Control method and system for vehicle abnormal behavior AI identification and early warning

Abstract

The embodiment of the invention relates to the technical field of traffic safety management, and discloses a vehicle abnormal behavior AI identification and early warning control method and system. The method comprises the following steps: firstly, acquiring video data (first target video data) of a vehicle entering a preset monitoring area, and inputting the video data into a cloud abnormal behavior recognition model to obtain a first abnormal degree value; if the value reaches or exceeds a first preset value, a temporary tactile or auditory early warning signal (first early warning signal) is triggered. And after a preset duration, re-acquiring video data (second target video data), and inputting the video data into the local abnormal behavior recognition model to obtain a second abnormal degree value. And if the value also reaches or exceeds the second preset value, triggering a continuous tactile or auditory early warning signal (second early warning signal). According to the technical scheme, the false alarm rate and response delay of a traditional vehicle early warning system are effectively reduced through cloud and local double-model collaborative decision making and graded early warning, and the accuracy of drunk driving or toxic driving abnormal driving behavior recognition is greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic safety management, and in particular to a control method and system for AI recognition and early warning of abnormal vehicle behavior. Background Art

[0002] With the continuous increase in the number of cars, road traffic safety issues are becoming increasingly prominent. Abnormal vehicle behavior, such as driver fatigue, drunk driving or drug driving, is an important cause of traffic accidents, seriously threatening people's lives and property safety.

[0003] Traditional vehicle safety monitoring mainly relies on manual monitoring or fixed sensor equipment, but these methods have many limitations in terms of efficiency, accuracy and real-time performance. Especially in complex and changing traffic environments, a single manual or fixed sensor often cannot efficiently and accurately capture abnormal vehicle behavior, and thus cannot issue early warning signals in time to effectively prevent traffic accidents. Summary of the invention

[0004] The main purpose of the present invention is to provide a control method and system for AI recognition and early warning of abnormal vehicle behavior, aiming to solve the technical problem of low efficiency and accuracy in the prior art of abnormal vehicle behavior recognition.

[0005] To achieve the above-mentioned purpose, in a first aspect, an embodiment of the present application provides a control method for AI recognition and early warning of abnormal vehicle behavior, which is applied to a vehicle early warning system, wherein the vehicle early warning system includes a camera device and a server connected to the camera device in communication, and the method includes: Acquire video data of a target vehicle entering a preset monitoring area to obtain first target video data; Inputting the first target video data into a pre-trained cloud-based abnormal behavior recognition model to obtain a first abnormality level value of the target vehicle; When the first abnormality level value is greater than or equal to a first preset level value, a first warning signal is triggered, wherein the first warning signal is used to provide a temporary tactile or auditory wake-up warning to the driver; After the first warning signal is triggered for a preset time, reacquiring the video data of the target vehicle to obtain second target video data; Inputting the second target video data into a pre-trained local abnormal behavior recognition model to obtain a second abnormality degree value of the target vehicle; When the second abnormality degree value is greater than or equal to a second preset degree value, a second warning signal is triggered, and the second warning signal is used to provide a continuous tactile or auditory wake-up warning to the driver, wherein the second preset degree value is greater than or equal to the first preset degree value.

[0006] In a possible implementation, reacquiring video data of the target vehicle includes: Determining the driving area of ​​the target vehicle according to the driving speed and driving direction of the vehicle; Determining a target camera device according to a driving area of ​​the target vehicle; The video data of the target vehicle is reacquired from the target camera device.

[0007] In a possible implementation, determining the driving area of ​​the target vehicle according to the driving speed and driving direction of the vehicle includes: Determine the virtual position range of the target vehicle according to the vehicle's driving speed and driving direction; The driving area of ​​the target vehicle is determined according to the virtual position range of the target vehicle.

[0008] In a possible implementation, determining a target camera device according to a driving area of ​​the target vehicle includes: Determine a plurality of candidate camera devices according to the driving area of ​​the target vehicle, wherein the distances between the plurality of candidate camera devices and the driving area of ​​the target vehicle are less than a preset distance value; Acquire the shooting angles of view FOV of the multiple candidate camera devices; A candidate camera device whose driving area of ​​the target vehicle is within the shooting angle of view FOV is selected as the target camera device.

[0009] In a possible implementation, the abnormal behavior recognition model is a vehicle trajectory recognition model, and the step of inputting the first target video data into a pre-trained cloud-based abnormal behavior recognition model to obtain a first abnormality degree value of the target vehicle includes: Determining the running track data of the target vehicle according to the first target video data; The running track data of the target vehicle is input into a pre-trained vehicle track recognition model to obtain a first abnormality degree value of the target vehicle.

[0010] In a possible implementation, the abnormal behavior recognition model is a vehicle trajectory thermal recognition model, and the step of inputting the first target video data into a pre-trained cloud-based abnormal behavior recognition model to obtain a first abnormality degree value of the target vehicle includes: Determine the running track thermal data of the target vehicle according to the first target video data, wherein the running track thermal data includes track data and lane change frequency data; The running track thermal data of the target vehicle is input into a pre-trained vehicle track thermal recognition model to obtain a first abnormality degree value of the target vehicle.

[0011] In a possible implementation, the abnormal behavior recognition model is a vehicle-associated speed recognition model, and the step of inputting the first target video data into a pre-trained cloud-based abnormal behavior recognition model to obtain a first abnormality degree value of the target vehicle includes: Determine the vehicle speed data of the target vehicle and the vehicle speed data of other surrounding vehicles according to the first target video data; The vehicle speed data of the target vehicle and the vehicle speed data of other surrounding vehicles are input into a pre-trained vehicle-associated speed recognition model to obtain a first abnormality degree value of the target vehicle.

[0012] In a possible implementation manner, before triggering the second warning signal, the method further includes: Acquire a historical behavior characteristic model of the target vehicle; Determining a historical abnormality value of the target vehicle according to the historical behavior characteristic model of the target vehicle; When the deviation between the second abnormality degree value and the historical abnormality degree value is greater than a preset value, the second warning signal is triggered.

[0013] In a possible implementation, inputting the first target video data into a pre-trained cloud-based abnormal behavior recognition model to obtain a first abnormality level value of the target vehicle includes: Acquiring a network status of a camera device, wherein the camera device is used to capture a driving status video of a target vehicle to obtain the first target video data; In the case where the network status of the camera device is abnormal, the first target video data is input into the edge server to perform abnormal behavior recognition to obtain a first abnormality degree value of the target vehicle; When the network status of the camera device is normal, the first target video data is input into a cloud server for abnormal behavior recognition to obtain a first abnormality degree value of the target vehicle, and the cloud server periodically updates the abnormal behavior recognition model of the edge server.

[0014] In a possible implementation, obtaining the network status of the camera device includes: Obtain signal strength parameters, bandwidth parameters, and delay parameters of the camera device network link; Inputting the signal strength parameter, bandwidth parameter and delay parameter of the network link of the camera device into the network status evaluation model to obtain a comprehensive network status score value, wherein the delay parameter is obtained by sending a preset test data packet to the server in advance and recording the round-trip time; If the network status comprehensive score value is determined to be less than a preset threshold, the camera device is determined to be in an abnormal network state; if the network status comprehensive score value is determined to be greater than or equal to a preset threshold, the camera device is determined to be in a normal network state.

[0015] In a possible implementation manner, after the second warning signal is triggered, the method further includes: Performing time series analysis on the second target video data to establish a time dynamic model of the target vehicle behavior; The behavior trend of the target vehicle in the future is predicted according to the time dynamic model. If the prediction result shows an increase in abnormal risk, the warning strategy is adjusted in advance to trigger a third warning signal, and the third warning signal is used to provide emergency braking warning for the target vehicle.

[0016] In a second aspect, an embodiment of the present application further provides a vehicle warning system, comprising: a memory and a processor, wherein the memory is used to store program code; the processor is used to call the program code to execute the method described in the first aspect.

[0017] Different from the prior art, the control method for AI recognition and early warning of abnormal vehicle behavior provided by the embodiment of the present application first obtains the video data of the target vehicle entering the preset monitoring area to obtain the first target video data; then the first target video data is input into the pre-trained cloud abnormal behavior recognition model to obtain the first abnormal degree value of the target vehicle; when the first abnormal degree value is greater than or equal to the first preset degree value, the first warning signal is triggered, and the first warning signal is used to temporarily tactilely or auditorily wake up the driver for early warning; after the first warning signal triggers the preset duration, the video data of the target vehicle is acquired again to obtain the second target video data; then the second target video data is input into the pre-trained local abnormal behavior recognition model to obtain the second abnormal degree value of the target vehicle; when the second abnormal degree value is greater than or equal to the second preset degree value, the second warning signal is triggered, and the second warning signal is used to continuously tactilely or auditorily wake up the driver for early warning. In this way, the technical solution of the present application solves the problems of high false alarm rate and response delay of the traditional vehicle early warning system through dual-model collaborative decision-making and hierarchical early warning mechanism, and greatly improves the accuracy of abnormal driving behavior recognition of drunk driving or drug driving. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0019] Figure 1 This is a schematic diagram of an application scenario of AI recognition and early warning of abnormal vehicle behavior in some embodiments of the present application; Figure 2 This is a flow chart of a control method for AI recognition and early warning of abnormal vehicle behavior in some embodiments of the present application; Figure 3 This is a flow chart of step S200 of a control method for AI recognition and early warning of abnormal vehicle behavior in some embodiments of the present application; Figure 4 This is a schematic diagram of the hardware structure of a vehicle warning system in some embodiments of the present application.

[0020] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0021] 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.

[0022] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0023] In addition, the descriptions of "first", "second", etc. in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, "and / or" in the full text includes three solutions. Taking A and / or B as an example, it includes technical solution A, technical solution B, and technical solution that satisfies both A and B. In addition, the technical solutions between the various embodiments can be combined with each other, but it must be based on the ability of ordinary technicians in this field to implement. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0024] With the continuous increase in the number of cars, road traffic safety issues are becoming increasingly prominent. Abnormal vehicle behavior, such as driver fatigue, drunk driving or drug driving, is an important cause of traffic accidents, seriously threatening people's lives and property safety.

[0025] Traditional vehicle safety monitoring mainly relies on manual monitoring or fixed sensor equipment, but these methods have many limitations in terms of efficiency, accuracy and real-time performance. Especially in complex and changing traffic environments, a single manual or fixed sensor often cannot efficiently capture abnormal vehicle behavior, and thus cannot issue early warning signals in time to effectively prevent traffic accidents.

[0026] like Figure 1-Figure 3 As shown, the following takes the early warning system executing the control method for vehicle abnormal behavior AI identification and early warning as an example for explanation. It should be noted that although the logical order is shown in the flow chart, in some cases, the steps shown or described can be performed in a different order than here. Please refer to the attached Figure 2 The method includes the following steps S100 to S600: Step S100, acquiring video data of a target vehicle entering a preset monitoring area to obtain first target video data; Specifically, a camera may be set in a preset monitoring area to capture video data of a vehicle. The target vehicle may refer to all vehicles entering the monitoring area, or may be a specific vehicle.

[0027] Step S200: input the first target video data into a pre-trained cloud-based abnormal behavior recognition model to obtain a first abnormality level value of the target vehicle; In the embodiment of the present application, the vehicle video data is collected in real time by the camera, and the cloud AI model (high computing power, complex algorithm) performs preliminary abnormality analysis to generate a first abnormality degree value. In this way, the initial screening can accurately screen out abnormal target vehicles using the high computing power of the cloud server and obtain the first abnormality degree value of the target vehicle. The first abnormality degree value refers to the abnormality degree value of the target vehicle determined by the cloud server, indicating the deviation or abnormality degree of the target vehicle relative to a normal vehicle.

[0028] To improve the efficiency and flexibility of vehicle abnormality recognition, in one embodiment, the step S200: inputting the first target video data into a pre-trained cloud abnormal behavior recognition model to obtain a first abnormality degree value of the target vehicle includes: Step S210: acquiring a network status of a camera device, wherein the camera device is used to capture a driving status video of a target vehicle to obtain the first target video data; Step S220: When the network status of the camera device is abnormal, the first target video data is input into the edge server to perform abnormal behavior recognition to obtain a first abnormality degree value of the target vehicle; Step S230: When the network status of the camera device is normal, the first target video data is input into a cloud server for abnormal behavior recognition to obtain a first abnormality level value of the target vehicle, and the cloud server periodically updates the abnormal behavior recognition model of the edge server.

[0029] Specifically, when the connection status between the camera device and the cloud server is normal, the embodiment of the present application preferentially uses the cloud server with high computing power for preliminary identification and judgment. If the connection status between the camera device and the cloud server is abnormal, the first target video data is input into the edge server for abnormal behavior identification to obtain the first abnormality degree value of the target vehicle. And the cloud server regularly updates the abnormal behavior identification model of the edge server. In this way, the interruption of vehicle abnormal identification caused by cloud server abnormality or network abnormality can be avoided, thereby improving the efficiency, accuracy and flexibility of vehicle abnormal identification.

[0030] Among them, to obtain the network status of the camera device, the signal strength parameters, bandwidth parameters and delay parameters of the network link of the camera device can be first obtained; then the signal strength parameters, bandwidth parameters and delay parameters of the network link of the camera device are input into the network status evaluation model to obtain a comprehensive network status score value, and the delay parameter is obtained by sending a preset test data packet to the server in advance and recording the round-trip time; when the comprehensive network status score value is less than a preset threshold value, it is determined that the camera device is in an abnormal network state, and when the comprehensive network status score value is greater than or equal to the preset threshold value, it is determined that the camera device is in a normal network state.

[0031] It should be noted that the abnormal behavior recognition model can be a vehicle trajectory recognition model, a lane change frequency recognition model, a speed recognition model, etc., or it can be an end-to-end recognition model that integrates vehicle trajectory, lane change frequency and speed.

[0032] In one embodiment, the abnormal behavior recognition model is a vehicle trajectory recognition model, and the step S200: inputting the first target video data into a pre-trained cloud-based abnormal behavior recognition model to obtain a first abnormality degree value of the target vehicle includes: determining the running trajectory data of the target vehicle according to the first target video data; inputting the running trajectory data of the target vehicle into the pre-trained vehicle trajectory recognition model to obtain the first abnormality degree value of the target vehicle.

[0033] Specifically, the first target video data is processed using a video processing algorithm (such as target detection and tracking algorithm) to identify the target vehicle. Then, based on the identified target vehicle, its position changes are tracked in continuous video frames to obtain the running trajectory data of the target vehicle. The extracted running trajectory data of the target vehicle is then input into a pre-trained vehicle trajectory recognition model. The model classifies and judges the trajectory of the target vehicle based on the input trajectory data and the normal and abnormal trajectory features learned internally. Finally, the first abnormality degree value of the target vehicle is output, which reflects the degree of deviation or abnormality of the target vehicle trajectory from the normal trajectory. In this way, by extracting the running trajectory data of the target vehicle and analyzing it using a special vehicle trajectory recognition model, it is possible to more accurately determine whether the target vehicle has abnormal behavior and the degree of abnormality.

[0034] In another embodiment, the abnormal behavior recognition model is a vehicle trajectory thermal recognition model, and the step S200: inputting the first target video data into a pre-trained cloud-based abnormal behavior recognition model to obtain a first abnormality degree value of the target vehicle, includes: determining the running trajectory thermal data of the target vehicle according to the first target video data, the running trajectory thermal data including trajectory data and lane change frequency data; inputting the running trajectory thermal data of the target vehicle into a pre-trained vehicle trajectory thermal recognition model to obtain a first abnormality degree value of the target vehicle.

[0035] Specifically, the embodiment of the present application first uses a video processing algorithm (such as target detection and tracking algorithm) to process the first target video data and identify the target vehicle. Then, in continuous video frames, the position change of the target vehicle is continuously tracked to accurately construct the running trajectory of the target vehicle. On this basis, the driving behavior of the target vehicle, such as lane change behavior, is further analyzed, and the lane change frequency data is counted. The trajectory data and the lane change frequency data are integrated to generate more detailed target vehicle running trajectory thermal data. Then the target vehicle running trajectory thermal data is input into the fully trained vehicle trajectory thermal recognition model. The model will comprehensively evaluate the target vehicle's driving trajectory based on the input trajectory thermal data and the normal and abnormal driving behavior characteristics learned internally. The model will eventually output a first abnormality degree value, which intuitively reflects the degree of deviation or abnormality of the target vehicle's driving behavior from normal behavior. In this way, by integrating multivariate information such as trajectory data and lane change frequency data, the vehicle trajectory thermal recognition model can more comprehensively capture the driving characteristics of the target vehicle, thereby significantly improving the recognition accuracy of abnormal behavior.

[0036] In other embodiments, the abnormal behavior recognition model is a vehicle-associated speed recognition model, and the step S200: inputting the first target video data into a pre-trained cloud-based abnormal behavior recognition model to obtain a first abnormality degree value of the target vehicle, including: determining the vehicle driving speed data of the target vehicle and the vehicle driving speed data of other surrounding vehicles based on the first target video data; inputting the vehicle driving speed data of the target vehicle and the vehicle driving speed data of other surrounding vehicles into the pre-trained vehicle-associated speed recognition model to obtain the first abnormality degree value of the target vehicle.

[0037] Specifically, the first target video data is first processed using video processing techniques and algorithms (such as target detection, tracking algorithms, and speed estimation algorithms). In continuous video frames, the position changes of the target vehicle and its surrounding vehicles are continuously tracked, and the speed of each vehicle is estimated using the frame difference and timestamp information. Then, the speed data of the target vehicle and its surrounding vehicles are integrated to form a complete vehicle-associated speed data set. The vehicle-associated speed data set is then input into a fully trained vehicle-associated speed recognition model. The model will comprehensively evaluate the driving behavior of the target vehicle based on the input associated speed data, combined with the normal and abnormal driving speed characteristics learned internally and the speed correlation between vehicles. The model will eventually output a first abnormality degree value, which intuitively reflects the degree of deviation or abnormality between the target vehicle's driving speed and the driving speed of the surrounding vehicles. In this way, by comprehensively considering the driving speed data of the target vehicle and its surrounding vehicles, the vehicle-associated speed recognition model can more accurately capture the abnormal driving behavior of the target vehicle.

[0038] Step S300: When the first abnormality level value is greater than or equal to a first preset level value, triggering a first warning signal, wherein the first warning signal is used to provide a temporary tactile or auditory wake-up warning to the driver; It can be understood that the first abnormality degree value represents the degree of deviation or abnormality of the target vehicle relative to the normal state. If the degree of deviation or abnormality is small, it means that the driving state of the vehicle is relatively normal; if the degree of deviation or abnormality is large, it means that the driving state of the vehicle is relatively abnormal, which may be due to driver fatigue driving, drunk driving or drug driving, etc.

[0039] Based on this, if it is identified in step S200 that the first abnormality degree value is greater than or equal to the first preset degree value, it means that the driving state of the vehicle is relatively abnormal, and the first warning signal is triggered at this time. For example, the seat is controlled to vibrate temporarily or a voice reminder is issued to provide a temporary tactile or auditory wake-up warning to the driver. If the driver is fatigued at this time, in this warning state, the temporary tactile or auditory wake-up warning can usually wake the driver up to restore the vehicle to normal. However, for abnormal conditions caused by serious violations such as drunk driving or drug driving, the warning signal may not be able to wake the driver up or restore his sanity. In this case, other emergency measures need to be taken to ensure road safety, such as automatic deceleration, parking, or requesting external assistance.

[0040] Step S400: after the first warning signal triggers a preset time, reacquire the video data of the target vehicle to obtain second target video data; In the embodiment of the present application, the video data of the target vehicle is reacquired after the first warning signal triggers a preset time, and the second target video data is obtained to detect whether the state of the vehicle has returned to normal after the temporary wake-up. The preset time can be adjusted according to the actual situation to ensure that the driver has enough time to respond to the warning signal and restore the vehicle state to normal.

[0041] In one embodiment, reacquiring video data of a target vehicle includes: determining a driving area of ​​the target vehicle according to a driving speed and a driving direction of the vehicle; determining a target camera device according to the driving area of ​​the target vehicle; and reacquiring video data of the target vehicle from the target camera device.

[0042] Specifically, the real-time driving data of the target vehicle, including driving speed and driving direction, can be collected first. Then, the virtual position range of the target vehicle is calculated using the map data and vehicle driving data (the virtual position range refers to the possible current position of the vehicle, such as Figure 1 This can be achieved by inputting the vehicle's speed and direction into a map matching algorithm that predicts the vehicle's likely location over a period of time in the future. The target vehicle's driving area is then determined based on the virtual location range. For example, the virtual location range is matched with a predefined area on the map to find the area that best matches (e.g., the center of the virtual location range is used as the target vehicle's driving area).

[0043] After determining the driving area of ​​the target vehicle, multiple candidate camera devices can be determined based on the driving area of ​​the target vehicle. These candidate camera devices should be located near the driving area and the distance should be less than a preset distance value. The preset distance value can be determined based on factors such as the coverage range, resolution, and driving speed of the camera device. Then, the shooting angle of view FOV (Field of View) of each candidate camera device is obtained. Finally, the candidate camera device whose driving area of ​​the target vehicle is within the shooting angle of view FOV is selected as the target camera device (such as Figure 1 The camera device P1 in the image processing apparatus.

[0044] After the target camera device is determined, the video data of the target vehicle is retrieved from the selected target camera device. For example, a real-time video stream or a recorded video file can be obtained by sending a request to the camera device. The acquired video data is preprocessed, including denoising, contrast enhancement, resolution adjustment, etc., to improve the video quality. Then, the AI ​​algorithm is used to identify abnormal behavior on the preprocessed video data.

[0045] In this way, by accurately acquiring the target camera device and sending a targeted request to the target camera device to obtain the real-time video stream or record the video file, the efficiency of data processing is greatly improved, thereby further improving the efficiency of abnormal behavior identification.

[0046] Step S500: input the second target video data into a pre-trained local abnormal behavior recognition model to obtain a second abnormality degree value of the target vehicle; After the cloud server initially determines that the target vehicle has a high degree of abnormality and triggers a temporary warning, the technical solution of the present application inputs the second target video data into a pre-trained local abnormal behavior recognition model to obtain the second abnormality value of the target vehicle in order to quickly determine whether the target vehicle has returned to a normal state so as to take follow-up measures in a timely manner. Since the model is deployed locally, real-time processing and analysis of the input data can be achieved, greatly improving the judgment speed.

[0047] In this way, through collaborative decision-making between the cloud and local dual models, both the accuracy and timeliness of abnormal behavior identification can be taken into account.

[0048] Step S600: When the second abnormality degree value is greater than or equal to a second preset degree value, trigger a second warning signal, wherein the second warning signal is used to provide a continuous tactile or auditory wake-up warning to the driver, wherein the second preset degree value is greater than or equal to the first preset degree value.

[0049] It can be understood that when the second abnormality value of the target vehicle is obtained through step S500, and it is identified that the second abnormality value is greater than or equal to the second preset value, it means that the target vehicle has not yet returned to a normal state. At this time, it can be judged that the driver cannot be simply awakened, and there is a high probability that he is driving under the influence of alcohol or drugs. Therefore, in order to improve the driving safety of the vehicle, the second warning signal is triggered. For example, the seat is controlled to vibrate continuously or a voice prompt warning is continuously issued to continuously wake up the driver by touch or hearing.

[0050] Based on this, a control method for AI recognition and early warning of abnormal vehicle behavior provided by an embodiment of the present application first obtains video data of a target vehicle entering a preset monitoring area to obtain first target video data; then the first target video data is input into a pre-trained cloud abnormal behavior recognition model to obtain a first abnormal degree value of the target vehicle; when the first abnormal degree value is greater than or equal to the first preset degree value, a first warning signal is triggered, and the first warning signal is used to temporarily tactilely or auditorily wake up the driver for early warning; after the first warning signal triggers a preset duration, the video data of the target vehicle is acquired again to obtain the second target video data; then the second target video data is input into a pre-trained local abnormal behavior recognition model to obtain a second abnormal degree value of the target vehicle; when the second abnormal degree value is greater than or equal to the second preset degree value, a second warning signal is triggered, and the second warning signal is used to continuously tactilely or auditorily wake up the driver for early warning. In this way, the technical solution of the present application can quickly and accurately identify the abnormal behavior of drunk driving or drug driving of a vehicle through a dual-model collaborative decision-making mechanism, and make timely warnings, greatly improving the safety of road traffic.

[0051] In one embodiment, before triggering the second warning signal, it also includes: acquiring a historical behavior characteristic model of the target vehicle; determining a historical abnormality degree value of the target vehicle based on the historical behavior characteristic model of the target vehicle; and triggering the second warning signal when the deviation between the second abnormality degree value and the historical abnormality degree value is greater than a preset value.

[0052] Specifically, by collecting and analyzing the behavior data of the target vehicle in the past period of time (such as driving speed, acceleration, steering angle, braking frequency, etc.), a historical behavior feature model can be established using machine learning or statistical methods. Using the established historical behavior feature model, the target vehicle's past behavior data is evaluated to determine its historical abnormality value. Then the severity of the current abnormal behavior and whether it deviates from the normal behavior pattern of the target vehicle are judged and the current abnormality value is obtained. Then the deviation between the second abnormality value (i.e., the current abnormality value) and the historical abnormality value is calculated. If the deviation is greater than the preset value, it means that there is a significant deviation between the current abnormal behavior and the historical normal behavior pattern, and an early warning needs to be triggered. In this way, by adding a judgment step based on the historical behavior feature model of the target vehicle before triggering the second early warning signal, the severity of the current abnormal behavior and whether an early warning needs to be triggered can be more accurately evaluated. This mechanism helps to improve the accuracy and reliability of the early warning system, thereby more effectively responding to potential safety risks.

[0053] In another embodiment, after triggering the second warning signal, it also includes: performing time series analysis on the second target video data to establish a time dynamic model of the target vehicle behavior; predicting the behavior trend of the target vehicle in a future period of time based on the time dynamic model, and if the prediction result shows an increased abnormal risk, adjusting the warning strategy in advance to trigger a third warning signal, and the third warning signal is used to provide emergency braking warning for the target vehicle.

[0054] Specifically, appropriate time series analysis techniques (such as ARIMA model, exponential smoothing method, machine learning algorithm, etc.) can be used to establish a time dynamic model based on key behavioral indicators (such as speed, acceleration, direction change, etc.) in the video data. Use the established time dynamic model to predict the behavior trend of the target vehicle in the future period of time. For example, speed change, driving trajectory, probability of abnormal behavior, etc. According to the prediction results, evaluate the risk of abnormal behavior of the target vehicle in the future time period. If the prediction results show that the abnormal risk increases, that is, the probability of abnormal behavior of the target vehicle in the future time period exceeds the preset threshold, it is determined that the risk increases. According to the risk assessment results, adjust the early warning strategy in advance to respond to potential risks more actively. When necessary, trigger the third early warning signal, which is used to warn the target vehicle of emergency braking. For example, this can be achieved by sending an emergency braking command to the vehicle and issuing a strong audio-visual alarm to the driver.

[0055] like Figure 4 As shown, Figure 4This is a schematic diagram of the hardware structure of the vehicle warning system in some embodiments of the present application. The vehicle warning system provided in the embodiments of the present application also includes a memory 1000 and a processor 2000, wherein the memory 1000 is used to store computer-readable instructions, and the processor 2000 is used to call the computer-readable instructions to execute the control method for AI recognition and warning of abnormal vehicle behavior as described above.

[0056] Among them, the processor 2000 is used to provide computing and control capabilities to control the vehicle warning system to perform corresponding tasks, for example, to control the vehicle warning system to execute the control method for AI recognition and warning of abnormal vehicle behavior in any of the above method embodiments, the method comprising: acquiring video data of a target vehicle entering a preset monitoring area to obtain first target video data; inputting the first target video data into a pre-trained cloud-based abnormal behavior recognition model to obtain a first abnormality degree value of the target vehicle; when the first abnormality degree value is greater than or equal to the first preset degree value, triggering a first warning signal, the first warning signal is used to perform a temporary tactile or auditory wake-up warning to the driver; after the first warning signal is triggered for a preset time, re-acquiring the video data of the target vehicle to obtain a second target video data; inputting the second target video data into a pre-trained local abnormal behavior recognition model to obtain a second abnormality degree value of the target vehicle; when the second abnormality degree value is greater than or equal to the second preset degree value, triggering a second warning signal, the second warning signal is used to perform a continuous tactile or auditory wake-up warning to the driver, wherein the second preset degree value is greater than or equal to the first preset degree value.

[0057] The processor 2000 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), a hardware chip, or any combination thereof; it may also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a programmable logic device (PLD), or any combination thereof. The above-mentioned PLD may be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof.

[0058] The memory 1000, as a non-transient computer-readable storage medium, can be used to store non-transient software programs, non-transient computer executable programs and modules, such as program instructions / modules corresponding to the control method for AI recognition and early warning of abnormal vehicle behavior in the embodiments of the present application. The processor 2000 can implement the control method for AI recognition and early warning of abnormal vehicle behavior in any of the above method embodiments by running the non-transient software programs, instructions and modules stored in the memory 1000.

[0059] Specifically, the memory 1000 may include a volatile memory (VM), such as a random access memory (RAM); the memory 1000 may also include a non-volatile memory (NVM), such as a read-only memory (ROM), a flash memory, a hard disk drive (HDD) or a solid-state drive (SSD) or other non-transitory solid-state storage device; the memory 1000 may also include a combination of the above-mentioned types of memory.

[0060] To sum up, the vehicle warning system of the present application adopts the technical solution of any one of the above-mentioned control method embodiments for AI identification and warning of abnormal vehicle behavior. Therefore, it has at least the beneficial effects brought by the technical solutions of the above-mentioned embodiments, which will not be repeated here one by one.

[0061] The present application also provides a computer-readable storage medium, such as a memory including a program code, which can be executed by a processor to complete the control method for AI recognition and early warning of abnormal vehicle behavior in the above embodiment. For example, the computer-readable storage medium can be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CDROM), a magnetic tape, a floppy disk, and an optical data storage device.

[0062] The present application also provides a computer program product, which includes one or more program codes stored in a computer-readable storage medium. The processor of the early warning system reads the program code from the computer-readable storage medium, and the processor executes the program code to complete the control method steps for AI recognition and early warning of abnormal vehicle behavior provided in the above embodiment.

[0063] A person skilled in the art will appreciate that all or part of the steps for implementing the above embodiments may be accomplished by hardware or by hardware associated with a program code, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.

[0064] It should be noted that the device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0065] Through the description of the above embodiments, it can be clearly understood by those skilled in the art that each embodiment can be implemented by means of software plus a general hardware platform, or by hardware. It can be understood by those skilled in the art that all or part of the processes in the above embodiment methods can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above methods. Among them, the storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0066] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. All equivalent structural changes made by using the contents of the present invention specification and drawings under the inventive concept of the present invention, or directly / indirectly applied in other related technical fields are included in the patent protection scope of the present invention.

Claims

1. A control method for AI recognition and early warning of abnormal vehicle behavior, characterized in that: Applied to a vehicle warning system, the vehicle warning system includes a camera device and a server connected to the camera device for communication, the method includes: Acquire video data of a target vehicle entering a preset monitoring area to obtain first target video data; Inputting the first target video data into a pre-trained cloud-based abnormal behavior recognition model to obtain a first abnormality degree value of the target vehicle; When the first abnormality level value is greater than or equal to a first preset level value, a first warning signal is triggered, wherein the first warning signal is used to provide a temporary tactile or auditory wake-up warning to the driver; After the first warning signal is triggered for a preset time, reacquiring the video data of the target vehicle to obtain second target video data; Inputting the second target video data into a pre-trained local abnormal behavior recognition model to obtain a second abnormality degree value of the target vehicle; When the second abnormality degree value is greater than or equal to a second preset degree value, a second warning signal is triggered, and the second warning signal is used to provide a continuous tactile or auditory wake-up warning to the driver, wherein the second preset degree value is greater than or equal to the first preset degree value.

2. The control method for AI recognition and early warning of abnormal vehicle behavior as claimed in claim 1, characterized in that: The reacquiring the video data of the target vehicle comprises: Determining the driving area of ​​the target vehicle according to the driving speed and driving direction of the vehicle; Determining a target camera device according to a driving area of ​​the target vehicle; The video data of the target vehicle is reacquired from the target camera device.

3. The control method for AI recognition and early warning of abnormal vehicle behavior as claimed in claim 2, characterized in that: Determining the driving area of ​​the target vehicle according to the driving speed and driving direction of the vehicle includes: Determine the virtual position range of the target vehicle according to the vehicle's driving speed and driving direction; The driving area of ​​the target vehicle is determined according to the virtual position range of the target vehicle.

4. The control method for AI recognition and early warning of abnormal vehicle behavior as claimed in claim 3, characterized in that: The step of determining a target camera device according to the driving area of ​​the target vehicle comprises: Determine a plurality of candidate camera devices according to the driving area of ​​the target vehicle, wherein the distances between the plurality of candidate camera devices and the driving area of ​​the target vehicle are less than a preset distance value; Acquire the shooting angles of view FOV of the multiple candidate camera devices; A candidate camera device whose driving area of ​​the target vehicle is within the shooting angle of view FOV is selected as the target camera device.

5. The control method for AI recognition and early warning of abnormal vehicle behavior as claimed in claim 1, characterized in that: The abnormal behavior recognition model is a vehicle trajectory recognition model, and the first target video data is input into a pre-trained cloud abnormal behavior recognition model to obtain a first abnormality degree value of the target vehicle, including: Determining the running track data of the target vehicle according to the first target video data; The running track data of the target vehicle is input into a pre-trained vehicle track recognition model to obtain a first abnormality degree value of the target vehicle.

6. The control method for AI recognition and early warning of abnormal vehicle behavior as claimed in claim 1, characterized in that: The abnormal behavior recognition model is a vehicle trajectory thermal recognition model, and the first target video data is input into a pre-trained cloud abnormal behavior recognition model to obtain a first abnormality degree value of the target vehicle, including: Determine the running track thermal data of the target vehicle according to the first target video data, wherein the running track thermal data includes track data and lane change frequency data; The running track thermal data of the target vehicle is input into a pre-trained vehicle track thermal recognition model to obtain a first abnormality degree value of the target vehicle.

7. The control method for AI recognition and early warning of abnormal vehicle behavior as claimed in claim 1, characterized in that: The abnormal behavior recognition model is a vehicle-associated speed recognition model, and the first target video data is input into a pre-trained cloud-based abnormal behavior recognition model to obtain a first abnormality degree value of the target vehicle, including: Determine the vehicle speed data of the target vehicle and the vehicle speed data of other surrounding vehicles according to the first target video data; The vehicle speed data of the target vehicle and the vehicle speed data of other surrounding vehicles are input into a pre-trained vehicle-associated speed recognition model to obtain a first abnormality degree value of the target vehicle.

8. The control method for AI recognition and early warning of abnormal vehicle behavior as claimed in claim 1, characterized in that: Before triggering the second warning signal, the method further includes: Acquire a historical behavior characteristic model of the target vehicle; Determining a historical abnormality value of the target vehicle according to the historical behavior characteristic model of the target vehicle; When the deviation between the second abnormality degree value and the historical abnormality degree value is greater than a preset value, the second warning signal is triggered.

9. The control method for AI recognition and early warning of abnormal vehicle behavior as claimed in claim 1, characterized in that: The step of inputting the first target video data into a pre-trained cloud-based abnormal behavior recognition model to obtain a first abnormality level value of the target vehicle includes: Acquiring a network status of a camera device, wherein the camera device is used to capture a driving status video of a target vehicle to obtain the first target video data; In the case where the network status of the camera device is abnormal, the first target video data is input into the edge server to perform abnormal behavior recognition to obtain a first abnormality degree value of the target vehicle; When the network status of the camera device is normal, the first target video data is input into a cloud server for abnormal behavior recognition to obtain a first abnormality degree value of the target vehicle, and the cloud server periodically updates the abnormal behavior recognition model of the edge server.

10. The control method for AI recognition and early warning of abnormal vehicle behavior as claimed in claim 9, characterized in that: Get the network status of the camera device, including: Obtain signal strength parameters, bandwidth parameters, and delay parameters of the camera device network link; Inputting the signal strength parameter, bandwidth parameter and delay parameter of the network link of the camera device into the network status evaluation model to obtain a comprehensive network status score value, wherein the delay parameter is obtained by sending a preset test data packet to the server in advance and recording the round-trip time; If the network status comprehensive score value is determined to be less than a preset threshold, the camera device is determined to be in an abnormal network state; if the network status comprehensive score value is determined to be greater than or equal to a preset threshold, the camera device is determined to be in a normal network state.

11. The control method for AI recognition and early warning of abnormal vehicle behavior as claimed in claim 1, characterized in that: After the second warning signal is triggered, the method further includes: Performing time series analysis on the second target video data to establish a time dynamic model of the target vehicle behavior; The behavior trend of the target vehicle in the future is predicted according to the time dynamic model. If the prediction result shows an increase in abnormal risk, the warning strategy is adjusted in advance to trigger a third warning signal, and the third warning signal is used to provide emergency braking warning for the target vehicle.

12. A vehicle warning system, characterized in that: include: A memory and a processor, wherein the memory is used to store program codes; The processor is used to call the program code to execute the method according to any one of claims 1 to 11.

Images