Driving behavior recognition method and device and vehicle

Abstract

The invention discloses a driving behavior recognition method and device and a vehicle, and is applied to the technical field of vehicles, and the method comprises the steps: obtaining the driving state data and travel index data of a target vehicle; performing feature extraction on the driving state data to obtain driving state features; performing feature extraction on the travel index data to obtain travel index features; performing feature fusion on the driving state feature and the travel index feature to obtain a target driving feature; and performing driving behavior recognition processing by using the target driving characteristics to obtain a driving behavior recognition result. Under the condition that the driving state of the vehicle and the index data associated with the driving journey of the vehicle are fully considered, the driving behavior of the vehicle is identified, so that the identification precision of the driving behavior is effectively improved.

Description

Technical Field

[0001] This application relates to the field of vehicle technology, and in particular to a driving behavior recognition method, device and vehicle. Background Technology

[0002] With the rapid development of intelligent connected vehicles, the analysis of driving behavior has become a key technology for improving traffic safety, optimizing vehicle services, and innovating business models. Related technologies identify vehicle driving behavior based on driving status data such as vehicle speed, combined with threshold judgments or machine learning methods. However, these technologies rely solely on this single dimension of driving status data to identify driving behavior, making it difficult to comprehensively and accurately describe actual driving behavior, resulting in low accuracy in driving behavior recognition. Summary of the Invention

[0003] This application provides a driving behavior recognition method, device, and vehicle, which can recognize the driving behavior of a vehicle by fully considering the vehicle's driving status and index data related to the vehicle's driving distance, thereby effectively improving the recognition accuracy of driving behavior.

[0004] On the one hand, embodiments of this application provide a driving behavior recognition method, including the following steps: Obtain the target vehicle's driving status data and trip index data; Feature extraction is performed on the driving status data to obtain driving status features; Feature extraction is performed on the travel indicator data to obtain travel indicator features; The driving status features and the trip index features are fused to obtain the target driving features; The driving behavior is identified by using the target driving features to obtain the driving behavior identification result.

[0005] Further, in one embodiment, the driving state data includes sensor data, on-board diagnostic data, and intelligent driving data at each preset time; the driving state features include sensor feature sequences, on-board diagnostic feature sequences, and intelligent driving feature sequences; the step of extracting features from the driving state data to obtain driving state features includes: Feature extraction is performed on the sensor data at each preset time to obtain the sensor feature sequence; Feature extraction is performed on the vehicle diagnostic data at each preset time to obtain the vehicle diagnostic feature sequence; Features are extracted from the intelligent driving data at each preset time point to obtain the intelligent driving feature sequence.

[0006] Further, in one embodiment, the trip indicator data includes travel time data and travel score data; the trip indicator features include travel time features and travel score features; the step of extracting features from the trip indicator data to obtain trip indicator features includes: Feature extraction is performed on the driving time data to obtain the driving time features; Feature extraction is performed on the driving score data to obtain the driving score features.

[0007] Furthermore, in one embodiment, the driving state features include sensor feature sequences, on-board diagnostic feature sequences, and intelligent driving feature sequences; the trip index features include driving time features and driving score features; and the target driving features include one-dimensional driving features. The step of fusing the driving state features and the trip index features to obtain the target driving features includes: The sensor feature sequence, the vehicle diagnostic feature sequence, and the intelligent driving feature sequence are averaged and spliced ​​to obtain the first spliced ​​feature; The first splicing feature and the driving time feature are fused together to obtain the first driving feature; The first splicing feature and the driving score feature are fused together to obtain the second driving feature; The first driving feature and the second driving feature are fused together to obtain the one-dimensional driving feature.

[0008] Furthermore, in one embodiment, the driving state features include sensor feature sequences, on-board diagnostic feature sequences, and intelligent driving feature sequences; the trip index features include driving time features and driving score features; and the target driving features include two-dimensional driving images. The step of fusing the driving state features and the trip index features to obtain the target driving features includes: The driving time feature and the driving score feature are concatenated to obtain the second concatenated feature; The second splicing feature and the sensor feature sequence are fused to obtain the first fused feature sequence; The second splicing feature and the vehicle diagnostic feature sequence are fused together to obtain the second fused feature sequence. The second splicing feature and the intelligent driving feature sequence are fused to obtain a third fused feature sequence; The two-dimensional driving image is obtained by performing image processing based on the first fusion feature sequence, the second fusion feature sequence and the third fusion feature sequence.

[0009] Further, in one embodiment, the image processing based on the first fused feature sequence, the second fused feature sequence, and the third fused feature sequence to obtain the two-dimensional driving image includes: A red channel map is generated based on the first fused feature sequence; A green channel map is generated based on the second fusion feature sequence; A blue channel image is generated based on the third fusion feature sequence; The two-dimensional driving image is obtained based on the red channel image, the green channel image, and the blue channel image.

[0010] Further, in one embodiment, the target driving features include one-dimensional driving features and a two-dimensional driving image; the step of using the target driving features to perform driving behavior recognition processing to obtain driving behavior recognition results includes: The two-dimensional driving image is input into a pre-trained neural network model to obtain a first recognition result and the confidence level of the first recognition result; The one-dimensional driving features are input into a pre-trained machine learning model to obtain a second recognition result and the confidence level of the second recognition result; Based on the confidence levels of the first and second identification results, decision processing is performed on the first and second identification results to obtain the driving behavior identification result.

[0011] Furthermore, in one embodiment, after acquiring the driving status data and trip index data of the target vehicle, the method further includes: The driving status data and the trip index data are cleaned. The driving status data and the trip index data are standardized.

[0012] On the other hand, embodiments of this application provide a driving behavior recognition device, including: The acquisition module is used to acquire the driving status data and trip index data of the target vehicle; The first processing module is used to extract features from the driving status data to obtain driving status features; The second processing module is used to extract features from the travel index data to obtain travel index features; The third processing module is used to fuse the driving status features and the travel index features to obtain the target driving features; The fourth processing module is used to perform driving behavior recognition processing using the target driving features to obtain driving behavior recognition results.

[0013] In another aspect, embodiments of this application provide a vehicle, including: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the above-described driving behavior recognition method.

[0014] According to an embodiment of this application, a driving behavior recognition method, apparatus, and vehicle are provided. The method acquires driving state data and travel indicator data of a target vehicle; extracts features from the driving state data to obtain driving state features; extracts features from the travel indicator data to obtain travel indicator features; fuses the driving state features and travel indicator features to obtain target driving features; and uses the target driving features for driving behavior recognition processing to obtain a driving behavior recognition result. This application, by fully considering the vehicle's driving state and indicator data associated with the vehicle's travel distance, achieves the recognition of the vehicle's driving behavior, thereby effectively improving the accuracy of driving behavior recognition.

[0015] Other features and advantages of this application will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures particularly pointed out in the description, claims and drawings. Attached Figure Description

[0016] Figure 1 This is a flowchart of a driving behavior recognition method provided in this application; Figure 2 This is a structural diagram of a driving behavior recognition device provided in this application; Figure 3 This is an example image of a vehicle provided in this application. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0018] The present application will be further described below with reference to the accompanying drawings and specific embodiments. The described embodiments should not be considered as limitations on the present application, and all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of the present application.

[0019] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0021] With the rapid development of intelligent connected vehicles, the analysis of driving behavior has become a key technology for improving traffic safety, optimizing vehicle services, and innovating business models. Among related technologies, driving behavior is identified based on driving status data such as vehicle speed, combined with threshold judgments or machine learning methods.

[0022] However, current technologies rely solely on single-dimensional data—driving state data—to identify driving behavior, which fails to comprehensively and accurately describe actual driving actions, resulting in low recognition accuracy. Specifically, driving behavior is a complex dynamic process influenced by numerous interacting factors. Simply relying on driving state data cannot encompass information such as shift timing, engine load changes, and vehicle posture adjustments, leading to overly incomplete assessments of driving behavior. For instance, when judging rapid acceleration, ignoring factors like engine speed and torque output may incorrectly classify normal acceleration as rapid acceleration, affecting the accuracy of driving behavior recognition.

[0023] Furthermore, the threshold judgment method used in these technologies is a fixed rule matching method, which lacks adaptability to individual driver differences and complex and ever-changing driving scenarios. Specifically, different drivers have different driving habits and styles, and fixed threshold judgments cannot meet personalized analysis needs, easily leading to misjudgments of driving behavior. Moreover, when encountering special scenarios such as traffic congestion or inclement weather, fixed rules are difficult to accurately handle and cannot effectively identify potentially dangerous driving behaviors.

[0024] Furthermore, the related technologies have failed to fully leverage the advantages of machine learning in processing complex data and pattern recognition. Specifically, as the amount of data generated by vehicles grows ever larger, the technologies are unable to extract deep, valuable information from this massive amount of data, making it difficult to achieve accurate driving behavior profiling and thus limiting their application in multiple fields such as traffic safety, vehicle insurance, and personalized services.

[0025] In summary, the relevant technologies, due to their limited data dimensions, simplistic analysis methods, and insufficient application of machine learning, cannot meet the need for comprehensive and accurate analysis of driving behavior.

[0026] In view of this, embodiments of this application provide a driving behavior recognition method, device, and vehicle, which, by fully considering the vehicle's driving state and index data associated with the vehicle's driving distance, realizes the recognition of the vehicle's driving behavior, thereby effectively improving the recognition accuracy of driving behavior.

[0027] First, the driving behavior recognition method provided in this application will be described in detail below with reference to the accompanying drawings.

[0028] This application provides a driving behavior recognition method that can be applied to a terminal, a server, or software running on either a terminal or a server. The terminal can be a tablet, laptop, desktop computer, etc., but is not limited to these. The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms. Furthermore, the server can be a node server in a blockchain network, but is not limited to these. Blockchain is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms.

[0029] Reference Figure 1 , Figure 1 This is a flowchart of a driving behavior recognition method provided in this application, which may include the following steps S101-S105.

[0030] S101, Obtain the target vehicle's driving status data and trip index data.

[0031] It should be noted that the target vehicle refers to a vehicle to which the driving behavior recognition method provided in this application is applicable.

[0032] As can be understood, driving status data refers to data associated with the driving status of the target vehicle. This data is time-series data and may include, but is not limited to, sensor data, on-board diagnostic data, and intelligent driving data at various preset times. Sensor data may include, but is not limited to, speed sensor data, acceleration sensor data, steering angle sensor data, brake sensor data, engine speed sensor data, fuel consumption sensor data, energy consumption sensor data, and position data. On-board diagnostic data refers to diagnostic signals output by the on-board diagnostic system (OBD). Intelligent driving data refers to signals output by advanced driver assistance systems (ADAS).

[0033] In addition, trip indicator data refers to data associated with the trip of the target vehicle. It is not time-series data and may include, but is not limited to, driving score data and driving time data. Among them, driving score data refers to the rating data given by users to the driving process. Its value range can be flexibly set according to the actual situation, for example, set to zero to ten. The higher the value, the higher the user's satisfaction with the driving process, and vice versa. Driving time data refers to the driving time of a single trip.

[0034] In this step, various specialized sensors are deployed on the target vehicle, including but not limited to speed sensors, acceleration sensors, steering angle sensors, brake sensors, engine speed sensors, fuel consumption sensors, and energy consumption sensors. These sensors are also connected to the data interfaces of the on-board diagnostic system, advanced driver assistance system, and intelligent driving system to acquire a portion of the vehicle's driving status data in real time. Optionally, data is collected at a set high frequency (e.g., 10 times per second or higher) to ensure data timeliness and completeness. Simultaneously, a dedicated mobile device application is developed to acquire another portion of the vehicle's driving status data via Bluetooth or wireless connection to the target vehicle, such as driving scores (which can be calculated by the on-board system and transmitted to the mobile phone) and driving time. Furthermore, the mobile device can utilize its own sensors (such as the Global Positioning System (GPS)) to acquire the vehicle's location data, providing richer data dimensions for driving behavior analysis. In this step, the data collected from the vehicle is aggregated to the on-board terminal device via the Controller Area Network (CAN) bus. The in-vehicle terminal device utilizes the Message Queuing Telemetry Transport (MQTT) protocol of the Internet of Things (IoT) platform to encapsulate data into data packets conforming to the MQTT message format, and sends them to the cloud server via 4G or 5G networks. The MQTT protocol is lightweight, low-power, and highly reliable, making it suitable for scenarios like vehicles with complex network environments and high requirements for data transmission stability. Data collected by the mobile device is also transmitted to the cloud server via the MQTT protocol or other suitable wireless communication protocols, ensuring accurate and timely aggregation of multi-source data.

[0035] S102, extract features from the driving status data to obtain driving status features.

[0036] In this step, after obtaining the driving status data, feature extraction is performed to extract key features associated with the driving status, thereby obtaining the driving status features.

[0037] S103, extract features from the travel indicator data to obtain travel indicator features.

[0038] In this step, after obtaining the travel indicator data, feature extraction is performed on it to extract key features associated with the travel indicator, thereby obtaining the travel indicator features.

[0039] S104, the driving status features and travel index features are fused to obtain the target driving features.

[0040] In this step, after obtaining the driving status features and travel index features, the two are fused together. The aim is to achieve the fusion and interaction between the key features of the driving status and the key features of the travel index, so as to obtain the target driving features.

[0041] S105, using the target driving characteristics to perform driving behavior recognition processing, and obtain driving behavior recognition results.

[0042] In this step, after obtaining the target driving characteristics, driving behavior recognition processing is performed based on these characteristics to obtain the driving behavior recognition result. Optionally, the driving behavior recognition result can be flexibly changed according to the actual recognition task. The recognition task can be divided into driver driving style classification tasks (such as aggressive, moderate, and economical), safe driving rating tasks (such as low rating, moderate rating, and high rating), potential dangerous behavior recognition tasks (such as dangerous type and safe type), and driving habit preference tasks (such as frequently used driving routes and driving time period preferences). For example, the driving behavior recognition result obtained by this method indicates that the driver's driving style in the past week was aggressive, or the safe driving rating was low.

[0043] The steps described above will be explained in further detail below.

[0044] In some embodiments, after obtaining the driving status data and trip index data of the target vehicle in step S101, the method may further include: Perform data cleaning on driving status data and trip index data; Standardize driving status data and trip index data.

[0045] In this embodiment, the collected multi-source data undergoes comprehensive cleaning to remove duplicate data, fill in missing values, and correct erroneous data. For example, for occasional anomalous jumps in sensor values, statistical methods (such as the 3σ criterion) or anomaly detection algorithms from machine learning are used for identification and correction. For missing fuel consumption data, linear regression or other data interpolation algorithms are used to fill in the missing data based on relevant data such as vehicle driving conditions and engine speed. Furthermore, appropriate standardization methods are used for different types of data to ensure they have uniform dimensions and scales. For numerical data, such as speed, acceleration, and fuel consumption, normalization or standardization is applied to map them to the [0, 1] interval or a standard normal distribution with a mean of 0 and a standard deviation of 1, to facilitate subsequent data analysis and model training.

[0046] In some embodiments, the aforementioned driving state data includes sensor data, on-board diagnostic data, and intelligent driving data at preset times; the aforementioned driving state features include sensor feature sequences, on-board diagnostic feature sequences, and intelligent driving feature sequences; in step S102, feature extraction is performed on the driving state data to obtain driving state features, including: Feature extraction is performed on sensor data at each preset time point to obtain sensor feature sequences; Feature extraction is performed on the on-board diagnostic data at each preset time point to obtain the on-board diagnostic feature sequence; Features are extracted from the intelligent driving data at each preset time point to obtain the intelligent driving feature sequence.

[0047] In this embodiment, feature extraction is performed on sensor data at each preset time to obtain a sensor feature sequence, which includes the sensor features at each preset time. For example, the sensor data includes speed sensor data, acceleration sensor data, and power consumption sensor data. For speed sensor data, features such as average speed, maximum speed, rate of change of speed, and speed standard deviation are extracted; for acceleration sensor data, features such as the number of rapid accelerations, the number of rapid decelerations, average acceleration, and peak acceleration are extracted; for power consumption sensor data, features such as charging frequency, charging duration, and charging location distribution are extracted. Furthermore, feature extraction is performed on vehicle diagnostic data at each preset time to obtain a vehicle diagnostic feature sequence, which includes the vehicle diagnostic features at each preset time. For example, for each preset time, if the signal output by the vehicle diagnostic system at the current preset time indicates that the target vehicle has any preset fault at the current preset time, then the vehicle diagnostic feature value at the current preset time is set to one; otherwise, the value is zero. Finally, feature extraction is performed on intelligent driving data at each preset time to obtain an intelligent driving feature sequence, which includes the intelligent driving features at each preset time. For example, features such as the number of lane-keeping assist activations and the frequency of adaptive cruise control adjustments can be extracted from intelligent driving data to comprehensively characterize the vehicle's intelligent driving behavior. In this way, key features associated with driving status can be effectively captured at a superficial level, thus contributing to a comprehensive portrayal of actual driving conditions.

[0048] In some embodiments, the aforementioned trip index data includes travel time data and travel score data; the aforementioned trip index features include travel time features and travel score features; in step S103, feature extraction is performed on the trip index data to obtain trip index features, including: Feature extraction is performed on the travel time data to obtain travel time features; Feature extraction is performed on the driving score data to obtain driving score features.

[0049] In this embodiment, feature extraction is performed on the driving time data. For example, driving duration distribution features for different time periods (such as weekdays, weekends, daytime, and nighttime) are extracted to obtain driving time features. Additionally, feature extraction is performed on the driving score data. For instance, driving score data ranges from zero to ten. If the driving score data is within the range of zero to three, the driving score feature value is set to zero, indicating a low score. If the driving score data is within the range of four to seven, the driving score feature value is set to one, indicating a medium score. If the driving score data is within the range of eight to ten, the driving score feature value is set to two, indicating a high score, thus obtaining the driving score features. In this way, key features associated with the driving journey can be effectively captured at a shallow level, thereby helping to comprehensively depict the actual driving situation.

[0050] In some embodiments, the aforementioned driving state features include sensor feature sequences, on-board diagnostic feature sequences, and intelligent driving feature sequences; the aforementioned trip index features include driving time features and driving score features; the aforementioned target driving features include one-dimensional driving features; in step S104, the driving state features and trip index features are fused to obtain the target driving features, including: The sensor feature sequence, vehicle diagnostic feature sequence, and intelligent driving feature sequence are averaged and spliced ​​to obtain the first spliced ​​feature. The first splicing feature and the driving time feature are fused together to obtain the first driving feature; The first splicing feature and the driving score feature are fused together to obtain the second driving feature; The first driving feature and the second driving feature are fused together to obtain a one-dimensional driving feature.

[0051] In this embodiment, firstly, the mean value of the sensor features at each preset time point in the sensor feature sequence is calculated to obtain the sensor mean feature; secondly, the mean value of the vehicle diagnostic features at each preset time point in the vehicle diagnostic feature sequence is calculated to obtain the vehicle diagnostic mean feature; and thirdly, the mean value of the intelligent driving features at each preset time point in the intelligent driving feature sequence is calculated to obtain the intelligent driving mean feature. The sensor mean feature, vehicle diagnostic mean feature, and intelligent driving mean feature are then concatenated to obtain the first concatenated feature. Thus, through mean concatenation processing, basic multimodal features covering the vehicle's state, health status, and intelligent driving behavior can be obtained.

[0052] Then, using the first concatenated feature as the query and the driving time feature as the key and value, a multi-head attention mechanism is used to fuse the first concatenated feature and the driving time feature to obtain the first driving feature. This promotes the interaction and fusion between the large dimension of driving state and the small dimension of driving time, allowing the feature information of driving state and the feature information of driving time to mutually perceive the information related to driving behavior contained in each other. Simultaneously, using the first concatenated feature as the query and the driving score feature as the key and value, a multi-head attention mechanism is used to fuse the first concatenated feature and the driving score feature to obtain the second driving feature. This promotes the interaction and fusion between the large dimension of driving state and the small dimension of driving score, allowing the feature information of driving state and the feature information of driving score to mutually perceive the information related to driving behavior contained in each other. In this way, by learning the association weights between multimodal dynamic features (i.e., feature information from sensors, on-board diagnostics, and intelligent driving) and static features (i.e., feature information from driving time and driving score) through a multi-head attention mechanism, we can focus on the interactive information that is more critical to the driving behavior recognition task, reduce irrelevant noise interference, and make the driving behavior recognition results more accurate.

[0053] Finally, the first and second driving features are concatenated to obtain a one-dimensional driving feature. Thus, through cross-modal feature association and fusion, the interactive features of the driving state dimension and the trip index dimension can be effectively integrated, thereby capturing key one-dimensional features related to the driving state at a deeper level. This facilitates the identification and processing of driving behavior from both one-dimensional and two-dimensional information perspectives in subsequent steps.

[0054] In some embodiments, the aforementioned driving state features include sensor feature sequences, on-board diagnostic feature sequences, and intelligent driving feature sequences; the aforementioned trip index features include driving time features and driving score features; the aforementioned target driving features include two-dimensional driving images; in step S104, the driving state features and trip index features are fused to obtain the target driving features, including: The driving time feature and the driving score feature are concatenated to obtain the second concatenated feature; The second splicing feature and the sensor feature sequence are fused to obtain the first fused feature sequence; The second splicing feature and the vehicle diagnostic feature sequence are fused to obtain the second fused feature sequence. The second splicing feature and the intelligent driving feature sequence are fused to obtain the third fused feature sequence; Image processing is performed based on the first fusion feature sequence, the second fusion feature sequence, and the third fusion feature sequence to obtain a two-dimensional driving image.

[0055] In this embodiment, firstly, the driving time feature and the driving score feature are concatenated to obtain a second concatenated feature. Thus, feature information covering driving time and driving score can be obtained through concatenation. Then, the second concatenated feature is inserted after the sensor features at each preset time in the sensor feature sequence to obtain a first fused feature sequence; and secondly, the second concatenated feature is inserted after the on-board diagnostic features at each preset time in the on-board diagnostic feature sequence to obtain a second fused feature sequence; and thirdly, the second concatenated feature is inserted after the intelligent driving features at each preset time in the intelligent driving feature sequence to obtain a third fused feature sequence. Thus, by inserting static features (i.e., driving time and driving score feature information) after the dynamic features (i.e., sequence values ​​of sensors, on-board diagnostics, and intelligent driving) at each preset time, the feature information at a single moment can simultaneously contain both dynamic and static information, thereby improving the diversity of driving behavior feature information. Afterwards, based on the first fused feature sequence, the second fused feature sequence, and the third fused feature sequence, image processing is performed to obtain a two-dimensional driving image, which is used to indicate the actual driving situation. In this way, we can better uncover the feature information hidden in the data of various dimensions that is highly related to driving behavior, capture more real state expressions, and facilitate the recognition and processing of driving behavior from both one-dimensional and two-dimensional information perspectives in subsequent steps.

[0056] In some embodiments, the above-mentioned image processing based on the first fusion feature sequence, the second fusion feature sequence, and the third fusion feature sequence to obtain a two-dimensional driving image includes: A red channel map is generated based on the first fusion feature sequence; A green channel map is generated based on the second fusion feature sequence; A blue channel image is generated based on the third fusion feature sequence; Based on the red channel image, green channel image, and blue channel image, a two-dimensional driving image is obtained.

[0057] In this embodiment, a relative position matrix is ​​constructed based on the first fusion feature sequence, and a red channel image is generated using the relative position matrix. The specific implementation of this method is existing technology and will not be elaborated further. Similarly, a Gram summation angle field matrix is ​​constructed based on the second fusion feature sequence, and a green channel image is generated using the Gram summation angle field matrix. The specific implementation of this method is existing technology and will not be elaborated further. Furthermore, a recursive graph matrix is ​​constructed based on the third fusion feature sequence, and a blue channel image is generated using the recursive graph matrix. The specific implementation of this method is existing technology and will not be elaborated further. Subsequently, the red, green, and blue channel images obtained through different image processing methods are superimposed to form a three-channel RGB image. This RGB image is the two-dimensional driving image. By using three different types of image processing methods, the feature information in each fusion feature sequence can be preserved to the maximum extent while converting the image, ensuring the accuracy of the two-dimensional driving image. This facilitates the full utilization of the advantages of deep learning models in image processing in subsequent steps, effectively improving the recognition accuracy of driving behavior.

[0058] In some embodiments, the target driving features include one-dimensional driving features and two-dimensional driving images; in step S105, driving behavior recognition processing is performed using the target driving features to obtain driving behavior recognition results, including: The two-dimensional driving image is input into a pre-trained neural network model to obtain the first recognition result and the confidence level of the first recognition result; One-dimensional driving features are input into a pre-trained machine learning model to obtain a second recognition result and the confidence level of the second recognition result; Based on the confidence levels of the first and second identification results, decision processing is performed on the first and second identification results to obtain the driving behavior identification results.

[0059] In this embodiment, a two-dimensional driving image is input into a pre-trained neural network model to obtain a first recognition result and its confidence level. This fully leverages the advantages of deep learning models in image processing, resulting in more accurate driving behavior recognition. Simultaneously, one-dimensional driving features are input into a pre-trained machine learning model to obtain a second recognition result and its confidence level. This fully utilizes the machine learning model's focus on quantifying statistical patterns, leading to accurate driving behavior recognition. Subsequently, the recognition result with the highest confidence level is selected as the driving behavior recognition result from the first and second recognition results. Here, decision processing is performed from both visual pattern-based and numerical statistical perspectives of driving behavior to obtain the final driving behavior recognition result. This allows the first and second recognition results to compensate for each other's blind spots, thereby improving the accuracy of driving behavior recognition.

[0060] It should be understood that the neural network model is trained using multiple two-dimensional driving image samples and their label information, while the machine learning model is trained using multiple one-dimensional driving feature samples and their label information. In one example, during training, the model parameters are continuously adjusted using backpropagation and gradient descent optimization methods to minimize the error between the model's predictions and the actual labels. Simultaneously, cross-validation is employed, dividing the training data into multiple subsets. One subset is used as the validation set, while the remaining subsets are used as the training set, to evaluate the model's performance and prevent overfitting. Through multiple experiments and parameter adjustments, the optimal model parameters are found, improving the model's accuracy and generalization ability.

[0061] Optionally, the type of neural network model can be set according to the actual situation. For example, the neural network model can be a Convolutional Neural Network (CNN), a Transformer, etc., but is not limited to this. In addition, the machine learning model can be set according to the actual situation. For example, the machine learning model can be a Random Forest (RF), a Decision Tree, a Support Vector Machine (SVM), etc., but is not limited to this.

[0062] In addition, refer to Figure 2 This application provides a driving behavior recognition device, including: The acquisition module 201 is used to acquire the driving status data and trip index data of the target vehicle; The first processing module 202 is used to extract features from the driving status data to obtain driving status features; The second processing module 203 is used to extract features from the travel indicator data to obtain travel indicator features; The third processing module 204 is used to fuse the driving status features and the travel index features to obtain the target driving features; The fourth processing module 205 is used to perform driving behavior recognition processing using the target driving features to obtain driving behavior recognition results.

[0063] The content of the above method embodiments is applicable to the device embodiments. The specific functions implemented by the device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0064] Finally, refer to Figure 3 This application also provides a vehicle, which may include: At least one processor 301; At least one memory 302 is used to store at least one program; When at least one program is executed by at least one processor 301, the at least one processor 301 implements the driving behavior recognition method described above.

[0065] The aforementioned vehicles can be private cars, such as sedans, sport utility vehicles (SUVs), multi-purpose vehicles (MPVs), or pickup trucks, or commercial vehicles, such as vans, buses, small trucks, or large trailers, or gasoline vehicles or new energy vehicles such as hybrid or pure electric vehicles.

[0066] The aforementioned memory 302, as a non-transitory network system, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory 302 may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory 302 may optionally include memory 302 remotely located relative to processor 301, and these remote memory 302s can be connected to processor 301 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0067] The aforementioned memory 302 can be implemented as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 302 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 302 and is called and executed by the processor 301.

[0068] The processor 301 described above can be implemented using a general-purpose central processing unit (CPU), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application.

[0069] In some embodiments, the vehicle may further include: Input / output interfaces are used to implement information input and output; The communication interface is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). A bus transmits information between various components of a device, such as processor 301, memory 302, input / output interfaces, and communication interfaces. The processor 301, memory 302, input / output interface, and communication interface can communicate with each other within the device via a bus.

[0070] The content of the above method embodiments is applicable to this vehicle embodiment. The specific functions implemented in this vehicle embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0071] In summary, the embodiments of this application have the following technical effects: On the one hand, this application integrates multi-source data and uses machine learning algorithms and deep learning methods to deeply explore the complex relationships and potential patterns between data, which can more accurately characterize the driver's behavioral characteristics and significantly improve the recognition accuracy of driving behavior compared with existing technologies.

[0072] On the other hand, machine learning algorithms possess self-learning and adaptive capabilities, enabling them to learn the behavioral patterns of different drivers from a large amount of historical data and adapt to individual differences. Furthermore, when faced with various complex driving scenarios, this application can accurately identify normal and abnormal driving behaviors based on real-time data, providing more accurate and personalized analysis results.

[0073] On the other hand, in terms of application, this application can bring rich application value to multiple fields.

[0074] In terms of traffic safety, it can monitor and warn of potential dangerous driving behaviors in real time, reducing the risk of traffic accidents.

[0075] In vehicle insurance, Usage-Based Insurance (UBI) is implemented based on driving behavior recognition results. Personalized insurance rates are then developed based on the driver's actual driving risk, achieving scientific and fair insurance pricing. Specifically, key information such as safe driving scores and driving styles from the driving behavior recognition results is provided to insurance companies. Insurance companies use this data to assess the driver's risk level and develop personalized insurance rates. Drivers with high safe driving scores and stable driving styles receive lower insurance rate discounts; while drivers with low safe driving scores and more potentially dangerous driving behaviors have their insurance rates appropriately increased. This approach achieves a precise match between insurance pricing and the driver's actual driving risk, improving the fairness and efficiency of the insurance market.

[0076] In terms of personalized services, by analyzing drivers' habits and preferences, the system provides users with precise recommendations, such as personalized vehicle setting suggestions, driver training course recommendations, and value-added services related to vehicle use, thereby improving user experience and satisfaction. Specifically, based on driving behavior recognition results, personalized recommendations are provided to users. If the profile shows that a driver frequently drives long distances and is concerned about fuel consumption, the system can recommend energy-saving vehicle accessories, such as high-efficiency air filters and low rolling resistance tires. If the driver frequently uses intelligent driving functions, upgrade services or training courses related to intelligent driving can be recommended to help them better master and use intelligent driving technology. In addition, based on the driver's driving habits and preferences, suitable driving routes, tourist attractions, and nearby car service facilities (such as gas stations, charging stations, and repair shops) can be recommended.

[0077] In terms of functional integration and optimization, the results of driving behavior recognition are deeply integrated and optimized with other vehicle functions. For example, the vehicle's power output mode, suspension system settings, or parameters of intelligent driving assistance systems are automatically adjusted based on the driver's driving style. For drivers with an aggressive driving style, the power response speed is appropriately increased; for drivers with a mild driving style, the suspension system is optimized to provide a more comfortable driving experience. At the same time, by combining ADAS information and intelligent driving information, intelligent driving functions are optimized, such as adjusting the following distance of adaptive cruise control and the sensitivity of lane keeping assist according to driving scenarios and driver habits, further improving vehicle safety and comfort.

[0078] In one example, taking a new energy vehicle as an example, high-precision current and voltage sensors are installed in the vehicle's power system to monitor battery charging and discharging and obtain energy consumption data; acceleration sensors are installed in the chassis to detect rapid acceleration and deceleration; and the onboard computer records data such as speed and driving time. Simultaneously, the driver connects to the vehicle via a mobile application to obtain driving scores and other information. Vehicle-side data is transmitted to the onboard terminal via the CAN bus, and then sent to the cloud server via the 4G network using the MQTT protocol. Data from the mobile phone is also transmitted to the cloud via the corresponding protocol.

[0079] On the cloud server, the acquired data is cleaned to remove abnormally high values ​​caused by sensor malfunctions, and occasionally missing data is filled in using linear interpolation. Speed ​​and acceleration data are standardized. Features are extracted from various data types and fusion processing is performed to obtain target fused features, including one-dimensional driving features and two-dimensional driving images.

[0080] A random forest model was chosen for driving style classification. Extensive historical driving data, including multi-source data from different drivers in various scenarios and corresponding driving style labels, was used to train the model. After training, real-time collected one-dimensional driving features were input into the random forest model, which output that the driver's driving style was "mild" with a confidence level of 90%. A convolutional neural network was also chosen for driving style classification. Extensive historical driving data, including multi-source data from different drivers in various scenarios and corresponding driving style labels, was used to train the model. After training, real-time collected two-dimensional driving images were input into the convolutional neural network model, which output that the driver's driving style was "aggressive" with a confidence level of 80%. Ultimately, "mild" was selected as the driver's driving style. Based on this profile, the insurance company offered the driver a relatively low insurance premium; simultaneously, the system recommended comfort-related vehicle upgrade accessories, such as high-quality seat cushions.

[0081] In summary, this application achieves efficient collection, deep fusion, and accurate analysis of multi-source driving behavior data. Leveraging the powerful pattern recognition and data analysis capabilities of machine learning and deep learning algorithms, it uncovers hidden driving behavior patterns and regularities within the data, generating high-precision driving behavior recognition results. These profiles provide rich and valuable information for various applications such as UBI insurance, user recommendation, and vehicle function optimization, thereby achieving beneficial effects such as improved analysis accuracy, adaptation to complex scenarios and individual differences, and expansion of diverse applications.

[0082] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this application are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is changed and sub-operations described as part of a larger operation are executed independently.

[0083] Furthermore, although this application is described in the context of functional modules, it should be understood that, unless otherwise stated to the contrary, one or more of the functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding this application. Rather, given the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed herein, the actual implementation of the module will be understood within the scope of conventional technology for an engineer. Therefore, those skilled in the art can implement the application set forth in the claims using ordinary techniques without excessive experimentation. It is also understood that the specific concepts disclosed are merely illustrative and not intended to limit the scope of this application, which is determined by the full scope of the appended claims and their equivalents.

[0084] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several programs to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0085] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequential list of executable programs for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, a program execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can retrieve and execute a program from or in conjunction with such a program execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can mean any means that can contain, store, communicate, propagate, or transmit a program for use by or in conjunction with a program execution system, apparatus, or device.

[0086] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0087] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable program execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0088] In the foregoing description of this specification, the references to terms such as "one embodiment," "another embodiment," or "some embodiments," etc., indicate that a specific feature, structure, material, or characteristic described in connection with an embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0089] Although embodiments of this application have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of this application, the scope of which is defined by the claims and their equivalents.

[0090] The above is a detailed description of the preferred embodiments of this application, but this application is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application, and these equivalent modifications or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A method for recognizing driving behavior, characterized in that, Includes the following steps: Obtain the target vehicle's driving status data and trip index data; Feature extraction is performed on the driving status data to obtain driving status features; Feature extraction is performed on the travel indicator data to obtain travel indicator features; The driving status features and the trip index features are fused to obtain the target driving features; The driving behavior is identified by using the target driving features to obtain the driving behavior identification result.

2. The method according to claim 1, characterized in that, The driving status data includes sensor data, vehicle diagnostic data, and intelligent driving data at each preset time point; the driving status features include sensor feature sequences, vehicle diagnostic feature sequences, and intelligent driving feature sequences; the feature extraction of the driving status data to obtain the driving status features includes: Feature extraction is performed on the sensor data at each preset time to obtain the sensor feature sequence; Feature extraction is performed on the vehicle diagnostic data at each preset time to obtain the vehicle diagnostic feature sequence; Features are extracted from the intelligent driving data at each preset time point to obtain the intelligent driving feature sequence.

3. The method according to claim 1, characterized in that, The trip indicator data includes travel time data and travel score data; the trip indicator features include travel time features and travel score features; the step of extracting features from the trip indicator data to obtain trip indicator features includes: Feature extraction is performed on the driving time data to obtain the driving time features; Feature extraction is performed on the driving score data to obtain the driving score features.

4. The method according to claim 1, characterized in that, The driving status characteristics include sensor feature sequences, on-board diagnostic feature sequences, and intelligent driving feature sequences; the trip index characteristics include driving time characteristics and driving score characteristics. The target driving characteristics include one-dimensional driving characteristics; The step of fusing the driving state features and the trip index features to obtain the target driving features includes: The sensor feature sequence, the vehicle diagnostic feature sequence, and the intelligent driving feature sequence are averaged and spliced ​​to obtain the first spliced ​​feature; The first splicing feature and the driving time feature are fused together to obtain the first driving feature; The first splicing feature and the driving score feature are fused together to obtain the second driving feature; The first driving feature and the second driving feature are fused together to obtain the one-dimensional driving feature.

5. The method according to claim 1, characterized in that, The driving status characteristics include sensor feature sequences, on-board diagnostic feature sequences, and intelligent driving feature sequences; the trip index characteristics include driving time characteristics and driving score characteristics. The target driving features include two-dimensional driving images; The step of fusing the driving state features and the trip index features to obtain the target driving features includes: The driving time feature and the driving score feature are concatenated to obtain the second concatenated feature; The second splicing feature and the sensor feature sequence are fused to obtain the first fused feature sequence; The second splicing feature and the vehicle diagnostic feature sequence are fused together to obtain the second fused feature sequence. The second splicing feature and the intelligent driving feature sequence are fused to obtain a third fused feature sequence; The two-dimensional driving image is obtained by performing image processing based on the first fusion feature sequence, the second fusion feature sequence and the third fusion feature sequence.

6. The method according to claim 5, characterized in that, The image processing based on the first fused feature sequence, the second fused feature sequence, and the third fused feature sequence to obtain the two-dimensional driving image includes: A red channel map is generated based on the first fused feature sequence; A green channel map is generated based on the second fusion feature sequence; A blue channel image is generated based on the third fusion feature sequence; The two-dimensional driving image is obtained based on the red channel image, the green channel image, and the blue channel image.

7. The method according to claim 1, characterized in that, The target driving features include one-dimensional driving features and two-dimensional driving images; the driving behavior recognition processing using the target driving features to obtain driving behavior recognition results includes: The two-dimensional driving image is input into a pre-trained neural network model to obtain a first recognition result and the confidence level of the first recognition result; The one-dimensional driving features are input into a pre-trained machine learning model to obtain a second recognition result and the confidence level of the second recognition result; Based on the confidence levels of the first and second identification results, decision processing is performed on the first and second identification results to obtain the driving behavior identification result.

8. The method according to claim 1, characterized in that, After acquiring the target vehicle's driving status data and trip index data, the method further includes: The driving status data and the trip index data are cleaned. The driving status data and the trip index data are standardized.

9. A driving behavior recognition device, characterized in that, include: The acquisition module is used to acquire the driving status data and trip index data of the target vehicle; The first processing module is used to extract features from the driving status data to obtain driving status features; The second processing module is used to extract features from the travel index data to obtain travel index features; The third processing module is used to fuse the driving status features and the travel index features to obtain the target driving features; The fourth processing module is used to perform driving behavior recognition processing using the target driving features to obtain driving behavior recognition results.

10. A vehicle, characterized in that, include: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements a driving behavior recognition method as described in any one of claims 1-8.

Images