A method and system for classifying and grading automotive data

Abstract

The application provides a vehicle data classification and grading method and system, extracts a label vector of vehicle data, preliminarily classifies the label vector of the vehicle data, fine-tunes the preliminary classification result based on the association between the label vector of the vehicle data and the vehicle data labels of a domain data set, obtains a target domain of the vehicle data, determines a target sub-domain of the vehicle data based on a neural network model corresponding to the target domain of the vehicle data; the multi-dimensional label vector is used for rough classification, so that the classification efficiency of the vehicle data is improved, the association between the domain data set and the label vector of the vehicle data is used for fine-tuning the rough classification result to obtain a relatively accurate target domain, and the corresponding neural network model is further used for subdivision on the basis of the target domain, so that the classification precision of the vehicle data is improved, and the application value of the vehicle data is effectively improved.

Description

Technical Field

[0001] This application relates to the field of data processing technology, specifically to a method and system for classifying and grading automobile data. Background Technology

[0002] With the deep application of emerging technologies such as big data, artificial intelligence, and cloud computing, data has become a core asset for enterprise operations and social development, making its full lifecycle management and security protection crucial. A scientific data classification and grading system can help enterprises clarify the context of their data assets, accurately define the importance and sensitivity of data, and thus implement differentiated management strategies and security protection measures. This effectively reduces security risks such as data leakage and tampering, ensures compliant data circulation and value release, and improves overall data governance efficiency.

[0003] In the automotive industry, data is widely distributed across the entire lifecycle, including R&D, manufacturing, sales and marketing, operation and maintenance, and end-of-life recycling. It encompasses multi-source, heterogeneous data such as vehicle operating status, driving behavior, passenger personal information, and fault diagnosis codes, exhibiting characteristics of wide data sources, diverse types, massive scale, and stringent compliance requirements. The classification and grading of automotive data not only needs to strictly protect personal privacy and trade secrets but also needs to effectively support diverse business scenarios such as vehicle performance optimization, intelligent driving algorithm iteration, after-sales customer service improvement, and market trend analysis. This necessitates the construction of a data classification and grading system that comprehensively covers data assets while possessing flexibility and dynamic adjustment capabilities.

[0004] However, traditional data classification and grading methods, whether based on fixed data structures, pre-defined rules and features, or relying on classic machine learning models, often require significant manual intervention for feature engineering, model training, and parameter tuning to adapt to new data sources or business scenarios. When dealing with highly complex, multimodal, and strongly correlated data in the automotive field, these methods struggle to maintain context awareness, are prone to classification errors, and cannot address the challenges posed by the scale, heterogeneity, and dynamism of automotive data. This results in low data governance efficiency and fails to meet the dual demands of data security and value extraction in the era of intelligent connected vehicles. Summary of the Invention

[0005] To address the aforementioned technical problems, this application is proposed. Embodiments of this application provide a method and system for classifying and grading vehicle data.

[0006] According to one aspect of this application, a method for classifying and grading automotive data is provided, comprising: acquiring automotive data; wherein the automotive data includes instruction data, text data, and image data; extracting all tags from the automotive data to obtain a tag vector of the automotive data; performing preliminary classification on the tag vector of the automotive data based on multiple domain data sets to obtain a preliminary classification result; wherein the domain data sets store multiple automotive data tags and multiple weight values ​​of automotive data tags, and different automotive data tags are stored in different domain data sets; fine-tuning the preliminary classification result based on the correlation between the tag vector of the automotive data and the automotive data tags in the domain data sets to obtain a target domain of the automotive data; and determining a target sub-domain of the automotive data based on the tag vector of the automotive data, the weights of the tags in the tag vector of the automotive data in the corresponding domain data sets, and the neural network model corresponding to the target domain of the automotive data.

[0007] In one embodiment, extracting all tags from the vehicle data to obtain the tag vector of the vehicle data includes: if the vehicle data is instruction data or text data, then for each tag in the multiple domain data sets, searching for whether the tag exists in the instruction data or text data; if the tag exists, then using the tag as the tag vector corresponding to the instruction data or text data.

[0008] In one embodiment, extracting all labels from the vehicle data to obtain the label vector of the vehicle data includes: if the vehicle data is image data, segmenting the target in the image data to obtain at least one sub-image; classifying the at least one sub-image using a classification model to obtain the classification result of the at least one sub-image, and using the classification result as the label vector of the image data.

[0009] In one embodiment, the preliminary classification of the label vector of the vehicle data based on multiple domain data sets to obtain the preliminary classification result includes: for each label in the label vector of the vehicle data, obtaining the domain data set to which the label belongs and the weight value of the label in the domain data set to which it belongs; adding the corresponding weight value to the association value between the label vector of the vehicle data and the domain data set to which it belongs; and selecting the domain corresponding to the domain dataset with the largest association value as the preliminary domain of the label vector of the vehicle data.

[0010] In one embodiment, fine-tuning the preliminary classification result based on the association between the label vector of the vehicle data and the vehicle data labels of the domain data set to obtain the target domain of the vehicle data includes: obtaining the association between multiple labels in multiple domain data sets; when there is an association between multiple labels, aggregating the multiple labels with association to form a label association set; wherein, each label association set corresponds to an intermediate domain, and the intermediate domain is the domain corresponding to multiple domain data sets; obtaining a label vector containing more than a preset number of labels; comparing the label vector with each label association set, and when the label vector contains all the labels in the label association set, determining that the label vector has an inclusion association with the label association set; and fine-tuning the preliminary classification result based on the intermediate domain corresponding to the label association set with the inclusion association with the label vector to obtain the target domain of the vehicle data.

[0011] In one embodiment, the step of fine-tuning the preliminary classification result based on the intermediate domains corresponding to the tag association sets that have an inclusion relationship with the tag vector to obtain the target domain of the vehicle data includes: if the tag vector of the vehicle data has an inclusion relationship with only one tag association set, and the intermediate domain corresponding to the one tag association set is different from the preliminary domain corresponding to the tag vector of the vehicle data, then the intermediate domain corresponding to the one tag association set is used as the template domain of the tag vector of the vehicle data; if the tag vector of the vehicle data has an inclusion relationship with multiple tag association sets, and the intermediate domains corresponding to the multiple tag association sets are different, then the frequency of occurrence of each intermediate domain in the intermediate domains corresponding to the multiple tag association sets is counted; the intermediate domain corresponding to the maximum frequency of occurrence is selected as the target domain of the tag vector of the vehicle data.

[0012] In one embodiment, determining the target sub-domain of the vehicle data based on the label vector of the vehicle data, the weights of the labels in the label vector of the vehicle data in the corresponding domain data set, and the neural network model corresponding to the target domain of the vehicle data includes: inputting the label vector of the vehicle data and the weights of the labels in the label vector of the vehicle data in the corresponding domain data set into the neural network model corresponding to the target domain of the vehicle data to determine the target sub-domain of the vehicle data; wherein, each target domain corresponds to a different neural network model.

[0013] In one embodiment, the vehicle data classification and grading method further includes: periodically and randomly sampling multiple data samples; wherein, the data samples include data pairs consisting of the labels of the vehicle data's label vectors and the weights of the labels in the target domain data set, as well as the target sub-domain of the vehicle data's label vectors; manually verifying the data samples to obtain a manual verification result; if the manual verification result differs from the target sub-domain of the vehicle data's label vectors, then the vehicle data's label vectors, the weights of the vehicle data's label vectors in their respective target domains, and the manual verification result are used as correction samples to retrain the neural network model.

[0014] In one embodiment, the vehicle data classification and grading method further includes: performing deduplication, invalid data filtering, and sensitive information removal operations on the instruction data; performing stop word processing and word form restoration operations on the text data; and performing detection and removal of damaged image files and filtering of low-resolution images on the image data.

[0015] According to another aspect of this application, a vehicle data classification and grading system is provided, comprising: a vehicle data acquisition module for acquiring vehicle data; wherein the vehicle data includes instruction data, text data, and image data; a label vector extraction module for extracting all labels of the vehicle data to obtain label vectors of the vehicle data; a preliminary result classification module for performing preliminary classification on the label vectors of the vehicle data based on multiple domain data sets to obtain preliminary classification results; wherein the domain data sets store multiple vehicle data labels and multiple weight values ​​of vehicle data labels, and different vehicle data labels are stored in different domain data sets; a preliminary result fine-tuning module for fine-tuning the preliminary classification results based on the correlation between the label vectors of the vehicle data and the vehicle data labels of the domain data sets to obtain the target domain of the vehicle data; and a target sub-domain determination module for determining the target sub-domain of the vehicle data based on the label vectors of the vehicle data, the weights of the labels in the label vectors of the vehicle data in the corresponding domain data sets, and the neural network model corresponding to the target domain of the vehicle data.

[0016] This application provides a method and system for classifying and grading automotive data. The method involves acquiring automotive data, including instruction data, text data, and image data; extracting all tags from the automotive data to obtain tag vectors; performing preliminary classification on the tag vectors based on multiple domain datasets to obtain preliminary classification results; wherein each domain dataset stores multiple automotive data tags and their weights, with different tags stored in different domain datasets; fine-tuning the preliminary classification results based on the correlation between the automotive data tag vectors and the automotive data tags in the domain datasets to obtain the target domain of the automotive data; determining the target sub-domain of the automotive data based on the automotive data tag vectors, the weights of the tags in the tag vectors within the corresponding domain datasets, and the neural network model corresponding to the target domain of the automotive data; performing coarse classification using multi-dimensional tag vectors to improve the classification efficiency of the automotive data, and fine-tuning the coarse classification results using the correlation between the domain datasets and the tag vectors of the automotive data to obtain a relatively accurate target domain; and further subdividing the target domain using the corresponding neural network model to improve the classification accuracy of the automotive data and effectively enhance the application value of the automotive data. Attached Figure Description

[0017] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0018] Figure 1 This is a flowchart illustrating an exemplary embodiment of the vehicle data classification and grading method provided in this application.

[0019] Figure 2 This is a schematic diagram of the structure of an exemplary embodiment of the vehicle data classification and grading system provided in this application.

[0020] Figure 3 This is a structural diagram of an electronic device provided in an exemplary embodiment of this application. Detailed Implementation

[0021] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0022] Figure 1This is a flowchart illustrating an exemplary embodiment of the vehicle data classification and grading method provided in this application. Figure 1 As shown, the vehicle data classification and grading method includes the following steps: Step 110: Obtain vehicle data.

[0023] The vehicle data includes command data, text data, and image data. This application can utilize sensors in the vehicle-side system to collect vehicle data in real time.

[0024] Step 120: Extract all labels from the vehicle data to obtain the label vector of the vehicle data.

[0025] This application extracts all the tags of the acquired vehicle data through the vehicle-side system and combines all the tags to obtain a tag vector of the vehicle data.

[0026] Step 130: Perform preliminary classification of the label vectors of the vehicle data based on multiple domain datasets to obtain preliminary classification results.

[0027] The domain dataset stores multiple vehicle data tags and their weight values; different domain datasets store different vehicle data tags. This application allows the vehicle-side system to perform preliminary classification of vehicle data based on the tag vectors after obtaining the tag vectors, thus obtaining preliminary classification results.

[0028] Step 140: Based on the correlation between the label vector of the vehicle data and the vehicle data labels of the domain data set, fine-tune the preliminary classification results to obtain the target domain of the vehicle data.

[0029] This application can utilize the vehicle-side system to fine-tune the preliminary classification results based on the correlation between the label vector of the vehicle data and the vehicle data labels in the domain data set, thereby obtaining the target domain of the vehicle data, that is, to achieve coarse classification of vehicle data on the vehicle side.

[0030] Step 150: Based on the label vector of the vehicle data, the weight of the label in the label vector of the vehicle data in the corresponding domain data set, and the neural network model corresponding to the target domain of the vehicle data, determine the target subdomain of the vehicle data.

[0031] This application can upload vehicle data and corresponding label vectors to a cloud system, and use the neural network model in the cloud system to combine the weights of the labels in the label vectors of the vehicle data in the corresponding domain dataset to determine the target subdomain of the vehicle data.

[0032] This application provides a method for classifying and grading automotive data. The method involves acquiring automotive data, including instruction data, text data, and image data; extracting all tags from the automotive data to obtain tag vectors; performing preliminary classification on the tag vectors based on multiple domain datasets to obtain preliminary classification results; wherein each domain dataset stores multiple automotive data tags and their weights, with different tags stored in different domain datasets; fine-tuning the preliminary classification results based on the correlation between the automotive data tag vectors and the automotive data tags in the domain datasets to obtain the target domain of the automotive data; determining the target sub-domain of the automotive data based on the automotive data tag vectors, the weights of the tags in the tag vectors within the corresponding domain datasets, and the neural network model corresponding to the target domain of the automotive data; performing coarse classification using multi-dimensional tag vectors to improve the classification efficiency of the automotive data, and fine-tuning the coarse classification results using the correlation between the domain datasets and the tag vectors of the automotive data to obtain a relatively accurate target domain; and further subdividing the target domain using the corresponding neural network model to improve the classification accuracy of the automotive data and effectively enhance the application value of the automotive data.

[0033] In one embodiment, the specific implementation of step 120 above may be as follows: if the vehicle data is instruction data or text data, then for each tag in the multiple domain data sets, search for whether the tag exists in the instruction data or text data; if the tag exists, then use the tag as the tag vector corresponding to the instruction data or text data.

[0034] When the vehicle data is instruction data or text data, for each label in multiple domain data sets, check if the label exists in the instruction data or text data. If the label exists, it is used as the label vector corresponding to the instruction data or text data. This process is repeated for each label in multiple domain data sets until all labels in the multiple domain data sets have been checked, thus finally obtaining the label vector corresponding to the instruction data or text data.

[0035] In one embodiment, the specific implementation of step 120 above may be as follows: if the car data is image data, the target in the image data is segmented to obtain at least one sub-image; a classification model is used to classify the at least one sub-image to obtain the classification result of the at least one sub-image, and the classification result is used as the label vector of the image data.

[0036] When the vehicle data is image data, the target in the image data is segmented to obtain one or more sub-images. A classification model (such as a neural network) is then used to classify these sub-images, obtaining the classification results. These classification results are used as the label vector for the image data. Image data labels include, but are not limited to: driver, seat, seatbelt, obstacle, lane line, pedestrian, road sign, traffic light, and dashboard.

[0037] In one embodiment, step 130 can be implemented as follows: for each label in the label vector of the vehicle data, obtain the domain data set to which the label belongs and the weight value of the label in the domain data set; add the corresponding weight value to the association value between the label vector of the vehicle data and the domain data set to which it belongs; select the domain corresponding to the domain dataset with the largest association value as the preliminary domain of the label vector of the vehicle data.

[0038] Specifically, initially, the association value between the vehicle data tag vector and each domain data set is set to 0. For each tag in the vehicle data tag vector, the domain data set to which the tag belongs is obtained, and the weight value of the tag in that domain data set is also obtained. The association value between the vehicle data tag vector and its domain data set is then increased by the corresponding weight value. For each tag in the vehicle data tag vector, the above tag lookup and weight accumulation are repeated to obtain the association value between the vehicle data tag vector and each domain data set. The domain data set corresponding to the maximum value of the association value is then selected, and the domain corresponding to this domain data set is used as the initial domain of the vehicle data tag vector. The domain data sets include, but are not limited to: operational data domain data set, infotainment data domain data set, and safety critical signal domain data set.

[0039] For example, the vehicle data label vector is {A, B, C, D, E, F}; the operation data domain dataset, infotainment data domain dataset, and safety critical signal domain dataset are the first domain dataset, the second domain dataset, and the third domain dataset, respectively; where labels A, B, and C belong to the first domain dataset, D and E belong to the second domain dataset, and F belongs to the third domain dataset; the weights of A, B, and C in the first domain dataset are 0.9, 0.8, and 0.8, respectively; the weights of D and E in the second domain dataset are 0.6 and 0.5, respectively; and the weight of F in the third domain dataset is 0.95. Therefore, the association value between the label vector and the first domain dataset is 0 + 0.9 + 0.8 + 0.8 = 2.5, the association value between the label vector and the second domain dataset is 0 + 0.6 + 0.5 = 1.1, and the association value between the label vector and the third domain dataset is 0 + 0.95 = 0.95. It is evident that the label vector has the highest association value with the first domain dataset, therefore, the label vector belongs to the first domain.

[0040] In one embodiment, step 140 can be implemented as follows: obtaining the association relationships of multiple tags in multiple domain data sets; when there is an association relationship between multiple tags, summarizing the tags with the association relationship to form a tag association set; wherein, each tag association set corresponds to an intermediate domain, and the intermediate domain is the domain corresponding to the multiple domain data sets; obtaining a tag vector containing more than a preset value of tags; comparing the tag vector with each tag association set, and when the tag vector contains all the tags in the tag association set, determining that the tag vector has an inclusion association relationship with the tag association set; and fine-tuning the preliminary classification results based on the intermediate domain corresponding to the tag association set that has an inclusion association relationship with the tag vector to obtain the target domain of the automotive data.

[0041] Specifically, the process involves obtaining the relationships between multiple tags from multiple domain datasets. When multiple tags are related, they are grouped into a set, known as a tag association set. Each tag association set stores the related tags, and the intermediate domain for each set is set by the user. Then, each tag in the multiple tag association sets is treated as a node in a topology graph. When two tags are related, a connection is established between the nodes corresponding to the two tags, forming a tag association topology graph that is then fed back to the user. The process also involves obtaining all tag vectors containing at least three tags. For each tag vector that meets this condition, it is compared with each tag association set. If a tag vector contains all the tags from its association set, a containment relationship is established between the tag vector and that association set. Based on the intermediate domain corresponding to the tag association set with this containment relationship, the initial classification results are fine-tuned to obtain the target domain for the automotive data.

[0042] In one embodiment, step 140 can be implemented as follows: if the label vector of the vehicle data has an inclusion relationship with only one label association set, and the intermediate domain corresponding to the one label association set is different from the preliminary domain corresponding to the label vector of the vehicle data, then the intermediate domain corresponding to the one label association set is used as the template domain of the label vector of the vehicle data; if the label vector of the vehicle data has an inclusion relationship with multiple label association sets, and the intermediate domains corresponding to the multiple label association sets are different, then the frequency of occurrence of each intermediate domain in the intermediate domains corresponding to the multiple label association sets is counted; the intermediate domain corresponding to the maximum frequency of occurrence is selected as the target domain of the label vector of the vehicle data.

[0043] Specifically, when the label vector of vehicle data has an inclusion relationship with only one label association set, and the intermediate domain corresponding to that label association set is different from the target domain corresponding to the label vector of vehicle data, the target domain corresponding to the label vector of vehicle data is modified to the intermediate domain corresponding to that label association set. When the label vector of vehicle data has an inclusion relationship with only one label association set, and the intermediate domain corresponding to that label association set is the same as the target domain corresponding to the label vector of vehicle data, the target domain corresponding to the label vector of vehicle data is not modified. When the label vector of vehicle data has an inclusion relationship with multiple label association sets, and the intermediate domains corresponding to these multiple label association sets are different, the frequency of each intermediate domain in the intermediate domains corresponding to these multiple label association sets is counted, the maximum frequency is selected, and the intermediate domain corresponding to the maximum frequency is obtained. The target domain corresponding to the label vector of vehicle data is then modified to the intermediate domain corresponding to the maximum frequency.

[0044] In one embodiment, step 150 can be implemented by inputting the label vector of the vehicle data and the weights of the labels in the label vector of the vehicle data in the corresponding domain data set into the neural network model corresponding to the target domain of the vehicle data to determine the target sub-domain of the vehicle data; wherein, each target domain corresponds to a different neural network model.

[0045] Specifically, multiple domain datasets and corresponding neural network model sets are obtained. Each target domain corresponds to a different neural network model. The labels of the label vectors of the car data and the weights of the labels in the label vectors of the car data in the target domain datasets are used as data pairs to input into the corresponding neural network models to obtain the target subdomain of the label vectors of the car data, thus obtaining the segmentation result.

[0046] In one embodiment, the above-mentioned vehicle data classification and grading method may further include: periodically randomly sampling multiple data samples; wherein, the data samples include data pairs consisting of the labels of the vehicle data label vectors and the weights of the labels in the target domain data set, as well as the target sub-domain of the vehicle data label vectors; manually calibrating the data samples to obtain the manual calibration result; if the manual calibration result is different from the target sub-domain of the vehicle data label vectors, then using the vehicle data label vectors, the weights of the vehicle data label vectors in their respective target domains, and the manual calibration result as correction samples to retrain the neural network model.

[0047] Specifically, this application involves periodically and randomly sampling multiple data samples. These samples include labels from vehicle data label vectors and data pairs consisting of the weights of those labels in the target domain dataset, as well as the target sub-domain of the vehicle data label vectors. The data samples are then manually verified to obtain a verification result; this verification result represents the target sub-domain corresponding to the vehicle data label vector. When the verification result differs from the target sub-domain obtained through the neural network model, the target sub-domain corresponding to the vehicle data label vector is corrected based on the verification result. The vehicle data label vector, its weights in its target domain, and the verification result are then used as correction samples to optimize and train the neural network model, resulting in an optimized neural network model.

[0048] In one embodiment, the above-mentioned vehicle data classification and grading method may further include: performing deduplication, invalid data filtering, and sensitive information removal operations on instruction data; performing stop word processing and word form restoration operations on text data; and performing detection and removal of damaged image files and filtering of low-resolution images on image data.

[0049] This application performs deduplication, invalid data filtering, and sensitive information removal operations on instruction data; performs stop word processing and word form restoration operations on text data; and performs detection and removal of damaged image files and filtering of low-resolution images on image data, thereby achieving preprocessing of vehicle data to improve data accuracy and the efficiency and effectiveness of subsequent data processing.

[0050] Figure 2 This is a schematic diagram of the structure of a vehicle data classification and grading system provided in an exemplary embodiment of this application. For example... Figure 2 As shown, the vehicle data classification and grading system 20 includes: a vehicle data acquisition module 21 for acquiring vehicle data, including instruction data, text data, and image data; a label vector extraction module 22 for extracting all labels from the vehicle data to obtain label vectors; a preliminary result classification module 23 for performing preliminary classification on the label vectors of the vehicle data based on multiple domain data sets to obtain preliminary classification results, wherein multiple vehicle data labels and their weight values ​​are stored in the domain data sets, and different vehicle data labels are stored in different domain data sets; a preliminary result fine-tuning module 24 for fine-tuning the preliminary classification results based on the correlation between the label vectors of the vehicle data and the vehicle data labels in the domain data sets to obtain the target domain of the vehicle data; and a target sub-domain determination module 25 for determining the target sub-domain of the vehicle data based on the label vectors of the vehicle data, the weights of the labels in the label vectors of the vehicle data in the corresponding domain data sets, and the neural network model corresponding to the target domain of the vehicle data.

[0051] This application provides a vehicle data classification and grading system, which acquires vehicle data through a vehicle data acquisition module 21, including instruction data, text data, and image data; a label vector extraction module 22 extracts all labels from the vehicle data to obtain label vectors; a preliminary result classification module 23 performs preliminary classification on the label vectors of the vehicle data based on multiple domain data sets to obtain preliminary classification results; wherein, each domain data set stores multiple vehicle data labels and their weight values, and different domain data sets store different vehicle data labels; and a preliminary result fine-tuning module 24 adjusts the system based on the correlation between the label vectors of the vehicle data and the vehicle data labels in the domain data sets. The initial classification results are fine-tuned to obtain the target domain of the vehicle data. The target sub-domain determination module 25 determines the target sub-domain of the vehicle data based on the label vector of the vehicle data, the weight of the label in the label vector of the vehicle data in the corresponding domain data set, and the neural network model corresponding to the target domain of the vehicle data. Coarse classification is performed through multi-dimensional label vectors to improve the classification efficiency of vehicle data. The correlation between the domain data set and the label vector of the vehicle data is used to fine-tune the coarse classification results to obtain a relatively accurate target domain. Based on the target domain, the corresponding neural network model is further used for subdivision to improve the classification accuracy of vehicle data and effectively enhance the application value of vehicle data.

[0052] In one embodiment, the label vector extraction module 22 can be further configured to: if the vehicle data is instruction data or text data, then for each label in the multiple domain data sets, search for whether the label exists in the instruction data or text data; if the label exists, then use the label as the label vector corresponding to the instruction data or text data.

[0053] In one embodiment, the label vector extraction module 22 can be further configured to: if the car data is image data, segment the target in the image data to obtain at least one sub-image; classify the at least one sub-image using a classification model to obtain the classification result of the at least one sub-image, and use the classification result as the label vector of the image data.

[0054] In one embodiment, the preliminary result classification module 23 can be further configured to: for each label in the label vector of the vehicle data, obtain the domain data set to which the label belongs and the weight value of the label in the domain data set to which it belongs; add the corresponding weight value to the association value between the label vector of the vehicle data and the domain data set to which it belongs; and select the domain corresponding to the domain dataset with the largest association value as the preliminary domain of the label vector of the vehicle data.

[0055] In one embodiment, the preliminary result fine-tuning module 24 can be further configured to: obtain the association relationship of multiple tags in multiple domain data sets; when there is an association relationship between multiple tags, summarize the multiple tags with the association relationship to form a tag association set; wherein, each tag association set corresponds to an intermediate domain, and the intermediate domain is the domain corresponding to the multiple domain data sets; obtain a tag vector containing more than a preset value of tags; compare the tag vector with each tag association set, and when the tag vector contains all the tags in the tag association set, determine that the tag vector has an inclusion association relationship with the tag association set; based on the intermediate domain corresponding to the tag association set that has an inclusion association relationship with the tag vector, fine-tune the preliminary classification result to obtain the target domain of the car data.

[0056] In one embodiment, the preliminary result fine-tuning module 24 can be further configured as follows: if the label vector of the vehicle data has an inclusion relationship with only one label association set, and the intermediate domain corresponding to the one label association set is different from the preliminary domain corresponding to the label vector of the vehicle data, then the intermediate domain corresponding to the one label association set is used as the template domain of the label vector of the vehicle data; if the label vector of the vehicle data has an inclusion relationship with multiple label association sets, and the intermediate domains corresponding to the multiple label association sets are different, then the frequency of occurrence of each intermediate domain in the intermediate domains corresponding to the multiple label association sets is counted; the intermediate domain corresponding to the maximum frequency of occurrence is selected as the target domain of the label vector of the vehicle data.

[0057] In one embodiment, the target sub-domain determination module 25 can be further configured to: input the label vector of the vehicle data and the weight of the label in the label vector of the vehicle data in the corresponding domain data set into the neural network model corresponding to the target domain of the vehicle data to determine the target sub-domain of the vehicle data; wherein, each target domain corresponds to a different neural network model.

[0058] In one embodiment, the above-mentioned vehicle data classification and grading system 20 can be further configured to: periodically and randomly sample multiple data samples; wherein, the data samples include data pairs consisting of the labels of the label vectors of vehicle data and the weights of the labels in the target domain data set, as well as the target sub-domain of the label vectors of vehicle data; manually calibrate the data samples to obtain the manual calibration result; if the manual calibration result is different from the target sub-domain of the label vectors of vehicle data, the label vectors of vehicle data, the weights of the label vectors of vehicle data in their respective target domains, and the manual calibration result are used as correction samples to retrain the neural network model.

[0059] In one embodiment, the above-mentioned vehicle data classification and grading system 20 can be further configured to: perform deduplication, invalid data filtering and sensitive information removal operations on instruction data; perform stop word processing and word form restoration operations on text data; and perform detection and removal of damaged image files and filtering of low-resolution images on image data.

[0060] Below, for reference Figure 3 This application describes an electronic device according to embodiments thereof. The electronic device may be either or both of a first device and a second device, or a standalone device independent of them, which may communicate with the first device and the second device to receive acquired input signals from them.

[0061] Figure 3 A block diagram of an electronic device according to an embodiment of this application is illustrated.

[0062] like Figure 3 As shown, the electronic device 10 includes one or more processors 11 and memory 12.

[0063] The processor 11 may be a central processing unit (CPU) or other form of processing unit with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 10 to perform desired functions.

[0064] The memory 12 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 11 may execute the program instructions to implement the methods of the various embodiments of this application described above and / or other desired functions. Various contents such as input signals, signal components, and noise components may also be stored in the computer-readable storage medium.

[0065] In one example, the electronic device 10 may also include an input device 13 and an output device 14, which are interconnected via a bus system and / or other forms of connection mechanism (not shown).

[0066] When the electronic device is a standalone device, the input device 13 can be a communication network connector for receiving the collected input signals from the first device and the second device.

[0067] In addition, the input device 13 may also include, for example, a keyboard, a mouse, etc.

[0068] The output device 14 can output various information to the outside, including determined distance information, direction information, etc. The output device 14 may include, for example, a display, a speaker, a printer, and a communication network and its connected remote output devices, etc.

[0069] Of course, for the sake of simplicity, Figure 3 Only some of the components of the electronic device 10 relevant to this application are shown in this illustration; components such as buses, input / output interfaces, etc., are omitted. In addition, the electronic device 10 may include any other suitable components depending on the specific application.

[0070] In addition to the methods and apparatus described above, embodiments of this application may also be computer program products, which include computer program instructions that, when executed by a processor, cause the processor to perform the steps in the methods according to various embodiments of this application described in the "Exemplary Methods" section above.

[0071] The computer program product can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this application. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0072] Furthermore, embodiments of this application may also be computer-readable storage media storing computer program instructions thereon, which, when executed by a processor, cause the processor to perform the steps in the methods according to various embodiments of this application described in the "Exemplary Methods" section above.

[0073] The computer-readable storage medium may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof.

[0074] The basic principles of this application have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this application are merely examples and not limitations, and should not be considered as essential features of each embodiment of this application. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the application to the necessity of employing the aforementioned specific details for implementation.

[0075] The block diagrams of devices, apparatuses, devices, and systems involved in this application are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0076] It should also be noted that in the apparatus, equipment, and methods of this application, the components or steps can be disassembled and / or recombined. These disassemblies and / or recombinations should be considered as equivalent solutions of this application.

[0077] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of this application. Therefore, this application is not intended to be limited to the aspects shown herein, but rather to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0078] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this application to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

Claims

1. A method for classifying and grading vehicle data, characterized in that, include: Acquire vehicle data; wherein, the vehicle data includes command data, text data, and image data; Extract all tags from the vehicle data to obtain the tag vector of the vehicle data; The label vectors of the vehicle data are initially classified based on multiple domain datasets to obtain preliminary classification results; wherein, each domain dataset stores multiple vehicle data labels and weight values ​​for each vehicle data label, and different domain datasets store different vehicle data labels. Based on the correlation between the label vector of the vehicle data and the vehicle data label of the domain data set, the preliminary classification result is fine-tuned to obtain the target domain of the vehicle data; Based on the label vector of the vehicle data, the weight of the label in the label vector of the vehicle data in the corresponding domain data set, and the neural network model corresponding to the target domain of the vehicle data, the target subdomain of the vehicle data is determined.

2. The vehicle data classification and grading method according to claim 1, characterized in that, The step of extracting all tags from the vehicle data to obtain the tag vector of the vehicle data includes: If the vehicle data is instruction data or text data, then for each tag in the multiple domain data sets, search for whether the tag exists in the instruction data or text data; If the label exists, then the label will be used as the label vector corresponding to the instruction data or text data.

3. The vehicle data classification and grading method according to claim 1, characterized in that, The step of extracting all tags from the vehicle data to obtain the tag vector of the vehicle data includes: If the vehicle data is image data, then the target in the image data is segmented to obtain at least one sub-image; A classification model is used to classify the at least one sub-image to obtain the classification result of the at least one sub-image, and the classification result is used as the label vector of the image data.

4. The vehicle data classification and grading method according to claim 1, characterized in that, The preliminary classification of the label vectors of the vehicle data based on multiple domain datasets yields the following preliminary classification results: For each label in the label vector of the vehicle data, obtain the domain data set to which the label belongs and the weight value of the label in the domain data set to which it belongs; Add a corresponding weight value to the association value between the label vector of the vehicle data and the domain data set to which it belongs; The domain corresponding to the domain dataset with the largest correlation value is selected as the initial domain of the label vector of the vehicle data.

5. The vehicle data classification and grading method according to claim 1, characterized in that, The association between the label vectors of the vehicle data and the vehicle data labels of the domain data set is used to fine-tune the preliminary classification results, resulting in the target domain of the vehicle data including: To obtain the relationships between multiple tags in multiple domain datasets; When multiple tags are related, the tags with the relationship are aggregated to form a tag association set; wherein, each tag association set corresponds to an intermediate domain, and the intermediate domain is the domain corresponding to multiple domain data sets; Get the tag vector containing more than a preset number of tags; The tag vector is compared with each tag association set. If the tag vector contains all the tags in the tag association set, it is determined that there is an inclusion relationship between the tag vector and the tag association set. Based on the intermediate domain corresponding to the set of labels that have an association relationship with the label vector, the preliminary classification result is fine-tuned to obtain the target domain of the vehicle data.

6. The vehicle data classification and grading method according to claim 5, characterized in that, The preliminary classification result is fine-tuned based on the intermediate domain corresponding to the tag association set that has an inclusion relationship with the tag vector, resulting in the target domain of the vehicle data, which includes: If the label vector of the vehicle data has an inclusion relationship with only one label association set, and the intermediate domain corresponding to the one label association set is different from the preliminary domain corresponding to the label vector of the vehicle data, then the intermediate domain corresponding to the one label association set is used as the template domain of the label vector of the vehicle data. If the label vector of the vehicle data has an inclusive relationship with multiple label association sets, and the intermediate domains corresponding to these multiple label association sets are different, then the number of times each intermediate domain appears in the intermediate domains corresponding to these multiple label association sets is counted. The intermediate region corresponding to the maximum frequency of occurrence is selected as the target region of the label vector of the vehicle data.

7. The vehicle data classification and grading method according to claim 1, characterized in that, The determination of the target sub-domain of the vehicle data based on the label vector of the vehicle data, the weights of the labels in the label vector of the vehicle data in the corresponding domain data set, and the neural network model corresponding to the target domain of the vehicle data includes: The label vector of the vehicle data and the weights of the labels in the label vector of the vehicle data in the corresponding domain data set are input into the neural network model corresponding to the target domain of the vehicle data to determine the target sub-domain of the vehicle data; wherein, each target domain corresponds to a different neural network model.

8. The vehicle data classification and grading method according to claim 1, characterized in that, The vehicle data classification and grading method also includes: Periodically and randomly sample multiple data samples; wherein, the data samples include data pairs consisting of the labels of the label vector of the vehicle data and the weights of the labels in the label vector of the vehicle data in the target domain data set, as well as the target subdomain of the label vector of the vehicle data; The data samples were manually proofread to obtain the proofreading results. If the result of the manual verification differs from the target sub-domain of the label vector of the vehicle data, then the label vector of the vehicle data, the weight value of the label vector of the vehicle data in its target domain, and the result of the manual verification are used as correction samples to retrain the neural network model.

9. The vehicle data classification and grading method according to claim 1, characterized in that, The vehicle data classification and grading method also includes: The instruction data is subjected to deduplication, invalid data filtering, and sensitive information removal operations. Perform stop word processing and lemmatization on the text data; The image data is processed to detect and remove corrupted image files and filter low-resolution images.

10. A vehicle data classification and grading system, characterized in that, include: A vehicle data acquisition module is used to acquire vehicle data; wherein, the vehicle data includes instruction data, text data, and image data; The label vector extraction module is used to extract all labels from the vehicle data to obtain the label vector of the vehicle data. The preliminary result classification module is used to perform preliminary classification on the label vector of the vehicle data based on multiple domain data sets to obtain preliminary classification results; wherein, the domain data sets store multiple vehicle data labels and multiple weight values ​​of vehicle data labels, and different vehicle data labels are stored in different domain data sets; The preliminary result fine-tuning module is used to fine-tune the preliminary classification result based on the correlation between the label vector of the vehicle data and the vehicle data labels of the domain data set, so as to obtain the target domain of the vehicle data; The target sub-domain determination module is used to determine the target sub-domain of the vehicle data based on the label vector of the vehicle data, the weight of the label in the label vector of the vehicle data in the corresponding domain data set, and the neural network model corresponding to the target domain of the vehicle data.

Images