Data closed-loop device and system for intelligent networked automobile
Through the data closed-loop device of intelligent connected vehicles, the combination of data acquisition module, labeling training module, simulation verification module and model deployment module is solved, and the problem of lack of data closed-loop in intelligent vehicle data acquisition and model training is realized, and the on-board computing power is realized The effective utilization and continuous iteration of the model have improved the perception and security of intelligent connected vehicles.
Patent Information
- Application Number
- CN202411269190.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-11
- Publication Date
- 2025-01-10
AI Technical Summary
The lack of data closed loop in the data acquisition and model training of smart cars has led to insufficient on-board computing power and cannot meet the needs of data growth. High computing power is a challenge for heat dissipation, which affects the iterative upgrade of AI models.
A closed-loop data device for intelligent connected vehicles was designed, including data acquisition module, labeling training module, simulation verification module and model deployment module. Data is collected through sensors, uploaded to the cloud for labeling and training, and generated models in a virtual environment. Perform simulation verification and deploy to on-board equipment through OTA technology to achieve continuous iteration of data and models.
Through the data closed-loop system, intelligent connected vehicles can continuously learn and optimize from actual driving scenarios, improve their perception of road environment and traffic conditions, and reduce the occurrence of traffic accidents.
Smart Images
Figure CN119294226A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of artificial intelligence technology, and more specifically, to a data closed-loop device and system for intelligent networked vehicles. Background Art
[0004] A strong transportation country is closely related to the intelligent road facilities and vehicles. The coordination between vehicles and roads not only promotes the development of intelligent connected vehicles, but also promotes the construction of a smart transportation system.
[0005] At present, smart cars use data collection vehicles to collect data separately, and then use the collected data offline to train models, and then deploy the trained models on the car. The whole process does not form a data closed loop, which is not conducive to the iteration and upgrade of AI models. In addition, the computing power on the vehicle side is limited. If the computing power of the background is not effectively used, the vehicle-mounted data will grow exponentially, and the vehicle-mounted computing power will be far from meeting the requirements. In addition, too high computing power is also a huge challenge for the heat dissipation of the vehicle side. Summary of the invention
[0006] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a data closed-loop device and system for intelligent connected vehicles to solve the problems raised in the above-mentioned background technology.
[0007] To achieve the above-mentioned purpose, the present invention provides the following technical solution: a data closed-loop device for intelligent connected vehicles, comprising a data acquisition module, a labeling training module, a simulation verification module, and a model deployment module;
[0008] Data collection module: collects data about the car's surroundings through sensors, and uploads the collected data to the annotation training module for processing and analysis. It also receives the new model delivered by the model deployment module, uses the newly deployed model to re-collect data, and re-uploads the collected data to the annotation training module.
[0009] Annotation training module: receives data from the data acquisition module, annotates and processes the collected data, uses the labeled data to train the machine learning model, and transmits the generated trained model to the simulation verification module;
[0010] Simulation verification module: receives the trained model from the annotation training module, performs simulation verification in a virtual environment, and uses OTA technology to send the model that has passed the simulation verification to the model deployment module;
[0011] Model deployment module: receives the model transmitted from the simulation verification module, performs engineering deployment based on TensorRT, and applies the redeployed model to the data acquisition module.
[0012] In a preferred embodiment, the data acquisition module is deployed on the vehicle-mounted device, collects data about the environment around the vehicle through sensors, uploads the collected data to the annotation training module for processing and analysis, receives the new model delivered by the model deployment module, re-collects data using the newly deployed model, and re-uploads the collected data to the annotation training module. The specific steps are as follows:
[0013] Step A1, data collection and preprocessing: Use on-board sensors to continuously collect environmental data around the car, including road conditions, the position and speed of nearby vehicles, images, videos, and radar data of pedestrian behavior, and add timestamps to record the time when the data was generated. Preprocess the collected data, including data denoising, coordinate system conversion, and data format conversion operations, for subsequent processing and analysis;
[0014] Step A2, data transmission: upload the preprocessed data to the cloud through the data transmission protocol for storage, so that the annotation training module can process and analyze it;
[0015] Step A3, data quality control: After uploading the data to the cloud, data quality control is performed, including data integrity check, abnormal data identification and repair, to ensure the accuracy and reliability of subsequent labeling and training work;
[0016] Step A4, receiving the newly deployed model: receiving the new model delivered by the model deployment module, re-performing data collection work according to the newly deployed model, and collecting updated environmental data;
[0017] Step A5, uploading data to the annotation training module: After the collected data is processed and compressed, the data is uploaded to the annotation training module again through a secure data transmission protocol.
[0018] In a preferred embodiment, the annotation training module is deployed in the cloud, receives data from the data acquisition module, annotates and processes the collected data, uses the labeled data to train the machine learning model, and transmits the generated trained model to the simulation verification module. The specific steps are as follows:
[0019] Step B1, data annotation: receiving images, videos, and radar data uploaded from the data acquisition module, wherein the received data includes vehicle, pedestrian, and road sign target information, and annotating the received data using a computer vision algorithm to mark the vehicles, pedestrians, and road sign targets therein, wherein the annotated data includes target category, location coordinates, and boundary information;
[0020] Step B2, feature extraction: After the labeling is completed, feature extraction is performed on the labeled data, including color, texture, shape and spatial features, and the extracted color, texture, shape and spatial features are combined into a feature vector as input data for training the model;
[0021] Step B3: Model training: Use the labeled data and extracted features to train the convolutional neural network model, and optimize the trained model to improve the performance and generalization ability of the model;
[0022] Step B4, transmitting to the simulation verification module: transmitting the trained and optimized model to the simulation verification module for simulation verification and testing in a virtual environment, and receiving the verification results from the simulation verification module to improve the training data.
[0023] In a preferred embodiment, in the feature extraction of step B2, after the labeling is completed, feature extraction is performed on the labeled data, including color, texture, shape and spatial features, further comprising the following steps:
[0024] Step B201, color features: convert the image into RGB color space using a histogram statistical method, count the pixel value distribution of each channel, and obtain a color histogram;
[0025] Step B202, texture features: using a local binary pattern to locally binarize the image, and counting the local texture pattern of each pixel to describe the texture information of the image;
[0026] Step B203, shape features: using edge detection algorithm to detect edge information in the image;
[0027] Step B204, spatial features: using the bounding box coordinates in the annotation information as spatial features.
[0028] In a preferred embodiment, in the step B3 model training, the convolutional neural network model is trained using the labeled data and the extracted features, further comprising the following steps:
[0029] Step B301, model construction: construct a convolutional neural network model consisting of multiple convolutional layers, pooling layers and fully connected layers for target detection;
[0030] Step B302, hyperparameter tuning: adjust the hyperparameters of the model, including learning rate, batch size, and optimizer type;
[0031] Step B303, model training: use the labeled data and extracted features to train the convolutional neural network model, and use the back propagation algorithm to update the model parameters;
[0032] Step B304, cross validation: Use cross validation technology to evaluate the trained model to improve the performance and generalization ability of the model.
[0033] In a preferred embodiment, the simulation verification module is deployed in the cloud, receives the trained model from the annotation training module, performs simulation verification in a virtual environment, and uses OTA technology to send the model that has passed the simulation verification to the model deployment module. The specific steps are as follows:
[0034] Step C1, simulation scenario design: select a virtual simulation environment and design simulation scenarios for various road conditions and traffic situations, including different weather conditions, road surface conditions, and vehicle behaviors;
[0035] Step C2, model integration: Integrate the model trained by the annotation training module into the simulation environment, use the integrated model to perform target detection and prediction in the virtual environment, and annotate the output results of the model;
[0036] Step C3, verification result analysis: Analyze the performance of the model in different simulation scenarios, including the accuracy of target detection and the stability of prediction, and record the verification results;
[0037] Step C4: Model delivery: Use OTA technology to deliver the model that has passed simulation verification to the model deployment module.
[0038] In a preferred embodiment, in the step C4, in sending the model, the model verified by simulation is sent to the model deployment module using OTA technology, further comprising the following steps:
[0039] Step C401, model push: Model push: package the model verified by simulation and create an OTA task on the OTA platform;
[0040] Step C402, data transmission: Use the secure channel provided by the OTA platform to transmit the packaged model to the model deployment module in an encrypted manner to ensure data integrity and confidentiality.
[0041] In a preferred embodiment, the model deployment module is deployed on the vehicle-mounted device, receives the model transmitted from the simulation verification module, performs engineering deployment based on TensorRT, and applies the redeployed model to the data acquisition module. The specific steps are as follows:
[0042] Step D1, device registration and authentication: register the model deployment module to the OTA platform and perform device authentication to ensure that only authorized devices can receive OTA updates;
[0043] Step D2, model update and rollback mechanism: After receiving the OTA push, the model deployment module updates the model and establishes a model rollback mechanism to deal with unexpected situations that occur during the update process;
[0044] Step D3, update confirmation and feedback: After the model deployment module completes the model update, it sends update confirmation information to the OTA platform;
[0045] Step D4, TensorRT engineering deployment: prepare the TensorRT operating environment, load the received model file, use TensorRT to parse, optimize and build the inference engine, and generate an optimized inference engine for the vehicle computing platform;
[0046] Step D5, apply the inference engine: deploy the TensorRT-optimized inference engine to the data acquisition module for real-time road scene perception and decision making.
[0047] The present application also provides a data closed-loop system for intelligent connected vehicles, which uploads the collected data to the labeling training module through the data acquisition module, and receives the new model transmitted by the model deployment module; the labeling training module receives the data from the data acquisition module, marks it for training the machine learning model, and transmits the trained model to the simulation training module; the simulation training module receives the model from the labeling training module, performs simulation verification, and uses OTA technology to send the model that passes the simulation verification to the model deployment module; the model deployment module receives the model transmitted from the simulation verification module, and redeploys it and applies it to the data acquisition module, and repeats the iteration of the data acquisition module, the labeling training module, the simulation verification module and the model deployment module.
[0048] The beneficial effects of the present invention are as follows: the intelligent connected vehicle continuously learns and optimizes from actual driving scenarios in a data closed-loop system, thereby improving the intelligent connected vehicle's perception of road environment, traffic conditions and other vehicles, and reducing the occurrence of traffic accidents. BRIEF DESCRIPTION OF THE DRAWINGS
[0049] Figure 1 is a system flow chart of the present invention;
[0050] Figure 2 It is a structural block diagram of the present invention. DETAILED DESCRIPTION
[0051] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.
[0052] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, "plurality" means two or more, unless otherwise clearly and specifically defined.
[0053] In the description of the present application, the term "for example" is used to mean "used as an example, illustration or description". Any embodiment described as "for example" in the present application is not necessarily to be construed as being more preferred or advantageous than other embodiments. In order to enable any technician in the field to implement and use the present invention, the following description is given. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes will not be elaborated in detail to avoid unnecessary details that make the description of the present invention obscure. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed in the present application.
[0054] Example 1
[0055] This embodiment provides Figure 1 A data closed-loop device for intelligent connected vehicles is shown, which specifically includes a data acquisition module, a labeling training module, a simulation verification module, and a model deployment module;
[0056] Data collection module: collects data about the car's surroundings through sensors, and uploads the collected data to the annotation training module for processing and analysis. It also receives the new model delivered by the model deployment module, uses the newly deployed model to re-collect data, and re-uploads the collected data to the annotation training module.
[0057] Annotation training module: receives data from the data acquisition module, annotates and processes the collected data, uses the labeled data to train the machine learning model, and transmits the generated trained model to the simulation verification module;
[0058] Simulation verification module: receives the trained model from the annotation training module, performs simulation verification in a virtual environment, and uses OTA technology to send the model that has passed the simulation verification to the model deployment module;
[0059] Model deployment module: receives the model transmitted from the simulation verification module, performs engineering deployment based on TensorRT, and applies the redeployed model to the data acquisition module.
[0060] In this embodiment, the data acquisition module specifically needs to be explained. The data acquisition module is deployed on the vehicle-mounted device, collects data about the environment around the car through sensors, uploads the collected data to the annotation training module for processing and analysis, receives the new model delivered by the model deployment module, re-collects data using the newly deployed model, and re-uploads the collected data to the annotation training module. The specific steps are as follows:
[0061] Step A1, data collection and preprocessing: Use on-board sensors to continuously collect environmental data around the car, including road conditions, the position and speed of nearby vehicles, images, videos, and radar data of pedestrian behavior, and add timestamps to record the time when the data was generated. Preprocess the collected data, including data denoising, coordinate system conversion, and data format conversion operations, for subsequent processing and analysis;
[0062] Step A2, data transmission: upload the preprocessed data to the cloud through the data transmission protocol for storage, so that the annotation training module can process and analyze it;
[0063] Step A3, data quality control: After uploading the data to the cloud, data quality control is performed, including data integrity check, abnormal data identification and repair, to ensure the accuracy and reliability of subsequent labeling and training work;
[0064] Step A4, receiving the newly deployed model: receiving the new model delivered by the model deployment module, re-performing data collection work according to the newly deployed model, and collecting updated environmental data;
[0065] Step A5, uploading data to the annotation training module: After the collected data is processed and compressed, the data is uploaded to the annotation training module again through a secure data transmission protocol.
[0066] In this embodiment, what needs to be explained specifically is the labeling training module, which is deployed in the cloud, receives data from the data acquisition module, labels and processes the collected data, uses the labeled data to train the machine learning model, and transmits the generated trained model to the simulation verification module. The specific steps are as follows:
[0067] Step B1, data annotation: receiving images, videos, and radar data uploaded from the data acquisition module, wherein the received data includes vehicle, pedestrian, and road sign target information, and annotating the received data using a computer vision algorithm to mark the vehicles, pedestrians, and road sign targets therein, wherein the annotated data includes target category, location coordinates, and boundary information;
[0068] Step B2, feature extraction: After the labeling is completed, feature extraction is performed on the labeled data, including color, texture, shape and spatial features, and the extracted color, texture, shape and spatial features are combined into a feature vector as input data for training the model;
[0069] Step B3: Model training: Use the labeled data and extracted features to train the convolutional neural network model, and optimize the trained model to improve the performance and generalization ability of the model;
[0070] Step B4, transmitting to the simulation verification module: transmitting the trained and optimized model to the simulation verification module for simulation verification and testing in a virtual environment, and receiving the verification results from the simulation verification module to improve the training data;
[0071] Among them, in the feature extraction of step B2, after the labeling is completed, feature extraction is performed on the labeled data, including color, texture, shape and spatial features, and further includes the following steps:
[0072] Step B201, color features: convert the image into RGB color space using a histogram statistical method, count the pixel value distribution of each channel, and obtain a color histogram;
[0073] Step B202, texture features: using a local binary pattern to locally binarize the image, and counting the local texture pattern of each pixel to describe the texture information of the image;
[0074] Step B203, shape features: using edge detection algorithm to detect edge information in the image;
[0075] Step B204, spatial features: using the bounding box coordinates in the annotation information as spatial features;
[0076] In the model training of step B3, the convolutional neural network model is trained using the labeled data and the extracted features, further comprising the following steps:
[0077] Step B301, model construction: construct a convolutional neural network model consisting of multiple convolutional layers, pooling layers and fully connected layers for target detection;
[0078] Step B302, hyperparameter tuning: adjust the hyperparameters of the model, including learning rate, batch size, and optimizer type;
[0079] Step B303, model training: use the labeled data and extracted features to train the convolutional neural network model, and use the back propagation algorithm to update the model parameters;
[0080] Step B304, cross validation: Use cross validation technology to evaluate the trained model to improve the performance and generalization ability of the model.
[0081] In this embodiment, the simulation verification module specifically needs to be explained. The simulation verification module is deployed in the cloud, receives the trained model from the annotation training module, performs simulation verification in a virtual environment, verifies the accuracy and reliability of the model by simulating various road conditions and traffic conditions, and uses OTA technology to send the model that has passed the simulation verification to the model deployment module. The specific steps are as follows:
[0082] Step C1, simulation scenario design: select a virtual simulation environment and design simulation scenarios for various road conditions and traffic situations, including different weather conditions, road surface conditions, and vehicle behaviors;
[0083] Step C2, model integration: Integrate the model trained by the annotation training module into the simulation environment, use the integrated model to perform target detection and prediction in the virtual environment, and annotate and record the accuracy and reliability of the model based on the output results of the model;
[0084] Step C3, verification result analysis: Analyze the performance of the model in different simulation scenarios, including the accuracy of target detection and the stability of prediction, and record the verification results;
[0085] Step C4, model delivery: Using OTA technology to deliver the model verified by simulation to the model deployment module, further including the following steps:
[0086] Step C401, model push: Model push: package the model verified by simulation and create an OTA task on the OTA platform;
[0087] Step C402, data transmission: Use the secure channel provided by the OTA platform to transmit the packaged model to the model deployment module in an encrypted manner to ensure data integrity and confidentiality.
[0088] In this embodiment, the model deployment module is specifically required to be explained. The model deployment module is deployed on the vehicle-mounted device, receives the model transmitted from the simulation verification module, performs engineering deployment based on TensorRT, and applies the redeployed model to the data acquisition module. The specific steps are as follows:
[0089] Step D1, device registration and authentication: register the model deployment module to the OTA platform and perform device authentication to ensure that only authorized devices can receive OTA updates;
[0090] Step D2, model update and rollback mechanism: After receiving the OTA push, the model deployment module updates the model and establishes a model rollback mechanism to deal with unexpected situations during the update process and ensure that the system can quickly recover to a stable state;
[0091] Step D3, update confirmation and feedback: After the model deployment module completes the model update, it sends update confirmation information to the OTA platform to ensure the integrity of the OTA task, test and verify the updated model, and ensure that the performance of the new model in the actual environment meets expectations;
[0092] Step D4, TensorRT engineering deployment: prepare the TensorRT operating environment, load the received model file, use TensorRT to parse, optimize and build the inference engine, and generate an optimized inference engine for the vehicle computing platform;
[0093] Step D5, apply the inference engine: deploy the TensorRT-optimized inference engine to the data acquisition module for real-time road scene perception and decision making.
[0094] Example 2
[0095] This embodiment provides Figure 2 A data closed-loop system for intelligent connected vehicles is shown, wherein the data closed-loop system refers to a data acquisition module uploading the collected data to a labeling training module, and receiving a new model transmitted by a model deployment module; the labeling training module receives data from the data acquisition module, labels it for training a machine learning model, and transmits the trained model to a simulation training module; the simulation training module receives the model from the labeling training module, performs simulation verification, and uses OTA technology to send the model that has passed the simulation verification to the model deployment module; the model deployment module receives the model transmitted from the simulation verification module, and redeploys it and applies it to the data acquisition module. The entire information flow of the data closed-loop system is closed-loop, and through repeated iterations, automated optimization and continuous improvement of the intelligent connected vehicle system are achieved.
[0096] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and for parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.
[0097] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0098] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0099] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0100] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0101] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.
[0102] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.
Claims
1. A data closed-loop device for intelligent connected vehicles, characterized in that: It includes data collection module, annotation training module, simulation verification module, and model deployment module; Data collection module: collects data about the car's surroundings through sensors, and uploads the collected data to the annotation training module for processing and analysis. It also receives the new model delivered by the model deployment module, uses the newly deployed model to re-collect data, and re-uploads the collected data to the annotation training module. Annotation training module: receives data from the data acquisition module, annotates and processes the collected data, uses the labeled data to train the machine learning model, and transmits the generated trained model to the simulation verification module; Simulation verification module: receives the trained model from the annotation training module, performs simulation verification in a virtual environment, and uses OTA technology to send the model that has passed the simulation verification to the model deployment module; Model deployment module: receives the model transmitted from the simulation verification module, performs engineering deployment based on TensorRT, and applies the redeployed model to the data acquisition module.
2. A data closed-loop device for intelligent networked vehicles according to claim 1, characterized in that: The data acquisition module collects data about the environment around the car through sensors, and uploads the collected data to the annotation training module for processing and analysis, and receives the new model delivered by the model deployment module, re-collects data using the newly deployed model, and re-uploads the collected data to the annotation training module. The specific steps are as follows: Step A1, data collection and preprocessing: Use vehicle-mounted sensors to collect environmental data around the car, including images, videos, and radar data, and add timestamps to record the time when the data was generated. Preprocess the collected data, including data denoising, coordinate system conversion, and data format conversion operations; Step A2, data transmission: uploading the preprocessed data to the cloud for storage through a data transmission protocol; Step A3, data quality control: after uploading the data to the cloud, data quality control is performed, including data integrity check, abnormal data identification and repair; Step A4, receiving the newly deployed model: receiving the new model delivered by the model deployment module, re-performing data collection according to the newly deployed model, and collecting updated environmental data; Step A5, uploading data to the annotation training module: after the collected data is processed and compressed, the data is uploaded to the annotation training module again through the data transmission protocol.
3. A data closed-loop device for intelligent networked vehicles according to claim 2, characterized in that: The labeling training module receives data from the data acquisition module, labels and processes the collected data, uses the labeled data to train the machine learning model, and transmits the generated trained model to the simulation verification module. The specific steps are as follows: Step B1, data labeling: receiving images, videos, and radar data uploaded from the data acquisition module, labeling the data using computer vision algorithms, marking vehicles, pedestrians, and road sign targets, and recording target categories, location coordinates, and boundary information; Step B2, feature extraction: After the labeling is completed, feature extraction is performed on the labeled data, including color, texture, shape and spatial features, and the extracted color, texture, shape and spatial features are combined into a feature vector as input data for training the model; Step B3, model training: Use the labeled data and extracted features to train the convolutional neural network model, and optimize the trained model; Step B4, transmitting to the simulation verification module: transmitting the trained and optimized model to the simulation verification module for simulation verification and testing in a virtual environment.
4. The data closed-loop device for intelligent networked vehicles according to claim 3, characterized in that: In the feature extraction of step B2, after the labeling is completed, feature extraction is performed on the labeled data, including color, texture, shape and spatial features, further comprising the following steps: Step B201, color features: convert the image into RGB color space using a histogram statistical method, count the pixel value distribution of each channel, and obtain a color histogram; Step B202, texture features: using a local binary pattern to locally binarize the image, and counting the local texture pattern of each pixel to describe the texture information of the image; Step B203, shape feature: using edge detection algorithm to detect edge information in the image; Step B204, spatial features: using the bounding box coordinates in the annotation information as spatial features.
5. The data closed-loop device for intelligent networked vehicles according to claim 4, characterized in that: In the model training of step B3, the convolutional neural network model is trained using the labeled data and the extracted features, further comprising the following steps: Step B301, model construction: construct a convolutional neural network model consisting of multiple convolutional layers, pooling layers and fully connected layers for target detection; Step B302, hyperparameter tuning: adjust the hyperparameters of the model, including learning rate, batch size, and optimizer type; Step B303, model training: use the labeled data and extracted features to train the convolutional neural network model, and update the model parameters through the back propagation algorithm; Step B304, cross validation: Use cross validation technology to evaluate the trained model to improve the performance and generalization ability of the model.
6. A data closed-loop device for intelligent networked vehicles according to claim 5, characterized in that: The simulation verification module receives the trained model from the annotation training module, performs simulation verification in a virtual environment, and uses OTA technology to send the model that has passed the simulation verification to the model deployment module. The specific steps are as follows: Step C1, simulation scene design: select a virtual simulation environment and design simulation scenes of various road conditions and traffic conditions; Step C2, model integration: Integrate the model trained by the annotation training module into the simulation environment, use the integrated model to perform target detection and prediction in the virtual environment, and annotate the output results of the model; Step C3, verification result analysis: analyzing the performance of the model in different simulation scenarios, including the accuracy of target detection and the stability of prediction; Step C4: Model delivery: Use OTA technology to deliver the model that has passed simulation verification to the model deployment module.
7. The data closed-loop device for intelligent networked vehicles according to claim 6, characterized in that: In the step C4, the model is sent down, using OTA technology to send the model verified by simulation to the model deployment module, further comprising the following steps: Step C401, model push: package the model verified by simulation and create an OTA task on the OTA platform; Step C402, data transmission: Use the secure channel provided by the OTA platform to transmit the packaged model to the model deployment module in an encrypted manner.
8. The data closed-loop device for intelligent networked vehicles according to claim 7, characterized in that: The model deployment module receives the model transmitted from the simulation verification module, performs engineering deployment based on TensorRT, and applies the redeployed model to the data acquisition module. The specific steps are as follows: Step D1, device registration and authentication: register the model deployment module to the OTA platform and perform device authentication; Step D2, model update and rollback mechanism: After receiving the OTA push, the model deployment module updates the model and establishes a model rollback mechanism to deal with unexpected situations that occur during the update process; Step D3, update confirmation and feedback: After the model deployment module completes the model update, it sends update confirmation information to the OTA platform; Step D4, TensorRT engineering deployment: prepare the TensorRT operating environment, load the received model file, use TensorRT to parse, optimize and build the inference engine, and generate an optimized inference engine for the vehicle computing platform; Step D5, apply the inference engine: deploy the TensorRT-optimized inference engine to the data acquisition module for real-time road scene perception and decision making.
9. A data closed-loop system for intelligent networked vehicles is applied to a data closed-loop device for intelligent networked vehicles as claimed in any one of claims 1 to 8, characterized in that: The data closed-loop system uploads the collected data to the annotation training module through the data acquisition module, and receives the new model transmitted by the model deployment module; the annotation training module receives the data from the data acquisition module, marks it for training the machine learning model, and transmits the trained model to the simulation training module; The simulation training module receives the model from the annotation training module, performs simulation verification, and uses OTA technology to send the model that has passed the simulation verification to the model deployment module; the model deployment module receives the model transmitted from the simulation verification module, and redeploys it and applies it to the data acquisition module, performing repeated iterations of the data acquisition module, annotation training module, simulation verification module and model deployment module.