Roadside sensing data quality monitoring method, system and equipment and storage medium

Abstract

The invention provides a roadside sensing data quality monitoring method, system and device and a storage medium, and the method comprises the steps: building and training a regression model according to historical data, and recognizing an abnormal point in test data through the regression model; determining an abnormal point threshold according to the identified multi-feature joint distribution of the abnormal points; and monitoring roadside sensing data in real time through the abnormal point threshold. According to the method, the threshold value of the abnormal point is determined based on abnormal point multi-feature joint distribution, the abnormal point is determined according to the threshold value of the abnormal point, the quality of the sensing data can be monitored in real time, and optimization of the roadside sensing model is assisted.

Description

Technical Field

[0001] The present invention relates to the field of vehicle-road cooperation, and more particularly, to a method, system, device and storage medium for monitoring the quality of roadside perception data. Background Art

[0002] In the field of vehicle-road cooperation, the quality of roadside perception data plays a crucial role in determining the ultimate success or failure of the vehicle-road cooperation system. Therefore, we need to comprehensively evaluate and monitor the quality of the original roadside perception data. Although there are some data quality evaluation methods in the Internet industry, most of them are too general and cannot fully meet the requirements of the autonomous driving field. Summary of the Invention

[0003] In view of this, an object of an embodiment of the present invention is to provide a method, system, electronic device and computer-readable storage medium for monitoring the quality of roadside perception data. The present invention determines the threshold of outliers based on the joint distribution of multiple features of outliers, and determines outliers according to the threshold of outliers, which can monitor the quality of perception data in real time and assist in optimizing the roadside perception model.

[0004] Based on the above object, an aspect of an embodiment of the present invention provides a method for monitoring the quality of roadside perception data, including the following steps: establishing and training a regression model according to historical data, and identifying outliers in test data through the regression model; determining an outlier threshold according to the joint distribution of multiple features of the identified outliers; and monitoring roadside perception data in real time through the outlier threshold.

[0005] In some embodiments, the step of establishing and training a regression model according to historical data includes: determining basic features and difference features of traffic participants according to importance, and using normal operation trajectories as a training data set; and establishing and training a regression model according to the training data set, the basic features and the difference features.

[0006] In some embodiments, the step of determining basic features and difference features of traffic participants according to importance includes: parsing basic attributes of traffic participants from roadside perception data as basic features, and taking the difference between features in two adjacent frames of images of the same traffic participant as difference features.

[0007] In some embodiments, the step of establishing and training a regression model according to the training data set, the basic features and the difference features includes: establishing a regression model through the relationship between the basic features and the difference features, and training the regression model according to the training data set to fit the relationship between the features and the abnormal distribution.

[0008] In some embodiments, the step of determining the outlier threshold according to the joint distribution of multiple features of the identified outliers includes: constructing a covariance matrix of multiple features, constructing a probability density function of a Gaussian distribution according to the covariance matrix; and calculating the threshold of each feature in the probability density function using a preset confidence level.

[0009] In some embodiments, the step of real-time monitoring the roadside perception data through the outlier threshold includes: determining the outliers in the roadside perception data according to the outlier threshold, and triggering an alarm mechanism in response to the number of outliers exceeding a number threshold.

[0010] In some embodiments, the step of real-time monitoring the roadside perception data through the outlier threshold includes: calculating a new outlier threshold according to the outliers to iteratively update the regression model, and drawing a real-time heat distribution statistical chart according to the distribution and number of the outliers.

[0011] On the other hand, an embodiment of the present invention provides a system for monitoring the quality of roadside perception data, including: a model module configured to establish and train a regression model according to historical data, and identify outliers in test data through the regression model; a threshold module configured to determine an outlier threshold according to the joint distribution of multiple features of the identified outliers; and a monitoring module configured to real-time monitor the roadside perception data through the outlier threshold.

[0012] In yet another aspect of the embodiments of the present invention, an electronic device is further provided, including: at least one processor; and a memory storing computer instructions executable on the processor, and when the instructions are executed by the processor, the steps of the above method are implemented.

[0013] In still another aspect of the embodiments of the present invention, a computer-readable storage medium is further provided, and the computer-readable storage medium stores a computer program that implements the steps of the above method when executed by a processor.

[0014] The present invention has the following beneficial technical effects: determining the threshold of outliers based on the joint distribution of multiple features of outliers, and determining outliers according to the outlier threshold, can real-time monitor the quality of perception data and assist in optimizing the roadside perception model. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention, and those of ordinary skill in the art can obtain other embodiments according to these drawings without creative efforts.

[0016] Figure 1 Schematic diagram of an embodiment of the method for monitoring the quality of roadside perception data provided by the present invention;

[0017] Figure 2 Architecture diagram of an embodiment of the method for monitoring the quality of roadside perception data provided by the present invention;

[0018] Figure 3 Flowchart of iterative optimization and application of the regression model provided by the present invention;

[0019] Figure 4 Schematic diagram of an embodiment of the system for monitoring the quality of roadside perception data provided by the present invention;

[0020] Figure 5 Schematic diagram of the hardware structure of an embodiment of the electronic device for monitoring the quality of roadside perception data provided by the present invention;

[0021] Figure 6 Schematic diagram of an embodiment of the computer storage medium for monitoring the quality of roadside perception data provided by the present invention. Detailed implementation manners

[0022] To make the objectives, technical solutions, and advantages of the present invention clearer and more understandable, the following further describes the embodiments of the present invention in detail with reference to specific embodiments and the accompanying drawings.

[0023] It should be noted that all the expressions using "first" and "second" in the embodiments of the present invention are used to distinguish two entities or parameters with the same name but different identities. It can be seen that "first" and "second" are only for the convenience of expression and should not be construed as a limitation on the embodiments of the present invention. This will not be elaborated one by one in the subsequent embodiments.

[0024] In the first aspect of the embodiments of the present invention, an embodiment of a method for monitoring the quality of roadside perception data is proposed. Figure 1 Shown is a schematic diagram of an embodiment of the method for monitoring the quality of roadside perception data provided by the present invention. As Figure 1 shown, the embodiments of the present invention include the following steps:

[0025] S1. Establish and train a regression model based on historical data, and identify abnormal points in the test data through the regression model;

[0026] S2. Determine the abnormal point threshold according to the multi-feature joint distribution of the identified abnormal points; and

[0027] S3. Monitor the roadside perception data in real time through the abnormal point threshold.

[0028] Figure 2This is the architecture diagram of the method for monitoring the quality of roadside perception data provided by the present invention. According to Figure 2 This will illustrate the embodiments of the present invention.

[0029] A regression model is established and trained based on historical data, and outliers in the test data are identified through the regression model.

[0030] In some embodiments, the step of establishing and training a regression model based on historical data includes: determining the basic features and difference features of traffic participants according to importance, and using the normal operation trajectory as the training data set; and establishing and training a regression model according to the training data set, the basic features and the difference features.

[0031] In some embodiments, the step of determining the basic features and difference features of traffic participants according to importance includes: parsing the basic attributes of traffic participants from the roadside perception data as the basic features, and taking the difference between the features of two adjacent frames of images of the same traffic participant as the difference features. Basic feature extraction: Parse the basic attributes of traffic participants from the roadside perception data, including information such as type, speed, acceleration, shape, orientation, position, etc. Difference feature calculation: For each traffic participant, calculate the differences in key features such as speed, acceleration, speed orientation, head orientation, shape, etc. between two adjacent frames of data to capture key motion change information.

[0032] In some embodiments, the step of establishing and training a regression model according to the training data set, the basic features and the difference features includes: establishing a regression model through the relationship between the basic features and the difference features, and training the regression model according to the training data set to fit the relationship between the features and the abnormal distribution.

[0033] Based on big data statistics, find vehicles with normal trajectories (which means there is no abnormality in the roadside perception data), select the corresponding roadside perception data as the marked data, and extract feature data from it; select a multiple regression model and model through the relationship between the normal data set and the combined multi-features; use the marked data to train the regression model to fit the relationship between the features and the abnormal distribution; jointly use the mean square error (MSE), R square (R 2 ) and accuracy to evaluate the trained regression model and check its fitting effect on the training data.

[0034] Determine the outlier threshold according to the joint distribution of multiple features of the outliers.

[0035] In some embodiments, the step of determining the anomaly point threshold according to the joint distribution of multiple features of the identified anomaly points includes: constructing a covariance matrix of multiple features, constructing a probability density function of a Gaussian distribution according to the covariance matrix; and calculating the threshold of each feature in the probability density function using a preset confidence level.

[0036] In some embodiments, the covariance matrix can be calculated in the following manner:

[0037] Given a data set X, assuming there are n samples and m features, the covariance matrix S of the data set X can be obtained based on the following formula (1):

[0038]

[0039] Where:

[0040] · S is the covariance matrix.

[0041] · n is the number of samples.

[0042] · m is the number of features.

[0043] · X i is the i-th sample, which is a column vector containing m features.

[0044] · is the mean vector of all samples, where the j-th element is the mean of all samples on the j-th feature.

[0045] Specifically, for each feature j, the mean is calculated as shown in the following formula (2):

[0046]

[0047] Then, the element S of the covariance matrix ij is calculated as shown in the following formula (3):

[0048]

[0049] Here, and are the means of the i-th feature and the j-th feature respectively. The finally obtained matrix S is an m×m matrix, where S ij represents the covariance between the i-th feature and the j-th feature.

[0050] In some embodiments, the probability density function of a multivariate Gaussian distribution can be constructed in the following manner:

[0051] For the probability density function of a multivariate Gaussian distribution, when extending the Gaussian distribution to a multi-dimensional space, it can be assumed that each space in the multi-dimensional space is completely independent. Then, an independent multivariate Gaussian distribution can be determined according to the following formula (4).

[0052]

[0053] Among them, x is the sample vector, μ is the mean vector, and ∑ is the covariance matrix.

[0054] With a 95% confidence level, the thresholds of each feature of the probability density function are calculated. Using the labeled normal trajectory data, a 99% (3 times the standard deviation) threshold range is determined. Using the joint threshold of each feature (outside the threshold range), bad-cases (abnormal points) are mined in real time from the roadside perception data.

[0055] The roadside perception data is monitored in real time through the abnormal point threshold.

[0056] In some embodiments, the step of monitoring the roadside perception data in real time through the abnormal point threshold includes: determining the abnormal points in the roadside perception data according to the abnormal point threshold, and triggering an alarm mechanism in response to the number of abnormal points exceeding the number threshold.

[0057] In some embodiments, the step of monitoring the roadside perception data in real time through the abnormal point threshold includes: calculating a new abnormal point threshold according to the abnormal points to iteratively update the regression model, and drawing a real-time heat distribution statistical chart according to the distribution and number of the abnormal points.

[0058] Figure 3 FIG. 300 is a flowchart of the iterative optimization and application of the regression model provided by the present invention. As Figure 3 shown, in step 301, a large amount of normal data is manually marked for the first time. In an embodiment of the present invention, the normal data refers to the roadside perception data of vehicles with normal trajectories. The roadside perception data of vehicles with normal trajectories means that there are no abnormalities in these roadside perception data. The manual marking method in this embodiment can be embodied as a process of obtaining multiple screening conditions based on big data statistics and screening the obtained roadside perception data through the multiple screening conditions to obtain a large amount of normal data. Those skilled in the art can understand that the above method of data screening by setting screening conditions is only an example rather than a limitation on the "manual marking" method.

[0059] Next, in step 302, a multi-feature regression analysis model is constructed using the marked normal data, and the combined threshold of the normal data is calculated. In this embodiment, the combined threshold can be understood as a range composed of multiple data thresholds. The multiple data can be, for example, the data related to the basic attributes of traffic participants extracted from the roadside perception data and the difference data of key features described above.

[0060] Subsequently, in step 303, abnormal points can be mined according to the combined threshold of normal data. Abnormal points usually refer to data points that do not conform to the normal data distribution or pattern, which may be caused by data errors (such as data collection errors or processing errors caused by equipment failures), abnormal events (such as abnormal vehicle running trajectories), or other reasons. In this embodiment, the points where the feature combination exceeds the combined threshold are regarded as abnormal points.

[0061] Next, in step 304, according to the abnormal points, the threshold of the abnormal points is calculated using the regression model. For the calculation of the threshold of the abnormal points, the covariance matrix of multiple features can be constructed as described above, and the probability density function of the Gaussian distribution can be constructed based on the covariance matrix; and the threshold of each feature in the probability density function can be calculated using the preset confidence level. In an embodiment of the present invention, the method for determining the threshold of the abnormal points in the above formula (1)-(4) can be adopted. Those skilled in the art can understand that the above method for determining the threshold of the abnormal points is only an example rather than a limitation, and other methods such as the method based on business rules or the method of machine learning can also be used to confirm the threshold of the abnormal points.

[0062] Subsequently, referring to step 305, in step 305, the abnormal point threshold determined in step 304 above is used to mine the abnormal point positions in a flink stream, and a large number of abnormal point positions are obtained based on the mining (that is, the points where the combined feature threshold is less than the abnormal point threshold are abnormal points). The mining of abnormal points through flink stream is because Flink supports real-time data processing and can capture and analyze abnormal points in the stream data in real time, which is very useful for scenarios that require real-time response. And the stream processing framework of Flink has high processing capabilities, can quickly process a large amount of data, and discover abnormal points in real time. This helps to improve data processing efficiency, reduce latency, and quickly respond to abnormal situations. The stream processing framework of Flink also has high flexibility and can easily expand and adjust the processing flow. This makes the mining of abnormal points more flexible and customizable and can be optimized according to different business requirements. In addition, the stream processing framework of Flink can also provide detailed abnormal point information, including the location, cause, impact, etc. of the abnormal points. This helps to better understand the abnormal situation and take corresponding measures for repair and optimization and has high reliability, which can ensure the integrity and consistency of the data.

[0063] Then, in step 306, taking a preset time period as a cycle, all abnormal points mined in the previously preset time period are used for threshold iteration, and the latest model is continuously iterated according to new data. By continuously iterating and updating the threshold periodically, the accuracy and robustness of the model for abnormal point detection are gradually improved.

[0064] After the construction and iteration of the above model are completed, the updated model can be applied to the detection of abnormal points. For example, in step 307, the abnormal points can be divided into regions. In one embodiment, according to the sensing range of the radar, the service area can be divided into multiple small regions, and the point positions are determined for each of the multiple small regions respectively. Taking a preset time period as a cycle, in a rolling window manner, Flink is used to calculate the number of bad cases in each small region of a 5-minute window in real time for time determination. And if the number of abnormal points (bad cases) within a 5-minute rolling window exceeds 1% of the number of objects, then in this case, it is considered that the abnormal points are relatively concentrated, thus triggering an abnormal alarm for the roadside sensing data.

[0065] Based on the abnormal point threshold obtained from the offline processing of the previous day, the roadside sensing data is processed in real-time streaming, and the results are pushed in real-time. When the sensing model has a relatively concentrated number of bad cases exceeding the quantity threshold, the alarm mechanism will be triggered; according to the distribution and quantity of bad cases, a real-time heat distribution statistic is drawn to monitor and optimize the roadside sensing model in real-time. According to the roadside sensing data in a previous period and the mined bad-case data, the parameters of the bad-case mining model are optimized and iterated.

[0066] It should be particularly noted that each step in each embodiment of the above method for monitoring the quality of roadside sensing data can be mutually crossed, replaced, added, or deleted. Therefore, these reasonable permutation and combination transformations for the method of monitoring the quality of roadside sensing data should also fall within the protection scope of the present invention, and the protection scope of the present invention should not be limited to the embodiments.

[0067] Based on the above purpose, the second aspect of the embodiments of the present invention proposes a system 400 for monitoring the quality of roadside sensing data. As Figure 4 shown, the system 400 includes the following modules: a model module configured to establish and train a regression model according to historical data and identify abnormal points in test data through the regression model; a threshold module configured to determine an abnormal point threshold according to the multi-feature joint distribution of the identified abnormal points; and a monitoring module configured to monitor the roadside sensing data in real time through the abnormal point threshold.

[0068] In some embodiments, the model module is further configured to: determine the basic features and difference features of traffic participants according to importance, and use the normal running trajectory as the training data set; and establish and train a regression model according to the training data set, the basic features and the difference features.

[0069] In some embodiments, the model module is further configured to: parse the basic attributes of traffic participants from the roadside perception data as the basic features, and use the difference between the features of two adjacent frames of the same traffic participant as the difference features.

[0070] In some embodiments, the model module is further configured to: establish a regression model through the relationship between the basic features and the difference features, and train the regression model according to the training data set to fit the relationship between the features and the abnormal distribution.

[0071] In some embodiments, the threshold module is further configured to: construct the covariance matrix of multiple features, construct the probability density function of the Gaussian distribution according to the covariance matrix; and calculate the thresholds of each feature in the probability density function using the preset confidence level.

[0072] In some embodiments, the monitoring module is further configured to: determine the abnormal points in the roadside perception data according to the abnormal point threshold, and trigger an alarm mechanism in response to the number of abnormal points exceeding the number threshold.

[0073] In some embodiments, the monitoring module is further configured to: calculate a new abnormal point threshold according to the abnormal points to iteratively update the parameters in the regression model, and draw a real-time heat distribution statistical chart according to the distribution and number of the abnormal points.

[0074] Based on the above object, in the third aspect of the embodiments of the present invention, an electronic device is proposed, including: at least one processor; and a memory storing computer instructions that can run on the processor, and the instructions are executed by the processor to implement the following steps: S1. Establish and train a regression model according to historical data, and identify abnormal points in the test data through the regression model; S2. Determine the abnormal point threshold according to the multi-feature joint distribution of the identified abnormal points; and S3. Real-time monitor the roadside perception data through the abnormal point threshold.

[0075] In some embodiments, the step of establishing and training a regression model according to historical data includes: determining the basic features and difference features of traffic participants according to importance, and using the normal running trajectory as the training data set; and establishing and training a regression model according to the training data set, the basic features and the difference features.

[0076] In some embodiments, the step of determining the basic features and difference features of traffic participants according to importance includes: parsing the basic attributes of traffic participants from roadside perception data as basic features, and taking the difference between the features of two adjacent frames of images of the same traffic participant as difference features.

[0077] In some embodiments, the step of establishing and training a regression model according to the training data set, the basic features and the difference features includes: establishing a regression model through the relationship between the basic features and the difference features, and training the regression model according to the training data set to fit the relationship between the features and the abnormal distribution.

[0078] In some embodiments, the step of determining the abnormal point threshold according to the joint distribution of multiple features of the identified abnormal points includes: constructing a covariance matrix of multiple features, constructing a probability density function of a Gaussian distribution according to the covariance matrix; and calculating the thresholds of each feature in the probability density function using a preset confidence level.

[0079] In some embodiments, the step of real-time monitoring of roadside perception data through the abnormal point threshold includes: determining abnormal points in the roadside perception data according to the abnormal point threshold, and triggering an alarm mechanism in response to the number of abnormal points exceeding a number threshold.

[0080] In some embodiments, the step of real-time monitoring of roadside perception data through the abnormal point threshold includes: calculating a new abnormal point threshold according to the abnormal points to iteratively update the parameters in the regression model, and drawing a real-time heat distribution statistical chart according to the distribution and number of the abnormal points.

[0081] As Figure 5 shown, it is a schematic hardware structure diagram of an embodiment of the above-mentioned electronic device for monitoring the quality of roadside perception data provided by the present invention.

[0082] Take the device as Figure 5 shown as an example. In this device, there is a processor 501 and a memory 502.

[0083] The processor 501 and the memory 502 can be connected through a bus or other means, Figure 5 and here take the connection through the bus as an example.

[0084] The memory 502, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules, such as the program instructions / modules corresponding to the method for monitoring the quality of roadside perception data in the embodiments of the present application. The processor 501 executes various functional applications and data processing of the server by running the non-volatile software programs, instructions, and modules stored in the memory 502, that is, implements the method for monitoring the quality of roadside perception data.

[0085] The memory 502 may include a program storage area and a data storage area. Among them, the program storage area can store an operating system and application programs required for at least one function; the data storage area can store data created according to the use of the method for monitoring the quality of roadside perception data, etc. In addition, the memory 502 may include a high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other non-volatile solid-state storage devices. In some embodiments, the memory 502 may optionally include a memory remotely provided relative to the processor 501, and these remote memories can be connected to the local module through a network. Examples of the above network include but are not limited to the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0086] One or more computer instructions 503 corresponding to the method for monitoring the quality of roadside perception data are stored in the memory 502. When executed by the processor 501, they execute the method for monitoring the quality of roadside perception data in any of the above method embodiments.

[0087] Any embodiment of the electronic device that executes the above method for monitoring the quality of roadside perception data can achieve the same or similar effects as any of the foregoing method embodiments corresponding thereto.

[0088] The present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, executes the method for monitoring the quality of roadside perception data.

[0089] As Figure 6 shown, it is a schematic diagram of an embodiment of the above computer storage medium for monitoring the quality of roadside perception data provided by the present invention. Taking the computer storage medium as shown in Figure 6 shown as an example, the computer-readable storage medium 601 stores a computer program 602 that, when executed by a processor, executes the above method.

[0090] Finally, it should be noted that those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The program for the method of monitoring the quality of roadside perception data can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above embodiments of each method. Among them, the storage medium of the program can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc. The embodiments of the above computer program can achieve the same or similar effects as the corresponding foregoing embodiments of any method.

[0091] The above are exemplary embodiments disclosed by the present invention. However, it should be noted that various changes and modifications can be made without departing from the scope of the embodiments disclosed by the present invention as defined by the claims. The functions, steps, and / or actions of the method claims according to the disclosed embodiments herein need not be performed in any specific order. In addition, although the elements disclosed in the embodiments of the present invention can be described or claimed in individual form, they can also be understood as plural unless explicitly limited to the singular.

[0092] It should be understood that, as used herein, unless the context clearly supports exceptions, the singular form "a" is also intended to include the plural form. It should also be understood that the "and / or" used herein refers to any and all possible combinations of one or more of the associated listed items.

[0093] The serial numbers of the disclosed embodiments of the present invention above are only for description and do not represent the superiority or inferiority of the embodiments.

[0094] Those of ordinary skill in the art can understand that all or part of the steps of the above embodiments can be completed by hardware or can be completed by instructing relevant hardware through a program. The program can be stored in a computer-readable storage medium. The storage medium mentioned above can be a read-only memory, a magnetic disk, an optical disk, etc.

[0095] Those of ordinary skill in the art should understand that the discussion of any of the above embodiments is only exemplary and is not intended to imply that the scope of the disclosure of the embodiments of the present invention (including the claims) is limited to these examples; under the concept of the embodiments of the present invention, the technical features between the above embodiments or different embodiments can also be combined, and there are many other variations in different aspects of the above embodiments of the present invention, which are not provided in detail for the sake of brevity. Therefore, any omissions, modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the embodiments of the present invention shall be included in the protection scope of the embodiments of the present invention.

Claims

1. A method for monitoring the quality of roadside perception data, characterized in that, It includes the following steps: Establish and train a regression model based on historical data, and identify outliers in the test data through the regression model; Determine the outlier threshold according to the multi-feature joint distribution of the identified outliers; And Monitor the roadside perception data in real time through the outlier threshold.

2. The method for monitoring the quality of roadside perception data according to claim 1, characterized in that, The step of establishing and training a regression model based on historical data includes: Determine the basic features and difference features of traffic participants according to importance, and use the normal operation trajectory as the training data set; and Establish and train a regression model according to the training data set, the basic features and the difference features.

3. The method for monitoring the quality of roadside perception data according to claim 2, characterized in that, The step of determining the basic features and difference features of traffic participants according to importance includes: Parse the basic attributes of traffic participants from the roadside perception data as the basic features, and use the difference between the features of two adjacent frames of the same traffic participant as the difference features.

4. The method for monitoring the quality of roadside perception data according to claim 2, characterized in that, The step of establishing and training a regression model according to the training data set, the basic features and the difference features includes: Establish a regression model through the relationship between the basic features and the difference features, and train the regression model according to the training data set to fit the relationship between the features and the abnormal distribution.

5. The method for monitoring the quality of roadside perception data according to claim 1, characterized in that, The step of determining the outlier threshold according to the multi-feature joint distribution of the identified outliers includes: Construct a covariance matrix of multi-features, and construct a probability density function of Gaussian distribution according to the covariance matrix; and Calculate the threshold of each feature in the probability density function using a preset confidence level.

6. The method for monitoring the quality of roadside perception data according to claim 1, characterized in that, The step of monitoring the roadside perception data in real time through the outlier threshold includes: Determine the outliers in the roadside perception data according to the outlier threshold, and trigger an alarm mechanism in response to the number of outliers exceeding the number threshold.

7. The method for monitoring the quality of roadside perception data according to claim 6, characterized in that, The step of monitoring the roadside perception data in real time through the outlier threshold includes: Calculate a new outlier threshold according to the outliers to iteratively update the regression model, and draw a real-time heat distribution statistical chart according to the distribution and number of the outliers.

8. A system for monitoring the quality of roadside perception data, characterized in that, It includes: A model module configured to establish and train a regression model based on historical data, and identify outliers in the test data through the regression model; A threshold module configured to determine the outlier threshold according to the multi-feature joint distribution of the identified outliers; And A monitoring module configured to monitor the roadside perception data in real time through the outlier threshold.

9. An electronic device, characterized in that, It includes: At least one processor; And A memory storing computer instructions that can be run on the processor, and when the instructions are executed by the processor, the steps of the method according to any one of claims 1-7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, the steps of the method according to any one of claims 1-7 are implemented.

Images