Driving behavior abnormality detection method, system, device and medium

Abstract

The present invention provides a method, system, device and medium for detecting abnormal driving behavior, which relates to the field of automobile driving technology. The method comprises: obtaining multiple sensor data of the object to be detected at the current moment, each of which is collected by different types of sensors; fusing the multiple sensor data to obtain a fused feature vector; generating a text description according to the fused feature vector; inputting the text description into a target model to obtain a driving behavior score; judging whether the current driving behavior is abnormal according to the driving behavior score, and issuing a warning indication if it is abnormal. The present invention improves the detection accuracy and warning accuracy of driving behavior.

Description

Technical Field

[0001] The present invention relates to the field of automobile driving technology, and in particular to a method, system, device and medium for detecting abnormal driving behavior. Background Art

[0002] Driving behavior detection systems have been widely used in the fields of intelligent driving, safety monitoring, and automobile management. Existing technologies analyze and detect driving behaviors based on collected sensor data through simple rules, but this approach has the problem of insufficient reliability. Summary of the invention

[0003] In view of this, the purpose of the present invention is to overcome the deficiencies in the prior art and provide a driving behavior detection method, system, device and medium. The present invention provides the following technical solutions:

[0004] In a first aspect, the present invention provides a method for detecting abnormal driving behavior, the method comprising: obtaining multiple sensor data of a test object at a current moment, each of the sensor data being collected by different types of sensors; fusing the multiple sensor data to obtain a fused feature vector; generating a text description based on the fused feature vector; inputting the text description into a target model to obtain a driving behavior score; judging whether the current driving behavior is abnormal based on the driving behavior score, and if abnormal, issuing a warning indication.

[0005] In one embodiment, before inputting the text description into the target model, it also includes: obtaining a sample data set, the sample data set including multiple sample data subsets, each of the sample data subsets corresponding to a different sample collection object, the sample data subsets including multiple groups of sensor data, and driving behavior labels corresponding to each group of the sensor data; preprocessing each group of the sensor data to obtain multiple groups of target sensor data; generating corresponding sample text descriptions according to each group of the target sensor data; dividing the multiple sample text descriptions and driving behavior labels corresponding to each group of the target sensor data into a training set, a test set and a validation set according to a preset ratio, and fine-tuning the pre-trained language model on the training set, the validation set and the test set to obtain the target model.

[0006] In one embodiment, each group of the sensor data includes multiple sensor data, and the preprocessing of each group of the sensor data includes: determining whether multiple sensor data in each group of the sensor data meet preset cleaning conditions, and if so, deleting the corresponding sensor data from the corresponding group of the sensor data to obtain a corresponding group of target sensor data.

[0007] In one embodiment, the multiple sensor data in each group of the sensor data are respectively collected by multiple different types of sensors, and the multiple sensor data collected by the same sensor at different times correspond to the same preset threshold range. The respectively judging whether the multiple sensor data in each group of the sensor data meet the preset cleaning condition includes: determining each sensor data as the sensor data to be determined in turn, the sensor data to be determined is collected by the sensor of the i-th type at the j-th time; judging whether the sensor data to be determined belongs to the preset threshold range corresponding to the sensor of the i-th type; if so, judging whether the absolute value of the difference between the sensor data to be determined and the sensor data collected by the sensor of the i-th type at the (j-1) time is greater than the preset difference threshold; if not, judging that the sensor data to be determined meets the preset cleaning condition.

[0008] In one embodiment, the fusion processing of the multiple sensor data includes: determining whether the number of the sensors of different types is greater than or equal to a preset number threshold; if so, using a high-level fusion method to splice the sensor data to obtain the fused feature vector; if less than, using a low-level fusion method to fuse the sensor data to obtain the fused feature vector.

[0009] In one embodiment, judging whether the current driving behavior is abnormal based on the driving behavior score includes: judging whether the driving behavior score is greater than or equal to a preset threshold, and if so, determining that the current driving behavior is abnormal.

[0010] In one embodiment, judging whether the current driving behavior is abnormal based on the driving behavior abnormality score further includes: if the current driving behavior abnormality score is less than the preset threshold, determining that the current driving behavior is normal.

[0011] In a second aspect, the present invention provides a driving behavior abnormality detection system, the system comprising:

[0012] An acquisition unit, used to acquire multiple sensor data of the object to be measured at the current moment, each of the sensor data is collected by different types of sensors;

[0013] A fusion unit, used for fusing the plurality of sensor data to obtain a fusion feature vector;

[0014] A text generation unit, used for generating a text description according to the fused feature vector;

[0015] A scoring unit, used for inputting the text description into a target model to obtain a driving behavior score;

[0016] The early warning unit is used to determine whether the current driving behavior is abnormal according to the driving behavior score, and if abnormal, issue an early warning indication.

[0017] In a third aspect, the present invention provides an electronic device including a memory and a processor, wherein the memory stores a computer program, and when the computer program runs on the processor, the method for detecting abnormal driving behavior described in the first aspect is executed.

[0018] In a fourth aspect, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the method for detecting abnormal driving behavior as described in the first aspect.

[0019] The present invention provides a method, system, device and medium for detecting abnormal driving behavior, the method comprising: obtaining multiple sensor data and environmental data of the object to be tested at the current moment, each of the sensor data is collected by different types of sensors; fusing the multiple sensor data to obtain a fused feature vector; generating a text description based on the fused feature vector and the environmental data; inputting the text description into a target model to obtain a driving behavior score; judging whether the current driving behavior is abnormal based on the driving behavior score, and if abnormal, issuing a warning indication. The present invention detects the current driving behavior through the fused multiple sensor data and environmental data, thereby improving the detection accuracy and warning accuracy of the driving behavior.

[0020] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and understandable, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0022] Figure 1 A schematic diagram of a process of a method for detecting abnormal driving behavior provided by an embodiment of the present invention is shown;

[0023] Figure 2 Another schematic diagram of the process of the abnormal driving behavior detection method provided by an embodiment of the present invention is shown;

[0024] Figure 3 A schematic diagram of the structure of a driving behavior abnormality detection system provided by an embodiment of the present invention is shown;

[0025] Figure 4 A schematic structural diagram of an electronic device provided by an embodiment of the present invention is shown.

[0026] Description of main component symbols:

[0027] 300-driving behavior abnormality detection system; 301-acquisition unit; 302-fusion unit; 303-text generation unit; 304-scoring unit; 305-warning unit; 400-electronic device; 401-transceiver; 402-processor; 403-memory. DETAILED DESCRIPTION

[0028] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0029] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying 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 one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0030] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein in the specification of the template are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0031] Example 1

[0032] The present invention provides a method for detecting abnormal driving behavior. Figure 1 , the method includes steps S110~S150.

[0033] Step S110, obtaining a plurality of sensor data of the object to be measured at the current moment, each of the sensor data is collected by a different type of sensor.

[0034] In this embodiment, the object to be tested refers to the current driving vehicle, and the sensor data are collected by multiple different types of sensors on the current driving vehicle. The sensor data includes but is not limited to the speed, acceleration, steering angle of the current driving vehicle and the environmental data of the current driving vehicle, etc., wherein the environmental data can also be obtained by calling external tools such as GPS, map services, weather forecast services and / or traffic information systems. The current driving behavior is comprehensively judged in combination with multiple sensor data at the current moment, thereby improving the accuracy of the driving behavior judgment.

[0035] Step S120, fusing the plurality of sensor data to obtain a fused feature vector.

[0036] In this embodiment, multiple sensor data are fused into a single output, which improves the efficiency of subsequent data processing. During the sensor data collection process, it is easy to be interfered by various factors, such as noise interference, signal interference, etc. By fusing multiple sensor data, signal errors can be reduced and data accuracy can be improved.

[0037] In one embodiment, the fusion processing of the multiple sensor data includes: determining whether the number of the sensors of different types is greater than or equal to a preset number threshold; if so, using a high-level fusion method to splice the sensor data to obtain the fused feature vector; if less than, using a low-level fusion method to fuse the sensor data to obtain the fused feature vector.

[0038] It should be noted that the low-level fusion method is suitable for simple scenarios with small data volume and fewer sensor types, while the high-level fusion method is suitable for complex scenarios with large data volume and more sensor types.

[0039] In this embodiment, by comparing the data volume of different types of sensors with the preset quantity threshold, it is determined whether to use the low-level fusion method or the high-level fusion method. The low-level fusion method is to splice multiple different types of sensor data with the same timestamp. For example, the preset quantity threshold is 4, and there are three different types of sensor data: speed v = 80 km / h, acceleration a = 3 m / s², steering angle θ = 15°, then the fusion feature vector obtained after splicing is [80, 3,15].

[0040] For another example, the preset quantity threshold is 2, and there are three different types of sensor data: speed v, acceleration and steering angle, then each sensor data is input into the multi-layer fusion sensor for nonlinear combination to obtain a fusion feature vector. With the current moment as the center, a time window is defined, for example, 5 seconds before the current moment to 5 seconds after the current moment, and multiple sensor data corresponding to different types of sensors within the time window are determined. Feature extraction is performed for multiple sensor data corresponding to different types of sensors, specifically, including determining the mean, standard deviation, maximum and minimum values, median and / or quartiles of multiple sensor data, and the extracted multiple features are combined to obtain a feature vector of the corresponding sensor. For example, for the speed sensor, multiple speed sensor data determined within the time window of t=0s to t=5s are v=60Km / h, v=65Km / h, v=70Km / h, v=75Km / h and v=80Km / h. Feature extraction is performed on the multiple speed sensor data to obtain the average value of the multiple sensor data μ_v = (60 + 65 + 70 + 75 + 80) / 5 = 70 km / h and the standard deviation: σ_v =sqrt(((60-70)² + (65-70)² + (70-70)² + (75-70)² + (80-70)²) / 5) ≈ 7.07 km / h. The average value and the standard deviation are combined to obtain the feature vector F_v = [μ_v, σ_v] = [70, 7.07] of the speed sensor. Multiple feature vectors corresponding to different types of sensors are input into a multilayer fusion perceptron for nonlinear fusion to obtain a fused feature vector. Among them, a multilayer perceptron is a feedforward neural network, usually composed of an input layer, one or more hidden layers, and an output layer. Each layer consists of several neurons, each neuron is connected to all neurons in the previous layer, and nonlinear transformation is performed through an activation function.

[0041] It should be noted that different types of sensors need to perform timestamp alignment before collecting corresponding sensor data. Specifically, the same sampling frequency is set for different types of sensors, for example, 10Hz, the time interval between each data point is 0.1 second, and the timestamps corresponding to each data are 0.1s, 0.2s, 0.3s, etc. The timestamps of the data collected by different types of sensors are aligned to ensure that each timestamp has corresponding sensor data. For example, the speed sensor collects a speed of 60 km / h at timestamp t = 0.1s, and a speed of 62 km / h at t = 0.2s, and so on; the acceleration sensor collects an acceleration of 2 m / s² at time point t = 0.1s, and an acceleration of 2.5 m / s² at t = 0.2s, and so on; the steering angle sensor collects a steering angle of 5° at time point t = 0.1s, and a steering angle of 6° at t = 0.2s, and so on.

[0042] Step S130, generating a text description according to the fused feature vector.

[0043] In this embodiment, a text description of the current vehicle driving behavior is generated based on the fused feature vector, for example, "The current time is 14:30 on October 5, 2023, and the vehicle is located at longitude 121.4737 and latitude 31.2304. The vehicle is currently traveling on a city road at a speed of 60 kilometers per hour, showing a relatively stable state. The acceleration sensor shows that the current acceleration is -2 meters per second squared, indicating that the vehicle is decelerating. The steering angle is 5 degrees, indicating that the vehicle is performing a slight steering operation. The road conditions ahead are good, with no obvious obstacles or traffic jams. The weather is clear, there are no obvious clouds in the sky, the temperature is 22 degrees Celsius, the humidity is 60%, and the light intensity is 800 lux, which is suitable for driving. The vehicle is braking, the braking force is moderate, and no emergency braking occurs."

[0044] Step S140: input the text description into the target model to obtain a driving behavior score.

[0045] In this embodiment, the target model scores the current driving behavior by extracting key information from the text description, and outputs the driving behavior score at the current moment.

[0046] It should be noted that different key information has different scoring standards, such as speed scoring: according to the current road type and speed limit, the speed is divided into low speed, medium speed, high speed and speeding, and different speeds correspond to different scores; acceleration scoring: according to the size of the acceleration, it is divided into steady, normal and sharp, and different accelerations correspond to different scores; steering angle scoring: according to the size of the steering angle, it is divided into slight, normal and sharp turn, and different steering angles correspond to different scores; road condition scoring: according to the description of the road condition, it is divided into good, general and complex, and different road conditions correspond to different scores; weather condition scoring: according to the weather conditions, it is divided into ideal, general and bad, and different weather conditions correspond to different scores; brake operation scoring: according to the braking force, it is divided into slight, normal and emergency braking, and different braking forces correspond to different scores. In addition, different key information is pre-assigned different weights according to the impression of driving safety of different key information. According to the scores corresponding to the extracted key information and their corresponding weights, the final driving behavior score can be determined. According to the driving behavior score, the driving behavior can be divided into several levels such as excellent, good, general, and needs improvement.

[0047] In one embodiment, before the text description is input into the target model, steps S210 to S230 are also included. Figure 2 .

[0048] Step S210, obtaining a sample data set, wherein the sample data set includes a plurality of sample data subsets, each of the sample data subsets corresponds to a different sample collection object, and the sample data subset includes a plurality of groups of sensor data and a driving behavior label corresponding to each group of the sensor data.

[0049] In this embodiment, the sample data includes sensor data collected by multiple vehicles within a preset time. The multiple sensor data collected by each vehicle at the same time is regarded as a group of sensor data, and a driving behavior label is pre-labeled for each group of sensor data, where the driving behavior label includes: normal driving or abnormal driving.

[0050] Step S220: pre-process each group of sensor data to obtain multiple groups of target sensor data.

[0051] In this embodiment, abnormal data is identified and removed from multiple sensor data. Specifically, multiple sensor data in a set of sensor data are collected by different types of sensors. To determine whether the sensor data is an abnormal value, it is necessary to compare the sensor data with a preset threshold range of the corresponding sensor. If the sensor data belongs to the corresponding preset threshold range, the sensor data is further compared with the sensor data collected by the corresponding sensor at the previous moment. It can be understood that if the value of the sensor data undergoes an abnormal mutation at a certain moment, the sensor data can be considered to be an abnormal value.

[0052] For example, the normal driving speed range of a vehicle is 0-200 km / h, and a speed change threshold Δv is set, such as 10 km / h. This means that within a sampling period, if the speed changes by more than 10 km / h, it is considered an outlier. Specifically, the speed difference between adjacent sampling points is calculated. If the speed at the current moment t is v(t) and the speed at the previous moment t-1 is v(t-1), the speed difference Δv = |v(t) - v(t-1)| is calculated. If Δv>10 km / h, the speed data v(t) at the current moment t is considered an outlier and should be removed. If at a certain moment t, v(t) = 150 km / h, and at the previous moment t-1, v(t-1) = 60 km / h, then Δv = |150 - 60| = 90 km / h>10 km / h, which is obviously unreasonable and v(t) should be removed.

[0053] In one embodiment, each group of the sensor data includes multiple sensor data, and the multiple sensor data in each group of the sensor data are preprocessed separately, including: separately judging whether the multiple sensor data in each group of the sensor data meet the preset cleaning conditions; if so, deleting the corresponding sensor data from the corresponding group of the sensor data to obtain a corresponding group of the target sensor data.

[0054] In this embodiment, the sensor data meeting the preset cleaning condition is deleted, and the remaining multiple sensor data are used as a group of target sensor data.

[0055] In one embodiment, the multiple sensor data in each group of the sensor data are respectively collected by multiple different types of sensors, and the multiple sensor data collected by the same sensor at different times correspond to the same preset threshold range. The respectively judging whether the multiple sensor data in each group of the sensor data meet the preset cleaning condition includes: determining each sensor data as the sensor data to be determined in turn, the sensor data to be determined is collected by the sensor of the i-th type at the j-th time; judging whether the sensor data to be determined belongs to the preset threshold range corresponding to the sensor of the i-th type; if so, judging whether the absolute value of the difference between the sensor data to be determined and the sensor data collected by the sensor of the i-th type at the (j-1) time is greater than the preset difference threshold; if not, judging that the sensor data to be determined meets the preset cleaning condition.

[0056] Step S230: Generate corresponding sample text descriptions according to each group of target sensor data.

[0057] In this embodiment, among the multiple sensor data collected by the same sensor at the same time, the multiple target sensor data remaining after cleaning constitute a group. According to the pre-set text rules, each group of target sensor data is converted into a corresponding sample text description, and each sample text description corresponds to a driving behavior label.

[0058] Step S240, dividing the multiple sample text descriptions and driving behavior labels corresponding to each group of the target sensor data into a training set, a test set and a validation set according to a preset ratio, and fine-tuning the pre-trained language model on the training set, the validation set and the test set to obtain a target model.

[0059] In this embodiment, the pre-trained language model includes: Qwen1.5-7b. By inputting annotated sample text descriptions into the pre-trained language model and setting fine-tuning parameters for the pre-trained language model, such as a learning rate of 1e-5, a batch size of 32, and a number of fine-tuning rounds of 3, the pre-trained language model is fine-tuned to obtain a target model. Specifically, multiple sample text descriptions and their corresponding driving behavior labels are divided into a training set, a validation set, and a test set according to a preset ratio, such as 70:15:15, and the pre-trained language model is fine-tuned on the training set, the validation set, and the test set, respectively, and the fine-tuned pre-trained language model is used as the target model.

[0060] Step S150, judging whether the current driving behavior is abnormal according to the driving behavior score, and if abnormal, issuing a warning indication.

[0061] In this embodiment, the driving behavior score ranges from 0 to 1. A higher driving behavior score indicates a higher risk of the current driving behavior. When the driving score is greater than or equal to a preset threshold, the current driving behavior is considered abnormal, and an early warning is issued to remind the current driver of the current driving risk. The early warning includes but is not limited to issuing voice prompt information, beep information, and / or indicator light information.

[0062] In one embodiment, judging whether the current driving behavior is abnormal based on the driving behavior score includes: judging whether the driving behavior score is greater than or equal to a preset threshold, and if so, determining that the current driving behavior is abnormal.

[0063] In one embodiment, judging whether the current driving behavior is abnormal based on the driving behavior abnormality score further includes: if the current driving behavior abnormality score is less than the preset threshold, determining that the current driving behavior is normal.

[0064] In this embodiment, if the driving behavior score is less than a preset threshold, the current driving behavior is determined to be normal. It should be noted that abnormal driving behavior indicates that the current driving behavior is high risk and a driving accident may occur.

[0065] The method for detecting abnormal driving behavior provided by an embodiment of the present invention comprises: obtaining multiple sensor data and environmental data of the object to be tested at the current moment, each of the sensor data is collected by different types of sensors; fusing the multiple sensor data to obtain a fused feature vector; generating a text description according to the fused feature vector and the environmental data; inputting the text description into a target model to obtain a driving behavior score; judging whether the current driving behavior is abnormal according to the driving behavior score, and if abnormal, issuing a warning indication. The present invention detects the current driving behavior through the fused multiple sensor data and environmental data, thereby improving the detection accuracy and warning accuracy of the driving behavior.

[0066] Example 2

[0067] The embodiment of the present invention also provides a driving behavior abnormality detection system. For details, see Figure 3 , the driving behavior abnormality detection system 300 includes:

[0068] An acquisition unit 301 is used to acquire multiple sensor data of the object to be measured at the current moment, each of which is collected by different types of sensors;

[0069] A fusion unit 302 is used to perform fusion processing on the plurality of sensor data to obtain a fusion feature vector;

[0070] A text generation unit 303, used to generate a text description according to the fused feature vector;

[0071] A scoring unit 304, configured to input the text description into a target model to obtain a driving behavior score;

[0072] The early warning unit 305 is used to determine whether the current driving behavior is abnormal according to the driving behavior score, and if abnormal, issue an early warning indication.

[0073] The abnormal driving behavior detection system 300 also includes a model training unit, which is used to obtain a sample data set, wherein the sample data set includes multiple sample data subsets, each of the sample data subsets corresponds to a different sample collection object, and the sample data subsets include multiple groups of sensor data, and driving behavior labels corresponding to each group of the sensor data; pre-processing each group of the sensor data to obtain multiple groups of target sensor data; generating corresponding sample text descriptions according to each group of the target sensor data; dividing the multiple sample text descriptions and driving behavior labels corresponding to each group of the target sensor data into a training set, a test set and a validation set according to a preset ratio, and fine-tuning the pre-trained language model on the training set, the validation set and the test set to obtain a target model.

[0074] Each group of sensor data includes multiple sensor data respectively. The model training unit is also used to determine whether the multiple sensor data in each group of sensor data meet the preset cleaning conditions. If so, the corresponding sensor data is deleted from the corresponding group of sensor data to obtain a corresponding group of target sensor data.

[0075] The multiple sensor data in each group of sensor data are respectively collected by multiple different types of sensors, and the multiple sensor data collected by the same sensor at different times correspond to the same preset threshold range. The model training unit is also used to sequentially determine each sensor data as sensor data to be determined, and the sensor data to be determined is collected by the i-th type of sensor at the j-th time; determine whether the sensor data to be determined belongs to the preset threshold range corresponding to the i-th type of sensor; if so, determine whether the absolute value of the difference between the sensor data to be determined and the sensor data collected by the i-th type of sensor at the (j-1) time is greater than the preset difference threshold; if not, determine that the sensor data to be determined meets the preset cleaning condition.

[0076] The fusion unit 302 is also used to determine whether the number of sensors of different types is greater than or equal to a preset number threshold. If so, the sensor data are spliced ​​using a high-level fusion method to obtain the fused feature vector; if less than, the sensor data are fused using a low-level fusion method to obtain the fused feature vector. The scoring unit 304 is also used to determine whether the driving behavior score is greater than or equal to a preset threshold. If so, the current driving behavior is determined to be abnormal; if not, the current driving behavior is determined to be normal.

[0077] Based on the same idea as the method in the above-mentioned embodiment, the system provided by the present application can implement the method in the above-mentioned embodiment. For the convenience of explanation, the structural diagram of the system embodiment only shows the parts related to the embodiment of the present application. Those skilled in the art can understand that the illustrated structure does not constitute a limitation on the system, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0078] The driving behavior abnormality detection system provided by the embodiment of the present invention obtains multiple sensor data of the object to be tested at the current moment, and each of the sensor data is collected by different types of sensors; the multiple sensor data are fused to obtain a fused feature vector; a text description is generated according to the fused feature vector; the text description is input into the target model to obtain a driving behavior score; and whether the current driving behavior is abnormal is determined according to the driving behavior score, and if abnormal, an early warning indication is issued. The present invention detects the current driving behavior through the fused multiple sensor data and environmental data, thereby improving the detection accuracy and early warning accuracy of the driving behavior.

[0079] Example 3

[0080] In addition, an embodiment of the present invention provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program runs on the processor, the driving behavior abnormality detection method provided in Example 4 is executed.

[0081] For details, see Figure 4 The electronic device 400 includes: a transceiver 401, a bus interface and a processor 402, wherein the processor 402 is used to obtain multiple sensor data of the object to be tested at the current moment, each of which is collected by different types of sensors; perform fusion processing on the multiple sensor data to obtain a fusion feature vector; generate a text description according to the fusion feature vector; input the text description into the target model to obtain a driving behavior score; determine whether the current driving behavior is abnormal according to the driving behavior score, and if abnormal, issue a warning indication.

[0082] In the embodiment of the present invention, the electronic device 400 further includes a memory 403. Figure 4 In the embodiment, the bus architecture may include any number of interconnected buses and bridges, specifically linking various circuits of one or more processors represented by processor 402 and memory represented by memory 403. The bus architecture may also link various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface. The transceiver 401 may be a plurality of components, i.e., including a transmitter and a receiver, providing a unit for communicating with various other devices on a transmission medium. The processor 402 is responsible for managing the bus architecture and general processing, and the memory 403 may store data used by the processor 402 when performing operations.

[0083] The electronic device 400 provided in the embodiment of the present invention can execute the abnormal driving behavior detection method provided in the above method embodiment 1, which will not be described again here to avoid repetition.

[0084] Example 4

[0085] In addition, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method for detecting abnormal driving behavior provided in Example 1 is implemented.

[0086] In this embodiment, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0087] The computer-readable storage medium provided in this embodiment can implement the abnormal driving behavior detection method provided in Example 1, and will not be described again here to avoid repetition.

[0088] In all examples shown and described herein, any specific values ​​should be interpreted as merely exemplary and not as limiting, and thus other examples of the exemplary embodiments may have different values.

[0089] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0090] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the present invention. It should be pointed out that for ordinary technicians in this field, several modifications and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention.

Claims

1. A method for detecting abnormal driving behavior, characterized in that: The method comprises: Acquire multiple sensor data of the object to be measured at the current moment, each of the sensor data is collected by different types of sensors; Performing fusion processing on the plurality of sensor data to obtain a fusion feature vector; Generate a text description according to the fused feature vector; Inputting the text description into a target model to obtain a driving behavior score; Determine whether the current driving behavior is abnormal according to the driving behavior score, and if abnormal, issue a warning indication; Before inputting the text description into the target model, the method further includes: obtaining a sample data set, the sample data set including a plurality of sample data subsets, each of the sample data subsets corresponding to a different sample collection object, the sample data subsets including a plurality of groups of sensor data, and a driving behavior label corresponding to each group of the sensor data; preprocessing each group of the sensor data to obtain a plurality of groups of target sensor data; generating a corresponding sample text description according to each group of the target sensor data; dividing the plurality of sample text descriptions and driving behavior labels corresponding to each group of the target sensor data into a training set, a test set and a validation set according to a preset ratio, and fine-tuning the pre-trained language model on the training set, the validation set and the test set to obtain a target model; Each group of the sensor data includes a plurality of sensor data, and the preprocessing of each group of the sensor data includes: determining whether a plurality of the sensor data in each group of the sensor data meets a preset cleaning condition, and if so, deleting the corresponding sensor data from a corresponding group of the sensor data to obtain a corresponding group of the target sensor data; The plurality of sensor data in each group of sensor data are respectively collected by a plurality of different types of sensors, and the plurality of sensor data collected by the same sensor at different times correspond to the same preset threshold range. The respectively judging whether the plurality of sensor data in each group of sensor data meet the preset cleaning condition comprises: sequentially determining each of the sensor data as the sensor data to be determined, the sensor data to be determined being collected by the sensor of the i-th type at the j-th time; judging whether the sensor data to be determined belongs to the preset threshold range corresponding to the sensor of the i-th type; if so, judging whether the absolute value of the difference between the sensor data to be determined and the sensor data collected by the sensor of the i-th type at the (j-1) time is greater than the preset difference threshold; if so, judging whether the sensor data to be determined meets the preset cleaning condition; The fusion processing of the multiple sensor data includes: judging whether the number of the different types of sensors is greater than or equal to a preset number threshold; if so, using a high-level fusion method to splice the sensor data to obtain the fused feature vector; if less than, using a low-level fusion method to fuse the sensor data to obtain the fused feature vector; and the different types of sensors need to perform timestamp alignment processing before respectively collecting the corresponding sensor data.

2. The method for detecting abnormal driving behavior according to any one of claim 1, characterized in that: The determining whether the current driving behavior is abnormal according to the driving behavior score includes: It is determined whether the driving behavior score is greater than or equal to a preset threshold, and if so, it is determined that the current driving behavior is abnormal.

3. The method for detecting abnormal driving behavior according to claim 2, characterized in that: The determining whether the current driving behavior is abnormal according to the driving behavior score further includes: If the current driving behavior abnormality score is less than the preset threshold, it is determined that the current driving behavior is normal.

4. A driving behavior abnormality detection system, characterized in that: The system comprises: An acquisition unit, used to acquire multiple sensor data of the object to be measured at the current moment, each of the sensor data is collected by different types of sensors; A fusion unit, used for fusing the plurality of sensor data to obtain a fusion feature vector; A text generation unit, used for generating a text description according to the fused feature vector; A scoring unit, used for inputting the text description into a target model to obtain a driving behavior score; An early warning unit, used to determine whether the current driving behavior is abnormal according to the driving behavior score, and if abnormal, issue an early warning indication; A model training unit is used to obtain a sample data set, wherein the sample data set includes multiple sample data subsets, each of which corresponds to a different sample collection object, and the sample data subset includes multiple groups of sensor data and driving behavior labels corresponding to each group of the sensor data; preprocess each group of the sensor data to obtain multiple groups of target sensor data; generate corresponding sample text descriptions according to each group of the target sensor data; divide the multiple sample text descriptions and driving behavior labels corresponding to each group of the target sensor data into a training set, a test set and a validation set according to a preset ratio, and fine-tune the pre-trained language model on the training set, the validation set and the test set to obtain a target model; Each group of sensor data includes a plurality of sensor data, and the model training unit is further used to determine whether the plurality of sensor data in each group of sensor data meets a preset cleaning condition, and if so, delete the corresponding sensor data from the corresponding group of sensor data to obtain a corresponding group of target sensor data; The plurality of sensor data in each group of sensor data are respectively collected by a plurality of different types of sensors, and the plurality of sensor data collected by the same sensor at different times correspond to the same preset threshold range. The model training unit is further used to sequentially determine each of the sensor data as sensor data to be determined, wherein the sensor data to be determined is collected by the sensor of the i-th type at the j-th time; determine whether the sensor data to be determined belongs to the preset threshold range corresponding to the sensor of the i-th type; if so, determine whether the absolute value of the difference between the sensor data to be determined and the sensor data collected by the sensor of the i-th type at the (j-1) time is greater than the preset difference threshold; if not, determine that the sensor data to be determined meets the preset cleaning condition; The fusion unit is further used to determine whether the number of different types of sensors is greater than or equal to a preset number threshold. If so, the sensor data are concatenated using a high-level fusion method to obtain the fused feature vector; if less than, the sensor data are fused using a low-level fusion method to obtain the fused feature vector; the scoring unit is further used to determine whether the driving behavior score is greater than or equal to a preset threshold. If so, it is determined that the current driving behavior is abnormal; if not, it is determined that the current driving behavior is normal; different types of sensors need to perform timestamp alignment before respectively collecting the corresponding sensor data.

5. An electronic device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and when the computer program is run on the processor, the method for detecting abnormal driving behavior according to any one of claims 1 to 3 is executed.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for detecting abnormal driving behavior according to any one of claims 1 to 3 is implemented.

Images