptability of the system to occlusion, rapid movement and large displacement, and also improves the overall prediction accuracy of the target vehicle, providing the driver with more accurate safety tips.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent vehicles, and in particular to a safe driving warning method based on a driving recorder. Background Art

[0002] With the continuous growth and complexity of urban traffic flow, ensuring road safety has become a core concern in the development of driving technology. As an important component of modern vehicles, dashcams not only provide evidence support for traffic accidents, but also show great potential in improving driving safety. However, traditional dashcams mainly focus on video recording, and the application of audio information is relatively limited. In view of this, it is particularly important to introduce sound recognition processors to enhance the functions of dashcams. In particular, for the recognition of emergency signals such as sirens, by analyzing the sounds of the surrounding environment in real time, the driver can be warned in advance that special vehicles such as ambulances, fire trucks or police cars are approaching. This safe driving warning method based on dashcams can promptly remind the driver to take appropriate avoidance measures without affecting normal driving, thereby effectively reducing the potential accident risks caused by failure to respond to emergency vehicles in time, and improving the overall safety and efficiency of road traffic.

[0003] Existing driving recorders usually focus on video recording, and most of them are limited to recording image data for later review. They do not have real-time analysis capabilities, especially when facing objects that are not in direct sight. It is impossible to accurately obtain the location information of the target object. When the target vehicle performs sudden acceleration, sudden deceleration, emergency lane change, etc., its speed and direction will change significantly during maneuvers. Path prediction will easily produce large prediction deviations. Therefore, a safe driving warning method based on a driving recorder is designed. Summary of the invention

[0004] The purpose of the present invention is to provide a safe driving warning method based on a driving recorder to solve the problem that the existing driving recorder proposed in the above background technology does not have real-time analysis capability, cannot accurately obtain the location information of the target object and the target vehicle, and the path prediction is also prone to large prediction deviations.

[0005] To achieve the above object, the present invention provides a safe driving warning method based on a driving recorder, comprising the following steps: S1. The driving recorder has a built-in sound recognition processor, which is used to monitor the sound source signal in the surrounding environment. When the volume of the sound source signal is greater than the preset volume threshold, the monitored sound source signal is compared with the target audio sample in the database. Once the target sound source is identified, the target sound source is located based on the hyperbolic intersection method of TDOA, and a first signal is generated; wherein the first signal carries the location of the target sound source; S2. After receiving the first signal, the driving recorder starts the camera to record the video, and uses the computer vision algorithm to analyze the image in real time, extract key features and identify based on hybrid decision-making. After identifying the target vehicle, the movement path of the target vehicle is predicted based on the positioning of the real-time target sound source. In the process of predicting the movement path of the target vehicle, the influence of occlusion, rapid movement and large displacement is introduced for optimization; S3. According to the movement path of the target vehicle, the driving assistance system of the self-driving vehicle collects surrounding traffic data, and uses a path planning algorithm to obtain an optimal avoidance plan based on the collected traffic data. After the avoidance plan is obtained, a second signal is generated; S4. After receiving the second signal, the driving recorder first checks the current state of the self-driving vehicle. If the self-driving vehicle is in a safe state, the driver is immediately warned and given an avoidance instruction through voice. Otherwise, the avoidance plan is adjusted.

[0006] As a further improvement of the technical solution, in S1, the target sound source is located based on the hyperbolic intersection method of TDOA, as follows: It is known that a driving recorder includes at least three microphones arranged in a collinear manner: , , , and the positions of the three microphones and the time delay difference between them are known: , , ,but: S11. Calculate the delay difference between microphones : ; in, is the generalized cross-correlation function; is the fast Fourier transform; is to represent the inverse fast Fourier transform; For Moment Microphone Time series signal of For Moment Microphone Time series signal of is the delay variable; is a small positive number; is the complex conjugate fast Fourier transform; ; in, For microphone and microphone The delay difference of the signal between them; S12, based on the delay difference Create a hyperbola between the three microphones: ; ; ; in, is the speed of sound propagation; is the time delay difference between the first microphone and the second microphone; is the time delay difference between the first microphone and the third microphone; is the time delay difference between the second microphone and the third microphone; is the position vector of the first microphone; is the position vector of the second microphone; is the position vector of the third microphone; is the position vector of the target sound source; The speed of sound after the influence of environmental factors is introduced; is temperature; is the relative humidity; is the atmospheric pressure; S13, using weighted least squares method for optimization, the optimization results are as follows: ; in, For the The difference between the actual distance from the microphone to the sound source and the predicted distance; For the The position vectors of the microphones; For the The position vector of the microphones; For the Microphone and The delay difference between the microphones; ; ; in, Minimize the weighted sum of squares of the position vector of the target sound source; It is Microphone and The weight of the delay difference between the microphones; is the environmental feature vector; is the weight predicted after introducing the environmental feature vector; S14. Finally, by adjusting Make When the minimum value is reached, the optimal target sound source coordinates can be obtained.

[0007] As a further improvement of the technical solution, in S2, key features are extracted and identified based on hybrid decision making, specifically: S21, using the color model to check whether the color of the vehicle conforms to the color spectrum range of the target vehicle in the known database, if it conforms, proceed to S22, otherwise, exclude the possibility that the vehicle is the target vehicle; S22, if it is in compliance, extract key features of the vehicle in the image, divide the key features into necessary key features and non-essential key features, first match the extracted necessary key features, if the match is successful, it is the target vehicle, otherwise proceed to S23; S23, matching the extracted non-essential key features, and only when all the identified non-essential key features are matched successfully, it is the target vehicle.

[0008] As a further improvement of the present technical solution, the necessary key features in S22 include vehicle markers and vehicle equipment configurations.

[0009] As a further improvement of the technical solution, in S2, after the target vehicle is identified, the movement path of the target vehicle is predicted according to the positioning of the real-time target sound source, as follows: ; in, is the target position at the next moment; for The target position at the moment; is the optical flow velocity vector; is the time interval between two frames of images; ; ; in, is the illumination compensation factor; is the proportionality coefficient; For the The average brightness of the frame image; For the The average brightness of the frame image; is the illumination error correction term; is the transpose operation.

[0010] As a further improvement of the technical solution, in S2, the influence of occlusion, rapid motion and large displacement is introduced to optimize the process of predicting the motion path of the target vehicle, and the optimization is as follows: in, is the target position at the next moment after optimization; It is the occlusion detection module; is the dynamic weight factor; For compensation items; Additional corrections provided for the global motion model.

[0011] As a further improvement of the present technical solution, the surrounding traffic data in S3 includes but is not limited to vehicle flow density, road type, and traffic light status.

[0012] As a further improvement of the technical solution, in S3, based on the collected traffic data, a path planning algorithm is used to obtain the optimal avoidance solution, specifically: ; in, is the stage cost; is the vehicle state vector, that is ; is the control input; is the reference state; is a diagonal matrix; is the transpose of the control input; is a diagonal matrix representing the cost of control input; ; in, is the terminal cost; is a terminal state; is the ideal final state; A weighting matrix for adjusting the influence of the terminal state error; is the length of the prediction time domain; Building the cost function: ; in, is the cost function; .

[0013] As a further improvement of the present technical solution, in S4, the current state of the self-driving vehicle includes checking whether the self-driving vehicle is stationary and whether the self-driving vehicle has enough space to operate. If the vehicle is stationary and the self-driving vehicle has enough space to operate, it is in a safe state.

[0014] As a further improvement of the present technical solution, in S4, the avoidance plan is adjusted specifically as follows: the driver is advised by voice to use the horn and high beam to alert the vehicle in front to avoid the vehicle in front, and the hazard warning lights are turned on to attract the attention of the vehicle behind, and the driver is reminded by voice not to accelerate suddenly or turn the steering wheel sharply.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. This safe driving warning method based on a dash cam introduces a special sound recognition processor to monitor and analyze sound source signals in the surrounding environment, which not only expands the functional scope of the dash cam, but also enables the system to obtain additional perceptual input in addition to visual information. Once the volume above the preset threshold is detected and matched to the target audio sample, the system will immediately start the TDOA hyperbolic intersection method to accurately locate the sound source. This sound-based warning mechanism, combined with high-precision sound source positioning, provides drivers with more comprehensive safety protection.

[0016] 2. In this safe driving warning method based on a driving recorder, the effects of occlusion, rapid motion and large displacement are introduced for optimization in the process of predicting the motion path of the target vehicle. This optimization not only enhances the system's adaptability to occlusion, rapid motion and large displacement, but also improves the overall prediction accuracy, so that the safe driving warning system can provide reliable assistance under a wider range of conditions. In this way, even in the face of complex and changing road environments, the system can maintain a high performance level and provide drivers with more accurate safety prompts. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flowchart of the overall process of the present invention. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] Example See also Figure 1 As shown, a safe driving warning method based on a driving recorder is provided, comprising the following steps: In this embodiment, the target vehicle is an ambulance, a fire truck, and a police car, and the target audio is the siren of the target vehicle; S1. The driving recorder has a built-in sound recognition processor, which is used to monitor the sound source signal in the surrounding environment. When the volume of the sound source signal is greater than the preset volume threshold, the monitored sound source signal is compared with the target audio sample in the database. Once the target sound source is identified, the target sound source is located based on the hyperbolic intersection method of TDOA, and a first signal is generated; wherein the first signal carries the location of the target sound source; The volume threshold is set to filter out background noise and other unimportant sounds, ensuring that further processing is triggered only when a sufficiently loud sound (which may be a potential danger signal) is detected. This effectively reduces the false alarm rate and avoids unnecessary waste of computing resources. It also improves the system's response speed because it will only initiate comparison when it is really necessary.

[0020] In S1, the target sound source is located based on the hyperbolic intersection method of TDOA, as follows: The hyperbolic intersection method of TDOA is a positioning technology based on time difference. It uses the time difference between radio signals or sound waves arriving at different receiving points to determine the location of the transmitting source. The core is to construct a series of hyperbolas with the receiving point as the focus and the distance difference corresponding to the time difference as the focal length, and lock the target position through the intersection of these hyperbolas; When multiple microphones in a dashcam receive sounds from the same source, the time it takes for the sounds to reach each microphone will be different due to the different distances between the microphones and the sound source. This time difference is called time delay difference TDOA. Using time delay difference TDOA, a series of hyperbolas can be constructed with the position of each microphone as the focus and the distance difference corresponding to the time difference as the focal length. The intersection of all these hyperbolas is the location of the sound source.

[0021] It is known that a driving recorder includes at least three microphones arranged in a collinear manner: , , , and the positions of the three microphones and the time delay difference between them are known: , , ,but: S11. Calculate the delay difference between microphones : ; in, is the generalized cross-correlation function; is the fast Fourier transform; is to represent the inverse fast Fourier transform; For Moment Microphone Time series signal of For Moment Microphone Time series signal of is the delay variable; It is a small positive number to prevent the denominator from being zero; is the complex conjugate fast Fourier transform; To convert the time domain signal into frequency domain representation for subsequent processing; Refers to the microphone The complex conjugate of the signal is taken after performing fast Fourier transform to calculate the cross-power spectrum between the two signals; the absolute value in the denominator normalizes the cross-power spectrum. This weighting method emphasizes phase information rather than amplitude information, thereby improving the robustness to noise and reverberation environments; ; in, For microphone and microphone The delay difference of the signal between them; Finding the generalized cross-correlation function The time offset corresponding to the maximum value of , which is the time delay difference of the signal between the two microphones This is because when the time alignment of the two signals is optimal, their correlation is strongest, which is manifested as the maximum peak of the cross-correlation function. This can not only effectively reduce the impact of environmental noise, but also maintain high positioning accuracy in the presence of reflected sound waves.

[0022] S12, based on the delay difference Create a hyperbola between the three microphones: ; ; ; in, is the speed of sound propagation; is the time delay difference between the first microphone and the second microphone; is the time delay difference between the first microphone and the third microphone; is the time delay difference between the second microphone and the third microphone; is the position vector of the first microphone; is the position vector of the second microphone; is the position vector of the third microphone; is the position vector of the target sound source; The speed of sound after the influence of environmental factors is introduced; is temperature; is the relative humidity; It is the atmospheric pressure; it can effectively deal with the uncertainty caused by environmental changes; S13. These three hyperbolas will intersect at a certain point in space, which is the location of the sound source. However, in practical applications, due to factors such as measurement errors, the three hyperbolas may not intersect at the same point exactly. Therefore, in order to minimize the errors under all these constraints, the weighted least squares method is used for optimization, and the optimization is as follows: ; in, For the The difference between the actual distance from the microphone to the sound source and the predicted distance; For the The position vector of the microphones; For the The position vector of the microphones; For the Microphone and The delay difference between the microphones; ; ; in, Minimize the weighted sum of squares of the position vector of the target sound source; It is Microphone and The weights of the delay differences between the microphones are usually based on their signal-to-noise ratio or other reliability indicators; is the environmental feature vector, including but not limited to environmental parameters (temperature, humidity, wind speed), signal-to-noise ratio between microphone pairs, and timestamp (taking into account that day and night changes may affect the speed of sound); is the weight predicted after introducing the environmental feature vector; S14. Finally, by adjusting Make When the minimum value is reached, the optimal target sound source coordinates can be obtained.

[0023] S2. After receiving the first signal, the driving recorder starts the camera to record the video, and uses the computer vision algorithm to analyze the image in real time, extract key features and identify based on hybrid decision-making. After identifying the target vehicle, the movement path of the target vehicle is predicted based on the positioning of the real-time target sound source. In the process of predicting the movement path of the target vehicle, the influence of occlusion, rapid movement and large displacement is introduced for optimization; In S2, key features are extracted and identified based on hybrid decisions, specifically: S21, using the color model to check whether the color of the vehicle conforms to the color spectrum range of the target vehicle in the known database, if it conforms, proceed to S22, otherwise, exclude the possibility that the vehicle is the target vehicle; By comparing the color of the vehicle to the color spectrum of the target vehicle in the known database, those objects that obviously do not meet the conditions can be quickly excluded. This method not only speeds up the recognition speed, but also reduces unnecessary computational burden, especially when facing a large amount of background information or non-target objects. For example, fire trucks are usually red, while ambulances are mostly white with specific logos. Such color characteristics make the color model an efficient pre-screening tool.

[0024] S22. If it is in compliance, extract key features of the vehicle in the image, divide the key features into necessary key features and non-essential key features, match the extracted necessary key features first, if the match is successful, it is the target vehicle, otherwise proceed to S23; because the necessary key features may not be extracted.

[0025] The essential key features of S22 include vehicle landmarks and vehicle equipment configuration; landmark positioning is to look for specific marks on the vehicle body, such as the word "AMBULANCE" or the fire department logo. These unique symbols can greatly increase the success rate of recognition; equipment configuration is to observe whether there are special equipment on the top of the vehicle, such as the ladder of a fire truck or the warning light device commonly found on an ambulance.

[0026] Differentiating key features into necessary key features and non-essential key features, and giving priority to matching necessary key features, can help improve the accuracy and reliability of recognition. In particular, for landmarks such as the word "AMBULANCE" or the fire department logo, as well as special equipment configurations such as the ladder on a fire truck or the warning light device commonly seen on an ambulance, their presence or absence can directly affect the final judgment result. This approach ensures that even if some non-deterministic features are not successfully extracted, as long as the core features are matched successfully, correct recognition can still be made. In addition, it allows the system to maintain high robustness in the face of partial occlusion or other adverse conditions.

[0027] S23. Match the extracted non-essential key features. Only when all the identified non-essential key features are matched successfully, it is the target vehicle. Although the non-essential key features alone are not sufficient to prove the identity of the vehicle, they can supplement and support the evidence provided by the necessary key features to a certain extent. Only when all the identified non-essential key features are matched successfully, it is finally confirmed as the target vehicle. This approach ensures that the system's decision is based on the most comprehensive information possible, thereby reducing the possibility of false alarms. At the same time, it also provides additional security for special situations that are difficult to identify with only a few significant features.

[0028] In S2, after the target vehicle is identified, the movement path of the target vehicle is predicted based on the real-time positioning of the target sound source, as follows: ; in, is the target position at the next moment; for The target position at the moment; is the optical flow velocity vector; is the time interval between two frames of images; ; ; in, is the illumination compensation factor; is the proportionality coefficient; For the The average brightness of the frame image; For the The average brightness of the frame image; It is the illumination error correction term; because the change of illumination conditions will affect the brightness value of pixels in the image, thus destroying the assumption of constant brightness in the optical flow method, resulting in inaccurate prediction; is the transpose operation; The introduction of the illumination compensation factor is to deal with the inevitable illumination change problem in the real world. Since the change of illumination conditions will directly affect the brightness value of the pixels in the image, it will destroy the assumption of constant brightness in the optical flow method, resulting in inaccurate prediction. By introducing the illumination compensation factor, the error caused by illumination change can be dynamically adjusted to ensure the robustness and accuracy of the prediction model. This not only improves the reliability of the target vehicle trajectory prediction, but also enhances the performance of the entire warning system.

[0029] In S2, the effects of occlusion, rapid motion, and large displacement are introduced to optimize the target vehicle's motion path prediction. The optimization results are as follows: in, is the target position at the next moment after optimization; The occlusion detection module is used to evaluate whether there is occlusion in the current frame and adjust the prediction results accordingly. Specifically, when occlusion is detected, the system should reduce the optical flow velocity vector Instead of relying on historical data or information provided by other sensors (such as radar); It is a dynamic weight factor. As the target speed increases, the traditional optical flow method may become insensitive. Therefore, it is necessary to adjust the weight to enhance the response to rapid changes. After introduction, it can increase with the increase of target speed, thereby enhancing the model's adaptability to rapid changes. It can be calculated based on the target's speed or acceleration to ensure that the target's position changes can be accurately captured even at high speeds. The compensation term is used to compensate for the error caused by occlusion. This compensation term can be determined based on the historical trajectory or the result of multi-sensor fusion. The occlusion compensation term is mainly used to compensate for the target position prediction error caused by occlusion. When the target vehicle is partially or completely occluded, the optical flow method may not be able to accurately estimate its speed and direction; Additional corrections to the global motion model to handle large displacements that are beyond the range of local optical flow estimates; This optimization not only enhances the system’s adaptability to occlusion, rapid movement, and large displacement, but also improves the overall prediction accuracy, enabling the safe driving warning system to provide reliable assistance under a wider range of conditions. In this way, even in the face of complex and changing road environments, the system can maintain a high level of performance and provide drivers with more accurate safety prompts.

[0030] S3. According to the movement path of the target vehicle, the driving assistance system of the self-driving vehicle collects surrounding traffic data, and uses a path planning algorithm to obtain an optimal avoidance plan based on the collected traffic data. After the avoidance plan is obtained, a second signal is generated; The surrounding traffic data in S3 includes but is not limited to traffic density, road type, and signal light status; In S3, based on the collected traffic data, the path planning algorithm is used to obtain the optimal avoidance solution, specifically: ; in, is the stage cost; is the vehicle state vector, that is ; Control input refers to the command applied to the vehicle to change its motion state. In the avoidance process, this mainly includes the steering angle and acceleration / acceleration. The steering angle determines the direction and degree of the vehicle's turn, and the acceleration / acceleration is used to adjust the vehicle's speed; is the reference state, usually the corresponding point on the desired safe trajectory. This means that, ideally, the vehicle should follow this trajectory to avoid obstacles or other potential threats. Therefore, Contains the same types of elements as above, but they represent target values ​​on the planned safe path; is a diagonal matrix, and the elements on the diagonal correspond to the importance weights of different state variables. For example, if you attach great importance to keeping the lane centered, you can give a larger weight to the position. Similarly, if you want to strictly control the speed, you should give a higher weight to the speed. is the transpose of the control input; The diagonal matrix represents the cost of control input, which reflects the cost of steering angle change and acceleration adjustment. This means that larger control actions are preferred to respond quickly to environmental changes, while larger They tend to adopt a gentler control strategy to reduce unnecessary mechanical stress; ; in, is the terminal cost; The terminal state refers to the state of the vehicle at the end of the prediction time domain. It should be as close as possible to a predefined safe end point, which may be a new location far away from the danger zone, or a stable point after returning to the normal driving state. is the desired final state, usually set to the target position the vehicle should reach after completing the avoidance maneuver. It can contain the same components as the state vector mentioned above, but these values ​​now represent the desired ideal result after the avoidance maneuver is completed; In order to adjust the weighting matrix of the influence of the terminal state error, by selecting the appropriate We can emphasize the importance of certain aspects, such as ensuring that the vehicle ends up in the right lane or maintaining an appropriate speed. If you are particularly concerned about the safe return of the vehicle, you can increase the weights related to position; on the contrary, if you are more concerned about a smooth transition, you can increase the weights related to speed and heading angle; is the length of the prediction horizon, that is, the number of time steps or sampling points considered in model predictive control. Specifically, Defines the time range from the current moment to a future moment, during which the controller will calculate the optimal control sequence based on the predicted state and control input; The selection of needs to be weighed according to the specific application scenario, which should ensure sufficient foresight to cope with possible changes, while ensuring computational efficiency and real-time requirements. It can help achieve quick and safe response, while ensuring that the vehicle can smoothly return to the normal driving path after completing the avoidance; Building the cost function: ; in, is the cost function; ; The benefit of this cost function in obtaining the optimal avoidance solution is that it not only takes into account the state tracking error and control cost at each moment in the entire prediction time domain, but also emphasizes the importance of the final state, thereby ensuring that the vehicle can quickly and effectively avoid obstacles during the entire avoidance process and smoothly return to the normal driving trajectory. By minimizing this cost function, we can find a safe and efficient avoidance strategy that allows the vehicle to make the best response in complex and changing road environments while maintaining the comfort and safety of the driving experience.

[0031] S4. After receiving the second signal, the driving recorder first checks the current state of the self-driving vehicle. If the self-driving vehicle is in a safe state, the driver is immediately warned and given an avoidance instruction by voice. Otherwise, the avoidance plan is adjusted. In S4, the current state of the self-driving vehicle includes checking whether the self-driving vehicle is stationary and whether the self-driving vehicle has enough space to operate. If the vehicle is stationary and the self-driving vehicle has enough space to operate, it is in a safe state; When the vehicle is at a complete stop, the avoidance procedure can be started more safely, because under static conditions, the vehicle will not produce unpredictable behavior due to sudden movements, reducing the risk caused by inertia or other physical effects; sufficient available space means that the vehicle can complete the predetermined avoidance action without colliding with surrounding objects, which is crucial to ensure the safety of the avoidance process, especially when a quick response is required in an emergency; by confirming the operating space in advance, the system can flexibly adjust the avoidance path according to the actual situation and select the optimal solution. For example, in narrow alleys, more cautious small turns may be required; on open roads, a faster way to avoid obstacles can be chosen.

[0032] In S4, the avoidance plan is adjusted as follows: the driver is advised to use the horn and high beam lights to alert the vehicle in front to avoid the vehicle in front, turn on the double flash warning lights to attract the attention of the vehicle behind, and remind the driver not to accelerate or swerve suddenly. The detailed operations are as follows: The first is that there is not enough space to perform the operation: Using the roadside or sidewalk: You can temporarily use the roadside or sidewalk to avoid traffic without affecting the safety of pedestrians. It should be noted that this must be done with caution to avoid inconvenience or danger to pedestrians; Entering the emergency lane: If there is an emergency lane available, you can enter it while ensuring safety, but you should return to the normal lane in time to continue driving, because the Road Traffic Safety Law stipulates that the emergency lane shall not be occupied in non-emergency situations; The second is that the vehicle is in a non-stationary state: Slow down immediately: Once you hear the alarm or see the flashing warning light, you should reduce your speed quickly but steadily. Avoid sudden braking to prevent the vehicle behind you from having a rear-end collision due to the vehicle not being able to react in time; Find a suitable parking location: Try to find a location that does not hinder the passage of other vehicles or the rapid passage of special vehicles. For example, you can park on a multi-lane road; if it is on a one-way street, you need to find a safe place to park as soon as possible; Pay attention to the situation in front, behind, left and right: When slowing down, keep paying attention to the surrounding traffic conditions, including the behavior of pedestrians, non-motor vehicles and other motor vehicles, to ensure that you will not cause new dangers by changing lanes or stopping suddenly; Use turn signals to indicate your direction of movement: Turning on your turn signals in advance will inform other road users of your direction of movement, which will help them make corresponding adjustments and reduce the possibility of accidents; Create space for passing: If the road is congested and conditions permit, you can try to turn sideways at a 45-degree angle to the left or right to make room in the middle for special vehicles. However, this approach requires the driver to have certain driving skills and ensure that there will be no conflict with other vehicles.

[0033] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments, and the above embodiments and descriptions are only preferred examples of the present invention, and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements all fall within the scope of the present invention to be protected.

Claims

1. A safe driving warning method based on a driving recorder, characterized in that: The steps include: S1. The driving recorder has a built-in sound recognition processor, which is used to monitor the sound source signal in the surrounding environment. When the volume of the sound source signal is greater than the preset volume threshold, the monitored sound source signal is compared with the target audio sample in the database. Once the target sound source is identified, the target sound source is located based on the hyperbolic intersection method of TDOA, and a first signal is generated; wherein the first signal carries the location of the target sound source; S2. After receiving the first signal, the driving recorder starts the camera to record the video, and uses the computer vision algorithm to analyze the image in real time, extract key features and identify based on hybrid decision-making. After identifying the target vehicle, the movement path of the target vehicle is predicted based on the positioning of the real-time target sound source. In the process of predicting the movement path of the target vehicle, the influence of occlusion, rapid movement and large displacement is introduced for optimization; S3. According to the movement path of the target vehicle, the driving assistance system of the self-driving vehicle collects surrounding traffic data, and uses a path planning algorithm to obtain an optimal avoidance plan based on the collected traffic data. After the avoidance plan is obtained, a second signal is generated; S4. After receiving the second signal, the driving recorder first checks the current state of the self-driving vehicle. If the self-driving vehicle is in a safe state, the driver is immediately warned and given an avoidance instruction through voice. Otherwise, the avoidance plan is adjusted.

2. The safe driving warning method based on a driving recorder according to claim 1, characterized in that: In S1, the target sound source is located based on the hyperbolic intersection method of TDOA, as follows: It is known that a driving recorder includes at least three microphones arranged in a collinear manner: , , , and the positions of the three microphones and the time delay difference between them are known: , , ,but: S11. Calculate the delay difference between microphones : ； in, is the generalized cross-correlation function; is the fast Fourier transform; is to represent the inverse fast Fourier transform; For Moment Microphone Time series signal of For Moment Microphone Time series signal of is the delay variable; is a small positive number; is the complex conjugate fast Fourier transform; ； in, For microphone and microphone The delay difference of the signal between them; S12, based on the delay difference Create a hyperbola between the three microphones: ； ； ； in, is the speed of sound propagation; is the time delay difference between the first microphone and the second microphone; is the time delay difference between the first microphone and the third microphone; is the time delay difference between the second microphone and the third microphone; is the position vector of the first microphone; is the position vector of the second microphone; is the position vector of the third microphone; is the position vector of the target sound source; The speed of sound after the influence of environmental factors is introduced; is temperature; is the relative humidity; is the atmospheric pressure; S13, using weighted least squares method for optimization, the optimization results are as follows: ； in, For the The difference between the actual distance from the microphone to the sound source and the predicted distance; For the The position vectors of the microphones; For the The position vectors of the microphones; For the Microphone and The delay difference between the microphones; ; ; in, Minimize the weighted sum of squares of the position vector of the target sound source; It is Microphone and The weight of the delay difference between the microphones; is the environmental feature vector; is the weight predicted after introducing the environmental feature vector; S14. Finally, by adjusting Make When the minimum value is reached, the optimal target sound source coordinates can be obtained.

3. The safe driving warning method based on a driving recorder according to claim 2 is characterized in that: In S2, key features are extracted and identified based on hybrid decision making, specifically: S21, using the color model to check whether the color of the vehicle conforms to the color spectrum range of the target vehicle in the known database, if it conforms, proceed to S22, otherwise, exclude the possibility that the vehicle is the target vehicle; S22, if it is in compliance, extract key features of the vehicle in the image, divide the key features into necessary key features and non-essential key features, first match the extracted necessary key features, if the match is successful, it is the target vehicle, otherwise proceed to S23; S23, matching the extracted non-essential key features, and only when all the identified non-essential key features are matched successfully, it is the target vehicle.

4. The safe driving warning method based on a driving recorder according to claim 3 is characterized in that: The necessary key features in S22 include vehicle markers and vehicle equipment configurations.

5. The safe driving warning method based on a driving recorder according to claim 4 is characterized in that: In S2, after the target vehicle is identified, the movement path of the target vehicle is predicted based on the real-time positioning of the target sound source, as follows: ； in, is the target position at the next moment; for The target position at the moment; is the optical flow velocity vector; is the time interval between two frames of images; ； ； in, is the illumination compensation factor; is the proportionality coefficient; For the The average brightness of the frame image; For the The average brightness of the frame image; is the illumination error correction term; is the transpose operation.

6. The safe driving warning method based on a driving recorder according to claim 5, characterized in that: In S2, in the process of predicting the moving path of the target vehicle, the influence of occlusion, rapid movement and large displacement is introduced for optimization, and the optimization is as follows: in, is the target position at the next moment after optimization; It is the occlusion detection module; is the dynamic weight factor; For compensation items; Additional corrections provided for the global motion model.

7. The safe driving warning method based on a driving recorder according to claim 6, characterized in that: The surrounding traffic data in S3 includes but is not limited to traffic density, road type, and traffic light status.

8. The safe driving warning method based on a driving recorder according to claim 7, characterized in that: In S3, based on the collected traffic data, a path planning algorithm is used to obtain the optimal avoidance solution, specifically: ； in, is the stage cost; is the vehicle state vector, that is ; is the control input; is the reference state; is a diagonal matrix; is the transpose of the control input; is a diagonal matrix representing the cost of control input; ； in, is the terminal cost; is a terminal state; is the ideal final state; A weighting matrix for adjusting the influence of the terminal state error; is the length of the prediction time domain; Building the cost function: ； in, is the cost function; .

9. The safe driving warning method based on a driving recorder according to claim 8, characterized in that: In S4, the current state of the self-driving vehicle includes checking whether the self-driving vehicle is stationary and whether the self-driving vehicle has enough space to operate. If the vehicle is stationary and the self-driving vehicle has enough space to operate, it is in a safe state.

10. The safe driving warning method based on a driving recorder according to claim 9, characterized in that: In S4, adjusting the avoidance plan specifically includes: suggesting the driver by voice to use the horn and high beam to remind the vehicle in front to avoid, turning on the double flash warning lights to attract the attention of the vehicle behind, and reminding the driver by voice not to accelerate suddenly or turn the steering wheel sharply.

Images