Data mining method, device, computer equipment, storage medium and program product

Abstract

The present application relates to a data mining method, apparatus, computer device, storage medium, and computer program product. The method comprises: determining multiple target decision models from multiple decision models deployed by an autonomous driving system, wherein the multiple target decision models have the same decision function; processing the same perception data using the multiple target decision models to obtain multiple decision results output by the multiple target decision models, wherein the multiple decision results are results corresponding to the same decision function in the multiple target decision models; determining first difference information between the multiple decision results, and determining whether the perception data is coarse-grained high-value data based on the first difference information. This method can improve the efficiency of data mining tasks.

Description

Technical Field

[0001] The present application relates to the field of data mining technology, and in particular to a data mining method, apparatus, computer equipment, storage medium, and computer program product. Background Art

[0002] In order to improve the accuracy of autonomous vehicles' perception of their surrounding environment, it is often necessary to use high-value data to train the decision-making model of autonomous vehicles.

[0003] In the prior art, mining of high-value data is generally achieved based on offline active learning data mining methods.

[0004] However, this data mining method can only realize data mining in an offline state and requires a lot of computing power, which leads to low efficiency of data mining tasks. Summary of the Invention

[0005] Based on this, it is necessary to provide a data mining method, device, computer equipment, computer-readable storage medium and computer program product with higher work efficiency to address the above technical problems.

[0006] In a first aspect, the present application provides a data mining method, comprising:

[0007] Determine multiple target decision models from multiple decision models deployed by the autonomous driving system, where the multiple target decision models have the same decision function; use the multiple target decision models to process the same perception data to obtain multiple decision results output by the multiple target decision models, where the multiple decision results are results corresponding to the same decision function in the multiple target decision models; determine first difference information between the multiple decision results, and determine whether the perception data is coarse-grained high-value data based on the first difference information.

[0008] In one embodiment, the perception data is determined to be coarse-grained high-value data based on the first difference information, including: if the first difference information represents that the degree of difference between the multiple decision results is greater than a first preset difference degree threshold, then the perception data is determined to be coarse-grained high-value data.

[0009] In one embodiment, the method further includes: when the perception data is coarse-grained high-value data, the perception data and the target decision result among the multiple decision results are stored correspondingly to obtain an initial high-value data group; and fine-grained high-value data mining is performed based on the initial high-value data group.

[0010] In one embodiment, the fine-grained high-value data mining based on the initial high-value data group includes: sending the high-value data group to a remote server, so that the remote server performs fine-grained high-value data mining based on a preset mining algorithm and the initial high-value data group.

[0011] In one embodiment, the mining process of the fine-grained high-value data mining includes: performing data enhancement processing on the perception data to obtain enhanced perception data; processing the enhanced perception data using a simulation decision model to obtain a simulated decision result output by the simulation decision model, wherein the simulation decision model is a target decision model for simulating the output of the target decision result; determining second difference information between the simulation decision result and the target decision result, and determining whether the perception data is fine-grained high-value data based on the second difference information.

[0012] In one embodiment, the second difference information is used to determine whether the perception data is fine-grained high-value data, including: if the second difference information represents that the degree of difference between the simulated decision result and the target decision result is greater than a second preset difference degree threshold, then the perception data is determined to be fine-grained high-value data.

[0013] In one embodiment, if the perception data is fine-grained high-value data, the perception data is used to iteratively train the target decision model.

[0014] In a second aspect, the present application further provides a data mining device, comprising:

[0015] a determination module, configured to determine a plurality of target decision models from a plurality of decision models deployed by the autonomous driving system, wherein the plurality of target decision models have the same decision function;

[0016] a first execution module, configured to process the same perception data using the multiple target decision models to obtain multiple decision results output by the multiple target decision models, wherein the multiple decision results are results corresponding to the same decision function in the multiple target decision models;

[0017] The second execution module is used to determine first difference information between the multiple decision results, and determine whether the perception data is coarse-grained high-value data based on the first difference information.

[0018] In one embodiment, the second execution module is specifically configured to determine that the perception data is coarse-grained high-value data if the first difference information represents that the difference between the multiple decision results is greater than a first preset difference threshold.

[0019] In one embodiment, the data mining device also includes a third execution module, which is used to store the perception data and the target decision results among the multiple decision results in correspondence when the perception data is coarse-grained high-value data, so as to obtain an initial high-value data group; and perform fine-grained high-value data mining based on the initial high-value data group.

[0020] In one embodiment, the third execution module is specifically configured to send the high-value data set to a remote server, so that the remote server performs fine-grained high-value data mining based on a preset mining algorithm and the initial high-value data set.

[0021] In one embodiment, the third execution module is specifically used to perform data enhancement processing on the perception data to obtain enhanced perception data; use a simulation decision model to process the enhanced perception data to obtain a simulated decision result output by the simulation decision model, and the simulation decision model is a target decision model used to simulate and output the target decision result; determine the second difference information between the simulation decision result and the target decision result, and determine whether the perception data is fine-grained high-value data based on the second difference information.

[0022] In one embodiment, the third execution module is specifically used to determine that the perception data is fine-grained high-value data if the second difference information represents that the difference between the simulated decision result and the target decision result is greater than a second preset difference threshold.

[0023] In one embodiment, if the perception data is fine-grained high-value data, the perception data is used to iteratively train the target decision model.

[0024] In a third aspect, the present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements any of the steps described in the first aspect when executing the computer program.

[0025] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the steps described in the first aspect.

[0026] In a fifth aspect, the present application further provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, implements any of the steps described in the first aspect.

[0027] The above-mentioned data mining method, device, computer equipment, storage medium and computer program product determine multiple target decision models from multiple decision models deployed by the autonomous driving system, and the multiple target decision models have the same decision function; use the multiple target decision models to process the same perception data to obtain multiple decision results output by the multiple target decision models, and the multiple decision results are the results corresponding to the same decision function in the multiple target decision models; determine the first difference information between the multiple decision results, and determine whether the perception data is coarse-grained high-value data based on the first difference information. The data mining method provided in this application first determines the target decision model, and then determines the decision results output by the target decision model for the same perception data, and determines whether the perception data is coarse-grained high-value data based on the difference information between the decision results. The data mining method provided in this application can not only realize online data mining to improve the flexibility of data mining, but also reduce the consumption of computing power and reduce costs, thereby effectively improving the work efficiency of data mining tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0029] Figure 1 1 is a flow chart of a data mining method according to an embodiment;

[0030] Figure 2 is a flow chart of a method in an embodiment when the perception data is coarse-grained high-value data;

[0031] Figure 3 1 is a flowchart of a mining process for fine-grained high-value data mining in one embodiment;

[0032] Figure 4 is a flow chart of a data mining method in another embodiment;

[0033] Figure 5 is a structural block diagram of a data mining device in one embodiment;

[0034] Figure 6 is a structural block diagram of a data mining device in another embodiment;

[0035] Figure 7 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0036] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0037] In order to improve the accuracy of autonomous vehicles' perception of their surrounding environment, it is often necessary to use high-value data to train the decision-making model of autonomous vehicles.

[0038] In the prior art, mining of high-value data is generally achieved based on offline active learning data mining methods.

[0039] However, this data mining method can only realize data mining in an offline state, and this data mining method also consumes a lot of computing power, which leads to low work efficiency of data mining tasks.

[0040] In view of this, the present application provides a data mining method that can effectively improve the work efficiency of data mining tasks.

[0041] The data mining method provided in the embodiment of the present application may be executed by a computer device, which may be a server.

[0042] In an exemplary embodiment, Figure 1 As shown, a data mining method is provided, which includes the following steps:

[0043] Step 101: Determine multiple target decision models from multiple decision models deployed by the autonomous driving system, where the multiple target decision models have the same decision function.

[0044] In one possible implementation, functional information of multiple decision models deployed by the autonomous driving system can be first obtained, and based on the functional information, a decision model with overlapping capabilities can be determined from the multiple decision models as a target decision model. For example, three decision models are deployed in the autonomous driving system, namely model A, model B, and model C. Assume that model A is used to determine whether there is an obstacle around the vehicle based on the image data around the vehicle collected by the camera, model B is used to determine whether there is an obstacle around the vehicle based on the image data around the vehicle collected by the camera, and to determine the type of obstacle if an obstacle exists, and model C is used to monitor whether there is an abnormality or fault in the vehicle. Therefore, based on the functional information of each model, model A and model B can be determined as decision models with overlapping capabilities. Therefore, model A and model B are determined as target decision models.

[0045] Step 102: Use the multiple target decision models to process the same perception data to obtain multiple decision results output by the multiple target decision models, where the multiple decision results are results corresponding to the same decision function in the multiple target decision models.

[0046] Optionally, the perception data may be data collected by sensors deployed in the perception layer of the autonomous driving system.

[0047] In one possible implementation, it is assumed that the target decision models to be determined are model A and model B, wherein model A is used to determine whether there is an obstacle around the vehicle based on the image data around the vehicle collected by the camera, and model B is used to determine whether there is an obstacle around the vehicle based on the image data around the vehicle collected by the camera, and to determine the type of obstacle if an obstacle exists. The same decision function of model A and model B is to determine whether there is an obstacle around the vehicle. Therefore, the results of determining whether there is an obstacle around the vehicle output by model A and model B can be determined as the decision result.

[0048] Step 103: Determine first difference information between the multiple decision results, and determine whether the perception data is coarse-grained high-value data based on the first difference information.

[0049] In one possible implementation, assuming that for the same perception data, the A model outputs that there is an obstacle around the vehicle, and the B model outputs that there is no obstacle around the vehicle, then the first difference information is determined to be 1.

[0050] In another possible implementation, assuming that for the same perception data, the A model outputs that there is no obstacle around the vehicle, and the B model outputs that there is an obstacle around the vehicle, then the first difference information is determined to be 1.

[0051] In another possible implementation, assuming that for the same perception data, the A model outputs that there are no obstacles around the vehicle, and the B model also outputs that there are no obstacles around the vehicle, then the first difference information is determined to be 0.

[0052] In another possible implementation, assuming that for the same perception data, the A model outputs that there is an obstacle around the vehicle, and the B model also outputs that there is an obstacle around the vehicle, then the first difference information is determined to be 0.

[0053] As described above, after determining the difference information, it is necessary to determine whether the perception data is coarse-grained high-value data based on the first difference information.

[0054] In one possible implementation, when the first difference information is 1, the perception data is determined to be coarse-grained high-value data; and when the first difference information is 0, the perception data is determined not to be coarse-grained high-value data.

[0055] In an optional embodiment of the present application, assuming that both the A model and the B model can output x-dimensional features, the features of corresponding dimensions can be subtracted, and the absolute values ​​can be obtained, and then all are added together to obtain the first difference information.

[0056] In one possible implementation, if the value indicated by the first difference information is greater than a preset threshold, it can be determined that the perception data is coarse-grained high-value data; otherwise, it is determined that the perception data is not coarse-grained high-value data.

[0057] In an optional embodiment of the present application, assuming that the overlapping function of the A model and the B model is to identify the type of obstacle, and the output of the A model is M and the output of the B model is N, the similarity between M and N can be determined, and the first difference information can be determined based on the similarity.

[0058] In an exemplary embodiment, the first difference information is used to determine whether the perception data is coarse-grained high-value data, including: if the first difference information represents that the degree of difference between the multiple decision results is greater than a first preset difference degree threshold, then the perception data is determined to be coarse-grained high-value data.

[0059] Optionally, the first preset difference degree threshold can be set by technical personnel according to actual needs.

[0060] In one possible implementation, if the degree of difference indicated by the first difference information is greater than the first preset difference degree threshold, the perception data is determined to be coarse-grained high-value data; if the degree of difference indicated by the first difference information is less than or equal to the first preset difference degree threshold, the perception data is determined not to be coarse-grained high-value data.

[0061] The above-mentioned data mining method determines multiple target decision models from multiple decision models deployed by the autonomous driving system, and the multiple target decision models have the same decision function; uses the multiple target decision models to process the same perception data to obtain multiple decision results output by the multiple target decision models, and the multiple decision results are the results corresponding to the same decision function in the multiple target decision models; determines the first difference information between the multiple decision results, and determines whether the perception data is coarse-grained high-value data based on the first difference information. The data mining method provided in this application first determines the target decision model, and then determines the decision results output by the target decision model for the same perception data, and determines whether the perception data is coarse-grained high-value data based on the difference information between the decision results. The data mining method provided in this application can not only realize online data mining to improve the flexibility of data mining, but also reduce the consumption of computing power and reduce costs, thereby effectively improving the work efficiency of data mining tasks.

[0062] In an exemplary embodiment, Figure 2 As shown, when the perception data is coarse-grained high-value data, the method further includes the following steps:

[0063] Step 201: Correspondingly storing the perception data and a target decision result among the multiple decision results to obtain an initial high-value data group;

[0064] In an optional embodiment of the present application, determining the target decision result among the multiple decision results includes: determining the decision result corresponding to the first difference information greater than the first preset difference degree threshold, and determining the decision result as the target decision result.

[0065] In one possible implementation, assuming that there are three target decision models, namely model A, model B and model C, and the degree of difference indicated by the first difference information between the decision results of model A and model B is greater than the first preset difference degree threshold, then the decision results of model A and model B are determined to be the target decision results, and the degree of difference indicated by the first difference information between the decision results of model A and model C is less than the first preset difference degree threshold, then the decision result of model C is determined not to be the target decision result.

[0066] In an optional embodiment of the present application, the perception data and the target decision result are desensitized, and the desensitized perception data and the target decision result are stored in a cloud server to obtain an initial high-value data group.

[0067] Optionally, the desensitization process refers to deforming certain sensitive information through desensitization rules to achieve reliable protection of sensitive privacy data.

[0068] Step 202: Perform fine-grained high-value data mining based on the initial high-value data group.

[0069] In an exemplary embodiment, the fine-grained high-value data mining based on the initial high-value data group includes: sending the high-value data group to a remote server, so that the remote server performs fine-grained high-value data mining based on a preset mining algorithm and the initial high-value data group.

[0070] In one possible implementation, the remote server may first send a request, and after receiving the request, obtain the initial high-value data group from the cloud server, and send the initial high-value data group to the remote server, so that the remote server can perform fine-grained high-value data mining based on a preset mining algorithm and the initial high-value data group.

[0071] In the above-mentioned case where the perception data is coarse-grained high-value data, the perception data and the target decision results among the multiple decision results are first stored correspondingly to obtain an initial high-value data group, and then fine-grained high-value data mining is performed based on the initial high-value data group. The high-value data obtained by the rough mining of online data is mined again, that is, the high-value data is finely mined offline based on the remote server to obtain high-value data with higher value, thereby effectively improving the work efficiency of the data mining task.

[0072] In an exemplary embodiment, Figure 3 As shown in FIG, the mining process of fine-grained high-value data mining includes the following steps:

[0073] Step 301: Perform data enhancement processing on the perception data to obtain enhanced perception data.

[0074] Optionally, the data augmentation can generate more equivalent (equally valid) data based on limited data to enrich the distribution of the data.

[0075] In one possible implementation, the data enhancement processing includes rotation / reflection transformation, flip transformation, scaling transformation, translation transformation, scale transformation, contrast transformation, noise perturbation, and color transformation. Any one or several of the above changes can be performed on the perception data to achieve data enhancement processing of the perception data.

[0076] Step 302: Process the enhanced perception data using a simulation decision model to obtain a simulation decision result output by the simulation decision model.

[0077] The simulation decision model is a target decision model used to simulate and output the target decision result.

[0078] In one possible implementation, the remote server stores various decision models in the autonomous driving system, and the target decision model stored in the remote server can be determined as the simulation decision model.

[0079] In another possible implementation, the cloud server stores parameter information of each decision model in the autonomous driving system, simulates a model identical to the target decision model based on the parameter information, and determines the model as the simulated decision model. Assuming that the target decision result includes the decision result of the A model and the decision result of the B model, the simulated decision model can be used to simulate the A model or the B model.

[0080] As mentioned above, after the simulation model is determined, the enhanced perception data needs to be processed using the simulation decision model.

[0081] In one possible implementation, the enhanced perception data may be input into the simulation decision model to obtain a simulation decision result output by the simulation decision model.

[0082] Step 303: Determine second difference information between the simulation decision result and the target decision result, and determine whether the perception data is fine-grained high-value data based on the second difference information.

[0083] In one possible implementation, if the simulation decision result is the same as the target decision result, the second difference information is determined to be 0.

[0084] In another possible implementation, if the simulated decision result is different from the target decision result, the second difference information is determined to be 1.

[0085] As described above, it is also necessary to determine whether the perception data is fine-grained high-value data based on the second difference information.

[0086] In one possible implementation, if the second difference information is 0, it is determined that the perception data is not fine-grained high-value data.

[0087] In another possible implementation, if the second difference information is 1, it is determined that the perception data is fine-grained high-value data.

[0088] The above method first performs data enhancement processing on the perception data to obtain enhanced perception data, then uses a simulation decision model to process the enhanced perception data to obtain a simulated decision result output by the simulation decision model, and finally determines the second difference information between the simulated decision result and the target decision result, and determines whether the perception data is fine-grained high-value data based on the second difference information. This method can obtain high-value data with higher value, thereby effectively improving the work efficiency of data mining tasks.

[0089] In an exemplary embodiment, the second difference information is used to determine whether the perception data is fine-grained high-value data, including: if the second difference information represents that the degree of difference between the simulated decision result and the target decision result is greater than a second preset difference degree threshold, then the perception data is determined to be fine-grained high-value data.

[0090] In one possible implementation method, the similarity between the simulated decision result and the target decision result can be determined first, and the second difference information can be determined based on the similarity. If the second difference information represents that the degree of difference between the simulated decision result and the target decision result is greater than a second preset difference degree threshold, then the perception data is determined to be fine-grained high-value data. If the second difference information represents that the degree of difference between the simulated decision result and the target decision result is less than or equal to the second preset difference degree threshold, then the perception data is determined not to be fine-grained high-value data.

[0091] In an exemplary embodiment, if the perception data is fine-grained high-value data, the perception data is used to iteratively train the target decision model.

[0092] In one possible practice, if the perception data is determined to be fine-grained and high-value data, the perception data can be manually annotated or pseudo-labeled to establish a training data set, and the target decision model can be iteratively updated and trained based on the training data set.

[0093] If the perception data is fine-grained and high-value data, the perception data is used to iteratively train the target decision model, which can optimize the target decision model and effectively improve the accuracy of the target decision model.

[0094] In an optional embodiment of the present application, if the perception data is coarse-grained high-value data, then the target decision model is iteratively trained using the coarse-grained high-value data when preset conditions are met.

[0095] Optionally, the preset condition can be pre-set by a technician.

[0096] In one possible implementation method, if the perception data is coarse-grained high-value data, and the value of the perception data meets the value requirements of the target decision model training data set, the perception data can be manually annotated or pseudo-labeled to establish a training data set, and the target decision model can be iteratively updated and trained based on the training data set.

[0097] In an optional embodiment of the present application, the data value requirement of the target decision model can be determined first to adjust the degree of data mining. That is, if the data value requirement of the target model is high, fine-grained high-value data mining is required, and it is trained based on the fine-grained high-value data. If the data value requirement of the target model is low, fine-grained high-value data mining is not required, and it is trained based on coarse-grained high-value data.

[0098] The above method of adjusting the degree of data mining according to the data value requirements of the target decision model meets the target decision model's demand for high-value data at a low cost, which not only improves the flexibility of the data mining task, but also improves the work efficiency of the data mining task.

[0099] In an exemplary embodiment, Figure 4 As shown, another data mining method is provided, which includes the following steps:

[0100] Step 401: Determine multiple target decision models from multiple decision models deployed by the autonomous driving system, where the multiple target decision models have the same decision function.

[0101] Step 402: Use the multiple target decision models to process the same perception data to obtain multiple decision results output by the multiple target decision models, where the multiple decision results are results corresponding to the same decision function in the multiple target decision models.

[0102] Step 403: Determine first difference information between the multiple decision results. If the first difference information indicates that the degree of difference between the multiple decision results is greater than a first preset difference degree threshold, determine that the perception data is coarse-grained high-value data.

[0103] Step 404: When the perception data is coarse-grained high-value data, the perception data and the target decision result among the multiple decision results are stored correspondingly to obtain an initial high-value data group.

[0104] Step 405: Send the high-value data group to a remote server, so that the remote server performs fine-grained high-value data mining based on a preset mining algorithm and the initial high-value data group. The mining process of the fine-grained high-value data mining includes: performing data enhancement processing on the perception data to obtain enhanced perception data; processing the enhanced perception data using a simulation decision model to obtain a simulated decision result output by the simulation decision model, where the simulation decision model is a target decision model for simulating the output of the target decision result; determining second difference information between the simulation decision result and the target decision result, and if the second difference information represents that the degree of difference between the simulation decision result and the target decision result is greater than a second preset difference degree threshold, then determining that the perception data is fine-grained high-value data.

[0105] In an optional embodiment of the present application, the technical solution of the present application is explained by taking the mining of high-quality splash LiDAR data as an example. There is splash of water at the rear of the target vehicle caused by the wheels. The LiDAR point output by the target decision model A at this position in this case is an unknown type, and the LiDAR point output by the target decision model B at this position in this case is splash of water. The outputs of the target decision model A and the target decision model B are also the decision results, and the first difference information is determined based on the decision result, and the perception data is determined to be coarse-grained high-value data based on the first difference information. The target decision model A or the target decision model B can be trained based on the coarse-grained perception data, or fine-grained high-value data mining can be performed according to the above method to obtain fine-grained high-value data, and the target decision model A or the target decision model B can be trained based on the fine-grained perception data.

[0106] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0107] Based on the same inventive concept, embodiments of the present application also provide a data mining device for implementing the aforementioned data mining method. The implementation solution provided by this device is similar to the implementation solution described in the aforementioned method. Therefore, the specific limitations in one or more data mining device embodiments provided below can be found in the above-mentioned limitations on the data mining method and will not be repeated here.

[0108] In an exemplary embodiment, Figure 5 As shown, a data mining device 500 is provided, comprising: a determination module 501, a first execution module 502 and a second execution module 503, wherein:

[0109] a determination module 501, configured to determine a plurality of target decision models from a plurality of decision models deployed by the autonomous driving system, wherein the plurality of target decision models have the same decision function;

[0110] A first execution module 502 is configured to process the same perception data using the multiple target decision models to obtain multiple decision results output by the multiple target decision models, where the multiple decision results are results corresponding to the same decision function in the multiple target decision models;

[0111] The second execution module 503 is configured to determine first difference information between the multiple decision results, and determine whether the perception data is coarse-grained high-value data based on the first difference information.

[0112] In one embodiment, the second execution module 503 is specifically configured to determine that the perception data is coarse-grained high-value data if the first difference information indicates that the difference between the multiple decision results is greater than a first preset difference threshold.

[0113] In one embodiment, Figure 6 As shown, another data mining device 600 is provided. In addition to the various modules included in the data mining device 500, the data mining device 600 also includes a third execution module 504.

[0114] In one embodiment, the third execution module 504 is used to store the perception data and the target decision results among the multiple decision results in correspondence to obtain an initial high-value data group when the perception data is coarse-grained high-value data; and perform fine-grained high-value data mining based on the initial high-value data group.

[0115] In one embodiment, the third execution module 504 is specifically configured to send the high-value data set to a remote server, so that the remote server performs fine-grained high-value data mining based on a preset mining algorithm and the initial high-value data set.

[0116] In one embodiment, the third execution module 504 is specifically used to perform data enhancement processing on the perception data to obtain enhanced perception data; use a simulation decision model to process the enhanced perception data to obtain a simulated decision result output by the simulation decision model, and the simulation decision model is a target decision model used to simulate and output the target decision result; determine the second difference information between the simulation decision result and the target decision result, and determine whether the perception data is fine-grained high-value data based on the second difference information.

[0117] In one embodiment, the third execution module 504 is specifically used to determine that the perception data is fine-grained high-value data if the second difference information represents that the degree of difference between the simulated decision result and the target decision result is greater than a second preset difference degree threshold.

[0118] In one embodiment, if the perception data is fine-grained high-value data, the perception data is used to iteratively train the target decision model.

[0119] Each module in the above-mentioned data mining device can be implemented in whole or in part through software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in the computer device in the form of software, so that the processor can call and execute the corresponding operations of each module.

[0120] In an exemplary embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. Figure 7 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a data mining method is implemented.

[0121] Those skilled in the art will understand that Figure 7 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0122] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0123] Determine multiple target decision models from multiple decision models deployed by the autonomous driving system, where the multiple target decision models have the same decision function; use the multiple target decision models to process the same perception data to obtain multiple decision results output by the multiple target decision models, where the multiple decision results are results corresponding to the same decision function in the multiple target decision models; determine first difference information between the multiple decision results, and determine whether the perception data is coarse-grained high-value data based on the first difference information.

[0124] In one embodiment, based on the first difference information, it is determined whether the perception data is coarse-grained high-value data, and the processor also implements the following steps when executing the computer program: if the first difference information represents that the degree of difference between the multiple decision results is greater than a first preset difference degree threshold, then the perception data is determined to be coarse-grained high-value data.

[0125] In one embodiment, when the processor executes the computer program, the following steps are also implemented: when the perception data is coarse-grained high-value data, the perception data and the target decision results among the multiple decision results are stored correspondingly to obtain an initial high-value data group; and fine-grained high-value data mining is performed based on the initial high-value data group.

[0126] In one embodiment, fine-grained high-value data mining is performed based on the initial high-value data group, and the processor further implements the following steps when executing the computer program: sending the high-value data group to a remote server, so that the remote server performs fine-grained high-value data mining based on a preset mining algorithm and the initial high-value data group.

[0127] In one embodiment, when the processor executes the computer program, it also implements the following steps: performing data enhancement processing on the perception data to obtain enhanced perception data; processing the enhanced perception data using a simulation decision model to obtain a simulated decision result output by the simulation decision model, where the simulation decision model is a target decision model used to simulate and output the target decision result; determining second difference information between the simulation decision result and the target decision result, and determining whether the perception data is fine-grained high-value data based on the second difference information.

[0128] In one embodiment, based on the second difference information, it is determined whether the perception data is fine-grained high-value data. When the processor executes the computer program, the following steps are also implemented: if the second difference information represents that the degree of difference between the simulated decision result and the target decision result is greater than a second preset difference degree threshold, then the perception data is determined to be fine-grained high-value data.

[0129] In one embodiment, if the perception data is fine-grained high-value data, the perception data is used to iteratively train the target decision model.

[0130] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0131] Determine multiple target decision models from multiple decision models deployed by the autonomous driving system, where the multiple target decision models have the same decision function; use the multiple target decision models to process the same perception data to obtain multiple decision results output by the multiple target decision models, where the multiple decision results are results corresponding to the same decision function in the multiple target decision models; determine first difference information between the multiple decision results, and determine whether the perception data is coarse-grained high-value data based on the first difference information.

[0132] In one embodiment, based on the first difference information, it is determined whether the perception data is coarse-grained high-value data. When the computer program is executed by the processor, the following steps are also implemented: if the first difference information represents that the degree of difference between the multiple decision results is greater than a first preset difference degree threshold, then the perception data is determined to be coarse-grained high-value data.

[0133] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: when the perception data is coarse-grained high-value data, the perception data and the target decision results among the multiple decision results are stored correspondingly to obtain an initial high-value data group; and fine-grained high-value data mining is performed based on the initial high-value data group.

[0134] In one embodiment, fine-grained high-value data mining is performed based on the initial high-value data group, and the computer program, when executed by the processor, also implements the following steps: sending the high-value data group to a remote server, so that the remote server performs fine-grained high-value data mining based on a preset mining algorithm and the initial high-value data group.

[0135] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: performing data enhancement processing on the perception data to obtain enhanced perception data; processing the enhanced perception data using a simulation decision model to obtain a simulated decision result output by the simulation decision model, where the simulation decision model is a target decision model used to simulate and output the target decision result; determining second difference information between the simulation decision result and the target decision result, and determining whether the perception data is fine-grained high-value data based on the second difference information.

[0136] In one embodiment, based on the second difference information, it is determined whether the perception data is fine-grained high-value data. When the computer program is executed by the processor, the following steps are also implemented: if the second difference information represents that the degree of difference between the simulated decision result and the target decision result is greater than a second preset difference degree threshold, then the perception data is determined to be fine-grained high-value data.

[0137] In one embodiment, if the perception data is fine-grained high-value data, the perception data is used to iteratively train the target decision model.

[0138] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0139] Determine multiple target decision models from multiple decision models deployed by the autonomous driving system, where the multiple target decision models have the same decision function; use the multiple target decision models to process the same perception data to obtain multiple decision results output by the multiple target decision models, where the multiple decision results are results corresponding to the same decision function in the multiple target decision models; determine first difference information between the multiple decision results, and determine whether the perception data is coarse-grained high-value data based on the first difference information.

[0140] In one embodiment, based on the first difference information, it is determined whether the perception data is coarse-grained high-value data. When the computer program is executed by the processor, the following steps are also implemented: if the first difference information represents that the degree of difference between the multiple decision results is greater than a first preset difference degree threshold, then the perception data is determined to be coarse-grained high-value data.

[0141] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: when the perception data is coarse-grained high-value data, the perception data and the target decision results among the multiple decision results are stored correspondingly to obtain an initial high-value data group; and fine-grained high-value data mining is performed based on the initial high-value data group.

[0142] In one embodiment, fine-grained high-value data mining is performed based on the initial high-value data group, and the computer program, when executed by the processor, also implements the following steps: sending the high-value data group to a remote server, so that the remote server performs fine-grained high-value data mining based on a preset mining algorithm and the initial high-value data group.

[0143] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: performing data enhancement processing on the perception data to obtain enhanced perception data; processing the enhanced perception data using a simulation decision model to obtain a simulated decision result output by the simulation decision model, where the simulation decision model is a target decision model used to simulate and output the target decision result; determining second difference information between the simulation decision result and the target decision result, and determining whether the perception data is fine-grained high-value data based on the second difference information.

[0144] In one embodiment, based on the second difference information, it is determined whether the perception data is fine-grained high-value data. When the computer program is executed by the processor, the following steps are also implemented: if the second difference information represents that the degree of difference between the simulated decision result and the target decision result is greater than a second preset difference degree threshold, then the perception data is determined to be fine-grained high-value data.

[0145] In one embodiment, if the perception data is fine-grained high-value data, the perception data is used to iteratively train the target decision model.

[0146] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0147] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processor involved in the various embodiments provided herein may be, but are not limited to, a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, and the like.

[0148] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0149] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A data mining method, characterized in that: The method comprises: determining a plurality of target decision models from a plurality of decision models deployed by the autonomous driving system, the plurality of target decision models corresponding to the same decision function; Processing the same perception data using the multiple target decision models to obtain multiple decision results output by the multiple target decision models, where the multiple decision results are results corresponding to the same decision function in the multiple target decision models; determining first difference information between the plurality of decision results, and determining whether the perception data is coarse-grained high-value data based on the first difference information; and Sending the coarse-grained high-value data to a remote server for the remote server to perform fine-grained high-value data mining; The fine-grained high-value data mining process includes: performing data enhancement processing on the perception data to obtain enhanced perception data; constructing a simulation decision model using model parameters of the target decision model stored in the remote server, and processing the enhanced perception data using the simulation decision model to obtain a simulation decision result output by the simulation decision model; determining second difference information between the simulation decision result and a target decision result among multiple decision results corresponding to the coarse-grained high-value data, and determining whether the perception data is fine-grained high-value data based on the second difference information, wherein the simulation decision model is used to simulate at least one target decision model corresponding to the target decision result; The fine-grained high-value data is provided for training at least one decision model among the multiple target decision models.

2. The method according to claim 1, characterized in that The determining, based on the first difference information, whether the perception data is coarse-grained high-value data includes: If the first difference information represents that the degree of difference between the multiple decision results is greater than a first preset difference degree threshold, the perception data is determined to be coarse-grained high-value data.

3. The method according to claim 1, characterized in that The method further comprises: In a case where the perception data is coarse-grained high-value data, the perception data and a target decision result from the plurality of decision results are stored in correspondence to obtain an initial high-value data group; Fine-grained high-value data mining is performed based on the initial high-value data group.

4. The method according to claim 3, characterized in that The performing fine-grained high-value data mining based on the initial high-value data group includes: The high-value data group is sent to the remote server, so that the remote server performs fine-grained high-value data mining based on a preset mining algorithm and the initial high-value data group.

5. The method according to claim 1, wherein The determining, based on the second difference information, whether the perception data is fine-grained high-value data includes: If the second difference information represents that the degree of difference between the simulated decision result and the target decision result is greater than a second preset difference degree threshold, the perception data is determined to be fine-grained high-value data.

6. A data mining device, characterized in that: The device comprises: a determination module, configured to determine a plurality of target decision models from a plurality of decision models deployed by the autonomous driving system, wherein the plurality of target decision models correspond to the same decision function; a first execution module, configured to process the same perception data using the multiple target decision models to obtain multiple decision results output by the multiple target decision models, wherein the multiple decision results are results corresponding to the same decision function in the multiple target decision models; A second execution module is used to determine first difference information between the multiple decision results, and determine whether the perception data is coarse-grained high-value data based on the first difference information; and send the coarse-grained high-value data to a remote server for the remote server to perform fine-grained high-value data mining; the fine-grained high-value data mining process includes: performing data enhancement processing on the perception data to obtain enhanced perception data; using the model parameters of the target decision model stored in the remote server to construct a simulation decision model, and using the simulation decision model to process the enhanced perception data to obtain a simulated decision result output by the simulation decision model; determining second difference information between the simulation decision result and a target decision result among the multiple decision results corresponding to the coarse-grained high-value data, and determining whether the perception data is fine-grained high-value data based on the second difference information, wherein the simulation decision model is used to simulate at least one target decision model corresponding to the target decision result; wherein the fine-grained high-value data is provided for training at least one decision model among the multiple target decision models.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

9. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images