Intelligent driving data multi-dimensional labeling method, system and device and storage medium

Abstract

The invention relates to the technical field of machine learning, and relates to an intelligent driving data multi-dimensional labeling method, system and device and a storage medium. The intelligent driving data multi-dimensional labeling method comprises the following steps: obtaining driving scene data to be labeled and environment information; determining an analysis rule of the driving scene data according to the environment information, and performing feature extraction on the driving scene data according to the determined analysis rule to obtain driving scene features; the driving scene features are input into a data labeling model, the data labeling model determines a target object of the driving scene and a label corresponding to the target object according to the driving scene features, and a first labeling result is obtained; determining a labeling confidence coefficient value of the target object according to the environment information; and adjusting the first labeling result according to the labeling confidence value to obtain a final labeling result. The accuracy of the target object labeling result can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of machine learning, and in particular to an intelligent driving data multi-dimensional labeling method, system, device and storage medium. BACKGROUND

[0002] With the development of automatic driving technology to L4 / L5 level, high-quality data labeling is the basis for training reliable perception models. The current industry generally adopts three types of technical solutions: traditional manual labeling relies on professional personnel to manually label targets in images or point clouds using LabelImg, CVAT and other tools; pure AI automatic labeling directly generates labeling results based on pre-trained models such as YOLO and Mask R-CNN; and multi-sensor labeling tools align camera and laser radar data through time stamp. However, with the development of automatic driving technology to L4 / L5 level, these existing solutions are facing increasingly severe challenges.

[0003] The wide application of multi-modal sensors leads to highly complex data dimensions, and heterogeneous data generated by cameras, laser radars and millimeter wave radars urgently need unified labeling standards. At the same time, dynamic environment adaptability has become an industry pain point - the labeling accuracy under extreme weather (such as heavy rain and sandstorm) is less than 50%, the mAP value of mainstream AI models in snowy scenes decreases by 62% compared to sunny days, and the false detection rate in foggy weather exceeds 30%. More seriously, automatic driving simulation testing requires real-time labeling capability of more than 100 frames per second, while the delay of traditional manual verification or offline batch processing exceeds 200 milliseconds, which cannot meet the real-time data flow requirements.

[0004] The existing technology has multiple structural defects. The multi-source heterogeneous data fusion capability is seriously insufficient, and mainstream tools such as LabelBox or Supervisely only support a single data type (2D image or 3D point cloud), lacking end-to-end joint labeling capability for multi-sensor data. Weak environmental dynamic adaptability leads to a sharp degradation in model performance under sudden changes in light or weather interference, for example, the labeling accuracy in foggy weather is only 70%. More prominent is the privacy compliance risk: 90% of labeling platforms do not have a built-in desensitization mechanism, directly transmitting raw video data containing license plates, faces and other sensitive information, violating the "traceable desensitization" principle of the European Union's GDPR. These defects together result in a labeling efficiency of multi-modal data that is only 40% of single-modal data, forming a bottleneck for technological upgrading.

[0005] In the face of the rapid development of the automatic driving industry, traditional labeling schemes have shown fundamental limitations. The fragmentation of single-modal tools hinders cross-sensor collaboration, static environment assumptions cannot cope with dynamic changes on real roads, and high labor costs restrict the speed of research and development iteration. SUMMARY

[0006] The application aims to at least solve the technical problems in the prior art, and provides an intelligent driving data multi-dimensional labeling method, system, device and storage medium.

[0007] In a first aspect, the application provides an intelligent driving data multi-dimensional labeling method, which comprises the following steps:

[0008] Obtaining driving scene data to be labeled and environment information, the environment information being used to represent weather conditions of the driving scene;

[0009] Determining an analysis rule for the driving scene data according to the environment information, and performing feature extraction on the driving scene data according to the determined analysis rule to obtain driving scene features;

[0010] Inputting the driving scene features into a data labeling model, the data labeling model determining target objects of the driving scene and labels corresponding to the target objects according to the driving scene features to obtain a first labeling result;

[0011] Determining a labeling confidence value of the target objects according to the environment information;

[0012] Adjusting the first labeling result according to the labeling confidence value to obtain a final labeling result.

[0013] In a second aspect, the application provides an intelligent driving data multi-dimensional labeling system, which comprises the following modules:

[0014] An obtaining module, which is used to obtain driving scene data to be labeled and environment information, the environment information being used to represent weather conditions of the driving scene;

[0015] A feature extraction module, which is used to determine an analysis rule for the driving scene data according to the environment information, and perform feature extraction on the driving scene data according to the determined analysis rule to obtain driving scene features;

[0016] A first output module, which is used to input the driving scene features into a data labeling model, the data labeling model determining target objects of the driving scene and labels corresponding to the target objects according to the driving scene features to obtain a first labeling result;

[0017] A processing module, which is used to determine a labeling confidence value of the target objects according to the environment information;

[0018] A second output module, which is used to adjust the first labeling result according to the labeling confidence value to obtain a final labeling result.

[0019] In a third aspect, the application provides an electronic device, which comprises the following modules:

[0020] At least one processor; and

[0021] A memory in communication connection with the at least one processor; wherein

[0022] The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to perform the intelligent driving data multi-dimensional labeling method described above.

[0023] In a fourth aspect, the present application further provides a computer readable storage medium, and the computer readable storage medium stores at least one computer program, and the at least one computer program is executed by a processor in an electronic device to implement the intelligent driving data multi-dimensional labeling method described above.

[0024] In summary, the present application includes the following beneficial technical effects:

[0025] According to the environmental information, the parsing rule of the driving scene data is determined, and the feature extraction is performed on the driving scene data according to the determined parsing rule, so that the extracted driving scene features contain more geometric information, texture information and contour information, thereby improving the recognition effect of the target object, and further improving the robustness of the complex scene labeling effect (the final labeling result); according to the environmental information, the labeling confidence value of the target object is determined, and the first labeling result is fine-tuned according to the labeling confidence value, which can further improve the accuracy of the final labeling result. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 A flowchart of the intelligent driving data multi-dimensional labeling method provided by an embodiment of the present application is shown;

[0027] Figure 2 A structural schematic diagram of an electronic device for implementing the intelligent driving data multi-dimensional labeling method provided by an embodiment of the present application is shown.

[0028] Reference signs: 10, processor; 11, memory; 12, communication bus; 13, communication interface.

[0029] The implementation of the present application, functional features and advantages will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0030] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference signs represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be understood as a limitation of the present application.

[0031] In the description of the present invention, it should be understood that the terms "longitudinal", "transverse", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention.

[0032] In the description of the present invention, unless otherwise specified and limited, it should be noted that the terms "installed", "connected" and "connected" should be understood in a broad sense. For example, it can be a mechanical connection or an electrical connection, or it can be the internal communication between two components. It can be a direct connection or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to the specific circumstances.

[0033] Reference Figure 1 FIG. 1 is a flow chart of a method for multi-dimensional labeling of intelligent driving data according to an embodiment of the present invention. In this embodiment, the method for multi-dimensional labeling of intelligent driving data includes:

[0034] S1. Obtain driving scene data and environmental information to be labeled.

[0035] Specifically, the driving scene data includes initial image data, initial point cloud data, and initial radar data.

[0036] In this embodiment, RGB image sequences are captured by multi-view cameras installed around the driving vehicle. The RGB image sequences are road scene images from different viewpoints. RGB images are a color image format based on the superposition of three primary color channels: red, green, and blue. The road scene images include at least one of vehicles, pedestrians, roads, traffic signs, sky, green belts, and obstacles. The number of cameras is usually more than one. During specific operations, the number and positions of cameras can be set according to actual needs, and this embodiment does not impose any restrictions.

[0037] The initial point cloud data is collected by a lidar device installed on the driving vehicle, and the initial radar data is collected by a millimeter-wave radar, which can collect distance and speed information of objects near the target vehicle.

[0038] In a preferred implementation of this embodiment, the intelligent driving data multi-dimensional labeling method further includes:

[0039] The initial image data, initial point cloud data and initial radar data are processed in time and space synchronously to obtain image data, point cloud data and radar data.

[0040] The space-time synchronization includes time synchronization and space calibration. In the time synchronization, an IEEE 1588v2 precision time protocol (PTP) is used to reduce the timestamp deviation of each sensor for collecting initial image data, initial point cloud data and initial radar data. In the space calibration, based on an improved iterative closest point (ICP) algorithm, a transformation matrix is calculated through feature point cloud matching (SIFT-3D), so as to reduce the calibration error of the driving scene data.

[0041] The environmental information is used to represent the weather condition of the driving scene, and includes an visibility parameter, an illumination parameter and a precipitation intensity parameter of the environment where the driving scene is located.

[0042] S2, determining an analysis rule for the driving scene data according to the environmental information, and performing feature extraction on the driving scene data according to the determined analysis rule to obtain driving scene features.

[0043] The initial analysis rule is to first extract features of the image data, the point cloud data and the radar data respectively by using a cross-modality feature extraction network (CMF-Net), and then perform feature-level fusion on the extracted features by using an attention mechanism (Cross-Modality Attention).

[0044] The specific operation rule of determining the analysis rule for the driving scene data according to the environmental information is as follows:

[0045] S201, comparing the visibility parameter with a visibility reference threshold value, and if the visibility parameter exceeds the visibility reference threshold value, enhancing the channel attention weight in the analysis rule.

[0046] In the existing attention mechanism, the normal channel weight of the channel type of reflectance intensity is 0.3, the normal channel weight of the channel type of height information is 0.72, and the normal channel weight of the channel type of density distribution is 0.21.

[0047] If the visibility parameter exceeds the visibility reference threshold value, the channel attention weight in the analysis rule is enhanced. Specifically, the channel weight of the channel type of reflectance intensity is increased from 0.3 to 0.8, so that the model can penetrate the surface problem of the target object in rain and fog to identify; the channel weight of the channel type of height information is increased from 0.72 to 0.75; and the channel weight of the channel type of density distribution is increased from 0.21 to 0.65, so as to improve the detection effect of the target object contour in rain and fog.

[0048] S202, comparing the illumination parameter with an illumination reference threshold value, and if the illumination parameter exceeds the illumination reference threshold value, activating a low-light enhancement branch in the analysis rule.

[0049] Specifically, the current light intensity is determined to be lower than a threshold value by a sensor or an algorithm, and then a low-light enhancement module in an image signal processor is triggered, which usually includes algorithms such as multi-frame noise reduction, brightness enhancement, local contrast optimization, etc. (such as deep learning-based models, multi-frame fusion, adaptive histogram equalization, noise suppression techniques, etc.), and finally outputs the enhanced image data. The entire process needs to dynamically adjust parameters to ensure the balance between picture quality and real-time performance.

[0050] S203, compare the precipitation intensity parameter and the precipitation intensity reference threshold value, and if the precipitation intensity parameter exceeds the precipitation intensity reference threshold value, enable the anti-interference feature selector in the analysis rule.

[0051] First, the meteorological sensor or algorithm detects whether the precipitation intensity exceeds the threshold value, and if it does, the anti-interference feature selector is triggered. The anti-interference feature selector dynamically selects and enhances effective signal features (such as radar echoes, optical imaging data), while suppressing rain, fog, noise and other interference factors, and finally outputs the optimized environmental perception data to improve system robustness. The entire process needs to combine adaptive filtering and machine learning models to achieve real-time adjustment. Through the rule dynamic adjustment mechanism of environmental perception, the feature extraction process is adapted to different weather conditions.

[0052] Specifically, the analysis rule for the driving scene data is determined according to the environmental information, and the driving scene features are extracted according to the determined analysis rule.

[0053] S210, determining the analysis rule for the driving scene data according to the environmental information.

[0054] S220, extracting features from the image data according to the determined analysis rule to obtain two-dimensional features of the driving scene.

[0055] Specifically, the two-dimensional features can be extracted by a ResNet-18 network.

[0056] S230, extracting features from the point cloud data according to the determined analysis rule to obtain three-dimensional features of the driving scene.

[0057] Specifically, the two-dimensional features can be extracted by a PointNet++ network.

[0058] S240, extracting features from the radar data according to the determined analysis rule to obtain the position features of the driving vehicle.

[0059] The radar data is processed by a 1D CNN network (One-Dimensional Convolutional Neural Network).

[0060] S250, fuse the two-dimensional features, the three-dimensional features and the position features through the preset attention mechanism to obtain driving scene features.

[0061] Specifically, the preset attention mechanism is determined according to the contents of steps S201 to S203. The weight setting rules of the attention mechanism are introduced in steps S201 to S203, which will not be repeated here.

[0062] S3, input the target object driving scene features into a data labeling model. The data labeling model determines the target object of the driving scene and the label corresponding to the target object according to the driving scene features, and obtains a first labeling result.

[0063] The data labeling model of the present application is generated by using neural architecture search (NAS) technology. In the present embodiment, the data labeling model is a 3D-RCNN labeling model. The English full name of 3D-RCNN is Point Region-based Convolutional Neural Network. After inputting the driving scene features into the pre-trained 3D-RCNN labeling model, the data labeling model outputs the target object recognition result and the category label, and generates the first labeling result containing the object bounding box.

[0064] S4, determine the labeling confidence value of the target object according to the environmental information.

[0065] The labeling confidence value ranges from 0 to 1. The labeling confidence value can be set according to actual conditions. For example, if the weather of the driving scene is normal, it is a regular scene, the labeling confidence value is set to between 0.7 and 1. If the environment of the driving scene is an extreme weather scene, the labeling confidence value is set to [0.5, 0.7).

[0066] S5, adjust the first labeling result according to the labeling confidence value to obtain a final labeling result.

[0067] The specific steps of adjusting the first labeling result according to the labeling confidence value to obtain the final labeling result include:

[0068] Compare the labeling confidence value with the first labeling confidence reference value:

[0069] If the labeling confidence value is greater than or equal to the labeling confidence reference value, the first labeling result is taken as the final labeling result of the driving scene data.

[0070] If the labeling confidence value is less than the labeling confidence reference value, the review information is generated based on the first labeling result and sent to the client where the labeling personnel is located, the labeling result correction information is received from the client where the labeling personnel is located, and the final labeling result of the driving scene data is determined according to the labeling result correction information and the first labeling result.

[0071] In the preferred embodiment of the present embodiment, the intelligent driving data multi-dimensional labeling method further comprises:

[0072] S6, sensitive targets are screened out from the target objects of the driving scene.

[0073] The sensitive target is a target object containing personal privacy information, and the sensitive target includes the face of a pedestrian and the license plate of a vehicle.

[0074] S7, the sensitive target in the final labeling result is blurred.

[0075] In the present embodiment, the blurring of the face and license plate information in the target object is achieved through a multi-modal collaborative technology: first, the RetinaFace model is used to locate the face area, and the CRNN and point cloud data are used to identify the three-dimensional license plate position; then, based on the environmental parameters (visibility / lighting / rain intensity), the Gaussian blur radius is dynamically calculated, and adaptive intensity blur is performed on the sensitive area; at the same time, the license plate point cloud space coordinates are disturbed (±0.2 meters) and the reflection intensity is zeroed, and the millimeter wave radar signal is set to a noise threshold (-120dB); finally, the FaceNet verification is performed to realize the face recognition rate from 98.7% to 0.3%, and the license plate reconstruction error is expanded to ±50cm, which meets the privacy compliance requirements while ensuring the availability of the labeling data.

[0076] Based on the same inventive concept, an embodiment of the present application provides an intelligent driving data multi-dimensional labeling system.

[0077] The intelligent driving data multi-dimensional labeling system can be loaded in an electronic device. According to the functions implemented, the intelligent driving data multi-dimensional labeling system comprises an acquisition module, a feature extraction module, a first output module, a processing module and a second output module. The acquisition module can acquire driving scene data to be labeled and environmental information, and the environmental information is used to represent weather conditions of the driving scene. The feature extraction module can determine an analysis rule for the driving scene data according to the environmental information, and perform feature extraction on the driving scene data according to the determined analysis rule to obtain driving scene features. The first output module can input the driving scene features into a data labeling model. The data labeling model determines target objects of the driving scene and labels corresponding to the target objects according to the driving scene features to obtain a first labeling result. The processing module can determine a labeling confidence value of the target objects according to the environmental information. The second output module can adjust the first labeling result according to the labeling confidence value to obtain a final labeling result.

[0078] The module of the present application can also be referred to as a unit, which refers to a series of computer program segments that can be executed by an electronic device processor and can complete a fixed function, and is stored in the memory of the electronic device.

[0079] The various variations and specific examples of the intelligent driving data multi-dimensional labeling method provided in the above embodiments are also applicable to the intelligent driving data multi-dimensional labeling system of the present embodiment. Through the foregoing detailed description of the intelligent driving data multi-dimensional labeling method, those skilled in the art can clearly understand the implementation method of the intelligent driving data multi-dimensional labeling system in the present embodiment. In order to make the description brief, the implementation method of the intelligent driving data multi-dimensional labeling system in the present embodiment will not be described in detail here.

[0080] The present application also discloses an electronic device, such as Figure 2 As shown in the figure, it is a structural schematic diagram of an electronic device for the intelligent driving data multi-dimensional labeling method provided by an embodiment of the present application. The electronic device can include at least one processor 10, a memory 11 in communication with the at least one processor, a communication bus 12 and a communication interface 13, and can also include a computer program stored in the memory 11 and executable on the processor 10, such as an intelligent driving data multi-dimensional labeling method program.

[0081] The processor 10 can be composed of integrated circuits in some embodiments, for example, can be composed of a single packaged integrated circuit, or can be composed of multiple packaged integrated circuits with the same function or different functions, including one or more combinations of central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor 10 is the control unit of the electronic device, connects various components of the entire electronic device through various interfaces and lines, executes programs or modules stored in the memory 11 (such as the method for multi-dimensional labeling of intelligent driving data, etc.), and calls data stored in the memory 11 to execute various functions and process data of the electronic device.

[0082] The memory 11 includes at least one type of readable storage medium, including flash memory, mobile hard disk, multimedia card, card-type memory (such as SD or DX memory, etc.), magnetic memory, disk, optical disk, etc. The memory 11 can be an internal storage unit of the electronic device in some embodiments, for example, a mobile hard disk of the electronic device. The memory 11 can also be an external storage device of the electronic device in other embodiments, for example, a plug-in mobile hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the electronic device. Further, the memory 11 can include both the internal storage unit and the external storage device of the electronic device. The memory 11 can be used not only to store application software and various data installed on the electronic device, such as the code of the method for multi-dimensional labeling of intelligent driving data, but also to temporarily store data that has been output or will be output.

[0083] The communication bus 12 can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. The bus is configured to realize the connection and communication between the memory 11, the processor 10, etc.

[0084] The communication interface 13 is used for communication between the electronic device and other devices, including a network interface and a user interface. Optionally, the network interface can include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is usually used to establish a communication connection between the electronic device and other electronic devices. The user interface can be a display, an input unit (such as a keyboard), and optionally, the user interface can also be a standard wired interface, a wireless interface. Optionally, in some embodiments, the display can be an LED display, a liquid crystal display, a touch liquid crystal display, an OLED (Organic Light-Emitting Diode) touch, etc. Among them, the display can also be appropriately referred to as a display screen or a display unit, which is used to display information processed in the electronic device and to display a visualized user interface.

[0085] Figure 2 Only the electronic device with components is shown, and those skilled in the art can understand that, Figure 2 The structure shown does not constitute a limitation on the electronic device, and can include fewer or more components than shown, or combine certain components, or different component arrangements. For example, although not shown, the electronic device can also include a power supply (such as a battery) to power each component. Preferably, the power supply can be logically connected to the at least one processor 10 through a power management device, so that functions such as charge management, discharge management, and power consumption management can be realized through the power management device. The power supply can also include one or more direct current or alternating current power sources, recharging devices, power failure detection circuits, power converters or inverters, power status indicators, and any other components. The electronic device can also include various sensors, Bluetooth modules, Wi-Fi modules, etc., which are not described here.

[0086] It should be understood that the embodiments are for illustration only and are not limited in scope by the structure.

[0087] Further, the modules / units integrated in the electronic device, if implemented in the form of software functional units and sold or used as independent products, can be stored in a computer readable storage medium. The computer readable storage medium can be volatile or non-volatile.

[0088] The embodiments of the present application provide a computer readable storage medium, for example, including any entity or device capable of carrying the computer program code, a recording medium, a U disk, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM). The computer readable storage medium stores a computer program capable of being loaded by a processor and executing the intelligent driving data multi-dimensional labeling method of the above-mentioned embodiments.

[0089] In the description of the present specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", "one implementation", "one preferred implementation" or "some examples" and the like means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Also, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in an appropriate manner.

[0090] Although the embodiments of the present application have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, alternatives and variations can be made thereto without departing from the principles and spirit of the present application, the scope of which is defined by the claims and their equivalents.

Claims

1. A multi-dimensional labeling method for intelligent driving data, characterized in that: The method comprises: Obtain driving scene data and environmental information to be labeled, where the environmental information is used to represent the weather conditions of the driving scene; Determine parsing rules for driving scene data based on environmental information, and extract features from the driving scene data based on the determined parsing rules to obtain driving scene features; Inputting the driving scene features into the data annotation model, the data annotation model determines the target object of the driving scene and the label corresponding to the target object according to the driving scene features, and obtains a first annotation result; Determine the labeling confidence value of the target object based on the environmental information; The first annotation result is adjusted according to the annotation confidence value to obtain the final annotation result.

2. The multi-dimensional labeling method for intelligent driving data according to claim 1, characterized in that: The step of adjusting the first annotation result according to the annotation confidence value to obtain the final annotation result includes: Compare the annotation confidence value with the first annotation confidence reference value: If the annotation confidence value is greater than or equal to the annotation confidence reference value, the first annotation result is used as the final annotation result of the driving scene data; If the annotation confidence value is less than the annotation confidence reference value, review information is generated based on the first annotation result and sent to the client where the annotation personnel is located, and the annotation result correction information sent by the client where the annotation personnel is located is received. The final annotation result of the driving scene data is determined based on the annotation result correction information and the first annotation result.

3. The multi-dimensional labeling method for intelligent driving data according to claim 1, characterized in that: The driving scene data includes initial image data, initial point cloud data, and initial radar data. The method further includes: The initial image data, initial point cloud data and initial radar data are processed in time and space synchronously to obtain image data, point cloud data and radar data.

4. The multi-dimensional labeling method for intelligent driving data according to claim 3, characterized in that: Determining a parsing rule for the driving scene data based on the environmental information, and extracting features from the driving scene data based on the determined parsing rule to obtain driving scene features, includes: Determine parsing rules for driving scene data based on environmental information; Perform feature extraction on the image data according to the determined analytical rules to obtain the two-dimensional features of the driving scene; Extract features from point cloud data according to the determined parsing rules to obtain the three-dimensional features of the driving scene; Extract features from radar data according to the determined parsing rules to obtain the position features of the driving vehicle; The driving scene features are obtained by fusing two-dimensional features, three-dimensional features and position features through a preset attention mechanism.

5. The multi-dimensional labeling method for intelligent driving data according to claim 4, characterized in that: The environmental information includes visibility parameters, lighting parameters, and precipitation intensity parameters of the driving scene environment. The method further includes: Compare the visibility parameter with the visibility reference threshold. If the visibility parameter exceeds the visibility reference threshold, enhance the channel attention weight in the parsing rule. Compare the lighting parameters with the lighting reference threshold. If the lighting parameters exceed the lighting reference threshold, activate the low-light enhancement branch in the parsing rule. The precipitation intensity parameter is compared with the precipitation intensity reference threshold. If the precipitation intensity parameter exceeds the precipitation intensity reference threshold, the anti-interference feature selector is enabled in the parsing rule.

6. The multi-dimensional labeling method for intelligent driving data according to any one of claims 1 to 5, characterized in that: The method further comprises: Filter out sensitive targets from the target objects in the driving scene. Sensitive targets are target objects that contain personal privacy information. Fuzzy processing is performed on sensitive targets in the final annotation results.

7. A multi-dimensional labeling system for intelligent driving data, used to implement the multi-dimensional labeling method for intelligent driving data according to any one of claims 1 to 6, characterized in that: include: An acquisition module is used to obtain driving scene data and environmental information to be labeled, where the environmental information is used to represent the weather conditions of the driving scene; A feature extraction module is used to determine an analysis rule for driving scene data based on environmental information, and extract features from the driving scene data based on the determined analysis rule to obtain driving scene features; A first output module is configured to input the driving scene features into a data annotation model, and the data annotation model determines a target object in the driving scene and a label corresponding to the target object based on the driving scene features to obtain a first annotation result; A processing module, configured to determine a labeling confidence value of a target object based on environmental information; The second output module is used to adjust the first annotation result according to the annotation confidence value to obtain the final annotation result.

8. An electronic device, characterized in that: The electronic device comprises: at least one processor (10); and, a memory (11) communicatively coupled to the at least one processor (10); The memory (11) stores a computer program that can be executed by the at least one processor (10), and the computer program is executed by the at least one processor (10) so that the at least one processor (10) can execute the multi-dimensional labeling method for intelligent driving data as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program; when the computer program is executed by a processor, the method for multi-dimensional labeling of intelligent driving data as described in any one of claims 1 to 6 is implemented.

Images