A Method for Analyzing the Confidence of Multi-Source Data of Intelligent Connected Vehicles

Abstract

The present invention proposes a method for analyzing the confidence level of multi-source data of intelligent connected vehicles, aiming to improve the data reliability of the autonomous driving system. The method includes obtaining multi-source data from multiple sensors of the vehicle, adding credibility to the data to generate vehicle environment cloud point data; establishing a database to store the historical information of all road data traveled by the vehicle; cleaning the current vehicle environment cloud point data to remove dynamic information and obtain road static cloud point data; comparing the current road static cloud point data obtained by the vehicle with the data in the database to obtain the confidence level of the multi-source data of the vehicle. By identifying outliers and removing dynamic information, the accuracy and reliability of the data are ensured. At the same time, the credibility is dynamically adjusted in combination with the historical performance of the sensor and environmental factors, improving the adaptability and generality of the system.

Description

Technical Field

[0001] The present invention belongs to the technical field of autonomous driving, and relates to a method for analyzing the confidence of multi-source data of an intelligent connected vehicle. Background Art

[0002] For safe operation, when planning a path through an environment, a vehicle should consider objects such as vehicles, people, trees, animals, buildings, signs, and poles. To do this, an object detector can be used to accurately detect in real time one or more objects depicted in an image (e.g., an image captured using one or more sensors mounted on an autonomous vehicle).

[0003] In the existing field of autonomous driving, the data obtained by sensors is processed on the premise of being trustworthy. However, in practical applications, the credibility of the data cannot be guaranteed. If the sensors obtain untrustworthy data, it may lead to missed or incorrect or delayed object detections, which may result in inappropriate paths or even collisions. Summary of the Invention

[0004] To solve the problems in the background art, the present invention proposes a method for analyzing the confidence of multi-source data of an intelligent connected vehicle.

[0005] To achieve the above object, the technical solution adopted by the present invention is as follows.

[0006] A method for analyzing the confidence of multi-source data of an intelligent connected vehicle includes: obtaining multi-source data acquired by multiple vehicle sensors, adding credibility to the data based on the attributes of the vehicle sensors, and generating current vehicle environment cloud point data through a point cloud detection algorithm; establishing a database to store the historical information of all road data traveled by the vehicle; cleaning the current vehicle environment cloud point data through a quartile algorithm, and removing dynamic information from the current vehicle environment cloud point data through a Harris corner detection algorithm to obtain current road static cloud point data; comparing the current road static cloud point data acquired by the vehicle with the data in the database, and adding a weighting algorithm based on the comparison result to obtain the confidence of the vehicle multi-source data.

[0007] Further, the specific method for adding credibility to the data based on the attributes of the vehicle sensors is: setting different attributes based on the sensor devices when the data source is acquired, setting the initial credibility of the sensor attributes according to the nominal accuracy of the sensors; dynamically adjusting the initial credibility according to the historical performance of the sensors to generate the credibility of the sensors; adding the influence of environmental factors on the sensor data, and making corresponding temporary adjustments to the credibility of a certain attribute of the sensor.

[0008] Further, the content saved in the database includes: vehicle ID, timestamp, location, speed, acceleration, sensor ID, sensor attribute, credibility field; road ID, historical road static cloud point data, and anchored road facilities.

[0009] Further, the specific method for cleaning the current vehicle environment cloud point data is as follows: First, perform outlier identification and calculate the first quartile of the current vehicle environment cloud point data and the third quartile , and the specific formula is:

[0010] , , where the second quartile is the median of the current vehicle environment cloud point data;

[0011] Based on the first quartile and the third quartile calculate the interquartile range , and the specific formula is:

[0012] ;

[0013] Based on the interquartile range determine the outlier threshold, and the specific formula is: ;

[0014] Clean the current vehicle environment cloud point data according to the outlier judgment method.

[0015] Further, the specific method for removing dynamic information from the current vehicle environment cloud point data through the Harris corner detection algorithm is as follows: Establish a cloud point data list based on multi-source data obtained by multiple vehicle sensors, establish a time series based on the timestamps in the cloud point data list, compare the data in the time series, and set multiple comparison windows for comparison; Judge the displacement of the information contained in any window in the time series, compare it based on the relative position of any cloud point and the location of the current vehicle, judge the displacement of the cloud point, if the displacement speed of the cloud point is consistent with the driving speed of the current vehicle and in the opposite direction, then judge that the cloud point is a static cloud point; Judge the movement of cloud points in all windows, remove the dynamic cloud points, and obtain the current road static cloud point data.

[0016] Further, the specific method for comparing the current road static cloud point data obtained by the vehicle with the data in the database is as follows: Extract the road static cloud point data corresponding to the road ID from the database based on the road ID of the current vehicle's driving road, obtain the current road static cloud point data obtained by the current vehicle, and use the Euclidean distance to judge the similarity between the current road static cloud point data and the data in the database; Generate a similarity list based on multiple roads and multiple vehicles to complete the comparison between the current road static cloud point data obtained by the current vehicle and the data in the database.

[0017] Further, the specific method for obtaining the confidence level of the multi-source data of the vehicle is as follows: Adopt a weighted algorithm to combine the credibility of the sensor with the matching result of the similarity list to calculate the comprehensive confidence level of the current data.

[0018] Compared with the prior art, the present invention has the following beneficial effects:

[0019] Through the steps of outlier identification and dynamic information removal, the present invention can effectively clean the data obtained by the vehicle, ensure the accuracy of the static cloud point data, and make the subsequent analysis more reliable. At the same time, the present invention allows for dynamic adjustment of the credibility based on the historical performance of the sensor and environmental factors, and can adapt to different driving environments and conditions in real time, thereby improving the data accuracy.

[0020] By adding credibility to the data of different sensors, the present invention can effectively identify and filter unreliable data sources, thereby improving the reliability of the overall system data. Adopting a weighted algorithm to combine the credibility of the sensor with the matching result of the similarity list provides a method for comprehensively evaluating the data confidence level, making the final result more credible.

[0021] The present invention is not only applicable to various vehicle models and sensor configurations, but also can be adjusted according to different driving scenarios, and has good versatility and adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 is a flowchart of the method of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0024] As Figure 1 shown, the technical solution adopted by the present invention is as follows: An intelligent network-connected vehicle multi-source data confidence level analysis method includes:

[0025] Obtain multi-source data acquired by multiple sensors of the vehicle, add credibility to the data based on the attributes of vehicle sensors, and generate current vehicle environment cloud point data through a point cloud detection algorithm.

[0026] Establish a database to save the historical information of all road data traveled by the vehicle.

[0027] Clean the current vehicle environment cloud point data through the quartile algorithm, and remove the dynamic information in the current vehicle environment cloud point data through the Harris corner detection algorithm to obtain the current road static cloud point data.

[0028] Compare the current road static cloud point data obtained by the vehicle with the data in the database, and add a weighting algorithm based on the comparison result to obtain the confidence of the vehicle multi-source data.

[0029] The present invention needs to establish a database to save the historical information of all road data traveled by the vehicle. The content saved in the database includes:

[0030] Vehicle ID, timestamp, location, speed, acceleration, sensor ID, sensor attribute, credibility field.

[0031] It contains vehicle-related information, specifically including vehicle ID, which is used to uniquely identify each vehicle; timestamp, which accurately records the time when the data is generated; location, which clearly indicates the specific geographical location where the vehicle is located; speed, which shows the driving speed of the vehicle at that time; acceleration, which reflects the change of the vehicle speed; sensor ID, which is used to distinguish different sensors; sensor attribute, which describes the characteristics of the sensor itself; and credibility field, which is used to evaluate the reliability of the data.

[0032] Road ID, historical road static cloud point data, anchored road facilities.

[0033] It involves data related to the road, including road ID, which is used to distinguish different roads; historical road static cloud point data, which is static information about road morphology, characteristics, etc.; and anchored road facilities, perhaps some information related to important facilities fixed on the road, which is of great significance for understanding the road conditions and the relationship between the vehicle and the road.

[0034] The following is a specific method for adding credibility to the data based on the attributes of vehicle sensors:

[0035] First, set different attributes according to the differences in the sensor devices involved when obtaining the data source. On this basis, set the initial credibility of the sensor attribute according to the nominal accuracy of the sensor itself. This means that sensors with different accuracies are given corresponding credibility base values at the initial stage.

[0036] Next, the initial credibility needs to be dynamically adjusted according to the historical performance of the sensor during actual use, so as to generate a more accurate credibility of the sensor. By analyzing and evaluating its past data, the credibility value is continuously optimized to make it more in line with the actual performance of the sensor.

[0037] In addition, the impact of environmental factors on sensor data also needs to be considered. When the environment changes, the credibility of a certain attribute of the sensor needs to be adjusted temporarily accordingly. For example, under harsh weather conditions, the performance of some sensors may be affected, and at this time, it is necessary to appropriately adjust their credibility to ensure the reliability and accuracy of the data.

[0038] After obtaining the multi-source data acquired by the vehicle sensors, it is necessary to clean the current vehicle environment cloud point data. The specific method is as follows:

[0039] First, perform outlier identification and calculate the first quartile of the current vehicle environment cloud point data and the third quartile , and the specific formula is:

[0040] ,

[0041] ,

[0042] where the second quartile is the median of the current vehicle environment cloud point data.

[0043] The interquartile range (IQR) is a statistic based on the position of the data. Its calculation is based on the quartiles of the data, is the 25th percentile, is the 50th percentile, is the 75th percentile. Compared with statistics such as the mean and standard deviation, it is not sensitive to outliers. For example, in a dataset containing extremely large or extremely small values, the mean is easily affected by these outliers and deviated, while the interquartile range focuses on the middle part of the data and will not change significantly due to a few extreme values.

[0044] Based on the first quartile and the third quartile calculate the interquartile range , and the specific formula is:

[0045] .

[0046] Based on the interquartile range determine the outlier threshold, and the specific formula is:

[0047] 。

[0048] Clean the current vehicle environment cloud point data according to the outlier judgment method.

[0049] The interquartile range helps to understand the dispersion of the middle 50% of the data. In data cleaning, by observing the size of the interquartile range, the dispersion of the data can be judged. If the interquartile range is small, it means that the middle part of the data is relatively concentrated; if the interquartile range is large, it indicates that the middle part of the data is relatively dispersed.

[0050] After cleaning the data, it is necessary to remove the dynamic data and rely only on the static data for subsequent analysis. The specific method for removing dynamic information is as follows:

[0051] First, construct a cloud point data list by using the multi-source data obtained by multiple sensors of the vehicle. On this basis, create a time series according to the timestamps in the cloud point data list to analyze and process the time characteristics of the data. Subsequently, carry out comparison work on the data in the time series, and set multiple comparison windows for detailed comparison operations.

[0052] During the comparison process, judge the displacement of the information contained in any window in the time series. This requires comparing based on the relative position of any cloud point and the current vehicle's location to accurately judge the displacement of the cloud point. Specifically, if the displacement speed of a certain cloud point is consistent with the current vehicle's driving speed and in the opposite direction, then it can be determined that this cloud point is a static cloud point. By judging the movement of cloud points in all windows one by one, the dynamic cloud points are identified and removed, and finally the current road static cloud point data is obtained. Such a processing process can effectively separate the static cloud point data from complex multi-source data, provide a more accurate and pure data basis for subsequent data analysis and applications, and help improve the accuracy and reliability of related research or applications.

[0053] The specific method for comparing the current road static cloud point data obtained by the vehicle with the data in the database is as follows:

[0054] First, accurately extract the road static cloud point data corresponding to the road ID from the database according to the road ID of the current vehicle's driving road. At the same time, obtain the current road static cloud point data collected by multiple sensors of the current vehicle. Then, use the Euclidean distance to measure the similarity between the current road static cloud point data and the data in the database.

[0055] The Euclidean distance is a commonly used method to measure the distance between two points in a metric space. Here it can effectively reflect the degree of difference in the spatial positions of two sets of cloud point data. By calculating the Euclidean distance, the similarity between two sets of data can be quantitatively judged.

[0056] Next, based on the relevant data of multiple roads and multiple vehicles, a similarity list is generated. For the data collected from different roads and vehicles, the similarity with the corresponding road data in the database is calculated respectively, and these similarity results are organized into a list form. In this way, a comprehensive comparison between the current road static cloud point data obtained by the vehicle and the data in the database is completed, providing a basis for subsequent analysis and applications.

[0057] Recently, the confidence of multi-source data of the vehicle is calculated. In obtaining the confidence of multi-source data of the vehicle, a weighted algorithm is adopted. Specifically, the credibility of the sensor is combined with the matching result of the above similarity list. The credibility of the sensor reflects the reliability of the sensor data, while the matching result of the similarity list reflects the degree of conformity between the data obtained by the vehicle and the standard data in the database. By reasonably setting the weights, these two factors are comprehensively calculated to obtain the comprehensive confidence of the current data. This comprehensive confidence can more comprehensively and accurately evaluate the quality and reliability of the vehicle multi-source data, providing strong support for relevant decisions and applications.

[0058] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A confidence analysis method for multi-source data of intelligent connected vehicles, characterized in that: Included are: Acquire multi-source data from multiple sensors of the car, add credibility to the data based on the attributes of the vehicle sensors, and generate cloud point data of the current vehicle environment through point cloud detection algorithms; Establish a database to save the historical information of all road data that vehicles have traveled; The current vehicle environment cloud point data is cleaned by using the quartile algorithm, and the dynamic information in the current vehicle environment cloud point data is removed by using the Harris corner detection algorithm to obtain the current road static cloud point data; Compare the current road static cloud point data obtained by the vehicle with the data in the database, and add a weighted algorithm based on the comparison results to obtain the confidence of the multi-source data of the vehicle; The specific method of adding credibility to data based on the attributes of vehicle sensors is: Different attributes are set based on the sensor device when the data source is acquired, and the initial credibility of the sensor attribute is set according to the nominal accuracy of the sensor; Dynamically adjust the initial credibility based on the historical performance of the sensor to generate the credibility of the sensor; Add the impact of environmental factors on sensor data and make corresponding temporary adjustments to the credibility of a certain attribute of the sensor; The specific method for cleaning the current vehicle environment cloud point data is: First, outliers are identified and the first quartile of the current vehicle environment cloud point data is calculated and the third quartile , the specific formula is: ； ； The second quartile is the median of the current vehicle environment cloud point data, based on the first quartile and the third quartile Calculate the interquartile range , the specific formula is: ； Based on interquartile range Determine the outlier threshold. The specific formula is: ； Clean the current vehicle environment cloud point data based on the judgment method of abnormal values; The specific method of removing dynamic information in the current vehicle environment cloud point data by using the Harris corner detection algorithm is: Establish a cloud point data list based on multi-source data acquired by multiple sensors of the car, establish a time series based on the timestamps in the cloud point data list, compare the data in the time series, and set multiple comparison windows for comparison; Determine the displacement of the information contained in any window in the time series, and compare the relative position of any cloud point with the current vehicle position to determine the displacement of the cloud point. If the displacement speed of the cloud point is consistent with the current vehicle speed and in the opposite direction, the cloud point is determined to be a static cloud point. Determine the movement of cloud points in all windows, remove dynamic cloud points, and obtain static cloud point data of the current road.

2. The method for analyzing the confidence of multi-source data of an intelligent connected vehicle according to claim 1, characterized in that: The contents stored in the database include: Vehicle ID, timestamp, location, speed, acceleration, sensor ID, sensor attributes, and credibility fields; Road ID, historical road static cloud point data, and anchored road facilities.

3. The method for analyzing the confidence of multi-source data of intelligent connected vehicles according to claim 1, characterized in that: The specific method of comparing the current road static cloud point data obtained by the vehicle with the data in the database is: Based on the road ID of the road on which the current vehicle is traveling, the road static cloud point data corresponding to the road ID is extracted from the database, the current road static cloud point data obtained by the current vehicle is obtained, and the Euclidean distance is used to determine the similarity between the current road static cloud point data and the data in the database; A similarity list is generated based on multiple roads and multiple vehicles to compare the static cloud point data of the current road obtained by the current vehicle with the data in the database.

4. The method for analyzing the confidence of multi-source data of an intelligent connected vehicle according to claim 3, characterized in that: The specific method for obtaining the confidence of automobile multi-source data is: A weighted algorithm is used to combine the sensor's credibility with the matching results of the similarity list to calculate the comprehensive confidence of the current data.

Images