Sample data mining method and device, electronic equipment and storage medium
By using vehicle data matching and confidence calculation in the roadside perception system, high-quality sample data is screened out, which solves the problem of sample data diversity and representativeness, and improves the training effect of deep learning models.
Patent Information
- Application Number
- CN202311865204.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-29
- Publication Date
- 2025-07-08
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
How to mine and filter high-quality sample data from a large amount of raw data to support the training of deep learning models, especially in roadside perception systems, to ensure the diversity, representation and balance of sample data.
By obtaining multi-frame data of the vehicle's on-board unit and the roadside calculation unit, using timestamps to match, construct matching pairs and calculate confidence, and filter out data with low confidence as high-quality sample data.
实现了从路侧感知系统中高效挖掘出高质量的样本数据,提高了深度学习模型的训练效果。
Smart Images
Figure CN120277404A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the technical field of vehicle-road cooperation, and particularly relates to a method, device, electronic device and storage medium for sample data mining. Background Art
[0002] As an important component unit of vehicle-road cooperation, the roadside perception system plays an irreplaceable role in various application scenarios. The roadside perception system constantly generates various types of data such as raw data, process data, and result data, and these data can support the training of deep learning models, that is, they can be used as sample data required for the training of deep learning models. When constructing a training data set, it is usually necessary to consider indicators such as the diversity, representativeness, and balance of sample data. Thus, how to mine and screen high-quality sample data has become a technical problem that needs to be considered by those skilled in the art. Summary of the Invention
[0003] In view of this, the embodiments of this application provide a method, device, electronic device and storage medium for sample data mining, which can mine and screen high-quality sample data.
[0004] The first aspect of the embodiments of this application provides a method for sample data mining, including:
[0005] Obtaining multiple frames of first vehicle data sent by in-vehicle units of each vehicle in a target area, and obtaining multiple frames of second vehicle data of all vehicles in the target area sent by a roadside computing unit;
[0006] Matching the multiple frames of first vehicle data and the multiple frames of second vehicle data according to time stamps to obtain multiple frames of data matching results; wherein, each frame of data matching result includes one frame of first vehicle data and one frame of second vehicle data corresponding to the time stamp;
[0007] For each frame of data matching result, determining the confidence level of this frame of data matching result according to the first vehicle data and the second vehicle data included in this frame of data matching result;
[0008] Selecting sample data from the multiple frames of data matching results according to the confidence level of each frame of data matching result.
[0009] In an embodiment of the present application, first, multiple frames of first vehicle data sent by in-vehicle units of various vehicles in a target area are obtained, and multiple frames of second vehicle data of all vehicles in the target area sent by a roadside computing unit are obtained; then, these first vehicle data and second vehicle data are matched according to timestamps, that is, each frame of first vehicle data and each frame of second vehicle data are associated according to timestamps, so as to obtain multiple frames of data matching results; next, for each frame of data matching result, the confidence of this frame of data matching result can be determined according to the first vehicle data and second vehicle data included in this frame of data matching result; finally, sample data can be screened and mined from these data matching results according to the confidence of each frame of data matching result. The confidence of each frame of data matching result can be used to characterize the error between the first vehicle data and the second vehicle data included therein, that is, if the confidence is lower, it means that the corresponding error is larger. Therefore, when selecting sample data from multiple frames of data matching results, data matching results with lower confidence can be selected as much as possible. The error between the first vehicle data and the second vehicle data included in these data matching results is larger, so it is more valuable as sample data, and thus high-quality sample data can be mined.
[0010] In an implementation manner of an embodiment of the present application, determining the confidence of this frame of data matching result according to the first vehicle data and second vehicle data included in this frame of data matching result includes:
[0011] Multiple matching pairs are constructed according to the first vehicle data and second vehicle data included in this frame of data matching result; wherein, each matching pair includes a first detection box included in the corresponding first vehicle data and a second detection box corresponding to the first detection box included in the corresponding second vehicle data.
[0012] The overlap degrees of the first detection box and the second detection box included in each matching pair are calculated respectively.
[0013] The confidence of this frame of data matching result is determined according to the overlap degrees corresponding to each matching pair.
[0014] In an implementation manner of an embodiment of the present application, determining the confidence of this frame of data matching result according to the overlap degrees corresponding to each matching pair includes:
[0015] The overlap degrees corresponding to each matching pair are weighted and summed to calculate the confidence of this frame of data matching result.
[0016] In an implementation manner of an embodiment of the present application, weighting and summing the overlap degrees corresponding to each matching pair to calculate the confidence of this frame of data matching result includes:
[0017] For each matching pair, determine the weight coefficient of the matching pair according to the target type of the detection box included in the matching pair;
[0018] For the overlap degrees corresponding to each matching pair, perform a weighted summation calculation according to the weight coefficient of each matching pair to obtain the confidence level of the matching result of this frame of data.
[0019] In one implementation manner of the embodiment of the present application, determining the weight coefficient of the matching pair according to the target type of the detection box included in the matching pair includes:
[0020] Determine the weight coefficient of the matching pair according to the accuracy rate of the target detection network to be trained for detecting targets of the target type; wherein, the determined weight coefficient is inversely proportional to the accuracy rate.
[0021] In one implementation manner of the embodiment of the present application, according to the first vehicle data and the second vehicle data included in the matching result of this frame of data, construct multiple matching pairs, including:
[0022] Through the Hungarian matching algorithm or the 3D IOU matching algorithm, determine the one-to-one correspondence between each detection box in the first vehicle data included in the matching result of this frame of data and each detection box in the second vehicle data included in the matching result of this frame of data, so as to obtain multiple matching pairs.
[0023] In one implementation manner of the embodiment of the present application, according to the confidence level of the matching result of each frame of data, select sample data from the matching results of multiple frames of data, including:
[0024] From the matching results of multiple frames of data, select the matching results of data with a confidence level less than the set threshold as the mined sample data.
[0025] The second aspect of the embodiment of the present application provides a sample data mining device, including:
[0026] A vehicle data acquisition module, configured to acquire multiple frames of first vehicle data sent by the on-vehicle units of each vehicle in the target area, and acquire multiple frames of second vehicle data of all vehicles in the target area sent by the roadside computing unit;
[0027] A vehicle data matching module, configured to match the multiple frames of first vehicle data and the multiple frames of second vehicle data according to the time stamp to obtain the matching results of multiple frames of data; wherein, the matching result of each frame of data includes one frame of first vehicle data and one frame of second vehicle data corresponding to the time stamp;
[0028] A confidence level determination module, configured to, for the matching result of each frame of data, determine the confidence level of the matching result of this frame of data according to the first vehicle data and the second vehicle data included in the matching result of this frame of data;
[0029] A sample data mining module, configured to select sample data from the multi-frame data matching results according to the confidence of the matching results of each frame of data.
[0030] A third aspect of the embodiments of the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the sample data mining method provided in the first aspect of the embodiments of the present application is implemented.
[0031] A fourth aspect of the embodiments of the present application provides a computer-readable storage medium storing a computer program, which when executed by a processor implements the sample data mining method provided in the first aspect of the embodiments of the present application.
[0032] A fifth aspect of the embodiments of the present application provides a computer program product, which when running on an electronic device causes the electronic device to execute the sample data mining method provided in the first aspect of the embodiments of the present application.
[0033] It can be understood that the beneficial effects of the above second aspect to the fifth aspect can refer to the relevant descriptions in the above first aspect and will not be elaborated here. Description of the Drawings
[0034] Figure 1 is a flowchart of a sample data mining method provided by an embodiment of the present application;
[0035] Figure 2 is a schematic structural diagram of a sample data mining device provided by an embodiment of the present application;
[0036] Figure 3 is a schematic diagram of an electronic device provided by an embodiment of the present application. Detailed Embodiments
[0037] In the following description, specific details such as specific system structures and technologies are proposed for the purpose of illustration rather than limitation, so as to thoroughly understand the embodiments of the present application. However, those skilled in the art should clearly understand that the present application can also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid unnecessary details from interfering with the description of the present application. In addition, in the description of the specification and claims of the present application, the terms "first", "second", "third", etc. are only used for distinguishing descriptions and cannot be understood as indicating or implying relative importance.
[0038] During the training process of a deep learning model, high-quality sample data is very important. How to mine and screen out high-quality sample data from a large amount of raw data has become a technical problem that those skilled in the art need to consider. In view of this, the embodiments of the present application provide a method, device, electronic device, and storage medium for mining sample data, which can mine and screen out high-quality sample data. For more specific technical implementation details of the embodiments of the present application, please refer to the method embodiments described below.
[0039] It should be understood that the execution subject of each method embodiment of the present application is various types of electronic devices. For example, it can be a mobile phone, a tablet computer, a wearable device, a vehicle-mounted device, an augmented reality (AR) / virtual reality (VR) device, a roadside unit, a laptop computer, an ultra-mobile personal computer (UMPC), a netbook, a personal digital assistant (PDA), a large-screen TV, and so on. The embodiments of the present application do not impose any restrictions on the specific type of this electronic device.
[0040] Please refer to Figure 1 , which shows a method for mining sample data provided by an embodiment of the present application, including:
[0041] 101. Obtain multiple frames of first vehicle data sent by on-vehicle units of each vehicle in a target area, and obtain multiple frames of second vehicle data of all vehicles in the target area sent by a roadside computing unit;
[0042] In the embodiments of the present application, there are two data sources, namely the on-vehicle units (OBUs) of each vehicle in the target area and the roadside computing unit (RSCU). Among them, the target area refers to a certain area within the detection range of the roadside system. For example, it can be a certain road section. The on-vehicle unit of each vehicle in the target area can send the vehicle data of its own vehicle, such as position, category, speed, acceleration, and heading angle data, to an electronic device for mining sample data (i.e., the execution subject of this method embodiment). The embodiments of the present application refer to the vehicle data sent by the on-vehicle units of each vehicle as the first vehicle data. Assuming there are n vehicles in the target area, the vehicle data sent by each vehicle are respectively represented as d OBU_1 , d OBU_2 , d OBU_3 , … d OBU_n , then one frame of the first vehicle data can be represented as [d OBU_1 , d OBU_2 , d OBU_3 , … d OBU_n, the first vehicle data of each frame has its time stamp, indicating the corresponding data acquisition time. Obviously, each on-vehicle unit will continuously send the acquired vehicle data to the electronic device, so that the electronic device obtains multiple frames of first vehicle data. The roadside computing unit RSCU can obtain the information of all traffic participants sensed by the roadside system, that is, it can obtain the vehicle data of all vehicles in the target area. This vehicle data can also include data such as the position, type, speed, acceleration, and heading angle of the vehicle, and send the data to the electronic device. In the embodiment of the present application, the vehicle data sent by the roadside computing unit is called the second vehicle data. Each frame of the second vehicle data has its time stamp, indicating the corresponding data acquisition time, and each frame of the second vehicle data contains the vehicle data of all vehicles in the target area. A frame of the second vehicle data can be expressed as d RSCU , obviously, the roadside computing unit will also continuously send the acquired vehicle data to the electronic device, so that the electronic device obtains multiple frames of the second vehicle data.
[0043] 102. Match the multiple frames of the first vehicle data and the multiple frames of the second vehicle data according to the time stamps to obtain multiple frame data matching results; wherein, each frame of the data matching result contains a frame of the first vehicle data and a frame of the second vehicle data corresponding to the time stamp;
[0044] After the electronic device obtains multiple frames of the first vehicle data and multiple frames of the second vehicle data, it will match these vehicle data according to the time stamps to obtain multiple frame data matching results, where each frame of the data matching result contains a frame of the first vehicle data and a frame of the second vehicle data corresponding to the time stamp. For example, a frame of the data matching result corresponding to the time stamp t1 can be expressed as t1[d OBU_1 , d OBU_2 , d OBU_3 , …d OBU_n , d RSCU , and a frame of the data matching result corresponding to the time stamp t2 can be expressed as t2[d OBU_1 , d OBU_2 , d OBU_3 , …d OBU_n , d RSCU , and so on.
[0045] 103. For each frame of the data matching result, determine the confidence level of the frame of the data matching result according to the first vehicle data and the second vehicle data included in the frame of the data matching result;
[0046] After obtaining the matching results of multiple frames of data, the electronic device can determine the confidence level of each frame of data matching result according to the first vehicle data and the second vehicle data included in each frame of data matching result. Since both the first vehicle data and the second vehicle data represent the data of the vehicles in the target area, for each frame of data matching result, the confidence level can be determined according to the error between the first vehicle data and the second vehicle data included therein. If the error is large, it means the confidence level is low; if the error is small, it means the confidence level is high.
[0047] In an implementation manner of the embodiment of the present application, determining the confidence level of the frame of data matching result according to the first vehicle data and the second vehicle data included in the frame of data matching result includes:
[0048] (1) Construct a plurality of matching pairs according to the first vehicle data and the second vehicle data included in the frame of data matching result; wherein, each matching pair includes a first detection frame included in the corresponding first vehicle data, and a second detection frame corresponding to the first detection frame included in the corresponding second vehicle data;
[0049] (2) Calculate the overlap degrees of the first detection frame and the second detection frame included in each matching pair respectively;
[0050] (3) Determine the confidence level of the frame of data matching result according to the overlap degrees corresponding to each matching pair.
[0051] During specific operation, data association processing can be performed first. Here, data association refers to finding the perception results corresponding to each vehicle in the first vehicle data from the second vehicle data. That is, it is necessary to find the perception result d RSCU from d OBU_i (i = 1, 2,... n) corresponding to each d RSCU_i , and then form a matching pair [d OBU_i , d RSCU_i . It can be seen that n matching pairs can be constructed. Taking a certain frame of data matching result [d OBU_1 , d OBU_2 , d OBU_3 ,... d OBU_n , d RSCU as an example, through the method of detection frame matching, the perception result d RSCU corresponding to d OBU_1 can be found from d RSCU_1 . Here, d OBU_1 can represent the vehicle position detection frame obtained by the on-vehicle unit of vehicle 1, and d RSCU_1 can represent the vehicle position detection frame of vehicle 1 sensed by the roadside system. Finally, the matching pair [d OBU_1 , d RSCU_1 is obtained. Similarly, other matching pairs [dOBU_2 , d RSCU_2 , [d OBU_3 , d RSCU_3 …[d OBU_n , d RSCU_n . In the embodiments of the present application, the vehicle position detection frame in the first vehicle data is referred to as the first detection frame, and the vehicle position detection frame in the second vehicle data is referred to as the second detection frame.
[0052] In one implementation manner of the embodiments of the present application, according to the first vehicle data and the second vehicle data included in the frame data matching result, a plurality of matching pairs are constructed, including:
[0053] By using the Hungarian matching algorithm or the 3D IOU matching algorithm, the one-to-one correspondence between each detection frame in the first vehicle data included in the frame data matching result and each detection frame in the second vehicle data included in the frame data matching result is determined, so as to obtain a plurality of matching pairs.
[0054] When finding the association relationship between detection frames, the Hungarian matching algorithm or the 3D IOU matching algorithm can be used to determine the one-to-one correspondence between each detection frame in the first vehicle data included in the frame data matching result and each detection frame in the second vehicle data included in the frame data matching result, and then the most suitable associated object can be found, so that each matching pair can be obtained. Among them, the specific principles of the Hungarian matching algorithm or the 3D IOU matching algorithm can refer to the prior art and will not be elaborated here.
[0055] After completing data association and obtaining each matching pair, the overlap degrees of the first detection frame and the second detection frame included in each matching pair can be calculated respectively. For example, for the matching pair [d OBU_1 , d RSCU_1 , calculate the overlap degree IOU1 of d OBU_1 and d RSCU_1 , for the matching pair [d OBU_2 , d RSCU_2 , calculate the overlap degree IOU2 of d OBU_2 and d RSCU_2 , and so on. After that, the confidence level of the frame data matching result can be determined according to the overlap degree corresponding to each matching pair.
[0056] In one implementation manner of the embodiments of the present application, determining the confidence level of the frame data matching result according to the overlap degree corresponding to each matching pair includes:
[0057] Perform a weighted sum calculation on the overlap degrees corresponding to each matching pair to obtain the confidence level of the frame data matching result.
[0058] When calculating the confidence of the matching result of the frame data, a method of weighted summation of the overlap degrees of each matching pair can be adopted, and the result of the weighted summation is used as the confidence of the matching result of the frame data, where the weight coefficient of each overlap degree can be reasonably set according to empirical values.
[0059] In an implementation manner of the embodiment of the present application, weighted summation calculation is performed on the overlap degrees respectively corresponding to each matching pair to obtain the confidence of the matching result of the frame data, including:
[0060] (1) For each matching pair, determine the weight coefficient of the matching pair according to the target type of the detection box included in the matching pair;
[0061] (2) Perform weighted summation calculation on the overlap degrees respectively corresponding to each matching pair according to the weight coefficients of each matching pair to obtain the confidence of the matching result of the frame data.
[0062] When determining the weight coefficient, different weight coefficients can be set respectively according to different target types of the detection box. For example, different target types such as cars, trucks, buses, and container trucks can be set with different weight coefficients respectively. For a certain matching pair [d OBU_i , d RSCU_i , the corresponding weight coefficient can be determined according to the target type of the detection box d OBU_i or the detection box d RSCU_i . For example, if the target type is a car, the weight coefficient is determined to be 0.1, and if the target type is a bus, the weight coefficient is determined to be 0.4, and so on. After determining the weight coefficients of each matching pair respectively, weighted summation calculation can be performed on the overlap degrees respectively corresponding to each matching pair according to the weight coefficients of each matching pair, so as to obtain the confidence of the data matching result.
[0063] For example, assume that the matching result of the i-th frame data contains m targets, that is, m matching pairs, then its confidence can be calculated according to the following formula:
[0064]
[0065] where, P i represents the confidence of the matching result of the i-th frame data, w j represents the weight coefficient of the j-th matching pair, and IOU j represents the overlap degree of the j-th matching pair. The higher the confidence, the closer the detection results of the two data sources (in-vehicle unit and roadside computing unit) are, and the higher the credibility of this frame of data. Otherwise, it means that the difference between the detection results of the two data sources is greater, and the credibility of this frame of data is lower.
[0066] In an implementation manner of the embodiment of the present application, determining the weight coefficient of the matching pair according to the target type of the detection box included in the matching pair includes:
[0067] Determining the weight coefficient of the matching pair according to the accuracy rate of the target detection network to be trained for detecting targets of the target type; wherein, the determined weight coefficient is inversely proportional to the accuracy rate.
[0068] When determining the weight coefficient of the matching pair, the accuracy rate of the target detection network to be trained for detecting targets of the target type can be obtained. If the accuracy rate is higher, it means that the accuracy rate of the target detection network for detecting targets of this type is originally relatively high, and obtaining the sample data of targets of this type to train the target detection network has limited improvement on the network performance. Therefore, a lower weight coefficient can be set, that is, the weight coefficient is inversely proportional to the accuracy rate. On the contrary, if the accuracy rate is lower, it means that the accuracy rate of the target detection network for detecting targets of this type is lower, and obtaining the sample data of targets of this type to train the target detection network has a greater improvement on the network performance. Therefore, a higher weight coefficient can be set.
[0069] 104. Selecting sample data from the matching results of multiple frames of data according to the confidence of the matching results of each frame of data.
[0070] After obtaining the confidence of the matching results of each frame of data, high-quality sample data can be selected from the matching results of multiple frames of data according to the size of the confidence. Here, usually the matching results of data with lower confidence can be selected as sample data.
[0071] In an implementation manner of the embodiment of the present application, selecting sample data from the matching results of multiple frames of data according to the confidence of the matching results of each frame of data includes:
[0072] Selecting the matching results of data with confidence less than the set threshold from the matching results of multiple frames of data as the mined sample data.
[0073] The confidence of the matching results of each frame of data can be used to characterize the error between the first vehicle data and the second vehicle data it contains, that is, if the confidence is lower, it means the corresponding error is greater. Therefore, when selecting sample data from the matching results of multiple frames of data, the matching results of data with lower confidence can be selected as much as possible. These matching results are more valuable when used as sample data. For example, a confidence screening threshold of 0.3 can be set, and all matching results of data with confidence higher than 0.3 are removed, that is, only the matching results of data with confidence not exceeding 0.3 are retained as the mined high-quality sample data.
[0074] In an embodiment of the present application, first, multiple frames of first vehicle data sent by in-vehicle units of each vehicle in a target area are acquired, and multiple frames of second vehicle data of all vehicles in the target area sent by a roadside computing unit are acquired; then, these first vehicle data and second vehicle data are matched according to timestamps, that is, each frame of first vehicle data and each frame of second vehicle data are associated according to timestamps, so as to obtain multiple frames of data matching results; next, for each frame of data matching result, the confidence of this frame of data matching result can be determined according to the first vehicle data and second vehicle data included in this frame of data matching result; finally, sample data can be screened and mined from these data matching results according to the confidence of each frame of data matching result. The confidence of each frame of data matching result can be used to characterize the error between the first vehicle data and the second vehicle data it includes, that is, if the confidence is lower, it means the corresponding error is larger. Therefore, when selecting sample data from multiple frames of data matching results, data matching results with lower confidence can be selected as much as possible. The error between the first vehicle data and the second vehicle data included in these data matching results is larger, so it is more valuable as sample data, and thus high-quality sample data can be mined.
[0075] It should be understood that the magnitudes of the sequence numbers of the steps in the above respective embodiments do not mean the order of execution. The order of execution of each process should be determined according to its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.
[0076] A method for mining sample data is mainly described above. Next, a device for mining sample data will be described.
[0077] Please refer to Figure 2 , an embodiment of a device for mining sample data in an embodiment of the present application includes:
[0078] A vehicle data acquisition module 201, configured to acquire multiple frames of first vehicle data sent by in-vehicle units of each vehicle in a target area, and acquire multiple frames of second vehicle data of all vehicles in the target area sent by a roadside computing unit;
[0079] A vehicle data matching module 202, configured to match multiple frames of first vehicle data and multiple frames of second vehicle data according to timestamps to obtain multiple frames of data matching results; wherein, each frame of data matching result includes one frame of first vehicle data and one frame of second vehicle data at the corresponding timestamp;
[0080] A confidence determination module 203, configured to, for each frame of data matching result, determine the confidence of this frame of data matching result according to the first vehicle data and second vehicle data included in this frame of data matching result;
[0081] A sample data mining module 204, configured to select sample data from the multi-frame data matching results according to the confidence of each frame of data matching result.
[0082] In an implementation manner of the embodiment of the present application, the confidence determination module includes:
[0083] A matching pair construction unit, configured to construct a plurality of matching pairs according to the first vehicle data and the second vehicle data included in the frame data matching result; wherein, each matching pair includes a first detection box included in the corresponding first vehicle data, and a second detection box corresponding to the first detection box included in the corresponding second vehicle data;
[0084] An overlap degree calculation unit, configured to calculate the overlap degree between the first detection box and the second detection box included in each matching pair respectively;
[0085] A confidence determination unit, configured to determine the confidence of the frame data matching result according to the overlap degree corresponding to each matching pair.
[0086] In an implementation manner of the embodiment of the present application, the confidence determination unit includes:
[0087] A weighted summation sub-unit, configured to perform a weighted summation calculation on the overlap degree corresponding to each matching pair to obtain the confidence of the frame data matching result.
[0088] In an implementation manner of the embodiment of the present application, the weighted summation sub-unit includes:
[0089] A weight coefficient determination sub-unit, configured to determine the weight coefficient of each matching pair according to the target type of the detection box included in the matching pair;
[0090] A confidence calculation sub-unit, configured to perform a weighted summation calculation on the overlap degree corresponding to each matching pair according to the weight coefficient of each matching pair to obtain the confidence of the frame data matching result.
[0091] In an implementation manner of the embodiment of the present application, the weight coefficient determination sub-unit includes:
[0092] A weight coefficient calculation sub-unit, configured to determine the weight coefficient of the matching pair according to the accuracy rate of the target detection network for detecting the target of the target type; wherein, the determined weight coefficient is inversely proportional to the accuracy rate.
[0093] In an implementation manner of the embodiment of the present application, the matching pair construction unit includes:
[0094] A matching pair construction subunit, which is used to determine the one-to-one correspondence between each detection box in the first vehicle data included in the matching result of this frame of data and each detection box in the second vehicle data included in the matching result of this frame of data through the Hungarian matching algorithm or the 3D IOU matching algorithm, so as to obtain multiple matching pairs.
[0095] In an implementation manner of the embodiment of the present application, the sample data mining module includes:
[0096] A sample data mining unit, which is used to select the data matching results with confidence levels less than the set threshold from the matching results of multiple frames of data as the mined sample data.
[0097] The embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the sample data mining method represented by any of the above embodiments.
[0098] The embodiment of the present application also provides a computer program product. When the computer program product runs on an electronic device, the electronic device is enabled to execute the sample data mining method represented by any of the above embodiments.
[0099] Figure 3 is a schematic diagram of an electronic device provided by an embodiment of the present application. As Figure 3 shown, the electronic device 3 of this embodiment includes: a processor 30, a memory 31, and a computer program 32 stored in the memory 31 and executable on the processor 30. When the processor 30 executes the computer program 32, it implements the steps in the embodiments of the above various sample data mining methods, such as Figure 1 the steps 101 to 104 shown. Or, when the processor 30 executes the computer program 32, it implements the functions of each module / unit in the above device embodiments, such as Figure 2 the functions of the modules 201 to 204 shown.
[0100] The computer program 32 can be divided into one or more modules / units. The one or more modules / units are stored in the memory 31 and executed by the processor 30 to complete the present application. The one or more modules / units can be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program 32 in the electronic device 3.
[0101] The so-called processor 30 may be a Central Processing Unit (CPU), or may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.
[0102] The memory 31 may be an internal storage unit of the electronic device 3, such as the hard disk or memory of the electronic device 3. The memory 31 may also be an external storage device of the electronic device 3, such as a plug-in hard disk equipped on the electronic device 3, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. Further, the memory 31 may also include both the internal storage unit of the electronic device 3 and the external storage device. The memory 31 is used to store the computer program and other programs and data required by the electronic device. The memory 31 may also be used to temporarily store data that has been output or is to be output.
[0103] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, only the above division of each functional unit and module is used as an example. In actual applications, the above functions can be allocated to different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiment may be integrated in a processing unit, or each unit may exist physically alone, or two or more units may be integrated in one unit. The above integrated unit may be implemented in the form of hardware or in the form of a software functional unit. In addition, the specific names of each functional unit and module are only for the convenience of mutual distinction and do not limit the protection scope of this application. The specific working processes of the units and modules in the above system can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated here.
[0104] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, the specific working processes of the above-described system, device, and unit can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated here.
[0105] In the above embodiments, the descriptions of the various embodiments have their own emphases. For parts not detailed or recorded in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.
[0106] Those of ordinary skill in the art will appreciate that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or in a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.
[0107] In the embodiments provided in this application, it should be understood that the disclosed apparatus and method can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of the modules or units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the couplings or direct couplings or communication connections shown or discussed with each other can be through some interfaces, and the indirect couplings or communication connections of the devices or units can be in electrical, mechanical or other forms.
[0108] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiments of this application.
[0109] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units.
[0110] When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, to implement all or part of the processes in the above-mentioned embodiment methods of this application, it can also be completed by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-mentioned various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file or some intermediate form, etc. The computer-readable medium can include: any entity or device that can carry the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content included in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.
[0111] The above-mentioned embodiments are only used to illustrate the technical solutions of this application, rather than to limit it; although this application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of this application, and should all be included in the protection scope of this application.
Claims
1. A method for sample data mining, characterized in that Including: Obtaining multiple frames of first vehicle data sent by in-vehicle units of each vehicle in the target area, and obtaining multiple frames of second vehicle data of all vehicles in the target area sent by the roadside computing unit; Matching the multiple frames of first vehicle data and the multiple frames of second vehicle data according to timestamps to obtain multiple frames of data matching results; wherein, each frame of the data matching result includes one frame of the first vehicle data and one frame of the second vehicle data corresponding to the timestamp; For each frame of the data matching result, determining the confidence level of this frame of data matching result according to the first vehicle data and the second vehicle data included in this frame of data matching result; Selecting sample data from the multiple frames of data matching results according to the confidence level of each frame of the data matching result.
2. The method according to claim 1, characterized in that, The determining the confidence level of this frame of data matching result according to the first vehicle data and the second vehicle data included in this frame of data matching result includes: Constructing multiple matching pairs according to the first vehicle data and the second vehicle data included in this frame of data matching result; wherein, each matching pair includes a first detection box included in the corresponding first vehicle data and a second detection box corresponding to the first detection box included in the corresponding second vehicle data; Calculating the overlap degrees of the first detection box and the second detection box included in each of the matching pairs respectively; Determining the confidence level of this frame of data matching result according to the overlap degrees corresponding to each of the matching pairs.
3. The method according to claim 2, wherein The determining the confidence level of this frame of data matching result according to the overlap degrees corresponding to each of the matching pairs includes: Performing a weighted summation calculation on the overlap degrees corresponding to each of the matching pairs to obtain the confidence level of this frame of data matching result.
4. The method according to claim 3, wherein The performing a weighted summation calculation on the overlap degrees corresponding to each of the matching pairs to obtain the confidence level of this frame of data matching result includes: For each matching pair, determining the weight coefficient of this matching pair according to the target type of the detection box included in this matching pair; Performing a weighted summation calculation on the overlap degrees corresponding to each of the matching pairs according to the weight coefficients of each of the matching pairs to obtain the confidence level of this frame of data matching result.
5. The method according to claim 4, wherein The determining the weight coefficient of this matching pair according to the target type of the detection box included in this matching pair includes: Determining the weight coefficient of this matching pair according to the accuracy rate of detecting the target of the target type by the target detection network to be trained; wherein, the determined weight coefficient is inversely proportional to the accuracy rate.
6. The method according to claim 2, wherein The constructing multiple matching pairs according to the first vehicle data and the second vehicle data included in this frame of data matching result includes: Determining the one-to-one correspondence between each detection box in the first vehicle data included in this frame of data matching result and each detection box in the second vehicle data included in this frame of data matching result through the Hungarian matching algorithm or the 3D IOU matching algorithm, thereby obtaining the multiple matching pairs.
7. The method according to any one of claims 1 to 6, characterized in that The selecting sample data from the multiple frames of data matching results according to the confidence level of each frame of the data matching result includes: From the multi-frame data matching results, select the data matching results with a confidence level less than the set threshold as the mined sample data.
8. A sample data mining device, characterized in that, It includes: A vehicle data acquisition module, configured to acquire multiple frames of first vehicle data sent by in-vehicle units of each vehicle in the target area, and acquire multiple frames of second vehicle data of all vehicles in the target area sent by the roadside computing unit; A vehicle data matching module, configured to match the multiple frames of first vehicle data and the multiple frames of second vehicle data according to timestamps to obtain multi-frame data matching results; wherein, each frame of the data matching results includes one frame of the first vehicle data and one frame of the second vehicle data corresponding to the timestamp; A confidence level determination module, configured to determine the confidence level of each frame of the data matching results according to the first vehicle data and the second vehicle data included in the frame of the data matching results; A sample data mining module, configured to select sample data from the multi-frame data matching results according to the confidence level of each frame of the data matching results.
9. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the sample data mining method according to any one of claims 1 to 7.
10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the sample data mining method according to any one of claims 1 to 7.