Driving behavior evaluation method and system based on data fusion and deep learning model
By fusing multi-source perception data and using a lightweight deep learning model, the accuracy and adaptability issues of existing driving behavior scoring systems are solved, achieving efficient and low-cost driving behavior scoring and style classification.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-27
- Publication Date
- 2026-03-10
AI Technical Summary
Existing driving behavior scoring systems lack accuracy, cannot comprehensively consider multiple driving factors, have poor adaptability to different driving scenarios, have high maintenance costs, and are difficult to adapt to changes in driver habits.
By employing multi-source perception data fusion and deep learning models, a lightweight deep learning network is constructed by collecting data from IMU, GPS, and CAN bus to automatically learn driving behavior characteristics and achieve driving behavior scoring and style classification.
It improves the accuracy and stability of driving behavior scoring, reduces false alarm rate, reduces manual rule maintenance costs, can adapt to changes in driver habits, has fewer model parameters and fast training speed, and is suitable for real-time scoring at the edge.
Smart Images
Figure CN121637166A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of driver assistance systems, and more particularly to a driving behavior evaluation method based on data fusion and deep learning models, a driving behavior evaluation system based on data fusion and deep learning models, electronic devices, and storage media. Background Technology
[0002] In existing technologies, driving behavior scoring systems generally use manual rules or simple statistical models to evaluate driver behavior. For example, they might count and score the number of instances of rapid acceleration, sudden braking, and sharp turns, or simply determine driving style (aggressive, normal, or stable) based on changes in vehicle speed. However, this simplistic approach has significant drawbacks: first, it lacks accuracy and cannot comprehensively consider multiple driving factors; second, it has poor adaptability to different driving scenarios, with a high false alarm rate in complex road conditions or inclement weather; and third, it relies on manually set rules, resulting in high maintenance costs and difficulty in adapting to changes in driver habits. Therefore, there is an urgent need for a more accurate, robust, and adaptive driving behavior scoring system. Summary of the Invention
[0003] In view of this, the purpose of the present invention is to provide a driving behavior evaluation method based on data fusion and deep learning models, a driving behavior evaluation system based on data fusion and deep learning models, an electronic device and a storage medium, aiming to solve the technical problems of existing technologies that cannot comprehensively consider multiple driving factors and have poor adaptability to different driving scenarios.
[0004] This invention provides the following solution:
[0005] According to one aspect of this application, a method for evaluating driving behavior based on data fusion and deep learning models is provided, comprising the following steps:
[0006] Collect multi-source perception data of the vehicle, including: IMU data, GPS speed and location data, and CAN bus data;
[0007] Synchronize and preprocess multi-source sensing data to form standardized temporal features;
[0008] A deep learning model is constructed, which learns from the standardized temporal features to obtain a deep learning model for driving behavior perception.
[0009] Obtain a public dataset, divide it into training and test sets, train and validate the deep learning model for driving behavior perception, and evaluate the model performance using a quantitative evaluation metric that measures the consistency between the model's predictions and the actual situation.
[0010] The trained and validated model is exported and deployed on the target vehicle device. The model outputs driving behavior scores or driving style classification results to obtain real-time evaluation of driving behavior.
[0011] Furthermore, including:
[0012] The quantitative evaluation indicators for the consistency between model predictions and reality include: quantifying the model's predictive performance and outputting comparable numerical results;
[0013] Calculate the deviation index between the model's predicted values and the actual values.
[0014] Furthermore, the IMU data includes: acceleration data and angular velocity data;
[0015] CAN bus data includes: pedal position data, steering angle data, and brake pressure data.
[0016] Furthermore, including:
[0017] Deep learning models are lightweight network structures; the network structure includes: input layer, convolutional layer, ReLU activation function, global average pooling layer, fully connected layer, and output layer.
[0018] Furthermore, including:
[0019] The input layer receives standardized temporal features;
[0020] Convolutional layers extract local temporal features;
[0021] The ReLU activation function performs a nonlinear transformation on the local temporal features extracted by the convolutional layer, preserving effective features and suppressing ineffective features;
[0022] Global average pooling layer compresses data dimensionality;
[0023] Fully connected layer fusion features;
[0024] The output layer outputs driving scores and driving style classification results.
[0025] Furthermore, the target devices in the vehicle include: in-vehicle edge computing devices or in-vehicle smart gateways;
[0026] Furthermore, including:
[0027] The deep learning model was trained and validated using the publicly available driving behavior dataset UAH-DriveSet.
[0028] According to two aspects of this application, a driving behavior evaluation system based on data fusion and deep learning models is provided, comprising:
[0029] The system includes a data acquisition module, a data preprocessing module, a model building module, a dataset training module, and a model application module.
[0030] The data acquisition module is used to collect multi-source perception data of the vehicle, including IMU data, GPS speed and location data, and CAN bus data.
[0031] The data preprocessing module is used to synchronize and preprocess multi-source sensing data to form standardized time-series features;
[0032] The model building module is used to build deep learning models. The deep learning models learn standardized temporal features to obtain a deep learning model for driving behavior perception.
[0033] The dataset training module is used to acquire public datasets, divide the public datasets into training and test sets, train and validate deep learning models, and evaluate model performance using a quantitative evaluation metric that measures the consistency between model predictions and the actual situation.
[0034] The model application module is used to export the trained and validated model and deploy it on the target vehicle device. The model outputs driving behavior scores or driving style classification results to obtain real-time evaluation of driving behavior.
[0035] According to three aspects of this application, an electronic device is provided, comprising: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;
[0036] The memory stores a computer program that, when executed by a processor, causes the processor to perform steps of a driving behavior assessment method based on data fusion and a deep learning model.
[0037] According to four aspects of this application, a computer-readable storage medium is provided that stores a computer program executable by an electronic device, which, when run on the electronic device, causes the electronic device to perform the steps of a driving behavior assessment method based on data fusion and a deep learning model.
[0038] Compared with the prior art, the present invention has the following advantages:
[0039] This application improves the accuracy and stability of driving behavior scoring, significantly reducing the false alarm rate compared to traditional methods; it automatically learns driving features, reducing the cost of manual rule maintenance, and can adapt to changes in driver habits; the model has few parameters, fast training speed, and convenient deployment, enabling real-time scoring at the edge, with low cost and high efficiency. Attached Figure Description
[0040] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0041] Figure 1 This is a flowchart of a driving behavior evaluation method based on data fusion and deep learning models provided by one or more embodiments of the present invention.
[0042] Figure 2 This is a structural diagram of a driving behavior evaluation system based on data fusion and deep learning models provided by one or more embodiments of the present invention.
[0043] Figure 3 This is a flowchart of a driving behavior evaluation method according to a specific embodiment of the present invention.
[0044] Figure 4 This is a block diagram of an electronic device structure for a driving behavior evaluation method based on data fusion and deep learning models, provided by one or more embodiments of the present invention. Detailed Implementation
[0045] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0046] The terminology used in the embodiments of this application is for the purpose of describing particular embodiments only and is not intended to limit the application. The singular forms “a,” “said,” and “the” used in the embodiments of this application and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.
[0047] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.
[0048] It should be understood that although the terms first, second, third, etc., may be used in the embodiments of this application, these descriptions should not be limited to these terms. These terms are only used to distinguish the descriptions. For example, first may also be referred to as second without departing from the scope of the embodiments of this application, and similarly, second may also be referred to as first.
[0049] Depending on the context, the words “if” or “suppose” as used here can be interpreted as “when” or “in response to determination” or “in response to detection.” Similarly, depending on the context, the phrases “if determination” or “if detection (of the stated condition or event)” can be interpreted as “when determination” or “in response to determination” or “when detection (of the stated condition or event)” or “in response to detection (of the stated condition or event).”
[0050] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that an article or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such an article or device. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the article or device that includes said element.
[0051] It should be noted that any symbols and / or numbers present in the specification that are not marked in the accompanying drawings are not reference numerals.
[0052] Figure 1 This is a flowchart of a driving behavior evaluation method based on data fusion and deep learning models provided by one or more embodiments of the present invention.
[0053] like Figure 1 As shown, it includes the following steps:
[0054] Step S1: Collect multi-source perception data of the vehicle, including: IMU data, GPS speed and location data, and CAN bus data;
[0055] Specifically, IMU data includes: acceleration data and angular velocity data;
[0056] CAN bus data includes: pedal position data, steering angle data, and brake pressure data.
[0057] Step S2: Synchronize and preprocess the multi-source sensing data to form standardized temporal features;
[0058] Specifically, using the CAN bus timestamp as a reference, linear interpolation is performed on the IMU and GPS data to ensure that all data are aligned on the same timeline.
[0059] The preprocessing includes noise reduction, which uses sliding window mean filtering to eliminate random noise from the sensor, such as minute jitter of the IMU and positioning drift of GPS.
[0060] Outlier removal involves eliminating data that deviates from the mean by three times the standard deviation. This process handles sensor fault data, such as CAN bus errors causing the pedal position to be 0 but the acceleration to be extremely high.
[0061] Standardization, achieved through Z-score normalization (mean = 0, standard deviation = 1), eliminates the influence of units, such as acceleration units m / s² vs. steering angle units, allowing the model to converge quickly.
[0062] Temporal window partitioning, by setting the window length to 3-5 seconds and the step size to 1 second, transforms continuous data into behavioral segments, such as an acceleration-uniform speed process within 3 seconds, to adapt to the model input.
[0063] The final output is a three-dimensional tensor: [number of samples, window length (e.g., 300 frames @ 100Hz), feature dimension (e.g., IMU 6D + GPS 2D + CAN 3D = 11D)], where each sample corresponds to a feature set of a driving behavior segment.
[0064] Step S3: Construct a deep learning model. The deep learning model learns the standardized temporal features to obtain a deep learning model for driving behavior perception.
[0065] Specifically, deep learning models can automatically extract high-level semantic features from temporal features (such as rapid acceleration = rapid increase in x-axis acceleration in a short period of time + significant increase in pedal position) without the need for manual rules.
[0066] Obtain a deep learning model for driving behavior perception, including performance optimization and generalization capability optimization of the deep learning model.
[0067] Step S4: Obtain the public dataset, divide the public dataset into training and test sets, train and validate the deep learning model for driving behavior perception, and evaluate the model performance using the consistency between the model prediction results and the real situation as a quantitative evaluation index.
[0068] Specifically, priority should be given to using authoritative datasets in the automotive field to ensure data diversity.
[0069] UAH-DriveSet: Includes 50+ drivers, 10+ scenarios (city / highway / rural), labeled with driving style, and includes IMU, GPS, and CAN data;
[0070] Segmentation strategy: Separately segment by driver, using 40 drivers' data for the training set and 10 for the test set to avoid model overfitting, such as only learning the operating habits of a certain driver;
[0071] Training parameters: Adam optimizer (adapts to sparse gradients), learning rate 0.001 (balances convergence speed and stability), epochs=20-30 (avoids overfitting), batch size=32 (adapts to GPU memory).
[0072] Monitor training set loss and validation set loss: If the training set loss continues to decrease but the validation set loss increases, it indicates overfitting;
[0073] Solutions include: adding data augmentation (such as injecting temporal noise), adding dropout layers (randomly discarding some neurons), and early stopping (stopping training when the validation set loss increases).
[0074] Step S5: Export the trained and validated model and deploy it on the target vehicle device. Obtain a real-time evaluation of driving behavior by outputting a driving behavior score or driving style classification result through the model.
[0075] Specifically, for export formats: prioritize lightweight formats such as ONNX (cross-platform compatible, supports C++ / Python deployment) and TFLite (optimized for mobile / MCU devices, with a size compression of 60%+).
[0076] Model optimization: quantization (INT8 quantization, converting 32-bit floating-point numbers to 8-bit integers, reducing size by 75%), pruning (removing redundant channels, reducing parameter count by 50%), ensuring compatibility with automotive MCUs (such as STM32H7) or edge boxes (such as NVIDIA Jetson Nano).
[0077] Furthermore, including:
[0078] The quantitative evaluation indicators for the consistency between model predictions and reality include: quantifying the model's predictive performance and outputting comparable numerical results;
[0079] Calculate the deviation index between the model's predicted values and the actual values.
[0080] Furthermore, including:
[0081] Deep learning models are lightweight network structures; the network structure includes: input layer, convolutional layer, ReLU activation function, global average pooling layer, fully connected layer, and output layer.
[0082] Furthermore, including:
[0083] The input layer receives standardized temporal features;
[0084] Convolutional layers extract local temporal features;
[0085] The ReLU activation function performs a nonlinear transformation on the local temporal features extracted by the convolutional layer, preserving effective features and suppressing ineffective features;
[0086] Global average pooling layer compresses data dimensionality;
[0087] Fully connected layer fusion features;
[0088] The output layer outputs driving scores and driving style classification results.
[0089] Furthermore, the target devices in the vehicle include: in-vehicle edge computing devices or in-vehicle smart gateways;
[0090] Furthermore, including:
[0091] The deep learning model was trained and validated using the publicly available driving behavior dataset UAH-DriveSet.
[0092] Specifically, it addresses the limitations of data from a single sensor: IMU captures dynamic responses, GPS provides spatiotemporal context, and CAN reflects operational intentions. These three complement each other to form a closed-loop perception of intention, action, and result, making driving behavior assessment more comprehensive.
[0093] Data synchronization and preprocessing ensure feature quality: timestamp alignment eliminates temporal misalignment, denoising and standardization reduce noise interference, enabling the model to focus on effective features and improving the evaluation accuracy compared to a single data source.
[0094] No manual feature design required: The model automatically learns high-level semantic features of driving behavior, such as rapid acceleration = sudden change in pedal position + increase in peak acceleration, avoiding the problem of poor adaptability of manual rules in traditional methods, further adapting to different vehicle models, and improving the generalization ability of driving habits.
[0095] Lightweight model design balances accuracy and computing power: Optimizations such as 1D CNN and quantization pruning reduce the number of model parameters to less than 100,000 and inference latency to less than 15ms, breaking through the computing power limitations of in-vehicle edge devices (MCU / smart gateways) and improving the deployment success rate compared to traditional deep learning models.
[0096] The preprocessing and inference pipelines run in parallel, and the lightweight model is optimized to achieve data acquisition, feature processing, and result output. The latency across the entire chain is reduced, and real-time feedback on driving behavior can be provided, such as immediate alerts when rapid acceleration occurs. This makes it more practical than traditional offline assessments.
[0097] Figure 2 This is a structural diagram of a driving behavior evaluation system based on data fusion and deep learning models provided by one or more embodiments of the present invention.
[0098] like Figure 2 As shown, it includes:
[0099] The system includes a data acquisition module, a data preprocessing module, a model building module, a dataset training module, and a model application module.
[0100] The data acquisition module is used to collect multi-source perception data of the vehicle, including: IMU data, GPS speed and location data, and CAN bus data;
[0101] The data preprocessing module is used to synchronize and preprocess multi-source sensing data to form standardized time-series features;
[0102] The model building module is used to build a deep learning model. The deep learning model learns from the standardized temporal features to obtain model data cognition of driving behavior patterns.
[0103] The dataset training module is used to acquire public datasets, divide the public datasets into training and test sets, train and validate deep learning models, and evaluate model performance using a quantitative evaluation metric that measures the consistency between model predictions and the actual situation.
[0104] The model application module is used to export the trained and validated model and deploy it on the target vehicle device. The model outputs driving behavior scores or driving style classification results to obtain real-time evaluation of driving behavior.
[0105] It is worth noting that although only some basic functional modules are disclosed in this embodiment, it does not mean that the composition of this system is limited to the above-mentioned basic functional modules. On the contrary, what this embodiment intends to express is that, based on the above-mentioned basic functional modules, those skilled in the art can arbitrarily add one or more functional modules in combination with existing technology to form an infinite number of embodiments or technical solutions. That is to say, this system is open rather than closed. The fact that this embodiment only discloses a few basic functional modules does not mean that the scope of protection of the claims of this invention is limited to the disclosed basic functional modules. At the same time, for the convenience of description, the above device is described separately according to its functions as various units and modules. Of course, in implementing this invention, the functions of each unit and module can be implemented in one or more software and / or hardware.
[0106] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0107] In another specific embodiment, such as Figure 3 As shown, it includes:
[0108] Step 1: Data Acquisition and Preprocessing. Acquire vehicle IMU data (acceleration, angular velocity), GPS speed and position data, and CAN bus data (e.g., pedal position, steering angle, braking pressure, etc.), perform synchronization and preprocessing, and form standardized timing characteristics.
[0109] Vehicle dynamic responses are captured by collecting IMU data, compensating for GPS lag. GPS provides absolute position and speed (error ±0.5m / s) for long-term driving behavior evaluation. CAN bus data directly reflects the driver's intentions, such as pedal travel (acceleration intention), braking pressure (deceleration intensity), and steering angle (driving aggressiveness).
[0110] Timestamp alignment: IMU / GPS data is linearly interpolated based on the CAN bus timestamp (1ms accuracy).
[0111] Preprocessing flow: Denoising, sliding window mean filtering, window size 5-10 frames → outlier removal, 3σ criterion → standardization, Z-score normalization → temporal window division, window length 3-5s, step size 1s. Obtain standardized temporal features.
[0112] Step 2: Build a lightweight deep learning model using a shallow 1D CNN or MLP network structure. Input the processed data window sequence, and the model automatically learns driving behavior features and outputs a driving score (0-100) or driving style classification (smooth, normal, aggressive).
[0113] Among them, 1D CNN includes: 1D convolutional layers, which extract local temporal features. Using multiple different convolutional kernels, it can extract 64 different types of local features in parallel.
[0114] The ReLU activation function performs a nonlinear transformation on the local temporal features extracted by the convolutional layer, preserving effective features and suppressing ineffective features. This introduces nonlinearity, enabling the model to learn complex nonlinear relationships, which is crucial for distinguishing different driving styles.
[0115] The global average pooling layer compresses data dimensionality; the fully connected layer fuses features; and the output layer outputs driving scores and driving style classification results.
[0116] The pooled feature vectors are mapped to the final output dimension (score or classification).
[0117] Step 3: Model Training and Validation. Use a public dataset (e.g., UAH-DriveSet), split it into training and testing sets, and perform model training and validation. Evaluate model performance using MSE or F1-score.
[0118] Step 4: Model Deployment. Export the trained model in a lightweight format, such as ONNX or TFLite, and deploy it on vehicle edge devices or in-vehicle smart gateways to achieve real-time scoring and feedback of driving behavior.
[0119] This invention employs the UAH-DriveSet public dataset for model training and evaluation. The dataset contains IMU and GPS data for different driving styles. The model structure is as follows: Input layer -> 1D convolutional layer (64 convolutional kernels, kernel size 3) -> ReLU activation function -> Global average pooling layer -> Fully connected layer (32 neurons) -> Output layer (score 0-100 or classification output). The model training process consists of 20 epochs, using the Adam optimizer with a learning rate of 0.001. The trained model achieves a driving style classification accuracy of over 90% on the test set.
[0120] Automatic driving behavior scoring is based on multi-source fusion of IMU, GPS and CAN data; a lightweight deep learning model is used to automatically extract driving behavior features, avoiding manual rule setting; the model is deployed at the edge to provide real-time and rapid driving score feedback.
[0121] In another embodiment, a GRU or LSTM model can be used instead of a 1D CNN or MLP, and the data input can also be expanded to more dimensional sensor data, such as wheel speed sensor data.
[0122] Figure 4 This is a block diagram of an electronic device structure for a driving behavior evaluation method based on data fusion and deep learning models, provided by one or more embodiments of the present invention.
[0123] like Figure 4As shown, this application provides an electronic device, including: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;
[0124] The memory stores a computer program that, when executed by a processor, causes the processor to perform steps of a driving behavior evaluation method based on data fusion and a deep learning model.
[0125] This application also provides a computer-readable storage medium storing a computer program executable by an electronic device, which, when run on the electronic device, causes the electronic device to perform steps of a driving behavior assessment method based on data fusion and a deep learning model.
[0126] For the sake of simplicity, the method embodiments are described as a series of actions. However, those skilled in the art should understand that the embodiments of the present invention are not limited to the described order of actions, because according to the embodiments of the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions involved are not necessarily essential to the embodiments of the present invention.
[0127] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of this application.
[0128] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A driving behavior evaluation method based on data fusion and deep learning model, characterized in that, The method comprises the following steps: Collecting multi-source perception data of a vehicle, the multi-source perception data comprising IMU data, GPS speed and position data, and CAN bus data; Synchronizing and preprocessing the multi-source perception data to form standardized time-series features; Building a deep learning model, the deep learning model learning the standardized time-series features to obtain a driving behavior perception deep learning model; Obtaining a public data set, dividing a training set and a test set based on the public data set, training and verifying the driving behavior perception deep learning model, and evaluating the performance of the model by using a consistency quantitative evaluation index of the prediction result of the model and the real situation; Exporting the trained and verified model to a vehicle target device, outputting a driving behavior score or a driving style classification result by the model, and obtaining real-time evaluation of driving behavior. 2.The driving behavior evaluation method based on data fusion and deep learning model according to claim 1, wherein, The consistency quantitative evaluation index of the prediction result of the model and the real situation comprises quantifying the prediction performance of the model and outputting comparable numerical results; Calculating the deviation index of the predicted value and the real value of the model. 3.The driving behavior evaluation method based on data fusion and deep learning model according to claim 1, characterized in that, The IMU data comprises acceleration data and angular velocity data; The CAN bus data comprises pedal position data, steering angle data, and brake pressure data. 4.The driving behavior evaluation method based on data fusion and deep learning model according to claim 1, wherein, The deep learning model is a lightweight network structure; The network structure comprises an input layer, a convolutional layer, an activation function ReLU, a global average pooling layer, a fully connected layer, and an output layer.
5. The driving behavior evaluation method based on data fusion and deep learning model according to claim 4, characterized in that, The input layer receives standardized time-series features; The convolutional layer extracts local time-series features; The activation function ReLU performs nonlinear transformation on the local time-series features extracted by the convolutional layer, retains effective features, and suppresses ineffective features; The global average pooling layer compresses the data dimension; The fully connected layer fuses features; The output layer outputs a driving score and a driving style classification result. The vehicle target device comprises a vehicle-mounted edge computing device or a vehicle-mounted intelligent gateway. 6.The driving behavior evaluation method based on data fusion and deep learning model according to claim 1, wherein, The public data set is specifically a UAH-DriveSet public driving behavior data set, which is used to train and verify the driving behavior perception deep learning model.
7. The driving behavior evaluation method based on data fusion and deep learning model according to claim 1, characterized in that, The method comprises the following steps:
8. A driving behavior evaluation system based on data fusion and deep learning model, characterized in that, A data acquisition module, a data preprocessing module, a model building module, a data set training module, and a model application module; The data acquisition module is configured to collect multi-source perception data of a vehicle, the multi-source perception data comprising IMU data, GPS speed and position data, and CAN bus data; The data preprocessing module is configured to synchronize and preprocess the multi-source perception data to form standardized time-series features; The model building module is configured to build a deep learning model, the deep learning model learning the standardized time-series features to obtain a driving behavior perception deep learning model; The data set training module is configured to obtain a public data set, divide a training set and a test set based on the public data set, train and verify the driving behavior perception deep learning model, and evaluate the performance of the model by using a consistency quantitative evaluation index of the prediction result of the model and the real situation. The model application module is used for exporting the model after training and verification, deploying the model on a vehicle target device, outputting a driving behavior score or a driving style classification result through the model, and obtaining a real-time evaluation of the driving behavior.
9. An electronic device, comprising: The method comprises the following steps: The processor, the communication interface, and the memory are in communication with each other through the communication bus. The memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the driving behavior evaluation method based on data fusion and a deep learning model according to any one of claims 1-7.
10. A computer-readable storage medium, characterized in that, The memory stores a computer program executable by the electronic device, and when the computer program runs on the electronic device, the electronic device executes the steps of the driving behavior evaluation method based on data fusion and a deep learning model according to any one of claims 1-7.