Driving behavior analysis method and device, automobile, storage medium and program product

Abstract

The invention provides a driving behavior analysis method and device, an automobile, a storage medium and a program product, and relates to the technical field of behavior analysis. The method comprises the following steps: acquiring driving data of an automobile; the driving data is obtained by carrying out data acquisition on the driving process of the automobile; determining driving behavior data from the driving data, and inputting the driving behavior data into a driving behavior classification model to obtain a driving behavior classification result output by the driving behavior classification model; the driving behavior data is used for analyzing driving behaviors of the driver; based on the driving behavior classification result, determining context data related to the driving behavior classification result from the driving data, and determining an occurrence reason of the driving behavior classification result; and determining a driving decision for the driving behavior classification result based on the occurrence reason. According to the invention, the background reason of the corresponding driving behavior is fully considered, so that the driving decision accuracy is improved, and finally the driving safety and the driving experience are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of behavior analysis, and in particular to a driving behavior analysis method, device, automobile, storage medium and program product. Background Art

[0002] With the rapid development of science and technology, people have higher and higher requirements for cars, so it is necessary to improve the level of automation and intelligence of cars. Among them, through accurate analysis of driving behavior and intelligent decision-making, the safety and driving experience of the vehicle can be enhanced. Therefore, it is necessary to accurately analyze driving behavior to facilitate accurate decision-making.

[0003] At present, a single driving data is input into the driving behavior analysis model to obtain the driving behavior analysis results output by the driving behavior analysis model, and the driving behavior analysis model is continuously improved to improve the accuracy of driving behavior analysis. However, due to the complexity of the actual driving environment and the actual driving scenes, driving decisions are made directly based on the driving behavior analysis results, that is, only knowing the driving behavior but not being able to make accurate driving decisions, which may lead to wrong decisions, thereby affecting driving safety and driving experience. Summary of the invention

[0004] The present invention provides a driving behavior analysis method, device, automobile, storage medium and program product, which are used to solve the defect of low driving decision accuracy in the prior art and realize high-accuracy driving decision.

[0005] The present invention provides a driving behavior analysis method, comprising: Acquiring driving data of the vehicle; the driving data is obtained by collecting data of the driving process of the vehicle; Determining driving behavior data from the driving data, inputting the driving behavior data into a driving behavior classification model, and obtaining a driving behavior classification result output by the driving behavior classification model; the driving behavior data is used to analyze the driving behavior of the driver; Determining a cause of the driving behavior classification result based on the driving behavior classification result and context data related to the driving behavior classification result determined from the driving data; Based on the occurrence cause, determining a driving decision for the driving behavior classification result; Wherein, the driving behavior classification model is obtained by training based on sample driving behavior data.

[0006] According to a driving behavior analysis method provided by the present invention, the determining the cause of the driving behavior classification result based on the driving behavior classification result and determining context data related to the driving behavior classification result from the driving data includes: Performing semantic analysis on the driving behavior classification result to obtain semantic information of the driving behavior classification result; Based on the semantic information and context data related to the semantic information determined from the driving data, a reason for the driving behavior classification result is determined.

[0007] According to a driving behavior analysis method provided by the present invention, determining a driving decision for the driving behavior classification result based on the occurrence cause includes: Based on the occurrence cause and the semantic information, a driving decision for the driving behavior classification result is determined.

[0008] According to a driving behavior analysis method provided by the present invention, the driving behavior data includes time series data, the context data includes time series data, and the driving behavior classification model is constructed based on a long short-term memory network LSTM feature extraction layer.

[0009] According to a driving behavior analysis method provided by the present invention, the driving behavior classification model is obtained based on the following method: Based on the sample driving behavior data, the training model is trained to obtain an initial training model; Determining the driving behavior classification accuracy of the initial training model in a preset driving scenario; Based on the driving behavior classification accuracy, the initial training model is optimized to obtain the driving behavior classification model.

[0010] According to a driving behavior analysis method provided by the present invention, after determining a driving decision for the driving behavior classification result based on the occurrence cause, the method further includes: Based on the driving decision, the automobile is controlled to execute an operation corresponding to the driving decision and / or output prompt information.

[0011] The present invention also provides a driving behavior analysis device, comprising: A data acquisition module is used to acquire driving data of the vehicle; the driving data is acquired by collecting data during the driving process of the vehicle; a behavior classification module, used to determine driving behavior data from the driving data, input the driving behavior data into a driving behavior classification model, and obtain a driving behavior classification result output by the driving behavior classification model; the driving behavior data is used to analyze the driving behavior of the driver; a cause determination module, configured to determine a cause of the driving behavior classification result based on the driving behavior classification result and context data related to the driving behavior classification result determined from the driving data; A decision determination module, used for determining a driving decision for the driving behavior classification result based on the occurrence cause; Wherein, the driving behavior classification model is obtained by training based on sample driving behavior data.

[0012] The present invention also provides an automobile, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, any one of the driving behavior analysis methods described above is implemented.

[0013] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the driving behavior analysis method as described in any one of the above is implemented.

[0014] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the driving behavior analysis method as described above is implemented.

[0015] The driving behavior analysis method, device, automobile, storage medium and program product provided by the present invention obtain driving data of the automobile, and the driving data is obtained by collecting data of the driving process of the automobile, so as to determine the driving behavior data from the driving data, and input the driving behavior data into a driving behavior classification model, accurately obtain the driving behavior classification result output by the driving behavior classification model, and accurately determine the cause of the driving behavior classification result based on the driving behavior classification result and context data related to the driving behavior classification result determined from the driving data, so as to accurately determine the driving decision for the driving behavior classification result based on the cause, that is, not only the driving behavior of the driver is analyzed, but also the cause of the driving behavior is analyzed, so as to fully consider the background cause of the corresponding driving behavior, and fully understand the cause of the driver's driving decision, and then comprehensively generate more accurate driving decisions, improve the accuracy of driving decisions, and ultimately improve driving safety and driving experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 This is one of the flow charts of the driving behavior analysis method provided by the present invention.

[0018] Figure 2 This is the second flow chart of the driving behavior analysis method provided by the present invention.

[0019] Figure 3 It is a structural schematic diagram of the driving behavior analysis device provided by the present invention.

[0020] Figure 4 It is a structural schematic diagram of the automobile provided by the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] With the rapid development of science and technology, people have higher and higher requirements for cars, so it is necessary to improve the level of automation and intelligence of cars. Among them, through accurate analysis of driving behavior and intelligent decision-making, the safety and driving experience of the vehicle can be enhanced. Therefore, it is necessary to accurately analyze driving behavior to facilitate accurate decision-making.

[0023] Traditional rule-based driving behavior analysis methods usually rely on predefined rules and thresholds. For example, specific speed, acceleration or braking force thresholds are set to determine whether driving behavior is abnormal. These rules are usually determined based on experience or standard driving behavior patterns, but they are difficult to adapt to complex or dynamically changing driving environments, that is, the rules need to be constantly updated and have poor flexibility.

[0024] At present, a single driving data is input into the driving behavior analysis model to obtain the driving behavior analysis results output by the driving behavior analysis model, and the driving behavior analysis model is continuously improved to improve the accuracy of driving behavior analysis. However, due to the complexity of the actual driving environment and the actual driving scenes, driving decisions are made directly based on the driving behavior analysis results, that is, only knowing the driving behavior but not being able to make accurate driving decisions, which may lead to wrong decisions, thereby affecting driving safety and driving experience.

[0025] In view of the above problems, the present invention proposes the following embodiments. Figure 1-Figure 2 The driving behavior analysis method of the present invention is described.

[0026] Figure 1 is one of the flow charts of the driving behavior analysis method provided by the present invention, such as Figure 1 As shown, the driving behavior analysis method includes the following steps 110, 120, 130 and 140.

[0027] Step 110, obtaining driving data of the car.

[0028] The driving data is obtained by collecting data about the driving process of the vehicle, and may include but is not limited to at least one of the following: vehicle travel data, vehicle external environment data, driver status monitoring data, driver operation data, and the like.

[0029] The driving data of the car may include but is not limited to at least one of the following: the driving speed, acceleration, braking information and steering information of the vehicle, etc. In one embodiment, the driving data of the car is collected by an onboard sensor of the car.

[0030] The external environment data of the car may include but is not limited to at least one of the following: lane marking data, obstacle data, traffic sign data, weather data, etc. In one embodiment, the external environment data of the car is collected by a camera of the car, that is, the camera provides visual data of the vehicle's surroundings. Of course, weather data can also be obtained through weather services.

[0031] Furthermore, lane change data may be determined based on the vehicle's external environment data, so that the driving data includes the lane change data. The lane change data may include but is not limited to at least one of the following: the vehicle's driving trajectory and lane keeping status, etc.

[0032] Furthermore, the surrounding environment information can be determined based on the vehicle's external environment data, so that the driving data includes the surrounding environment information. The surrounding environment information may include but is not limited to at least one of the following: distance information of the obstacle ahead, traffic signal status, weather conditions, and the like.

[0033] The driver's state monitoring data may include but is not limited to at least one of the following: heart rate data, eye movement data, attention data, etc. In one embodiment, the driver's state monitoring data is obtained by monitoring a driver's physiological state monitoring device.

[0034] The driver's operation data may include but is not limited to at least one of the following: steering angle information of the driver, and usage of the accelerator pedal and the brake pedal, etc. In one embodiment, the driver's operation data is collected by an onboard sensor of the vehicle.

[0035] Furthermore, the collected driving data is preprocessed to ensure the accuracy of the subsequent driving behavior analysis based on the driving data. The data preprocessing may include but is not limited to at least one of the following: denoising, normalization, data completion, and outlier processing. For example, denoising may include signal filtering and smoothing.

[0036] For example, the collected raw driving data is denoised to reduce the impact of sensor errors and environmental interference on the data, improve the accuracy of driving data, and then improve the accuracy of driving behavior analysis, and ultimately improve the accuracy of driving decisions. Secondly, the data is normalized to convert data from different sources to a unified scale for subsequent analysis and modeling. That is, normalization helps to eliminate dimensional differences and improve data consistency. In addition, data completion and outlier processing are required to ensure the integrity and accuracy of driving data.

[0037] It should be understood that data collection of the driving process of a car is the basis of driving behavior analysis, and subsequent driving behavior analysis needs to be conducted based on actual data.

[0038] Further, the driving data includes time series data, thereby improving the accuracy of subsequent driving behavior analysis. For example, the driving data includes data within a preset time period, and the last moment of the preset time period is the current moment, and the duration of the preset time period is the preset duration.

[0039] Step 120 , determining driving behavior data from the driving data, inputting the driving behavior data into a driving behavior classification model, and obtaining a driving behavior classification result output by the driving behavior classification model.

[0040] Wherein, the driving behavior data is used to analyze the driving behavior of the driver. It should be understood that not all data in the driving data can be used to analyze the driving behavior of the driver. Based on this, the driving behavior data used to analyze the driving behavior of the driver is determined from the driving data. The driving behavior data can be data directly included in the driving data, or data obtained by analyzing the data in the driving data, that is, feature extraction is performed on the driving data to screen and construct feature data for driving behavior classification from the driving data, that is, information useful for driving behavior classification is extracted from the driving data.

[0041] In one embodiment, the driving behavior data is determined from the driving data based on a preset rule. In another embodiment, the driving data is input into a feature extraction model to obtain the driving behavior data output by the feature extraction model, where the feature extraction model is trained based on the sample driving data and its corresponding driving behavior data labels.

[0042] For example, driving behavior data includes the driver's reaction time, which is used to indicate the delay time from the driver's response to the appearance of an obstacle. For example, driving behavior data includes the strength and frequency of the acceleration and brake pedals. These operation intensity data reflect the driver's driving habits and coping strategies, which facilitates the classification of driving behavior. For example, driving behavior data includes the driver's operation mode data, which includes the change trend data of the steering angle and the fluctuation data of the vehicle speed, etc., which facilitates the classification of driving behavior.

[0043] Wherein, the driving behavior classification model is trained based on sample driving behavior data. Further, the driving behavior classification model is trained based on sample driving behavior data and driving behavior classification result labels corresponding to the sample driving behavior data. The driving behavior classification model is used to perform driving behavior analysis on the driving behavior data, that is, to perform driving behavior classification to obtain a driving behavior classification result. The driving behavior classification model may include a feature extraction layer and a classification layer connected in sequence.

[0044] Considering that the rule-based driving behavior analysis method may perform poorly when dealing with multivariate and nonlinear driving behaviors and has low flexibility; and using the decision tree model to classify driving behaviors, classifying data through a series of decision nodes, it is suitable for processing structured data. However, when faced with time series data and complex driving behaviors, decision trees may not be able to capture long-term dependencies and dynamic changes, and their performance may also be affected by the amount of training data and data quality; and the driving behavior analysis method based on convolutional neural networks, which perform well in image processing and visual data analysis. Some intelligent car systems use convolutional neural networks. The network processes the visual data from the camera to identify lane markings, traffic signs, obstacles, etc., but it has limitations in processing time series data and contextual analysis of driving behavior patterns. It mainly focuses on spatial features and has limited ability to capture dynamic changes in time series; and the driving behavior analysis method based on ensemble learning improves the accuracy and robustness of driving behavior analysis by combining the prediction results of multiple models. For example, multiple classifiers are integrated to classify driving behavior. Although this method can improve accuracy, when processing time series data and dynamic driving behavior, the model complexity and computational overhead are large, and real-time performance may be affected. In other words, the rule-based driving behavior analysis method has poor flexibility and is difficult to adapt to complex and dynamically changing driving environments. The rule setting relies on experience, lacks self-learning capabilities, and has limited processing capabilities for abnormal driving behaviors. The decision tree model used to classify driving behaviors performs poorly when processing time series data and cannot effectively capture long-term dependencies and dynamic changes. The decision tree may be too simple to handle high-dimensional data and complex driving behavior patterns. Convolutional neural networks are mainly used for static image data and cannot effectively process dynamic information in time series. They have limited contextual analysis capabilities for driving behavior patterns and cannot achieve a deep understanding of time series data. Ensemble learning has high computational overhead, complex models, and may limit real-time performance. When processing time series data, the computational complexity and response time of the ensemble model may affect the real-time performance of the system.

[0045] Based on this, driving behavior data includes time series data, and the driving behavior classification model is built based on the LSTM (Long Short Term Memory) feature extraction layer. Among them, LSTM performs well in processing time series data, can capture long-term dependencies and complex dynamic changes in driving behavior, and improve the ability to understand driver behavior. That is, based on this, the driver's operating mode and the changes of these modes over time can be learned, and then the driving behavior can be more accurately predicted and classified, thereby improving the accuracy of driving behavior classification and the accuracy of driving decisions.

[0046] For ease of understanding, for example, the driving behavior classification results include "normal driving", "sudden braking" or "quick turning" and so on.

[0047] It should be understood that driving behavior data is determined from driving data, and driving data is obtained by collecting data on the driving process of the car, so that the data sources of driving behavior data are diverse, and then driving behavior classification is performed based on multiple data sources to improve the accuracy of driving behavior classification, thereby improving the accuracy of driving decisions.

[0048] Step 130 : determining a cause of the driving behavior classification result based on the driving behavior classification result and context data related to the driving behavior classification result determined from the driving data.

[0049] Here, the context data may be directly data in the driving data, or may be data obtained by further analyzing the data in the driving data.

[0050] Here, the cause of occurrence is the reason for the occurrence of the driving behavior corresponding to the driving behavior classification result, that is, understanding the background in which the driving behavior occurs. In other words, by combining driving behavior with contextual information, the driver's driving decision can be more accurately explained and the reasons behind his behavior can be understood. Therefore, based on the contextual data related to the driving behavior classification result, the cause of the driving behavior classification result can be accurately determined, thereby improving the accuracy of determining subsequent driving decisions.

[0051] For example, context data includes vehicle environment data, which includes front obstacle information, traffic signal information, road condition information, etc., which can be used to explain driving behavior.

[0052] For example, assuming that the driving behavior classification result is "sudden braking", the context data related to "sudden braking" (including information about obstacles ahead and traffic signals) is determined from the driving data, so that context analysis can determine whether the "sudden braking" is caused by an obstacle suddenly appearing ahead or the traffic light turning red; for example, the obstacle ahead is close to the vehicle, and the driver's braking operation is also sudden, which indicates that an emergency situation that requires a quick response is encountered. Based on this, it can be determined that the cause of the "sudden braking" is "an obstacle appears ahead, causing emergency braking."

[0053] Exemplarily, assuming that the driving behavior classification result is "quick turn", the context data related to the "quick turn" (including obstacle information, front vehicle information, driving scene information and traffic condition information) is determined from the driving data, so as to determine through context analysis whether the "quick turn" is caused by an obstacle suddenly appearing in front or a car that suddenly brakes in front. For example, in a driving scenario on a highway, the traffic conditions around the vehicle are good, which may be a normal high-speed driving demand rather than dangerous driving. Based on this, it can be determined that the cause of the "quick turn" is "high-speed driving scenario, good surrounding traffic conditions, leading to a quick turn".

[0054] Exemplarily, assuming that the context data related to the driving behavior classification result includes vehicle environment data and driving parameter data, the cause of the driving behavior classification result is determined based on the driving behavior classification result, the vehicle environment data and the driving parameter data. The vehicle environment data may include information about obstacles ahead, traffic signal information and road condition information, etc. The obstacle ahead information may include information about the distance of obstacles from the vehicle and obstacle type (stationary or moving), etc. The traffic signal information may include information about the state of red lights and traffic signs, etc. The road condition information may include lane marking information, road surface condition information and weather condition information, etc. The driving parameter data may include vehicle speed information, operation behavior data and driver status information, etc. The vehicle speed information may include the current speed of the vehicle, the acceleration state of the vehicle and the deceleration state of the vehicle, etc. The operation behavior data may include steering angle, the use of the accelerator pedal and the use of the brake pedal, etc. The driver status information may include heart rate, eye movement information and attention status, etc. Based on this, in the context analysis process, assuming that the driving behavior classification result is "sudden braking", the context data related to "sudden braking" (including front obstacle information and traffic signal information) is determined from the driving data, so that the context analysis determines whether the "sudden braking" is caused by the sudden approach of an obstacle in front or the traffic light turning red. For example, the obstacle in front is close to the vehicle, and the vehicle speed and acceleration and deceleration at the time of emergency braking are determined based on the driving parameters. It is determined that the driver's braking operation is also very sudden, which indicates that an emergency situation that requires a quick response is encountered. That is, through environmental detection, it can be determined whether there is an emergency situation that causes emergency braking. Based on this, it can be determined that the cause of the "sudden braking" is "an obstacle appears in front, causing emergency braking." It should be understood that driving parameters (such as braking force, speed changes, etc.) can also help to further judge. If the vehicle suddenly brakes at a higher speed, this may mean that the driver overreacts to the emergency. Based on this, a driving decision is made to suggest the driver to adjust the braking method to avoid unnecessary emergency braking. If the braking operation is too abrupt and the speed drops rapidly, it can be inferred that the driver may not be able to identify environmental changes in time. Driving decisions are made to prompt the driver to maintain a reasonable speed or suggest appropriate driving strategies (such as slowing down in advance); in context analysis, driving parameters help understand the driver's reaction method, judge whether the behavior is excessive or correct, and generate optimized driving decisions based on this; in other words, when analyzing the "sudden braking" driving behavior, it is necessary not only to see whether there are obstacles, but also to use the vehicle's braking force, speed and other driving parameters to determine whether it is necessary to remind the driver to adjust the behavior or recommend slowing down.

[0055] In one embodiment, the driving behavior classification results and context data are input into a cause determination model to obtain the cause of occurrence output by the cause determination model. The cause determination model is trained based on the sample driving behavior classification results, sample context data, and the cause labels corresponding to the two sample data. That is, an artificial intelligence model is used to automatically determine the cause of the behavior. A causal analysis model can be constructed, and the driving behavior classification results and related context data are input into the model. The model learns the relationship between driving behavior and environmental factors, and outputs the most likely cause of the behavior. In addition, the cause determination model can be a classification model, a probabilistic graph model, or a multimodal model based on a Transformer. This embodiment is suitable for complex and dynamic scenarios.

[0056] In another embodiment, a set of rules are predefined to determine the cause of the driving behavior classification result based on the driving behavior classification result and the context data. For example, if the driving behavior classification result is "sudden braking", the rule will check the context data such as the distance to the obstacle ahead and the status of the traffic signal to determine whether the behavior is related to the emergency. Rule-based context analysis is usually simple and highly interpretable, suitable for some clear scene judgments. In addition, the two methods of determining the cause can be used in combination to combine the advantages of both to improve the accuracy and robustness of the context analysis.

[0057] Step 140: Determine a driving decision for the driving behavior classification result based on the occurrence cause.

[0058] Specifically, based on the driving behavior classification results and the causes thereof, a driving decision for the driving behavior classification results is determined. Thus, in addition to making a driving decision based on the driving behavior, a driving decision is also made based on the causes of the driving behavior, thereby improving the accuracy of the driving decision.

[0059] Furthermore, after the above step 140, the method further includes: based on the driving decision, controlling the car to perform the operation corresponding to the driving decision and / or outputting prompt information. That is, the driving decision includes decision information for automatically controlling the car to perform the operation, and decision information for prompting the driver how to drive. Based on this, driving safety and driving experience are improved.

[0060] For example, assuming that the driving behavior classification result is "sudden braking", the cause of "sudden braking" is "an obstacle appears in front, causing emergency braking". Based on this, it is determined that the driving decision "requires rapid adjustment of vehicle speed" and / or "the driver is advised to take evasive action". If the driving decision is "the driver is advised to take evasive action", a prompt message can be output to prompt the driver to change lanes or slow down.

[0061] For example, assuming that the driving behavior classification result is "quick turn", the reason for the "quick turn" is "complex road conditions, resulting in quick turn", based on this, the driving decision is determined to "suggest the driver to slow down", and then a prompt message can be output to prompt the driver to increase attention to the surrounding environment.

[0062] For example, assuming that the driving behavior classification result is "normal driving" and the reason for "normal driving" is "normal situation", based on this, the driving decision is determined to be "maintain the current driving strategy and continue to monitor the surrounding environment", and then the car can be controlled to perform the operation corresponding to the driving decision.

[0063] Exemplarily, assuming that the contextual data related to the driving behavior classification result includes vehicle environment data and driving parameter data, the cause of the driving behavior classification result is determined based on the driving behavior classification result, the vehicle environment data and the driving parameter data, and the cause includes the driving scenario (such as urban roads, rural roads, expressways, etc.), so that the driving decision is adjusted based on the real-time monitored driving data, that is, the cause is obtained by performing contextual analysis based on the vehicle environment data and the driving parameter data to generate the optimal driving decision.

[0064] In one embodiment, the driving behavior classification results and the reasons for their occurrence are input into the driving decision model to obtain the driving decision output by the driving decision model. The driving decision model is trained based on the sample driving behavior classification results and the reasons for the sample occurrence. Based on this, a more flexible decision-making scheme is automatically generated through the artificial intelligence model, which is more suitable for complex scenarios. The model can integrate multiple information and output personalized optimal driving decisions.

[0065] In another embodiment, a set of driving decision rules are predefined to determine the driving decision of the driving behavior classification result based on the cause. For example, a corresponding response plan is directly given according to the cause. For example, when it is recognized that the sudden braking is caused by the approach of an obstacle, the driving decision of "slowing down and reminding the driver to avoid" can be triggered based on the driving decision rules. In addition, the two methods of determining the driving decision can be used in combination to improve the accuracy and robustness of the driving decision by combining the advantages of both.

[0066] Furthermore, whether it is a driving decision based on an artificial intelligence model or a driving decision based on rules, it has the ability to self-learn and adapt. By continuously collecting and analyzing driving data and optimizing driving decision strategies, the accuracy of driving decisions can be further improved, thereby improving the intelligence level of the driving decision system and driving safety.

[0067] Furthermore, based on the semantic information of the cause and the driving behavior classification result, a driving decision for the driving behavior classification result is determined. Based on this, a driving decision is accurately made based on the semantic information with stronger representation ability.

[0068] Furthermore, in the process of generating driving decisions, more data sources or driving strategies can be combined to determine driving decisions for the driving behavior classification results. The driving strategies can include speed adjustment strategies, lane keeping strategies, and traffic rule compliance strategies.

[0069] It should be understood that based on the driving behavior classification results, and determining the context data related to the driving behavior classification results from the driving data, determining the cause of the driving behavior classification results, the vehicle status and environmental factors can be analyzed in real time, so that before making a driving decision, multiple factors can be comprehensively considered to generate the best driving decision. And the above driving data are all real-time data, so that accurate and reasonable driving decisions can be generated in a timely manner, ensuring that the driver can make appropriate responses in a complex traffic environment, and improving overall driving safety and efficiency.

[0070] The driving behavior analysis method provided by the embodiment of the present invention obtains driving data of a car, and the driving data is obtained by collecting data of the driving process of the car, so as to determine the driving behavior data from the driving data, and input the driving behavior data into a driving behavior classification model, accurately obtain the driving behavior classification result output by the driving behavior classification model, and accurately determine the cause of the driving behavior classification result based on the driving behavior classification result and context data related to the driving behavior classification result determined from the driving data, so as to accurately determine the driving decision for the driving behavior classification result based on the cause, that is, not only the driving behavior of the driver is analyzed, but also the cause of the driving behavior is analyzed, so as to fully consider the background cause of the corresponding driving behavior, and fully understand the cause of the driver's driving decision, and then comprehensively generate more accurate driving decisions, improve the accuracy of driving decisions, and ultimately improve driving safety and driving experience.

[0071] Based on any of the above embodiments, in the method, the above step 130 includes: Performing semantic analysis on the driving behavior classification result to obtain semantic information of the driving behavior classification result; Based on the semantic information and context data related to the semantic information determined from the driving data, a reason for the driving behavior classification result is determined.

[0072] In a specific embodiment, semantic analysis is performed on the driving behavior classification results, that is, semantic mapping is performed on the driving behavior classification results to convert the driving behavior classification results into a more specific semantic level, that is, further semantic analysis is performed into semantics that have practical significance for subsequent analysis, thereby facilitating the subsequent accurate determination of the cause of the driving behavior classification results, so as to further improve the accuracy of driving decisions.

[0073] For example, if the driving behavior classification result is "sudden braking", the corresponding semantic information includes "emergency situation", indicating that the vehicle may have encountered an emergency situation that requires a quick response. If the driving behavior classification result is "fast turning", the corresponding semantic information includes "high-speed driving" or "challenging driving", indicating that the driver is operating at a high speed. If the driving behavior classification result is "normal driving", the corresponding semantic information includes "normal driving state", indicating that there are no special environmental factors affecting the vehicle operation.

[0074] In one embodiment, the semantic information of the driving behavior classification result and the context data are input into the cause determination model to obtain the occurrence cause output by the cause determination model. The cause determination model is trained based on the semantic information of the sample driving behavior classification result, the sample context data and the occurrence cause labels corresponding to the two sample data.

[0075] In another embodiment, a set of rules is predefined to determine the cause of the driving behavior classification result based on the semantic information of the driving behavior classification result and the context data.

[0076] It should be understood that semantic analysis of driving behavior classification results not only provides a basic semantic description of the driving behavior classification results, but also provides more specific semantic interpretation information (background information) for subsequent context analysis to determine the cause, thereby improving the accuracy of determining the cause of the driving behavior; and based on the semantic information, context data related to the semantic information can be determined from the driving data, and the required context data can be more accurately determined based on semantic information with stronger representation capabilities, thereby improving the accuracy of determining the cause of the driving behavior, and further improving the accuracy of driving decisions.

[0077] The driving behavior analysis method provided in the embodiment of the present invention performs semantic analysis on the driving behavior classification results to obtain semantic information of the driving behavior classification results, so as to provide more specific semantic interpretation information of the driving behavior classification results, and then more accurately determine the cause of the driving behavior classification results based on the semantic information with stronger representation ability and context data related to the semantic information, thereby further improving the accuracy of driving decisions.

[0078] Based on any of the above embodiments, in the method, the above step 140 includes: Based on the occurrence cause and the semantic information, a driving decision for the driving behavior classification result is determined.

[0079] It should be understood that semantic analysis of driving behavior classification results not only provides a basic semantic description of the driving behavior classification results, but also provides more specific semantic interpretation information (background information) for subsequent driving decisions, thereby improving the accuracy of driving decisions.

[0080] In one embodiment, the semantic information of the driving behavior classification results and the reasons for their occurrence are input into a driving decision model to obtain a driving decision output by the driving decision model. The driving decision model is trained based on the semantic information of the sample driving behavior classification results and the reasons for the sample occurrence.

[0081] In another embodiment, a set of driving decision rules is predefined, and the driving decision of the driving behavior classification result is determined based on the occurrence cause and the semantic information of the driving behavior classification result.

[0082] The driving behavior analysis method provided in the embodiment of the present invention performs semantic analysis on the driving behavior classification results to obtain semantic information of the driving behavior classification results, so as to provide more specific semantic interpretation information of the driving behavior classification results, and then accurately determine the driving decision for the driving behavior classification results based on the occurrence cause and the semantic information with stronger representation ability, so as to further improve the accuracy of the driving decision.

[0083] Based on any of the above embodiments, in the method, the driving behavior data includes time series data, the context data includes time series data, and the driving behavior classification model is constructed based on a long short-term memory network LSTM feature extraction layer.

[0084] It should be noted that driving behavior data includes time series data, and of course it can also include non-time series data, and the driving behavior classification model is constructed based on the LSTM feature extraction layer, so that it can capture the long-term dependencies and complex dynamic changes in driving behavior, and improve the ability to understand driver behavior. That is, based on this, the driver's operating mode and the changes of these modes over time can be learned, and then the driving behavior can be more accurately predicted and classified, thereby improving the accuracy of driving behavior classification. In other words, each driving behavior category has its own unique time series characteristics, and the LSTM feature extraction layer can effectively extract these time series features to accurately distinguish each driving behavior category.

[0085] In a specific embodiment, the driving behavior classification model includes an LSTM feature extraction layer and a classification layer connected in sequence. The classification layer accurately distinguishes each driving behavior category based on the time series features extracted by the LSTM feature extraction layer to obtain an accurate driving behavior classification result.

[0086] In the driving behavior analysis method provided by the embodiment of the present invention, the driving behavior data includes time series data, and the driving behavior classification model is constructed based on the LSTM feature extraction layer, so that the long-term dependencies and complex dynamic changes in the driving behavior can be captured, and the ability to understand the driver's behavior can be improved, so as to more accurately predict and classify the driving behavior, thereby improving the accuracy of driving behavior classification, and further improving the accuracy of driving decisions; at the same time, the context data also includes time series data, so as to more accurately determine the cause of the driving behavior classification result, and the driving behavior classification result is also determined based on the time series data, so that the context data based on the time series data can more accurately determine the cause of the driving behavior classification result, and finally further improve the accuracy of driving decisions.

[0087] Based on any of the above embodiments, Figure 2 This is the second flow chart of the driving behavior analysis method provided by the present invention, such as Figure 2 As shown, the driving behavior classification model is obtained based on the following method.

[0088] Step 210: Based on the sample driving behavior data, the model to be trained is trained to obtain an initial training model.

[0089] Specifically, an initial training model for driving behavior classification is obtained by preliminary training based on the sample driving behavior data. Further, based on the sample driving behavior data and its corresponding driving behavior classification result label, the to-be-trained model is trained to obtain an initial training model.

[0090] Step 220, determining the driving behavior classification accuracy of the initial training model in a preset driving scenario.

[0091] Here, the preset driving scenario is a pre-set driving scenario, which is a specific driving scenario, such as various dangerous driving scenarios (such as high-speed turning scenarios and driving scenarios in rainy and snowy days, etc.), so that the model is further optimized through error analysis in specific driving scenarios to ensure that the model can make correct driving decisions in various preset driving scenarios. In other words, considering that it is particularly necessary to make correct driving decisions for specific driving scenarios, such as dangerous driving scenarios, the initial training model is further optimized based on the accuracy of the driving behavior classification of the initial training model in the preset driving scenarios. In addition, the number of the preset driving scenarios can be multiple, so as to ensure the accuracy of driving decisions in multiple specific driving scenarios.

[0092] In a specific embodiment, the accuracy of driving behavior classification can be evaluated through error analysis to determine the performance differences of the model in different driving scenarios by comparing the model prediction results and the actual driving data, so as to better optimize the model for each specific driving scenario and improve the decision-making accuracy for each specific driving scenario.

[0093] Step 230: Based on the driving behavior classification accuracy, optimize the initial training model to obtain the driving behavior classification model.

[0094] Here, the model optimization method may include but is not limited to: adjusting the hyperparameters of the model (such as learning rate, number of hidden layers and number of units, etc.) and adjusting the model parameters, etc.

[0095] In a specific embodiment, it is determined that the driving behavior classification accuracy is less than a preset accuracy threshold, and the initial training model is optimized. For example, if the prediction accuracy of the model is low when turning at high speed, the hyperparameters of the initial training model can be adjusted to optimize the performance of the model. Further, cross-validation and hyperparameter tuning are performed to select the optimal model configuration to improve the classification accuracy of the driving behavior classification model.

[0096] Furthermore, the sample driving behavior data may be optimized.

[0097] In one embodiment, the diversity of sample data is enhanced, and sample driving behavior data in different driving scenarios is added to improve the adaptability of the model to various driving scenarios. Furthermore, each driving scenario is a specific preset driving scenario, so as to improve the driving decision accuracy of the model for a specific driving scenario, such as a dangerous driving scenario.

[0098] In one embodiment, data preprocessing is performed on the collected sample driving behavior data, such as denoising and normalization, to improve the model training effect and ultimately improve the accuracy and robustness of the model.

[0099] In one embodiment, data enhancement is used to generate more training data and increase the generalization ability of the model.

[0100] Furthermore, the optimized initial training model is verified, and the model is further optimized based on the verification results to obtain a more accurate driving behavior classification model.

[0101] Exemplarily, the verification methods mainly include real-scene testing and simulation experiments to comprehensively evaluate the performance of the model in practical applications. Testing in a real driving environment can provide intuitive feedback on the actual performance of the model, thereby improving the accuracy and robustness of the model. These tests usually include actual driving tests under different weather conditions, traffic conditions and driving modes. By monitoring the real-time performance of the model and analyzing its responsiveness to various complex driving scenarios, the optimization model can be further verified, thereby improving the accuracy and robustness of the model. For example, the performance of the model is tested in different driving scenarios such as urban roads, rural roads and highways. These measured data help to discover the shortcomings of the system and provide a basis for further optimization. Experiments using driving simulators can test the performance of the model in a controlled environment and quickly verify various specific driving scenarios. Simulation experiments can reproduce various extreme driving conditions, such as high-speed rear-end collisions, sharp turns, bad weather, etc., and evaluate the performance of the model in these situations, so as to optimize the model for specific driving scenarios, such as dangerous driving scenarios, and improve the accuracy of the model's driving decisions for the driving scenario. Through the analysis of simulation data, the stability and reliability of the model can be evaluated, and necessary model optimization adjustments can be made. Simulation experiments can also be used to validate the effectiveness of new features and improvements, ensuring that the model meets the expected performance standards before release.

[0102] Furthermore, the model can be optimized based on user feedback data.

[0103] For example, user feedback data is an important source for continuous model improvement. User feedback data can include the driver's experience and opinions during use. These feedback data can be obtained through questionnaires, user interviews, data log analysis, etc. Key feedback data include ease of use, decision accuracy, response speed, etc. By analyzing user feedback data, problems in the system, such as false positives, missed negatives, or decision delays, are identified, and targeted model optimization and improvement are carried out to improve the robustness and accuracy of the model.

[0104] Furthermore, we continuously monitor the performance of the model in actual use, collect and analyze long-term usage data, and discover potential performance degradation issues and new problems, so as to continuously optimize the model to improve its robustness and accuracy.

[0105] It should be understood that the above-mentioned optimization model approach is not limited to the driving behavior classification model, but can also be applied to other models, such as a cause determination model for determining the cause of occurrence, and a driving decision model for making driving decisions.

[0106] The driving behavior analysis method provided in the embodiment of the present invention trains a training model to obtain an initial training model based on sample driving behavior data, determines the driving behavior classification accuracy of the initial training model in a preset driving scenario, and optimizes the initial training model based on the driving behavior classification accuracy of the initial training model in the preset driving scenario to obtain a driving behavior classification model with excellent performance for the preset driving scenario, that is, better optimizes the model for each specific driving scenario to improve the classification accuracy for each specific driving scenario, thereby further improving the decision accuracy of driving decisions.

[0107] Based on the above embodiments, the present invention provides an efficient, flexible and intelligent driving behavior analysis and driving decision system, which can better cope with complex driving environments and real-time driving decision needs.

[0108] The driving behavior analysis device provided by the present invention is described below. The driving behavior analysis device described below and the driving behavior analysis method described above can be referenced to each other.

[0109] Figure 3 is a schematic diagram of the structure of the driving behavior analysis device provided by the present invention, such as Figure 3 As shown, the driving behavior analysis device includes a data acquisition module 310, a behavior classification module 320, a cause determination module 330 and a decision determination module 340.

[0110] The data acquisition module 310 is used to acquire driving data of the vehicle; the driving data is obtained by collecting data on the driving process of the vehicle.

[0111] The behavior classification module 320 is used to determine driving behavior data from the driving data, input the driving behavior data into a driving behavior classification model, and obtain a driving behavior classification result output by the driving behavior classification model; the driving behavior data is used to analyze the driver's driving behavior.

[0112] The cause determination module 330 is used to determine the cause of the driving behavior classification result based on the driving behavior classification result and context data related to the driving behavior classification result determined from the driving data.

[0113] The decision determination module 340 is used to determine a driving decision for the driving behavior classification result based on the occurrence cause.

[0114] Wherein, the driving behavior classification model is obtained by training based on sample driving behavior data.

[0115] The driving behavior analysis device provided by the embodiment of the present invention obtains driving data of a car, and the driving data is obtained by collecting data of the driving process of the car, so as to determine the driving behavior data from the driving data, and input the driving behavior data into a driving behavior classification model, accurately obtain the driving behavior classification result output by the driving behavior classification model, and accurately determine the cause of the driving behavior classification result based on the driving behavior classification result and context data related to the driving behavior classification result determined from the driving data, so as to accurately determine the driving decision for the driving behavior classification result based on the cause, that is, not only the driving behavior of the driver is analyzed, but also the cause of the driving behavior is analyzed, so as to fully consider the background cause of the corresponding driving behavior, and fully understand the cause of the driver's driving decision, and then comprehensively generate more accurate driving decisions, improve the accuracy of driving decisions, and ultimately improve driving safety and driving experience.

[0116] Based on any of the above embodiments, the cause determination module 330 is specifically used to: Performing semantic analysis on the driving behavior classification result to obtain semantic information of the driving behavior classification result; Based on the semantic information and context data related to the semantic information determined from the driving data, a reason for the driving behavior classification result is determined.

[0117] Based on any of the above embodiments, the decision determination module 340 is specifically used to: Based on the occurrence cause and the semantic information, a driving decision for the driving behavior classification result is determined.

[0118] Based on any of the above embodiments, the driving behavior data includes time series data, the context data includes time series data, and the driving behavior classification model is constructed based on a long short-term memory network LSTM feature extraction layer.

[0119] Based on any of the above embodiments, the device further includes a model optimization module, which is used to: Based on the sample driving behavior data, the training model is trained to obtain an initial training model; Determining the driving behavior classification accuracy of the initial training model in a preset driving scenario; Based on the driving behavior classification accuracy, the initial training model is optimized to obtain the driving behavior classification model.

[0120] Based on any of the above embodiments, the device further includes: A decision execution module is used to control the vehicle to execute operations corresponding to the driving decision and / or output prompt information based on the driving decision.

[0121] Figure 4 An example of a physical structure diagram of a car is shown below. Figure 4 As shown, the car may include: a processor 410, a communication interface 420, a memory 430 and a communication bus 440, wherein the processor 410, the communication interface 420 and the memory 430 communicate with each other through the communication bus 440. The processor 410 may call the logic instructions in the memory 430 to execute the driving behavior analysis method, which includes: obtaining driving data of the car; the driving data is obtained by collecting data of the driving process of the car; determining driving behavior data from the driving data, inputting the driving behavior data into a driving behavior classification model, and obtaining a driving behavior classification result output by the driving behavior classification model; the driving behavior data is used to analyze the driving behavior of the driver; based on the driving behavior classification result and context data related to the driving behavior classification result determined from the driving data, determining the cause of the driving behavior classification result; based on the cause of the occurrence, determining a driving decision for the driving behavior classification result; wherein the driving behavior classification model is trained based on sample driving behavior data.

[0122] In addition, the logic instructions in the above-mentioned memory 430 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0123] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the driving behavior analysis method provided by the above methods, which method includes: obtaining driving data of a car; the driving data is obtained by collecting data of the driving process of the car; determining driving behavior data from the driving data, inputting the driving behavior data into a driving behavior classification model, and obtaining a driving behavior classification result output by the driving behavior classification model; the driving behavior data is used to analyze the driver's driving behavior; based on the driving behavior classification result and context data related to the driving behavior classification result determined from the driving data, determining the cause of the driving behavior classification result; based on the cause of occurrence, determining a driving decision for the driving behavior classification result; wherein the driving behavior classification model is trained based on sample driving behavior data.

[0124] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the driving behavior analysis method provided by the above-mentioned methods, the method comprising: obtaining driving data of a car; the driving data is obtained by collecting data of the driving process of the car; determining driving behavior data from the driving data, inputting the driving behavior data into a driving behavior classification model, and obtaining a driving behavior classification result output by the driving behavior classification model; the driving behavior data is used to analyze the driver's driving behavior; based on the driving behavior classification result and context data related to the driving behavior classification result determined from the driving data, determining the cause of the driving behavior classification result; based on the cause of the occurrence, determining a driving decision for the driving behavior classification result; wherein the driving behavior classification model is trained based on sample driving behavior data.

[0125] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0126] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0127] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A driving behavior analysis method, characterized in that: include: Get the car's driving data; The driving data is obtained by collecting data on the driving process of the automobile; Determining driving behavior data from the driving data, inputting the driving behavior data into a driving behavior classification model, and obtaining a driving behavior classification result output by the driving behavior classification model; The driving behavior data is used to analyze the driver's driving behavior; Determining a cause of the driving behavior classification result based on the driving behavior classification result and context data related to the driving behavior classification result determined from the driving data; Based on the occurrence cause, determining a driving decision for the driving behavior classification result; Wherein, the driving behavior classification model is obtained by training based on sample driving behavior data.

2. The driving behavior analysis method according to claim 1, characterized in that: The determining, based on the driving behavior classification result and determining context data related to the driving behavior classification result from the driving data, a cause of the driving behavior classification result, comprises: Performing semantic analysis on the driving behavior classification result to obtain semantic information of the driving behavior classification result; Based on the semantic information and context data related to the semantic information determined from the driving data, a reason for the driving behavior classification result is determined.

3. The driving behavior analysis method according to claim 2, characterized in that: The determining of a driving decision for the driving behavior classification result based on the occurrence cause includes: Based on the occurrence cause and the semantic information, a driving decision for the driving behavior classification result is determined.

4. The driving behavior analysis method according to claim 1, characterized in that: The driving behavior data includes time series data, the context data includes time series data, and the driving behavior classification model is constructed based on a long short-term memory network (LSTM) feature extraction layer.

5. The driving behavior analysis method according to claim 1, characterized in that: The driving behavior classification model is obtained based on the following method: Based on the sample driving behavior data, the training model is trained to obtain an initial training model; Determining the driving behavior classification accuracy of the initial training model in a preset driving scenario; Based on the driving behavior classification accuracy, the initial training model is optimized to obtain the driving behavior classification model.

6. The driving behavior analysis method according to any one of claims 1 to 5, characterized in that: After determining the driving decision for the driving behavior classification result based on the occurrence cause, the method further includes: Based on the driving decision, the automobile is controlled to execute an operation corresponding to the driving decision and / or output prompt information.

7. A driving behavior analysis device, characterized in that: include: A data acquisition module, used to acquire driving data of the car; The driving data is obtained by collecting data on the driving process of the automobile; A behavior classification module, used to determine driving behavior data from the driving data, input the driving behavior data into a driving behavior classification model, and obtain a driving behavior classification result output by the driving behavior classification model; The driving behavior data is used to analyze the driver's driving behavior; a cause determination module, configured to determine a cause of the driving behavior classification result based on the driving behavior classification result and context data related to the driving behavior classification result determined from the driving data; A decision determination module, used for determining a driving decision for the driving behavior classification result based on the occurrence cause; Wherein, the driving behavior classification model is obtained by training based on sample driving behavior data.

8. An automobile, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the driving behavior analysis method according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the driving behavior analysis method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the driving behavior analysis method according to any one of claims 1 to 6 is implemented.

Images