Driving data processing method and device, electronic equipment and readable storage medium
By collecting and processing a variety of driving data, identifying and labeling long-tail data, and optimizing the decision-making model of the autonomous driving system, the problem of insufficient collection and processing of long-tail data in the existing technology is solved, and the system's decision-making ability in complex scenarios is improved and security.
Patent Information
- Application Number
- CN202411072840.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-06
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2044-08-06
AI Technical Summary
The existing autonomous driving system has shortcomings in the collection and processing of long-tail data, resulting in limited decision-making capabilities and inability to effectively deal with rare situations, and the value of long-tail data has not been fully explored.
Vehicle status data, environmental perception data, driver behavior data and communication data are collected through vehicle detection equipment, and data processing and feature processing are carried out, including coarse-grained screening, spatio-temporal alignment, fine-grained screening, data standardization processing, feature extraction and feature screening. Identify and label long-tail data to optimize the decision model.
It improves data quality and consistency, enhances the generalization and adaptability of decision-making models, can better respond to rare situations, reduce accident risks, and improves the decision-making capabilities of autonomous driving systems in complex scenarios.
Smart Images

Figure CN119028045A_ABST
Abstract
Description
Technical Field
[0001] Embodiments of the present invention relate to the technical field of autonomous driving data mining, and in particular, to a method for processing driving data, a device for processing driving data, an electronic device, and a computer-readable storage medium. Background Art
[0002] With the rapid development of autonomous driving technology, more and more vehicles are equipped with autonomous driving systems. During the autonomous driving process, a large amount of driving data will be generated, including some abnormal situations, such as vehicle failures, sensor failures, etc. These data are usually called "long-tail data". Although they occur less frequently, they have an important impact on the safety and stability of the autonomous driving system. However, due to the low frequency of long-tail data, the collected data may have problems such as high noise and inaccurate labeling. These problems often affect the quality of long-tail data and limit the decision-making ability of the vehicle.
[0003] At present, the collection and processing of long-tail data is a crucial link in the development of autonomous driving technology. Existing autonomous driving systems often ignore the importance of this part of data, resulting in limited decision-making capabilities of vehicles when faced with rare situations. In addition, existing data processing technologies may not be able to effectively cope with these challenges, resulting in the value of long-tail data not being fully tapped.
[0004] Specifically, in the existing long-tail data mining process, long-tail data is usually obtained based on perception data, and perception data is often concentrated on common driving scenarios and normal operations, which results in a very small proportion of long-tail data in the overall data set. This deviation makes the long-tail data insufficiently weighted in model training, making it difficult to effectively influence the vehicle's decision-making model. Summary of the invention
[0005] An embodiment of the present invention provides a method, device, electronic device and computer-readable storage medium for processing driving data to solve the problem in the prior art that the decision-making ability of the autonomous driving system is limited due to the inadequacy of long-tail data collection and processing being easily ignored.
[0006] An embodiment of the present invention discloses a method for processing driving data, the method comprising:
[0007] Collecting driving data of the vehicle through vehicle detection equipment; the driving data includes: vehicle status data, environment perception data, driver behavior data and communication data;
[0008] Processing the driving data to obtain preprocessed data; the data processing includes coarse-grained screening, spatiotemporal alignment, fine-grained screening and data standardization processing;
[0009] Performing feature processing on the preprocessed data to obtain an optimal feature data set; the feature processing includes feature extraction and feature screening;
[0010] According to the driving data and the optimal feature data set, a plurality of data results and weights respectively corresponding to the plurality of data results are obtained; wherein the weights are determined based on the correlation between the plurality of data results;
[0011] Obtaining long-tail data according to the plurality of data results and the weights respectively corresponding to the plurality of data results;
[0012] Annotating the long-tail data according to the characteristics of the long-tail data to obtain annotated long-tail data;
[0013] The decision model of the vehicle is optimized according to the labeled long-tail data; the decision model is used to guide the vehicle to make a decision.
[0014] Optionally, the performing data processing on the driving data to obtain pre-processed data includes:
[0015] Performing the coarse-grained screening on the driving data according to a data quality threshold to obtain first data;
[0016] Performing the spatiotemporal alignment on the spatiotemporal information of the first data to obtain second data; the spatiotemporal information includes time information and space information;
[0017] Performing the fine-grained screening on the second data to obtain third data;
[0018] The third data is subjected to the data standardization process to obtain preprocessed data; the data standardization process is used to convert the third data into a target data format.
[0019] Optionally, performing feature processing on the preprocessed data to obtain an optimal feature data set includes:
[0020] Performing the feature extraction on the preprocessed data to obtain feature data sets corresponding to a plurality of feature types;
[0021] The feature screening is performed on the feature data sets respectively corresponding to the multiple feature types to obtain the optimal feature data set, wherein the optimal feature data set includes the optimal feature data respectively corresponding to the multiple feature types.
[0022] Optionally, obtaining a plurality of data results and weights corresponding to the plurality of data results respectively according to the driving data and the optimal feature data set includes:
[0023] Constructing a long-tail data recognition model according to the optimal feature data set;
[0024] Inputting the driving data into the long-tail data recognition model for data recognition to obtain a plurality of data results;
[0025] Constructing the plurality of data results into a graph structure; wherein the graph nodes in the graph structure are the data results;
[0026] Determine the correlation coefficient between adjacent graph nodes in the graph structure, and determine the weights corresponding to the plurality of data results based on the correlation coefficient.
[0027] Optionally, obtaining long-tail data according to the plurality of data results and the weights respectively corresponding to the plurality of data results includes:
[0028] Multiplying the data result by the weight corresponding to the data result to obtain weighted data corresponding to the data result;
[0029] The weighted data corresponding to the plurality of data results are aggregated to obtain the long-tail data.
[0030] Optionally, after obtaining the long-tail data according to the plurality of data results and the weights respectively corresponding to the plurality of data results, the method further includes:
[0031] Determine the similarity between the plurality of long-tail data, and determine the association relationship between the plurality of long-tail data based on the similarity between the plurality of long-tail data;
[0032] Analyze the similarities between the plurality of long-tail data to determine a plurality of data groups;
[0033] Determine a data pattern corresponding to the data group; the data pattern is used to characterize a pattern corresponding to long-tail data that appears repeatedly in the data group;
[0034] The long-tail data recognition model is optimized and updated according to the long-tail data, the association relationship between the long-tail data, the data group and the data pattern.
[0035] Optionally, optimizing the decision model of the vehicle according to the labeled long-tail data includes:
[0036] Determining, in the annotated long-tail data, annotated feature data corresponding to the annotated information according to the annotated information in the annotated long-tail data; the annotated information includes driving scenes, driving difficulty, and driving risk factors;
[0037] Training and adjusting the decision model according to the annotated feature data;
[0038] Evaluate the decision model obtained through training and adjustment to obtain evaluation results;
[0039] The decision model obtained through training and adjustment is optimized according to the evaluation results, so that the decision model outputs corresponding decisions when dealing with different driving scenarios, driving difficulties and driving risk factors.
[0040] Optionally, after labeling the long-tail data according to the characteristics of the long-tail data to obtain the labeled long-tail data, the method further includes:
[0041] The annotated long-tail data is stored in a target database; the annotated long-tail data is displayed through a visual interface;
[0042] Searching the target database according to the keywords corresponding to the annotated long-tail data;
[0043] and / or,
[0044] The target database is searched according to the conditional keywords corresponding to the annotated long-tail data; the conditional keywords are obtained by conditionally combining multiple keywords.
[0045] The embodiment of the present invention further discloses a driving data processing device, the device comprising:
[0046] A collection module, used to collect the driving data of the vehicle through the detection equipment of the vehicle; the driving data includes: vehicle status data, environment perception data, driver behavior data and communication data;
[0047] A data processing module, used for processing the driving data to obtain pre-processed data; the data processing includes coarse-grained screening, spatiotemporal alignment, fine-grained screening and data standardization processing;
[0048] A feature processing module, used to perform feature processing on the preprocessed data to obtain an optimal feature data set; the feature processing includes feature extraction and feature screening;
[0049] A weight module, used to obtain a plurality of data results and weights corresponding to the plurality of data results respectively according to the driving data and the optimal feature data set; wherein the weights are determined based on the correlation between the plurality of data results;
[0050] A long-tail data module, used for obtaining long-tail data according to the plurality of data results and the weights respectively corresponding to the plurality of data results;
[0051] A labeling module, used for labeling the long-tail data according to the characteristics of the long-tail data to obtain labeled long-tail data;
[0052] An optimization module is used to optimize the decision model of the vehicle according to the labeled long-tail data; the decision model is used to guide the vehicle to make decisions.
[0053] The embodiment of the present invention further discloses an electronic device, comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus;
[0054] The memory is used to store computer programs;
[0055] The processor is used to implement the method described in the embodiment of the present invention when executing the program stored in the memory.
[0056] The embodiment of the present invention further discloses a computer program product, which is stored in a storage medium and is executed by at least one processor to implement the method described in the embodiment of the present invention.
[0057] The embodiment of the present invention further discloses a computer-readable storage medium having instructions stored thereon, which, when executed by one or more processors, enables the processors to execute the method described in the embodiment of the present invention.
[0058] The embodiments of the present invention include the following advantages:
[0059] In an embodiment of the present invention, driving data of the vehicle is collected by a detection device of the vehicle; the driving data includes: vehicle status data, environmental perception data, driver behavior data and communication data; the driving data is processed to obtain pre-processed data; the data processing includes coarse-grained screening, spatiotemporal alignment, fine-grained screening and data standardization processing; feature processing is performed on the pre-processed data to obtain an optimal feature data set; the feature processing includes feature extraction and feature screening; based on the driving data and the optimal feature data set, multiple data results and weights corresponding to the multiple data results are obtained; wherein the weights are determined based on the correlation between the multiple data results; long-tail data is obtained based on the multiple data results and the weights corresponding to the multiple data results; the long-tail data is labeled according to the characteristics of the long-tail data to obtain labeled long-tail data; the decision model of the vehicle is optimized based on the labeled long-tail data; the decision model is used to guide the vehicle to make decisions. The embodiment of the present invention collects a variety of driving data including vehicle status data, environmental perception data, driver behavior data and communication data through the vehicle detection equipment, which not only expands the scope of data collection, but also increases the diversity of data. This wide data collection method helps to capture more rare long-tail data, thereby providing a richer information source for the optimization of subsequent decision models. In the data processing and feature processing stages, the quality and consistency of the data are ensured through coarse-grained screening, spatiotemporal alignment, fine-grained screening and data standardization processing. Through feature extraction and feature screening, the most valuable features for the decision model can be extracted from a large amount of data, which not only improves the efficiency of data processing, but also enhances the generalization ability and adaptability of the decision model. Traditional autonomous driving systems often ignore the importance of long-tail data, while the embodiment of the present invention can effectively identify and utilize these long-tail data through refined data processing, feature processing and other processing methods, so that the autonomous driving system can better cope with rare situations and reduce accident risks. In addition, through automated data collection and labeling processes, the efficiency and accuracy of data processing are improved, and the decision model is optimized using long-tail data to improve the decision-making ability of the autonomous driving system in complex and rare scenarios. BRIEF DESCRIPTION OF THE DRAWINGS
[0060] Figure 1 is a flowchart of a method for processing driving data provided in an embodiment of the present invention;
[0061] Figure 2 is a structural block diagram of a driving data processing device provided in an embodiment of the present invention;
[0062] Figure 3 is a schematic diagram of the hardware structure of an electronic device for implementing various embodiments of the present invention;
[0063] Figure 4 is a schematic diagram of a computer-readable medium provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0064] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.
[0065] Reference Figure 1 , shows a flowchart of a method for processing driving data provided in an embodiment of the present invention, which may specifically include the following steps:
[0066] In an embodiment of the present invention, a method for processing driving data provided in an embodiment of the present invention is implemented by an efficient mining and processing system for long-tail data. The system specifically includes a data collection module, a data preprocessing module, a long-tail data identification module, a data annotation module, a data storage and retrieval module, a model optimization module, and a user interface module. The details are as follows:
[0067] 1) Data collection module: Comprehensive driving data is collected in real time through the on-board sensor network and communication system, including vehicle status data, environmental perception data, driver behavior data and communication data; this comprehensive data collection method helps to capture more rare long-tail data, providing a rich data source for subsequent data processing and model training.
[0068] 2) Data preprocessing module: Through coarse-grained screening, spatiotemporal alignment, fine-grained screening and data standardization processing, the quality and availability of long-tail data are effectively improved; these data processing steps ensure the consistency and accuracy of the data, laying a solid foundation for subsequent feature processing and model training.
[0069] 3) Long-tail data identification module: Use deep learning models (such as CNN, LSTM) to extract features and obtain the optimal feature data set through feature screening; these optimal feature data sets contain the most relevant and influential features, which can more accurately reflect the key information of long-tail data, thereby improving the prediction accuracy and performance of the model.
[0070] 4) Data labeling module: Label long-tail data manually or automatically to clarify the characteristics and importance of the data; the labeled data is used as high-quality training samples for subsequent training and optimization of decision-making models, improving the generalization ability and robustness of the model.
[0071] 5) Data storage and retrieval module: Store the annotated long-tail data in a high-performance database and provide keyword and conditional keyword-based retrieval functions; this efficient retrieval method greatly improves the efficiency of data retrieval and provides strong support for subsequent data analysis and model training.
[0072] 6) Model optimization module: Use the mined long-tail data to optimize the decision model of the autonomous driving system. By incorporating these rare situations into the training set, the accuracy and robustness of the decision model in dealing with similar situations can be improved; the optimization process uses advanced machine learning algorithms (such as deep reinforcement learning) combined with the characteristics and annotation information of long-tail data to iteratively train and adjust the decision model.
[0073] 7) User interface module: provides an intuitive and user-friendly operating interface to display the mined long-tail data and its annotation information; users can view, analyze and process this data through the interface to further understand the performance and optimization direction of the autonomous driving system in rare situations.
[0074] Step 101: collecting driving data of the vehicle through a vehicle detection device; the driving data includes: vehicle status data, environment perception data, driver behavior data and communication data;
[0075] In an embodiment of the present invention, the data collection module is used to collect the driving data of the vehicle, and the original driving data is collected from the sensors and systems of the autonomous driving vehicle. Specifically, the data collection module collects the driving data of the autonomous driving vehicle in real time through the vehicle-mounted sensor network (such as GPS, radar, camera, inertial measurement unit, etc.) and the communication system (such as vehicle-to-everything (V2X) technology).
[0076] In the existing long-tail data acquisition process, in the existing long-tail data mining process, long-tail data is usually acquired based on perception data. However, perception data is often concentrated on common driving scenarios and normal operations, which results in a very small proportion of long-tail data in the overall data set. This deviation makes the weight of long-tail data in model training insufficient, and it is difficult to effectively affect the vehicle's decision model. Therefore, in order to more effectively mine and utilize long-tail data, and thereby improve the safety and stability of the autonomous driving system, the embodiment of the present invention collects driving data such as vehicle status data, environmental perception data, driver behavior data, and communication data through the vehicle's detection equipment, so that the collected data is more comprehensive, not only covering the vehicle's own operating status, but also including the vehicle's environmental information, the driver's behavior pattern, and communication data between vehicles, etc. This diversity and comprehensiveness makes the data set richer, which helps to capture more rare long-tail data in the future.
[0077] The data collection module has a high data capture capability to ensure that all types of data generated during the autonomous driving process, regardless of size, frequency and type, can be fully recorded. At the same time, the data collection module also has a data cache function, which can temporarily store data when data transmission is limited or interrupted, and then transmit it after the network is restored.
[0078] Step 102: Process the driving data to obtain preprocessed data; the data processing includes coarse-grained screening, spatiotemporal alignment, fine-grained screening and data standardization processing;
[0079] In the prior art, due to the low frequency of long-tail data, the collected data may have problems such as large noise, which often affect the quality of long-tail data and limit the decision-making ability of the vehicle. Therefore, in the embodiment of the present invention, the driving data is processed mainly through four processing steps: coarse-grained screening, spatiotemporal alignment, fine-grained screening, and data standardization processing, so as to solve the problem of low quality of the existing collected long-tail data.
[0080] Specifically, problems such as large noise and many outliers that may exist in long-tail data can be solved by coarse-grained screening to remove obviously erroneous or irrelevant data, such as outliers caused by sensor failure and data loss caused by communication interruption, thus solving the data quality problem and ensuring the accuracy and effectiveness of subsequent processing steps; since data collected by different sensors may have time delays and differences in spatial positions, the timestamps and spatial coordinates of the data are unified by time-space alignment, solving the data synchronization problem, so that data from different sources can be accurately associated and analyzed, and improving the comprehensive utilization value of the data; on the basis of coarse-grained screening, fine-grained screening is used to further eliminate data that does not meet specific standards or conditions, such as behavioral data that does not conform to driving logic and noise data in environmental perception, solving the data accuracy problem and ensuring that every piece of data in the data set is highly relevant and available; since data collected by different sensors and systems often have different dimensions and ranges, data standardization is used to convert the data into a unified format and range, eliminating the dimensional differences between the data and facilitating subsequent feature extraction and model training.
[0081] The embodiments of the present invention effectively solve the quality problem of long-tail data through coarse-grained screening, spatiotemporal alignment, fine-grained screening and data standardization processing, and improve the quality and availability of long-tail data sets.
[0082] Step 103: performing feature processing on the preprocessed data to obtain an optimal feature data set; the feature processing includes feature extraction and feature screening;
[0083] In actual application, the original pre-processed data may contain a large number of dimensions, many of which do not contribute much to the decision model. High-dimensional data may also make the model too complex, increase the computational cost and the risk of overfitting. In the embodiment of the present invention, feature extraction and feature screening can effectively reduce the dimension of the data, improve the efficiency of data processing and the training speed of the model.
[0084] The embodiment of the present invention extracts features that are valuable to the decision model from the preprocessed data through feature extraction, and converts a large amount of preprocessed data into a more concise and representative feature set, which not only reduces the complexity of data processing and simplifies the model structure, but also improves the training efficiency and prediction accuracy of the model. Subsequently, the most relevant and influential features are further selected from the extracted feature set through feature screening to construct an optimal feature data set, removing redundant and irrelevant features, reducing the dimension of the data, reducing the complexity of the model, and avoiding the risk of overfitting.
[0085] Specifically, the long-tail data recognition module first uses a deep learning model (such as a convolutional neural network CNN, a long short-term memory network LSTM) to extract features from the preprocessed data to obtain feature data sets corresponding to multiple feature types. In the embodiments of the present invention, the deep learning model used for feature extraction is not limited. Next, feature selection is implemented through a feature selection algorithm specific to long-tail data, and the most representative optimal feature data is selected from the feature data sets corresponding to each feature type. Then, the optimal feature data set corresponding to each feature type is aggregated to obtain the optimal feature data set.
[0086] Since long-tail data occurs infrequently, feature extraction and feature screening can extract valuable features from these rare data, enhancing the model's ability to process long-tail data and improving the system's decision-making ability and security when facing rare situations.
[0087] Step 104: obtaining a plurality of data results and weights corresponding to the plurality of data results respectively according to the driving data and the optimal feature data set; wherein the weights are determined based on the correlation between the plurality of data results;
[0088] In the embodiment of the present invention, the identification of long-tail data is realized by a long-tail data identification module, which is the core part of the system. First, in the long-tail data identification module, a long-tail data identification model based on a support vector machine (SVM) or a random forest is constructed according to the optimal feature data set, mainly using a machine learning algorithm to classify and identify the optimal feature data in the optimal feature data set, so that in the subsequent long-tail data identification process, the long-tail data can be accurately identified from a large amount of data.
[0089] Subsequently, the driving data is input into the long-tail data recognition model for data recognition, and multiple data results are obtained. These data results represent different types of long-tail data recognized by the model, and the multiple data results are constructed into a graph structure, in which the graph nodes represent the data results. The system determines the correlation coefficients between adjacent graph nodes in the graph structure, and determines the weights corresponding to each data result based on these correlation coefficients. These weights reflect the correlation and importance between the data results.
[0090] The efficient mining and processing system for long-tail data of autonomous driving in the embodiment of the present invention quantifies the correlation between data results and assigns corresponding weights to each data result, so that the data can be more reasonably used in the subsequent decision-making process, and the safety and reliability of the autonomous driving system can be effectively improved. The determination of weights improves the intelligence level of data processing and improves data processing efficiency, so that the system can more effectively utilize long-tail data and optimize model performance. The system has broad application prospects and market value.
[0091] Step 105: Obtain long-tail data according to the plurality of data results and the weights respectively corresponding to the plurality of data results;
[0092] In an embodiment of the present invention, the system uses weights to quantify the importance of each data result, thereby screening out the most valuable long-tail data for the decision model from among numerous data results. Specifically, each data result is multiplied by its corresponding weight to obtain weighted data corresponding to each data result, so as to ensure that in subsequent data aggregation, data results with high importance have a greater impact. The weighted data corresponding to multiple data results are aggregated to obtain the final long-tail data.
[0093] By combining data results with weights, the embodiment of the present invention enables the system to identify and extract long-tail data that have a significant impact on the autonomous driving system. These data usually involve rare but critical driving scenarios, such as extreme weather conditions, complex traffic conditions, vehicle failures, etc., which solves the problem of ignoring rare events in traditional data processing and provides more comprehensive data support for vehicle decision model training and optimization.
[0094] The efficient mining and processing system for long-tail data of autonomous driving in the embodiment of the present invention realizes efficient identification and extraction of long-tail data by comprehensively considering all data results and their weights, thereby ensuring the accuracy and representativeness of the long-tail data.
[0095] Step 106: labeling the long-tail data according to the characteristics of the long-tail data to obtain labeled long-tail data;
[0096] In practical applications, the labeling of long-tail data requires a combination of professional knowledge and algorithms to ensure the accuracy and consistency of the labeling. However, the long-tail data collected through existing technologies often have inaccurate labeling, resulting in low labeling quality for the long-tail data, which affects the subsequent data processing and model training effects.
[0097] In an embodiment of the present invention, the long-tail data is annotated by a data annotating module. Specifically, the data annotating module manually or automatically annotates the identified long-tail data. The annotating process includes clarifying the characteristics and importance of the data, such as the type of driving scene, difficulty level, and risk factors.
[0098] Labeling can be done manually or automatically. For long-tail data that is difficult to identify automatically, the module provides manual labeling, which is usually accurately labeled by a team of experts to ensure the accuracy and reliability of the data. Automatic labeling may involve the use of machine learning algorithms or rule engines to automatically identify and label data. These algorithms can identify the characteristics of data based on predefined rules or patterns and label them accordingly.
[0099] The efficient mining and processing system for long-tail data of autonomous driving in the embodiment of the present invention uses the labeled data as high-quality training samples for the training and optimization of subsequent decision-making models.
[0100] Step 107: Optimize the decision model of the vehicle according to the labeled long-tail data; the decision model is used to guide the vehicle to make a decision.
[0101] In an embodiment of the present invention, the decision model of the autonomous driving system is optimized through a model optimization module. Specifically, the model optimization module optimizes the decision model of the autonomous driving system using the mined long-tail data. By incorporating these rare situations (long-tail data) into the training set, the accuracy and robustness of the decision model in dealing with similar situations can be improved. The optimization process uses advanced machine learning algorithms (such as deep reinforcement learning) and combines the characteristics and annotation information of long-tail data to iteratively train and adjust the decision model. The optimized decision model can better cope with complex and rare driving scenarios and improve the safety and reliability of the autonomous driving system.
[0102] In an embodiment of the present invention, driving data of the vehicle is collected by a detection device of the vehicle; the driving data includes: vehicle status data, environmental perception data, driver behavior data and communication data; the driving data is processed to obtain pre-processed data; the data processing includes coarse-grained screening, spatiotemporal alignment, fine-grained screening and data standardization processing; feature processing is performed on the pre-processed data to obtain an optimal feature data set; the feature processing includes feature extraction and feature screening; based on the driving data and the optimal feature data set, multiple data results and weights corresponding to the multiple data results are obtained; wherein the weights are determined based on the correlation between the multiple data results; long-tail data is obtained based on the multiple data results and the weights corresponding to the multiple data results; the long-tail data is labeled according to the characteristics of the long-tail data to obtain labeled long-tail data; the decision model of the vehicle is optimized based on the labeled long-tail data; the decision model is used to guide the vehicle to make decisions. The embodiment of the present invention collects a variety of driving data including vehicle status data, environmental perception data, driver behavior data and communication data through the vehicle detection equipment, which not only expands the scope of data collection, but also increases the diversity of data. This wide data collection method helps to capture more rare long-tail data, thereby providing a richer information source for the optimization of subsequent decision models. In the data processing and feature processing stages, the quality and consistency of the data are ensured through coarse-grained screening, spatiotemporal alignment, fine-grained screening and data standardization processing. Through feature extraction and feature screening, the most valuable features for the decision model can be extracted from a large amount of data, which not only improves the efficiency of data processing, but also enhances the generalization ability and adaptability of the decision model. Traditional autonomous driving systems often ignore the importance of long-tail data, while the embodiment of the present invention can effectively identify and utilize these long-tail data through refined data processing, feature processing and other processing methods, so that the autonomous driving system can better cope with rare situations and reduce accident risks. In addition, through automated data collection and labeling processes, the efficiency and accuracy of data processing are improved, and the decision model is optimized using long-tail data to improve the decision-making ability of the autonomous driving system in complex and rare scenarios.
[0103] In one embodiment of the present invention, in step 102, the driving data is processed to obtain pre-processed data, including:
[0104] Performing the coarse-grained screening on the driving data according to a data quality threshold to obtain first data;
[0105] Performing the spatiotemporal alignment on the spatiotemporal information of the first data to obtain second data; the spatiotemporal information includes time information and space information;
[0106] Performing the fine-grained screening on the second data to obtain third data;
[0107] The third data is subjected to the data standardization process to obtain preprocessed data; the data standardization process is used to convert the third data into a target data format.
[0108] In the embodiment of the present invention, the driving data is preprocessed by a data preprocessing module, and the data preprocessing generally needs to be implemented through four stages: coarse-grained screening, spatiotemporal alignment, fine-grained screening, and standardized processing.
[0109] Specifically, in the coarse-grained screening stage: the data preprocessing module will first clean the collected raw data, and quickly remove obvious unqualified data such as noise, invalid and erroneous data by setting data quality thresholds. By setting simple data quality thresholds for judgment, the processing speed is fast, which can effectively reduce the amount of data and improve the efficiency of subsequent processing.
[0110] Spatiotemporal alignment stage: Advanced sensor data calibration technology is used to ensure the time and space synchronization between different sensors to ensure data accuracy and consistency. Spatiotemporal alignment after coarse-grained screening can ensure that the alignment operation is performed on a relatively clean data set, reducing errors and complexity in the alignment process.
[0111] Fine-grained screening stage: further remove outliers and interference in the data through noise filtering algorithms (such as Kalman filters). This usually involves more complex algorithms and calculations, and the processing speed is relatively slow. Fine-grained screening after coarse-grained screening and spatiotemporal alignment can ensure that the algorithm runs on a higher quality data set and improve the accuracy and effect of screening.
[0112] Standardization processing stage: Implement data standardization and convert all data to the same measurement range to lay the foundation for subsequent data mining and analysis. The same measurement range here can be understood as a data format that the system can recognize and process, that is, the target data format. Standardization processing after fine-grained screening can ensure that the standardization operation is performed on high-quality data sets after multiple rounds of screening, thereby improving the efficiency and effectiveness of standardization processing.
[0113] Coarse-grained screening can quickly remove a large amount of unqualified data and reduce the amount of data, thereby saving time and computing resources for subsequent processing steps. If fine-grained screening is performed directly, computing resources and time may be wasted on a large amount of unqualified data. The combined use of coarse-grained screening and fine-grained screening can simplify the complexity of subsequent processing steps (such as spatiotemporal alignment and standardization) to ensure that they can be performed on relatively clean and consistent data sets, improving processing efficiency and effectiveness.
[0114] In one embodiment of the present invention, the performing feature processing on the preprocessed data to obtain an optimal feature data set includes:
[0115] Performing the feature extraction on the preprocessed data to obtain feature data sets corresponding to a plurality of feature types;
[0116] The feature screening is performed on the feature data sets respectively corresponding to the multiple feature types to obtain the optimal feature data set; the optimal feature data set includes the optimal feature data respectively corresponding to the multiple feature types.
[0117] In the existing technology, feature extraction often only focuses on some common features, while ignoring other features that may have an important impact on decision-making. This incomplete feature extraction method limits the performance and generalization ability of the model. When screening features, there is often a lack of refined screening methods, resulting in some redundant or irrelevant features being retained, which increases the complexity and computational cost of the model. It may also introduce noise and affect the accuracy of the model.
[0118] At present, due to the inadequacy of feature extraction and feature screening, there is often a lack of sufficient high-quality feature support during model training, which limits the generalization ability and robustness of the model when facing complex and rare situations. Therefore, in an embodiment of the present invention, the most valuable features for the decision model are extracted from a large amount of preprocessed data through feature extraction, and redundant and irrelevant features are removed through feature screening, thereby constructing an efficient and streamlined optimal feature data set.
[0119] 1) Feature extraction: First, a deep learning model (such as CNN and LSTM) is used to extract feature data sets under each feature type from the preprocessed driving data to capture the complex patterns and relationships in the data. The extracted feature types may include but are not limited to: vehicle speed change pattern, acceleration distribution, directional stability, environmental perception information (such as obstacle distance, traffic sign recognition), etc.
[0120] These extracted features are classified into different feature types according to their nature and purpose, such as: dynamic features (speed, acceleration), static features (position, direction), environmental features (obstacles, traffic signs), etc.
[0121] 2) Feature screening: Under each feature type, use specific feature selection algorithms (such as principal component analysis PCA, correlation analysis, information gain, etc.) to evaluate and select the most representative and relevant features to identify the most critical features for long-tail data, thereby reducing the dimension of the feature space and improving the processing efficiency and accuracy of the model. Through this step, the optimal feature data corresponding to each feature type can be obtained, thereby forming the optimal feature data set.
[0122] The optimal feature data set in the embodiment of the present invention contains the most relevant and influential feature data, which can more accurately reflect key information, thereby improving the prediction accuracy and performance of the model. The optimal feature data set can be used to train the model in the future so that the model can better adapt to various driving scenarios, including rare situations, and enhance the robustness and reliability of the model.
[0123] The embodiment of the present invention ensures that the features used by the recognition model are the most representative and relevant through feature extraction and screening, thereby improving the recognition accuracy of long-tail data. The optimal feature data set in the embodiment of the present invention can provide more accurate and reliable information support for the decision-making of the autonomous driving system, thereby optimizing the vehicle's decision-making process, so that the efficient mining and processing system of autonomous driving long-tail data can make more reasonable and safer decisions when facing complex and rare situations.
[0124] In one embodiment of the present invention, the step 104 of obtaining a plurality of data results and weights corresponding to the plurality of data results respectively according to the driving data and the optimal feature data set includes:
[0125] Constructing a long-tail data recognition model according to the optimal feature data set;
[0126] Inputting the driving data into the long-tail data recognition model for data recognition to obtain a plurality of data results;
[0127] Constructing the plurality of data results into a graph structure; wherein the graph nodes in the graph structure are the data results;
[0128] Determine the correlation coefficient between adjacent graph nodes in the graph structure, and determine the weights corresponding to the plurality of data results based on the correlation coefficient.
[0129] In an embodiment of the present invention, a long-tail data recognition model is constructed based on the optimal feature data set. The long-tail data recognition model can be based on machine learning algorithms such as support vector machines (SVM) or random forests. These algorithms can process high-dimensional data and have good generalization capabilities, and are suitable for identifying long-tail data from massive data. Subsequently, the driving data is input into the long-tail data recognition model, and the model analyzes and processes the input driving data to obtain multiple data results. These results represent different types of long-tail data recognized by the model, reflecting rare situations that the vehicle may encounter during driving.
[0130] In order to better understand and analyze the relationship between these data results, multiple data results are constructed into a graph structure, where the graph nodes represent the data results. By constructing the graph structure, the system can intuitively display the associations and patterns between the data results, which is convenient for subsequent analysis and processing. Determine the correlation coefficients between adjacent graph nodes in the graph structure, and determine the weights corresponding to each data result based on these correlation coefficients. These weights reflect the correlation and importance between the data results, which helps the system to use the data more reasonably in the subsequent decision-making process. In addition, by quantifying the relationship between the data results, the system can assign corresponding weights to each data result, thereby making more effective use of long-tail data in the decision-making process.
[0131] The long-tail data recognition model constructed by the embodiment of the present invention is based on the optimal feature data set, which can better generalize to unseen data and improve the model's ability to handle new situations. By determining the graph structure and correlation coefficient, the system can more effectively process and analyze data results, improving the efficiency of data processing. The identified data results and their weights provide a high-quality data foundation for subsequent long-tail data annotation and model optimization, which helps to improve the decision support capabilities of the autonomous driving system.
[0132] In one embodiment of the present invention, the step 105 of obtaining long-tail data according to the plurality of data results and the weights respectively corresponding to the plurality of data results includes:
[0133] Multiplying the data result by the weight corresponding to the data result to obtain weighted data corresponding to the data result;
[0134] The weighted data corresponding to the plurality of data results are aggregated to obtain the long-tail data.
[0135] Long-tail data generally refers to data that occurs less frequently but has important value, which is easily overlooked in traditional data processing methods. The embodiments of the present invention enable the autonomous driving system to better cope with rare situations and reduce the risk of accidents by mining and processing long-tail data.
[0136] In the embodiment of the present invention, first, each data result is multiplied by its corresponding weight to obtain weighted data corresponding to each data result, so as to adjust the influence of the data result according to the importance of the data result (represented by the weight). The weight is usually determined based on the relevance and importance between the data results, and the data results with high weights will have a greater influence in the subsequent data aggregation. Subsequently, the weighted data corresponding to the multiple data results are aggregated, and all the data results and their weights are comprehensively considered to obtain the final long-tail data.
[0137] Data aggregation can adopt a variety of methods, such as weighted average, weighted summation or other statistical methods. The specific method depends on the nature of the data and the application scenario, and is not limited in the embodiments of the present invention.
[0138] In one embodiment of the present invention, after obtaining the long-tail data according to the plurality of data results and the weights respectively corresponding to the plurality of data results in step 105, the method further includes:
[0139] Determine the similarity between the plurality of long-tail data, and determine the association relationship between the plurality of long-tail data based on the similarity between the plurality of long-tail data;
[0140] Analyze the similarities between the plurality of long-tail data to determine a plurality of data groups;
[0141] Determine a data pattern corresponding to the data group; the data pattern is used to characterize a pattern corresponding to long-tail data that appears repeatedly in the data group;
[0142] The long-tail data recognition model is optimized and updated according to the long-tail data, the association relationship between the long-tail data, the data group and the data pattern.
[0143] In the embodiment of the present invention, in order to continuously improve the accuracy and generalization ability of the recognition model, the long-tail data recognition module also adopts incremental learning technology, so that the long-tail data recognition model can continuously learn and adapt to new long-tail data.
[0144] In addition, similarity metrics (such as cosine similarity, Euclidean distance, etc.) are used to calculate the similarity between multiple long-tail data. Based on the similarity, association rule mining algorithms (such as Apriori or FP-Growth) are used to discover the association relationship between long-tail data and reveal the potential connection between data. Cluster analysis (such as K-means or DBSCAN) is also used to discover hidden patterns and groups (data patterns and data groups) in long-tail data, providing strong support for subsequent model optimization.
[0145] Specifically, association rule mining algorithms are used to discover the associations between long-tail data. These algorithms can identify frequently occurring item sets and rules in a data set, thereby revealing the potential connections between long-tail data. For example, in the field of autonomous driving, assuming that long-tail data includes "sudden traffic jams" and "foggy weather", through association rule mining, it can be found that these two events often occur at the same time, forming an association rule "If you encounter a sudden traffic jam, it is likely that foggy weather will occur at the same time."
[0146] In the process of discovering data groups and data patterns in long-tail data by cluster analysis based on the similarity between long-tail data, cluster analysis is an unsupervised learning method that groups objects in a data set according to similarity to form different clusters. Each cluster represents a data group. Objects within a cluster have high similarity, while objects between different clusters have low similarity. The data patterns corresponding to each data group are determined. These patterns characterize the common characteristics of long-tail data that recur in the group. For example, in the field of autonomous driving, assume that long-tail data includes the performance of "tire burst" under different driving conditions. Through cluster analysis, it can be found that the performance patterns of "tire burst" are different under high-speed driving and low-speed driving conditions, forming two different clusters, representing tire burst under high-speed driving conditions and tire burst under low-speed driving conditions, respectively.
[0147] Optimizing and updating the long-tail data recognition model based on long-tail data, the correlation between long-tail data, data groups and data patterns can improve the accuracy and generalization ability of the long-tail data recognition model and enable it to better adapt to new long-tail data.
[0148] In one embodiment of the present invention, the step 107 of optimizing the decision model of the vehicle according to the labeled long-tail data includes:
[0149] Determining, in the annotated long-tail data, annotated feature data corresponding to the annotated information according to the annotated information in the annotated long-tail data; the annotated information includes driving scenes, driving difficulty, and driving risk factors;
[0150] Training and adjusting the decision model according to the annotated feature data;
[0151] Evaluate the decision model obtained through training and adjustment to obtain evaluation results;
[0152] The decision model obtained through training and adjustment is optimized according to the evaluation results, so that the decision model outputs corresponding decisions when dealing with different driving scenarios, driving difficulties and driving risk factors.
[0153] In an embodiment of the present invention, the decision model of the autonomous driving system is optimized through a model optimization module. Specifically, the model optimization module optimizes the decision model of the autonomous driving system using the mined long-tail data. By incorporating these rare situations (long-tail data) into the training set, the accuracy and robustness of the decision model in dealing with similar situations can be improved. The optimization process uses advanced machine learning algorithms (such as deep reinforcement learning) and combines the characteristics and annotation information of long-tail data to iteratively train and adjust the decision model. The optimized decision model can better cope with complex and rare driving scenarios and improve the safety and reliability of the autonomous driving system.
[0154] In an embodiment of the present invention, first, based on the annotation information in the annotated long-tail data, the annotated feature data corresponding to the annotated information is determined. The annotated information may include driving scenarios (such as night driving, rainy and snowy weather, etc.), driving difficulty (such as easy, medium, difficult, etc.) and driving risk factors (such as high-speed driving, emergency braking, etc.). These annotated information provide key supervisory signals for the training of the decision model. Subsequently, the decision model is trained and adjusted based on the annotated feature data. In this training and adjustment process, it is usually involved to use a machine learning algorithm (such as deep reinforcement learning) to train the model so that it can output corresponding decisions based on the input feature data.
[0155] After the model training and adjustment is completed, the decision model obtained through training and adjustment needs to be evaluated to obtain the evaluation results. The evaluation usually involves using a validation set or a test set to test the performance of the model. The evaluation indicators may include accuracy, recall rate, F1 score, etc. The evaluation results reflect the performance of the model in dealing with different driving scenarios, driving difficulties and driving risk factors.
[0156] The decision model obtained by training and adjustment is optimized according to the evaluation results. If the evaluation results show that the model performs poorly in some cases, the decision model can be optimized by adjusting model parameters, increasing training data, improving feature engineering, etc. The optimized decision model can better cope with complex and rare driving scenarios and improve the safety and reliability of the autonomous driving system.
[0157] In practical applications, data annotation is usually performed on image data. However, the embodiments of the present invention can not only perform data annotation on image data, but also perform data annotation on other types of driving data, such as vehicle status data, environmental perception data, etc.
[0158] In an embodiment of the present invention, during the training and adjustment process of the decision model, the annotation information of various driving scenarios, driving difficulty and driving risk factors is combined, so that the model can better adapt to various complex driving environments, enhance the adaptability and robustness of the model, and by combining machine learning algorithms and long-tail data, the safety and reliability of the autonomous driving system can be further improved.
[0159] In an embodiment of the present invention, after the long-tail data is labeled according to the characteristics of the long-tail data in step 106 to obtain labeled long-tail data, the method further includes:
[0160] The annotated long-tail data is stored in a target database; the annotated long-tail data is displayed through a visual interface;
[0161] Searching the target database according to the keywords corresponding to the annotated long-tail data;
[0162] and / or,
[0163] The target database is searched according to the conditional keywords corresponding to the annotated long-tail data; the conditional keywords are obtained by conditionally combining multiple keywords.
[0164] In the field of autonomous driving, long-tail data usually involves rare but critical driving scenarios. The collection and management of this data is complex, and traditional data management methods often have difficulty in effectively processing and utilizing this data. By annotating long-tail data, we can clarify the category, status, importance and other information of the data, and improve the quality and availability of the data. However, how to efficiently manage and retrieve this annotated data is a problem that has not been fully solved in existing technologies.
[0165] The embodiment of the present invention achieves efficient management and utilization of the annotated long-tail data by storing the annotated long-tail data in a target database and providing a search function based on keywords and conditional keywords.
[0166] In an embodiment of the present invention, the storage and retrieval of the annotated long-tail data are realized through the data storage and retrieval module. Specifically, the data storage and retrieval module is responsible for storing the annotated long-tail data in a high-performance database and providing an efficient retrieval function. The database is designed with full consideration of data security and scalability to ensure the stability and reliability of the annotated long-tail data during storage and retrieval.
[0167] In an embodiment of the present invention, the display of annotated long-tail data is realized through a user interface module. Specifically, the user interface module provides an intuitive and friendly operation interface to display the mined long-tail data and its annotated information. Users can view, analyze and process these data through the interface to further understand the performance and optimization direction of the autonomous driving system in rare situations. The interface design fully considers the user's operating habits and needs, and provides a variety of data display methods and interactive functions to facilitate users to conduct efficient data management and decision analysis.
[0168] In addition, based on the data storage and retrieval module and the user interface module, users can quickly retrieve the required long-tail data through keywords, condition combinations, etc., providing strong support for subsequent data analysis and model training. For example, users can quickly find relevant long-tail data by entering "foggy weather". This retrieval method is simple and direct, and is suitable for quickly locating specific types of data. Users can also combine the keywords "foggy weather" and "highway" to retrieve long-tail data under foggy weather on highways. This retrieval method is more flexible and can meet more complex data retrieval needs.
[0169] The embodiment of the present invention uses a retrieval function based on keywords and conditional keywords, so that the system can quickly locate and obtain the required annotated long-tail data. This efficient retrieval method greatly improves the efficiency of data retrieval and saves time and resources. The efficient retrieval function provides more accurate and reliable information support for the decision support of the autonomous driving system. The system can quickly obtain relevant long-tail data based on the retrieval results, optimize the decision-making process, and improve the accuracy and safety of the decision. By effectively utilizing the annotated long-tail data, the system can better respond to potential risks and challenges, thereby improving the performance and reliability of the autonomous driving system. The introduction of the retrieval function enables the system to more accurately identify and process various driving scenarios, thereby improving the overall performance of the system.
[0170] The embodiments of the present invention facilitate users to view, analyze and process long-tail data by providing an intuitive user interface, thereby providing strong support for the research and development and optimization of autonomous driving systems.
[0171] In an embodiment of the present invention, driving data of the vehicle is collected by a detection device of the vehicle; the driving data includes: vehicle status data, environmental perception data, driver behavior data and communication data; the driving data is processed to obtain pre-processed data; the data processing includes coarse-grained screening, spatiotemporal alignment, fine-grained screening and data standardization processing; feature processing is performed on the pre-processed data to obtain an optimal feature data set; the feature processing includes feature extraction and feature screening; based on the driving data and the optimal feature data set, multiple data results and weights corresponding to the multiple data results are obtained; wherein the weights are determined based on the correlation between the multiple data results; long-tail data is obtained based on the multiple data results and the weights corresponding to the multiple data results; the long-tail data is labeled according to the characteristics of the long-tail data to obtain labeled long-tail data; the decision model of the vehicle is optimized based on the labeled long-tail data; the decision model is used to guide the vehicle to make decisions. The embodiment of the present invention collects a variety of driving data including vehicle status data, environmental perception data, driver behavior data and communication data through the vehicle detection equipment, which not only expands the scope of data collection, but also increases the diversity of data. This wide data collection method helps to capture more rare long-tail data, thereby providing a richer information source for the optimization of subsequent decision models. In the data processing and feature processing stages, the quality and consistency of the data are ensured through coarse-grained screening, spatiotemporal alignment, fine-grained screening and data standardization processing. Through feature extraction and feature screening, the most valuable features for the decision model can be extracted from a large amount of data, which not only improves the efficiency of data processing, but also enhances the generalization ability and adaptability of the decision model. Traditional autonomous driving systems often ignore the importance of long-tail data, while the embodiment of the present invention can effectively identify and utilize these long-tail data through refined data processing, feature processing and other processing methods, so that the autonomous driving system can better cope with rare situations and reduce accident risks. Through automated data collection and labeling processes, the efficiency and accuracy of data processing are improved, and the decision model is optimized using long-tail data to improve the decision-making ability of the autonomous driving system in complex and rare scenarios.
[0172] It should be noted that, for the sake of simplicity, the method embodiments are described as a series of action combinations, but those skilled in the art should be aware that the embodiments of the present invention are not limited by the order of the actions described, because according to the embodiments of the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present invention.
[0173] Reference Figure 2, shows a structural block diagram of a driving data processing device provided in an embodiment of the present invention, which may specifically include the following modules:
[0174] The acquisition module 201 is used to acquire the driving data of the vehicle through the detection equipment of the vehicle;
[0175] The data processing module 202 is used to process the driving data to obtain pre-processed data; the data processing includes coarse-grained screening, spatiotemporal alignment, fine-grained screening and data standardization processing;
[0176] The feature processing module 203 is used to perform feature processing on the preprocessed data to obtain an optimal feature data set; the feature processing includes feature extraction and feature screening;
[0177] A weight module 204 is used to obtain a plurality of data results and weights corresponding to the plurality of data results according to the driving data and the optimal feature data set; wherein the weights are determined based on the correlation between the plurality of data results;
[0178] A long-tail data module 205 is used to obtain long-tail data according to the plurality of data results and the weights respectively corresponding to the plurality of data results;
[0179] A labeling module 206, configured to label the long-tail data according to the characteristics of the long-tail data to obtain labeled long-tail data;
[0180] The optimization module 207 is used to optimize the decision model of the vehicle according to the labeled long-tail data; the decision model is used to guide the vehicle to make a decision.
[0181] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.
[0182] In addition, an embodiment of the present invention further provides an electronic device, including: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the various processes of the above-mentioned driving data processing method embodiment are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be described here.
[0183] The embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, each process of the above-mentioned driving data processing method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it is not repeated here. The computer-readable storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.
[0184] An embodiment of the present invention also provides a computer program product, which is stored in a storage medium. The program product is executed by at least one processor to implement the various processes of the driving data processing method embodiment as described above, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.
[0185] Figure 3 A schematic diagram of the hardware structure of an electronic device for implementing various embodiments of the present invention.
[0186] The electronic device 300 includes but is not limited to: a radio frequency unit 301, a network module 302, an audio output unit 303, an input unit 304, a sensor 305, a display unit 306, a user input unit 307, an interface unit 308, a memory 309, a processor 310, and a power supply 311. Those skilled in the art will appreciate that Figure 3 The electronic device structure shown in the figure does not constitute a limitation on the electronic device, and the electronic device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently. In the embodiments of the present invention, the electronic device includes but is not limited to a mobile phone, a tablet computer, a laptop computer, a PDA, a vehicle-mounted terminal, a wearable device, and a pedometer.
[0187] It should be understood that in the embodiment of the present invention, the radio frequency unit 301 can be used for receiving and sending signals during information transmission or communication. Specifically, after receiving downlink data from the base station, it is sent to the processor 310 for processing; in addition, uplink data is sent to the base station. Generally, the radio frequency unit 301 includes but is not limited to an antenna, at least one amplifier, a transceiver, a coupler, a low noise amplifier, a duplexer, etc. In addition, the radio frequency unit 301 can also communicate with the network and other devices through a wireless communication system.
[0188] The electronic device provides users with wireless broadband Internet access through the network module 302, such as helping users to send and receive emails, browse web pages, and access streaming media.
[0189] The audio output unit 303 can convert the audio data received by the RF unit 301 or the network module 302 or stored in the memory 309 into an audio signal and output it as sound. Moreover, the audio output unit 303 can also provide audio output related to a specific function performed by the electronic device 300 (for example, a call signal reception sound, a message reception sound, etc.). The audio output unit 303 includes a speaker, a buzzer, a receiver, etc.
[0190] The input unit 304 is used to receive audio or video signals. The input unit 304 may include a graphics processor (GPU) 3041 and a microphone 3042, and the graphics processor 3041 processes the image data of a static picture or video obtained by an image capture device (such as a camera) in a video capture mode or an image capture mode. The processed image frame can be displayed on the display unit 306. The image frame processed by the graphics processor 3041 can be stored in the memory 309 (or other storage medium) or sent via the radio frequency unit 301 or the network module 302. The microphone 3042 can receive sound and can process such sound into audio data. The processed audio data can be converted into a format output that can be sent to a mobile communication base station via the radio frequency unit 301 in the case of a telephone call mode.
[0191] The electronic device 300 also includes at least one sensor 305, such as a light sensor, a motion sensor, and other sensors. Specifically, the light sensor includes an ambient light sensor and a proximity sensor, wherein the ambient light sensor can adjust the brightness of the display panel 3061 according to the brightness of the ambient light, and the proximity sensor can turn off the display panel 3061 and / or the backlight when the electronic device 300 is moved to the ear. As a type of motion sensor, the accelerometer sensor can detect the magnitude of acceleration in each direction (generally three axes), and can detect the magnitude and direction of gravity when stationary, which can be used to identify the posture of the electronic device (such as horizontal and vertical screen switching, related games, magnetometer posture calibration), vibration recognition related functions (such as pedometer, tapping), etc.; the sensor 305 can also include a fingerprint sensor, a pressure sensor, an iris sensor, a molecular sensor, a gyroscope, a barometer, a hygrometer, a thermometer, an infrared sensor, etc., which will not be repeated here.
[0192] The display unit 306 is used to display information input by the user or information provided to the user. The display unit 306 may include a display panel 3061, which may be configured in the form of a liquid crystal display (LCD), an organic light-emitting diode (OLED), or the like.
[0193] The user input unit 307 can be used to receive input digital or character information, and to generate key signal input related to user settings and function control of the electronic device. Specifically, the user input unit 307 includes a touch panel 3071 and other input devices 3072. The touch panel 3071, also known as a touch screen, can collect the user's touch operation on or near it (such as the user's operation on the touch panel 3071 or near the touch panel 3071 using any suitable object or accessory such as a finger, stylus, etc.). The touch panel 3071 may include two parts: a touch detection device and a touch controller. Among them, the touch detection device detects the user's touch orientation, detects the signal brought by the touch operation, and transmits the signal to the touch controller; the touch controller receives the touch information from the touch detection device, converts it into the contact point coordinates, and then sends it to the processor 310, receives the command sent by the processor 310 and executes it. In addition, the touch panel 3071 can be implemented using multiple types such as resistive, capacitive, infrared and surface acoustic waves. In addition to the touch panel 3071, the user input unit 307 may also include other input devices 3072. Specifically, other input devices 3072 may include but are not limited to a physical keyboard, function keys (such as volume control keys, switch keys, etc.), a trackball, a mouse, and a joystick, which are not described in detail here.
[0194] Furthermore, the touch panel 3071 may be overlaid on the display panel 3061. When the touch panel 3071 detects a touch operation on or near it, it is transmitted to the processor 310 to determine the type of touch event. Then, the processor 310 provides corresponding visual output on the display panel 3061 according to the type of touch event. Figure 3 In the figure, the touch panel 3071 and the display panel 3061 are used as two independent components to realize the input and output functions of the electronic device. However, in some embodiments, the touch panel 3071 and the display panel 3061 can be integrated to realize the input and output functions of the electronic device, which is not limited here.
[0195] The interface unit 308 is an interface for connecting an external device to the electronic device 300. For example, the external device may include a wired or wireless headset port, an external power supply (or battery charger) port, a wired or wireless data port, a memory card port, a port for connecting a device with an identification module, an audio input / output (I / O) port, a video I / O port, a headphone port, etc. The interface unit 308 may be used to receive input (e.g., data information, power, etc.) from an external device and transmit the received input to one or more elements within the electronic device 300 or may be used to transmit data between the electronic device 300 and an external device.
[0196] The memory 309 can be used to store software programs and various data. The memory 309 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory 309 can include a high-speed random access memory, and can also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.
[0197] The processor 310 is the control center of the electronic device. It uses various interfaces and lines to connect various parts of the entire electronic device. By running or executing software programs and / or modules stored in the memory 309 and calling data stored in the memory 309, it performs various functions of the electronic device and processes data, thereby monitoring the electronic device as a whole. The processor 310 may include one or more processing units; preferably, the processor 310 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface, and application programs, and the modem processor mainly processes wireless communications. It is understandable that the above-mentioned modem processor may not be integrated into the processor 310.
[0198] The electronic device 300 may also include a power source 311 (such as a battery) for supplying power to each component. Preferably, the power source 311 may be logically connected to the processor 310 through a power management system, thereby implementing functions such as charging, discharging, and power consumption management through the power management system.
[0199] In addition, the electronic device 300 includes some functional modules not shown, which will not be described in detail here.
[0200] like Figure 4 As shown, in another embodiment provided by the present invention, a computer-readable storage medium 401 is also provided, in which instructions are stored. When the computer-readable storage medium 401 is run on a computer, the computer executes the method for processing driving data described in the above embodiment.
[0201] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.
[0202] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for enabling a terminal (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present invention.
[0203] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are within the protection of the present invention.
[0204] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in the embodiments of the present invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.
[0205] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.
[0206] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.
[0207] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0208] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.
[0209] If the functions are implemented in the form of 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 the present invention, 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. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or 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: various media that can store program codes, such as USB flash drives, mobile hard disks, ROM, RAM, magnetic disks, or optical disks.
[0210] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.
Claims
1. A method for processing driving data, characterized in that: The method comprises: Collecting driving data of the vehicle through vehicle detection equipment; the driving data includes: vehicle status data, environment perception data, driver behavior data and communication data; Processing the driving data to obtain preprocessed data; the data processing includes coarse-grained screening, spatiotemporal alignment, fine-grained screening and data standardization processing; Performing feature processing on the preprocessed data to obtain an optimal feature data set; the feature processing includes feature extraction and feature screening; According to the driving data and the optimal feature data set, a plurality of data results and weights respectively corresponding to the plurality of data results are obtained; wherein the weights are determined based on the correlation between the plurality of data results; Obtaining long-tail data according to the plurality of data results and the weights respectively corresponding to the plurality of data results; Annotating the long-tail data according to the characteristics of the long-tail data to obtain annotated long-tail data; The decision model of the vehicle is optimized according to the labeled long-tail data; the decision model is used to guide the vehicle to make a decision.
2. The method according to claim 1, characterized in that The step of processing the driving data to obtain pre-processed data includes: Performing the coarse-grained screening on the driving data according to a data quality threshold to obtain first data; Performing the spatiotemporal alignment on the spatiotemporal information of the first data to obtain second data; the spatiotemporal information includes time information and space information; Performing the fine-grained screening on the second data to obtain third data; The third data is subjected to the data standardization process to obtain preprocessed data; the data standardization process is used to convert the third data into a target data format.
3. The method according to claim 1, characterized in that The performing feature processing on the preprocessed data to obtain an optimal feature data set includes: Performing the feature extraction on the preprocessed data to obtain feature data sets corresponding to a plurality of feature types; The feature screening is performed on the feature data sets respectively corresponding to the multiple feature types to obtain the optimal feature data set; the optimal feature data set includes the optimal feature data respectively corresponding to the multiple feature types.
4. The method according to claim 1, characterized in that: The step of obtaining a plurality of data results and weights respectively corresponding to the plurality of data results according to the driving data and the optimal feature data set includes: Constructing a long-tail data recognition model according to the optimal feature data set; Inputting the driving data into the long-tail data recognition model for data recognition to obtain a plurality of data results; Constructing the plurality of data results into a graph structure; wherein the graph nodes in the graph structure are the data results; Determine the correlation coefficient between adjacent graph nodes in the graph structure, and determine the weights corresponding to the plurality of data results based on the correlation coefficient.
5. The method according to claim 1, characterized in that The obtaining of long-tail data according to the plurality of data results and the weights respectively corresponding to the plurality of data results includes: Multiplying the data result by the weight corresponding to the data result to obtain weighted data corresponding to the data result; The weighted data corresponding to the plurality of data results are aggregated to obtain the long-tail data.
6. The method according to claim 1, characterized in that The optimizing the decision model of the vehicle according to the labeled long-tail data includes: Determining, in the annotated long-tail data, annotated feature data corresponding to the annotated information according to the annotated information in the annotated long-tail data; the annotated information includes driving scenes, driving difficulty, and driving risk factors; Training and adjusting the decision model according to the annotated feature data; Evaluate the decision model obtained through training and adjustment to obtain evaluation results; The decision model obtained through training and adjustment is optimized according to the evaluation results, so that the decision model outputs corresponding decisions when dealing with different driving scenarios, driving difficulties and driving risk factors.
7. The method according to claim 1, characterized in that After labeling the long-tail data according to the characteristics of the long-tail data to obtain the labeled long-tail data, the method further includes: The annotated long-tail data is stored in a target database; the annotated long-tail data is displayed through a visual interface; Searching the target database according to the keywords corresponding to the annotated long-tail data; and / or, The target database is searched according to the conditional keywords corresponding to the annotated long-tail data; the conditional keywords are obtained by conditionally combining multiple keywords.
8. A driving data processing device, characterized in that: The device comprises: A collection module, used to collect the driving data of the vehicle through the detection equipment of the vehicle; the driving data includes: vehicle status data, environment perception data, driver behavior data and communication data; A data processing module, used for processing the driving data to obtain pre-processed data; the data processing includes coarse-grained screening, spatiotemporal alignment, fine-grained screening and data standardization processing; A feature processing module, used to perform feature processing on the preprocessed data to obtain an optimal feature data set; the feature processing includes feature extraction and feature screening; A weight module, used to obtain a plurality of data results and weights corresponding to the plurality of data results respectively according to the driving data and the optimal feature data set; wherein the weights are determined based on the correlation between the plurality of data results; A long-tail data module, used for obtaining long-tail data according to the plurality of data results and the weights respectively corresponding to the plurality of data results; A labeling module, used for labeling the long-tail data according to the characteristics of the long-tail data to obtain labeled long-tail data; An optimization module is used to optimize the decision model of the vehicle according to the labeled long-tail data; the decision model is used to guide the vehicle to make decisions.
9. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; The memory is used to store computer programs; The processor is used to implement the method according to any one of claims 1 to 7 when executing the program stored in the memory.
10. A computer-readable storage medium having instructions stored thereon, which, when executed by one or more processors, cause the processors to perform the method according to any one of claims 1 to 7.
Citation Information
Patent Citations
Optimization method and system for long-tail data recognition model
CN116310590A
Data processing method and system for autonomous vehicle
CN118289029A
Decision model training and using method and device in automatic driving vehicle and vehicle
CN118323198A
Automated iteration method for target detection model, device and storage medium
WO2023045935A1
Data processing method and apparatus, computer device, vehicle, and readable storage medium
WO2024119412A1
Cited By
Method and system for storing driving data of autonomous vehicle, and medium
CN121191236A
