Gross error detection method and system for RTK positioning results

Abstract

The present application relates to the field of satellite positioning technology, and discloses a method and system for checking gross errors of RTK positioning results, the method comprising: obtaining gross error marking results of RTK positioning results and the value of at least one gross error related feature; training a gross error checking model by using a classification algorithm, as well as the gross error marking results and gross error related features of each RTK positioning result, so as to determine a reference bias value of the gross error checking model and a reference weight coefficient corresponding to each gross error related feature; substituting the gross error related features corresponding to the RTK positioning result to be checked into the trained gross error checking model, so as to perform gross error checking on the RTK positioning result to be checked.

Description

Technical Field

[0001] The present application relates to the field of satellite positioning technology, and in particular to positioning result verification technology. Background Art

[0002] High-precision GNSS (Global Navigation Satellite System) positioning technology must use carrier phase observations, and RTK (Real-time kinematic) positioning technology is a real-time kinematic positioning technology based on carrier phase observations.

[0003] Specifically, the standard mode of RTK positioning technology is to set up a base station in a certain area. The base station encodes the observed carrier, pseudorange observation values, base station coordinates, and antenna information into RTCM (International Radio Technical Committee for Maritime Affairs) differential format data through a data link and sends it to the mobile station user; the user end receives the differential format data from the base station, combines its own observation data, and constructs inter-station and inter-satellite double difference observation values, eliminating the receiver carrier pseudorange bias, receiver clock error, and satellite clock error, and greatly weakens the influence of satellite orbit error, atmospheric error, etc. on the integer ambiguity search, so that the integer ambiguity can be quickly fixed to obtain real-time centimeter-level positioning results.

[0004] Among them, RTK positioning gross error refers to the positioning result that the positioning accuracy fails to meet the preset standard due to the influence of various deviations that have not been eliminated in the observation values, satellite geometric configuration, and integer ambiguity fixation errors in the positioning algorithm during the solution process.

[0005] Furthermore, the purpose of the gross error test is to identify and eliminate the solution results that do not meet the positioning accuracy standards, so as to improve the positioning accuracy of the RTK algorithm.

[0006] The positioning result of RTK is obtained through least square estimation. The traditional gross error detection method only relies on the residual after the ambiguity is fixed. When the residual is greater than a certain threshold or fails the chi-square test, the positioning result is judged as a gross error.

[0007] However, there are some problems with the current technology for gross error detection of RTK positioning results, such as insufficient real-time performance, insufficient speed, insufficient accuracy, etc. Summary of the invention

[0008] The purpose of this application is to provide a method and system for checking gross errors of RTK positioning results, which can check gross errors of RTK positioning results in a more real-time, faster and more accurate manner.

[0009] The present application discloses a method for detecting gross errors of RTK positioning results, comprising:

[0010] Obtaining a gross error marking result of the RTK positioning result and a value of at least one gross error related feature, wherein the gross error related feature includes: one or more of a GDOP value, a number of satellites fixedly eliminated for partial ambiguity, a number of fixed satellites, a residual of an ionosphere-free combination, a residual of a wide-lane combination, a positioning difference between a first-frequency positioning result and a second-frequency positioning result, a positioning difference between a positioning result of an ionosphere-free combination and a RTK positioning result, and a positioning difference between a positioning result of a wide-lane combination and a RTK positioning result;

[0011] Using a classification algorithm, as well as the gross error marking result of each RTK positioning result and the gross error related features, a gross error detection model is trained to determine a reference bias value of the gross error detection model and a reference weight coefficient corresponding to each gross error related feature;

[0012] The gross error related features corresponding to the RTK positioning result to be tested are substituted into the trained gross error test model to perform a gross error test on the RTK positioning result to be tested.

[0013] In a preferred example, the gross error related features also include ratio values ​​and / or residuals of RTK positioning.

[0014] In a preferred example, the RTK positioning result is a single-frequency RTK positioning result, and the gross error related features include: residual error of RTK positioning, GDOP value, number of satellites fixedly eliminated for partial ambiguity, and number of fixed satellites.

[0015] In a preferred example, the RTK positioning result is a dual-frequency RTK positioning result, and the gross error related features include: the residual of RTK positioning, the GDOP value, the number of satellites fixedly eliminated for partial ambiguity, the number of fixed satellites, the positioning difference between the positioning result of the ionosphere-free combination and the RTK positioning result, and the residual of the ionosphere-free combination.

[0016] In a preferred example, the step of training the gross error detection model by using the classification algorithm, the gross error marking result of each RTK positioning result and the gross error related features to determine the reference bias value of the gross error detection model and the reference weight coefficient corresponding to each gross error related feature further includes:

[0017] Performing nonlinear mapping on the values ​​of the one or more gross error related features;

[0018] The gross error detection model is trained using the gross error related features after nonlinear mapping.

[0019] In a preferred example, when the gross error related feature includes a fixed number of satellites, in the step of performing nonlinear mapping on the value of the gross error related feature, the fixed number of satellites is nonlinearly mapped in the following manner:

[0020] N fix =10lg(n fix )+1

[0021] Among them, n fix is the number of fixed satellites; N fix is the mapped sample feature corresponding to the fixed number of satellites.

[0022] In a preferred example, when the gross error related feature includes the number of satellites that are fixedly excluded with partial ambiguity, in the step of performing nonlinear mapping on the value of the gross error related feature, the number of satellites that are fixedly excluded with partial ambiguity is nonlinearly mapped in the following manner:

[0023] N kick =n kick / (n kick +n fix )

[0024] Among them, n kick N is the number of satellites removed when the ambiguity of the RTK positioning algorithm is fixed; kick is the corresponding sample feature after mapping, n fix is a fixed number of satellites.

[0025] In a preferred example, when the gross error related feature includes the residual of the ionosphere-free combination or the residual of the wide lane combination, in the step of performing non-linear mapping on the value of each gross error related feature, the residual of the ionosphere-free combination or the residual of the wide lane combination is non-linearly mapped in the following manner:

[0026]

[0027] in, is the residual of the wide-lane combination or the residual of the ionosphere-free combination of the i-th fixed satellite; is the mean of the corresponding residual; Std * is the sample feature after mapping corresponding to the residual, n fix is a fixed number of satellites.

[0028] In a preferred example, when the gross error related feature includes a positioning difference between a first frequency positioning result and a second frequency positioning result, in the step of performing nonlinear mapping on the value of each gross error related feature, the positioning difference between the first frequency positioning result and the second frequency positioning result is nonlinearly mapped in the following manner:

[0029]

[0030] Among them, (x1, y1, z1) is the first frequency positioning result, and (x2, y2, z2) is the second frequency positioning result; is the corresponding sample feature after mapping.

[0031] In a preferred example, when the gross error related feature includes the positioning difference between the positioning result of the ionosphere-free combination and the RTK positioning result, in the step of performing non-linear mapping on the value of each gross error related feature, the positioning difference between the positioning result of the ionosphere-free combination and the RTK positioning result is non-linearly mapped in the following manner:

[0032]

[0033] Among them, (x iono-free ,y iono-free ,z iono-free ) is the positioning result of the ionosphere-free combination, (x, y, z) is the RTK positioning result to be tested; dP iono-free It is the sample feature after mapping of the positioning difference between the positioning result of the ionosphere-free combination and the RTK positioning result;

[0034] and / or

[0035] When the gross error related feature includes the positioning difference between the positioning result of the wide lane combination and the RTK positioning result, in the step of performing nonlinear mapping on the value of each gross error related feature, the positioning difference between the positioning result of the wide lane combination and the RTK positioning result is nonlinearly mapped in the following manner:

[0036]

[0037] Among them, (x wl ,y wl ,z wl ) is the positioning result of the wide lane combination, (x, y, z) is the RTK positioning result to be tested; dP wl It is the sample feature after mapping the positioning difference between the positioning result of the wide lane combination and the positioning result of the RTK positioning.

[0038] In a preferred example, the step of training the gross error detection model by using the classification algorithm, the gross error marking result of each RTK positioning result and the gross error related features to determine the reference bias value of the gross error detection model and the reference weight coefficient corresponding to each gross error related feature further includes:

[0039] Performing nonlinear mapping on the values ​​of one or more of the gross error related features;

[0040] The gross error detection model is trained using the gross error related features after nonlinear mapping.

[0041] In a preferred example, when the gross error related feature includes a ratio value, in the step of performing nonlinear mapping on the value of the gross error related feature, the ratio value is nonlinearly mapped in the following manner:

[0042] Ratio=10lg(ratio)

[0043] Among them, ratio is the ratio value of the ambiguity fixation of the RTK positioning algorithm; Ratio is the corresponding sample feature after mapping.

[0044] In a preferred example, when the gross error related feature includes the residual of RTK positioning, in the step of performing nonlinear mapping on the value of each gross error related feature, the residual of RTK positioning is nonlinearly mapped in the following manner:

[0045]

[0046] in, is the residual error of the RTK positioning of the i-th fixed satellite; Std is the mean of the residuals of RTK positioning; rtk The sample features after mapping corresponding to the residual of RTK positioning, n fix is a fixed number of satellites.

[0047] In a preferred example, the method further comprises: cleaning the values ​​of the gross error related features to remove abnormal data, wherein the abnormal data comprises at least one of erroneous observation data, erroneous ephemeris data, erroneous reference coordinate data and gross error feature missing data.

[0048] The present application also discloses a gross error detection system for RTK positioning results, comprising:

[0049] an acquisition unit, configured to acquire a gross error marking result of an RTK positioning result and a value of at least one gross error related feature, wherein the gross error related feature comprises one or more of a GDOP value, a number of satellites fixedly eliminated for partial ambiguity, a number of fixed satellites, a residual of an ionosphere-free combination, a residual of a wide-lane combination, a positioning difference between a first-frequency positioning result and a second-frequency positioning result, a positioning difference between a positioning result of an ionosphere-free combination and a RTK positioning result, and a positioning difference between a positioning result of a wide-lane combination and a RTK positioning result;

[0050] A training unit is used to train a gross error detection model by using a classification algorithm, a gross error marking result of each RTK positioning result and the gross error related features, so as to determine a reference bias value of the gross error detection model and a reference weight coefficient corresponding to each gross error related feature;

[0051] The gross error detection unit is used to substitute the gross error related features corresponding to the RTK positioning result to be checked into the trained gross error detection model to perform gross error detection on the RTK positioning result to be checked.

[0052] In the implementation mode of the present application, firstly, the classification method in machine learning, as well as the gross error marking result of each observation data and the large sample data of the gross error related features after nonlinear mapping, are used to train and test the gross error detection model containing the corresponding weight coefficient and bias value of each gross error related feature, so as to determine more appropriate reference weight coefficient and reference bias value of the gross error related features, so as to perform gross error detection on the RTK positioning results in a more real-time, faster and more accurate manner.

[0053] In addition, through multiple gross error related features such as the number of fixed satellites, the number of satellites fixedly eliminated with partial ambiguity, GDOP value, ratio value, residual of RTK positioning, residual of ionosphere-free combination, positioning difference between the first-frequency positioning result and the second-frequency positioning result, positioning difference between the positioning result of ionosphere-free combination and the RTK positioning result, positioning difference between the positioning result of wide-lane combination and the RTK positioning result, the gross error of the observation data is evaluated more comprehensively, so that the gross error test result obtained is more accurate.

[0054] Moreover, by first performing nonlinear mapping on the gross error related features to reduce the influence of the nonlinear characteristics between them and the gross error, and then training the above model through a sample feature matrix composed of the mapped gross error related features and the gross error marking results, the accuracy of the above model can be effectively improved, and more appropriate reference weight coefficients and reference bias values ​​of the gross error related features can be determined, which can further make the gross error detection results of the RTK positioning results more real-time, faster and more accurate.

[0055] A large number of technical features are recorded in the specification of this application, which are distributed in various technical solutions. If all possible combinations of technical features of this application (i.e., technical solutions) are to be listed, the specification will be too long. In order to avoid this problem, the various technical features disclosed in the above-mentioned invention content of this application, the various technical features disclosed in the various embodiments and examples below, and the various technical features disclosed in the accompanying drawings can be freely combined with each other to form various new technical solutions (these technical solutions are all deemed to have been recorded in this specification), unless the combination of such technical features is technically infeasible. For example, in one example, feature A+B+C is disclosed, and in another example, feature A+B+D+E is disclosed, and features C and D are equivalent technical means that play the same role. Technically, only one can be used, and it is impossible to use them at the same time. Feature E can be combined with feature C technically. Then, the solution of A+B+C+D should not be deemed to have been recorded because it is technically infeasible, and the solution of A+B+C+E should be deemed to have been recorded. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a flowchart of a method for checking gross errors of RTK positioning results according to the first embodiment of the present application;

[0057] Figure 2 is another flow chart of the gross error detection method of the RTK positioning result according to the first embodiment of the present application;

[0058] Figure 3 It is a structural schematic diagram of a gross error detection system for RTK positioning results according to the second embodiment of the present application. DETAILED DESCRIPTION

[0059] In the following description, many technical details are provided to help readers better understand the present application. However, those skilled in the art can understand that the technical solution claimed in the present application can be implemented even without these technical details and various changes and modifications based on the following embodiments.

[0060] The inventor of this application has found through research and analysis of a large amount of gross error result data that residual is only one dimension of the gross error feature, and the gross errors that can be detected only by residual account for about 20% of the total gross errors. Furthermore, the features that affect whether the positioning result is a gross error may also include one of the following or any combination thereof:

[0061] 1) Fixed number of satellites;

[0062] 2) The number of satellites eliminated by fixing some ambiguities;

[0063] 3) GDOP value;

[0064] 4) ratio value when the ambiguity is fixed;

[0065] 5) Residuals of different combinations of observations, e.g. wide lane and ionosphere-free combinations;

[0066] 6) The difference between the positioning results of different observation value combinations and the RTK positioning results;

[0067] 7) Differences in positioning results between different frequencies;

[0068] In other words, the gross error of RTK positioning is the result of the combined influence of the above characteristics. However, the relationship between the above characteristics and the gross error is a fuzzy relationship that is difficult to be accurately described by artificially designed formulas.

[0069] Therefore, the inventor of the present application proposes a method for detecting gross errors in RTK positioning based on machine learning. The classification algorithm in machine learning is used to train and learn the relationship between each feature and whether the RTK positioning result is a gross error under large sample data, so as to obtain a more accurate classification model for RTK positioning results.

[0070] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below in conjunction with the accompanying drawings.

[0071] First, the machine learning involved in the embodiments of this specification is explained.

[0072] Machine learning is an algorithm that can learn from data. It obtains models that can be applied to large-scale data sets by learning from small-scale data sets, helping us solve some problems that are difficult to solve with human-designed and deterministic programs. Machine learning algorithms can be divided into two categories: supervised learning and unsupervised learning. Supervised learning is used to process labeled data and learn the relationship between data and labels, such as classification and regression problems; unsupervised learning is used to mine the relationship between data, such as clustering and principal component analysis problems. The basic principles of machine learning are common knowledge for those skilled in the art and will not be elaborated here.

[0073] The first embodiment of the present application relates to a method for checking gross errors of RTK positioning results, the process of which is as follows: Figure 1 As shown, the method comprises the following steps:

[0074] Step 110: Obtain a gross error marking result of the RTK positioning result and a value of at least one gross error related feature, wherein the gross error related feature includes one or more of a GDOP value, a number of satellites fixedly eliminated for partial ambiguity, a fixed number of satellites, a residual of an ionosphere-free combination, a residual of a wide-lane combination, a positioning difference between a first-frequency positioning result and a second-frequency positioning result, a positioning difference between a positioning result of an ionosphere-free combination and a RTK positioning result, and a positioning difference between a positioning result of a wide-lane combination and a RTK positioning result.

[0075] Preferably, the gross error related features may also include ratio value and / or residual R of RTK positioning rtk .

[0076] Preferably, the gross error related features may include residuals of RTK positioning, GDOP values, the number of satellites fixedly eliminated for partial ambiguity, and the number of fixed satellites.

[0077] Preferably, the gross error related features may include the residual of RTK positioning, the GDOP value, the number of satellites fixedly eliminated for partial ambiguity and the number of fixed satellites, the positioning difference between the positioning result of the ionosphere-free combination and the RTK positioning result, and the residual of the ionosphere-free combination.

[0078] Furthermore, in other embodiments of the present specification, the gross error related feature may be one of the following or any combination thereof:

[0079] 1) Fixed number of satellites n fix

[0080] For example, if there are 3 fixed satellites, then the number of fixed satellites n is fix That’s 3.

[0081] 2) Partial ambiguity fixed elimination of satellite number n kick

[0082] For example, if three satellites are excluded from the partial ambiguity, then the number of satellites excluded from the partial ambiguity is n. kick That’s 3.

[0083] 3) GDOP value

[0084] GDOP, or Geometric Dilution Precision, is a parameter that characterizes the relative influence of media in different parts of space on the measurement results.

[0085] The GDOP can be calculated by ambiguity-fixing the position of the satellite.

[0086] 4) Ratio value

[0087] That is, the ratio of the suboptimal ambiguity to the optimal ambiguity during the ambiguity search. For example, if the ratio of the suboptimal ambiguity to the optimal ambiguity is 4, then ratio is 4.

[0088] 5) RTK positioning residual R rtk

[0089] RTK, or Real-time kinematic, is a carrier phase differential technology that processes the differential method of carrier phase observations of two measuring stations in real time. The carrier phase collected by the base station is sent to the user receiver to calculate the difference in coordinates.

[0090] The residual error R of RTK positioning rtk It refers to the residual error of the satellite carrier observation after the ambiguity is fixed. For example, if the residual error of the satellite carrier observation after the ambiguity is fixed is 0.03 meters, then R rtk It is 0.03 meters.

[0091] 6) Residual R of ionosphere-free combination iono-free

[0092] That is, the residual of the ionosphere-free combination of satellite carrier observations after the ambiguity is fixed. For example, the residual of the ionosphere-free combination of satellite carrier observations after the ambiguity is fixed is 0.03 meters, then Riono-free It is 0.03 meters.

[0093] 7) Residual R of wide lane combination wl

[0094] That is, the residual of the wide-lane combination of satellite carrier observations after the ambiguity is fixed. For example, the residual of the wide-lane combination of satellite carrier observations after the ambiguity is fixed is 0.03 meters, then R wl It is 0.03 meters.

[0095] 8) The positioning difference between the first frequency positioning result and the second frequency positioning result

[0096] The first frequency positioning result refers to the RTK positioning result after the first frequency ambiguity is fixed; the second frequency positioning result refers to the RTK positioning result after the second frequency ambiguity is fixed.

[0097] 9) Positioning difference dP between the positioning result of the ionosphere-free combination and the RTK positioning result iono-free , the positioning difference dP between the positioning result of the wide lane combination and the RTK positioning result wl

[0098] The positioning result of the ionosphere-free combination refers to the positioning result of the ionosphere-free combination of satellite carrier observation values ​​after the ambiguity is fixed; the positioning result of the wide-lane combination refers to the positioning result of the wide-lane combination of satellite carrier observation values ​​after the ambiguity is fixed; the RTK positioning result refers to the RTK positioning result to be verified.

[0099] The combination of the above ionosphere-free combination and wide lane combination are shown in equations (1) and (2) respectively:

[0100]

[0101] Among them, Φiono-free is the ionosphere-free combined observation value of the carrier, f1, f2 are the frequencies of the first and second frequencies respectively, λ1, λ2 are the wavelengths of the first and second frequencies respectively, φ1, φ2 are the carrier observation values ​​of the first and second frequencies respectively, among which the first frequency and the second frequency are the two frequencies of the navigation satellite signal respectively.

[0102]

[0103] Among them, Φwide is the wide-lane combined observation value of the carrier, f1, f2 are the frequencies of the first and second frequencies respectively, λ1, λ2 are the wavelengths of the first and second frequencies respectively, φ1, φ2 are the carrier observation values ​​of the first and second frequencies respectively, among which the first frequency and the second frequency are the two frequencies of the navigation satellite signal respectively.

[0104] Preferably, the positioning result of the ionosphere-free combination is the positioning result obtained after the ionosphere-free combination residual is adjusted by least squares.

[0105] Preferably, the positioning result of the wide lane combination is the positioning result obtained after the residual of the wide lane combination is adjusted by least squares.

[0106] It should be pointed out that the inventors of this application have found through long-term research and experiments that in addition to the GDOP value, ratio value, and non-combined residuals such as the residual R of RTK positioning rtk In addition to the gross error related features that affect the positioning results, the other features mentioned above will also affect the positioning results. In this case, by further considering and analyzing the newly added gross error related features, it can help to more effectively implement the gross error inspection of RTK positioning results.

[0107] Preferably, in the embodiments of the present specification, the above-mentioned gross error related features can be divided into important gross error related features and less important gross error related features.

[0108] Among them, the important gross error related features refer to the features that have a greater impact on the gross error detection performance, such as the residual R of RTK positioning rtk , the residual R of the ionosphere-free combination inno-free , GDOP value, number of satellites to be excluded for partial ambiguity kick , the positioning difference dP between the positioning result of the ionosphere-free combination and the RTK positioning result inno-free Or the positioning difference between the first frequency positioning result and the second frequency positioning result; the second important gross error related feature refers to the feature that has less impact on the gross error detection performance compared with the above important gross error related features, such as: the number of fixed satellites n fix , the positioning difference dP between the positioning result of the wide lane combination and the RTK positioning result wl , the residual R of the wide lane combination wl The optional feature is the ratio value.

[0109] Furthermore, in this step, it is necessary to obtain the gross error marking results of the RTK positioning results, that is, the gross error marking results corresponding to the observation data of each mobile station and reference station with reference coordinates, that is, the gross error marking results of the sample.

[0110] Preferably, it can be achieved by the following formula:

[0111]

[0112] Among them, OutMark is the gross error mark result; dPreal is the difference between the RTK positioning result and the reference coordinates of the mobile station; (x, y, z) is the coordinate value of the RTK positioning result; (x real ,y real ,zreal ) is the reference coordinate value of the mobile station given in advance; Th is the preset gross error threshold, which can be set or adjusted. For example, according to the centimeter-level positioning accuracy requirement, it can be set to 10cm. When dPreal is greater than 10cm, Outmark is 1, which is marked as a gross error.

[0113] Step 120: Cleaning the values ​​of the gross error related features to remove abnormal data, wherein the abnormal data includes at least one of erroneous observation data, erroneous ephemeris data, erroneous reference coordinate data and gross error feature missing data.

[0114] The advantage of doing this is that more reliable sample data can be obtained through data cleaning.

[0115] In other words, the abnormal data may include one or any combination of the following:

[0116] 1) The coordinates of the reference station or mobile station are incorrect. This abnormal data will affect the calibration of whether the observation data sample is a gross error; 2) The observation data or ephemeris data is incorrect. This abnormal data will affect the reliability of the gross error-related feature information output by the solution; 3) The gross error-related features are missing. This abnormal situation will cause the dimension of the feature vector to be inconsistent during model training.

[0117] It can be understood that this step is an optional step, and therefore, this step can be omitted in other embodiments of this specification. Further, the specific method of cleaning data is common knowledge of those skilled in the art and will not be described in detail here.

[0118] Step 130: Using the classification algorithm, the gross error marking result of each RTK positioning result and the gross error related features, the gross error detection model is trained to determine the reference bias value of the gross error detection model and the reference weight coefficient corresponding to each gross error related feature.

[0119] Among them, the reference bias value of the gross error detection model is the bias value shared by all relevant features in a gross error detection model. In other words, a gross error detection model contains a bias value and a corresponding weight coefficient for each gross error related feature.

[0120] Specifically, this step may include the following sub-steps:

[0121] Step 1302: performing nonlinear mapping on the values ​​of the one or more gross error related features;

[0122] It should be pointed out that in the embodiments of this specification, nonlinear mapping may be performed only on the value of one gross error related feature, while other values ​​are not; nonlinear mapping may be performed on the values ​​of multiple gross error related features; or nonlinear mapping may be performed on the values ​​of all gross error related features.

[0123] The advantage of doing this is that nonlinear mapping can reduce the loss of model accuracy caused by the nonlinear characteristics between gross error related features and gross errors, but nonlinear mapping is not necessary. In the specific implementation process, features can be selected for nonlinear mapping as needed.

[0124] Specifically, since RTK is a real-time algorithm, a linear classification model is used to take into account the efficiency of algorithm execution. In this step, a corresponding nonlinear mapping function is designed for each gross error-related feature to reduce the model accuracy loss caused by the nonlinear characteristics between each gross error-related feature and the gross error.

[0125] Preferably, different gross error related features can be mapped accordingly by the following mapping function formulas (3)-(7), as follows:

[0126] Fixed satellite number n fix Nonlinear mapping of:

[0127] N fix =10lg(n fix )+1 Formula (3)

[0128] n fix The actual number of fixed satellites for the RTK positioning algorithm;

[0129] N fix is the corresponding sample feature after mapping, that is, the original gross error feature is obtained after nonlinear mapping and can be used for training.

[0130] The basis for designing the above mapping function formula (3) is that according to experience, the correlation between the number of fixed satellites and the positioning gross error decreases as the number increases, and its changing trend is similar to formula (3).

[0131] The number of satellites n that are eliminated for partial ambiguity fixation kick Nonlinear mapping of:

[0132] N kick =n kick / (n kick +n fix ) Formula (4)

[0133] n kick The number of satellites that are removed when the RTK positioning algorithm has some ambiguities fixed;

[0134] N kickis the corresponding sample feature after mapping, that is, the sample feature that can be used for training is obtained after the original gross error feature is nonlinearly mapped;

[0135] n fix is a fixed number of satellites.

[0136] The basis for designing the above mapping function (4) is that according to experience, the proportion of satellites that are fixedly eliminated with partial ambiguity is approximately linearly related to the possibility of gross errors in the positioning results.

[0137] Nonlinear mapping of ratio values:

[0138] Ratio=10lg(ratio) Formula (5)

[0139] ratio is the ratio value of the ambiguity fixation of the RTK positioning algorithm;

[0140] Ratio is the corresponding sample feature after mapping, that is, the original gross error feature is obtained after nonlinear mapping and can be used for training.

[0141] The basis for designing the above mapping function formula (5) is that according to experience, the correlation between the ratio value and the positioning gross error decreases as the value increases, and its change trend is similar to formula (5).

[0142] The residual error R of RTK positioning rtk , the residual R of the ionosphere-free combination inno-free , the residual R of the wide lane combination wl Nonlinear mapping of:

[0143]

[0144] is the positioning residual of the i-th fixed satellite;

[0145] * can be replaced by RTK positioning, wide-lane combined positioning, and ionosphere-free combined positioning. In other words, the above nonlinear mapping function can be applied to the residual R of RTK positioning respectively. rtk , the residual R of the ionosphere-free combination iono-free , the residual R of the wide lane combination wl Non-linear mapping.

[0146] is the mean of the positioning residuals;

[0147] Std * is the corresponding sample feature after mapping, that is, the original gross error feature is obtained after nonlinear mapping and can be used for training.

[0148] For example, when the gross error related feature includes the residual of RTK positioning, in the step of performing nonlinear mapping on the value of each gross error related feature, the residual of RTK positioning is nonlinearly mapped in the following manner:

[0149]

[0150] in, is the residual error of the RTK positioning of the i-th fixed satellite; Std is the mean of the residuals of RTK positioning; rtk The sample features after mapping corresponding to the residual of RTK positioning, n fix is a fixed number of satellites.

[0151] The basis for designing the above mapping function (6) is that since the number of fixed satellites in different epochs is different, the number of residuals is also different. In order to make the dimension of the sample feature matrix consistent, it is necessary to convert the different numbers of residual features in different epochs into a unified feature index through standard deviation.

[0152] The positioning difference between the first frequency positioning result and the second frequency positioning result The positioning difference dP between the positioning result of the ionosphere-free combination and the RTK positioning result iono-free , the positioning difference dP between the positioning result of the wide lane combination and the RTK positioning result wl Nonlinear mapping of:

[0153]

[0154] (x * ,y * ,z * ) is the combined positioning result, where * can be replaced by wide-lane positioning and ionosphere-free combined positioning;

[0155] (x, y, z) is the RTK positioning result to be tested;

[0156] dP * is the corresponding sample feature after mapping, that is, the original gross error feature is obtained after nonlinear mapping and can be used for training.

[0157] The basis for designing the above mapping (7) is to unify the positioning errors in different directions into a characteristic index so that its linear relationship with the gross error is more significant.

[0158] More specifically, when the gross error related feature includes a positioning difference between a first frequency positioning result and a second frequency positioning result, in the step of performing nonlinear mapping on the value of each gross error related feature, the positioning difference between the first frequency positioning result and the second frequency positioning result is nonlinearly mapped in the following manner:

[0159]

[0160] Among them, (x1, y1, z1) is the first frequency positioning result, and (x2, y2, z2) is the second frequency positioning result; is the corresponding sample feature after mapping.

[0161] and / or

[0162] When the gross error related feature includes the positioning difference between the positioning result of the ionosphere-free combination and the RTK positioning result, in the step of performing non-linear mapping on the value of each gross error related feature, the positioning difference between the positioning result of the ionosphere-free combination and the RTK positioning result is non-linearly mapped in the following manner:

[0163]

[0164] Among them, (x iono-free ,y iono-free ,z iono-free ) is the positioning result of the ionosphere-free combination, (x, y, z) is the RTK positioning result to be tested; dP iono-free It is the sample feature after mapping of the positioning difference between the positioning result of the ionosphere-free combination and the RTK positioning result.

[0165] and / or

[0166] When the gross error related feature includes the positioning difference between the positioning result of the wide lane combination and the RTK positioning result, in the step of performing nonlinear mapping on the value of each gross error related feature, the positioning difference between the positioning result of the wide lane combination and the RTK positioning result is nonlinearly mapped in the following manner:

[0167]

[0168] Among them, (x wl ,y wl ,z wl ) is the positioning result of the wide lane combination, (x, y, z) is the RTK positioning result to be tested; dP wl It is the sample feature after mapping the positioning difference between the positioning result of the wide lane combination and the positioning result of the RTK positioning.

[0169] Step 1304: train a gross error detection model using the gross error related features after nonlinear mapping.

[0170] In this step, firstly, a corresponding sample feature matrix is ​​generated according to the mapped gross error related features and gross error marking results corresponding to each of the observation data.

[0171] Specifically, for each of the observation data of the mobile station and the reference station with reference coordinates, a sample feature matrix is ​​generated according to each mapped gross error related feature of the observation data and the gross error marking result of the observation data.

[0172] Preferably, the sample feature matrix can be expressed as:

[0173] [id,N fix ,N kick ,Ratio,Gdop,Std * ,dP * ,OutMark] Formula (9)

[0174] Among them, id is the number of the sample feature vector, which represents the observation data of each mobile station and reference station with reference coordinates; Gdop is the gdop value of the fixed satellite of the RTK positioning algorithm; other features have been explained above and will not be repeated here.

[0175] It should be pointed out that in the embodiments of this specification, the features and order in the sample feature matrix are not limited to this, and can be increased, decreased and adjusted as needed. For example, it can be any combination of all gross error related features mentioned in this specification, and the corresponding gross error marking results, which will not be elaborated here.

[0176] Then, the sample feature matrix and classification algorithm are used to train and test the above-mentioned gross error detection model to determine the reference weight coefficient w corresponding to each gross error related feature. i And the reference bias value bias.

[0177] In this step, the gross error detection model is trained and tested by using the sample feature matrix composed of the mapped gross error related features of each observation data and the corresponding gross error marking results, and the weight coefficients w of each gross error related feature corresponding to the optimal gross error detection model can be screened out. i and the bias value bias, which is used as the reference weight coefficient w corresponding to each gross error related feature i And the reference bias value bias.

[0178] More specifically, the sample feature matrix is ​​divided into a training set and a test set. Through a classification algorithm, the gross error test model is used to obtain the weight coefficient w corresponding to each gross error related feature using the training set. i and bias value bias, then, the test set is used to test to screen out the optimal gross error detection model, and the reference weight coefficient w of each gross error related feature is determined according to the optimal gross error detection model i And the reference bias value bias.

[0179] In other words, the weight coefficients w of each gross error related feature are obtained by training in the training set. i and bias value bias, verify the effect of the gross error detection model in the test set, and select the optimal gross error detection model and its corresponding weight coefficient w of each gross error related feature i And the bias value bias, that is, the reference weight coefficient w corresponding to each gross error related feature i And the reference bias value bias.

[0180] Preferably, in this step, classification algorithms such as logistic regression, Bayesian classification, SVM (Support Vector Machine), decision tree, K nearest neighbor, etc. can be used for classification, among which a linear SVM classification algorithm can be further selected.

[0181] It should be noted that the above “optimal” may refer to the maximum value of the screening index of the gross error detection model. For example, when the screening index of the gross error detection model is the sum of the recall rate and the specificity, that is:

[0182] max(TPR+TNR) Formula (10)

[0183] TPR is the recall rate, which represents the proportion of samples whose true values ​​are gross errors that are correctly tested by the gross error test model; TNR is the specificity, which represents the proportion of samples whose true values ​​are not gross errors that are correctly tested by the gross error test model; wherein the above true value refers to the true positioning result.

[0184] In this case, the weight coefficients w of each gross error-related feature of the gross error test model with the largest sum of recall and specificity can be i and the bias value bias as the reference weight coefficient w corresponding to each gross error related feature i And the reference bias value bias.

[0185] Step 140: Substitute the gross error related features corresponding to the RTK positioning result to be tested into the trained gross error detection model to perform a gross error detection on the RTK positioning result to be tested.

[0186] Specifically, in this step, the value of the gross error related feature corresponding to the RTK positioning result to be tested is substituted into the reference weight coefficient w corresponding to the reference bias value bias and each gross error related feature. i A gross error detection model is used to perform a gross error detection on the RTK positioning result and output a gross error detection result.

[0187] More specifically, in real-time RTK positioning, the reference bias value bias and the reference weight coefficients w corresponding to the gross error related features are iAs a constant, it is implanted into the RTK positioning algorithm. After obtaining the RTK positioning result, the gross error test of the RTK positioning result can be performed according to the following formula:

[0188]

[0189] Among them, s i is the value of the gross error related feature; N is the total number of the gross error related features; w i is the reference weight coefficient of the i-th gross error related feature; bias is the reference bias value of the gross error detection model.

[0190] It can be understood that, as described above, step 120 in this embodiment is an optional step. If step 120 is not performed in this embodiment, the overall process can be referred to in Figure 2 .

[0191] Compared with the prior art, this embodiment has the following technical advantages and effects:

[0192] Firstly, by using the classification method in machine learning, the gross error marking results of each observation data and the large sample data of the gross error related features after nonlinear mapping, a gross error detection model containing the corresponding weight coefficient and bias value of each gross error related feature is trained and tested to determine more appropriate reference weight coefficients and reference bias values ​​of the gross error related features, so as to perform gross error detection on the RTK positioning results in a more real-time, faster and more accurate manner.

[0193] And, by fixing the number of satellites n fix 、Number of satellites to be excluded for partial ambiguity fixation kick , GDOP value, ratio value, residual R of RTK positioning rtk , the residual R of the ionosphere-free combination inno-free , the positioning difference dP between the first frequency positioning result and the second frequency positioning result f1f2 , the positioning difference dP between the positioning result of the ionosphere-free combination and the RTK positioning result iono-free , the positioning difference dP between the positioning result of the wide lane combination and the RTK positioning result wl The gross error related features such as the above can be used to more comprehensively evaluate the gross error of the observed data, making the gross error test results more accurate.

[0194] Moreover, by first performing nonlinear mapping on the gross error related features to reduce the influence of the nonlinear characteristics between them and the gross error, and then training the above model through a sample feature matrix composed of the mapped gross error related features and the gross error marking results, the accuracy of the above model can be effectively improved, and more appropriate reference weight coefficients and reference bias values ​​of the gross error related features can be determined, which can further make the gross error detection results of the RTK positioning results more real-time, faster and more accurate.

[0195] The second embodiment of the present application relates to a gross error detection system for RTK positioning results, the structure of which is as follows: Figure 3 As shown, the gross error detection system of the RTK positioning result includes:

[0196] an acquisition unit, configured to acquire a gross error marking result of an RTK positioning result and a value of at least one gross error related feature, wherein the gross error related feature comprises one or more of a GDOP value, a number of satellites fixedly eliminated for partial ambiguity, a number of fixed satellites, a residual of an ionosphere-free combination, a residual of a wide-lane combination, a positioning difference between a first-frequency positioning result and a second-frequency positioning result, a positioning difference between a positioning result of an ionosphere-free combination and a RTK positioning result, and a positioning difference between a positioning result of a wide-lane combination and a RTK positioning result;

[0197] A training unit is used to train a gross error detection model by using a classification algorithm, a gross error marking result of each RTK positioning result and the gross error related features, so as to determine a reference bias value of the gross error detection model and a reference weight coefficient corresponding to each gross error related feature;

[0198] The gross error detection unit is used to substitute the gross error related features corresponding to the RTK positioning result to be checked into the trained gross error detection model to perform gross error detection on the RTK positioning result to be checked.

[0199] The first implementation manner is a method implementation manner corresponding to the present implementation manner. The technical details in the first implementation manner can be applied to the present implementation manner, and the technical details in the present implementation manner can also be applied to the first implementation manner.

[0200] It should be noted that those skilled in the art should understand that the implementation functions of each module shown in the implementation of the above-mentioned gross error inspection system for RTK positioning results can be understood by referring to the relevant description of the gross error inspection method for the above-mentioned RTK positioning results. The functions of each module shown in the implementation of the above-mentioned gross error inspection system for RTK positioning results can be implemented by a program (executable instruction) running on a processor, or by a specific logic circuit. If the gross error inspection system for RTK positioning results in the embodiment of the present application is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the embodiment of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a U disk, a mobile hard disk, a read-only memory (ROM, Read Only Memory), a disk or an optical disk. In this way, the embodiment of the present application is not limited to any specific combination of hardware and software.

[0201] Accordingly, an embodiment of the present application further provides a computer storage medium, in which computer executable instructions are stored. When the computer executable instructions are executed by a processor, the various method embodiments of the present application are implemented.

[0202] In addition, the embodiment of the present application also provides a gross error inspection device for RTK positioning results, which includes a memory for storing computer executable instructions, and a processor; the processor is used to implement the steps in the above-mentioned method implementation methods when executing the computer executable instructions in the memory. Among them, the processor can be a central processing unit (Central Processing Unit, referred to as "CPU"), or other general-purpose processors, digital signal processors (Digital Signal Processor, referred to as "DSP"), Application Specific Integrated Circuit (Application Specific Integrated Circuit, referred to as "ASIC"), etc. The aforementioned memory can be a read-only memory (read-only memory, referred to as "ROM"), a random access memory (random access memory, referred to as "RAM"), a flash memory (Flash), a hard disk or a solid-state hard disk, etc. The steps of the method disclosed in each embodiment of the present invention can be directly embodied as a hardware processor to be executed, or a combination of hardware and software modules in the processor to be executed.

[0203] It should be noted that in the application documents of this patent, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "including one" do not exclude the existence of other identical elements in the process, method, article or device including the elements. In the application documents of this patent, if it is mentioned that an action is performed according to an element, it means that the action is performed at least according to the element, which includes two situations: performing the action only according to the element, and performing the action according to the element and other elements. Expressions such as multiple, multiple, and multiple include 2, 2 times, 2 kinds, and more than 2, more than 2 times, and more than 2 kinds.

[0204] All documents mentioned in this application are considered to be included in the disclosure of this application as a whole, so that they can be used as the basis for modification when necessary. In addition, it should be understood that after reading the above disclosure of this application, those skilled in the art can make various changes or modifications to this application, and these equivalent forms also fall within the scope of protection claimed in this application.

Claims

1. A method for detecting gross errors of RTK positioning results, characterized in that: Include: Obtaining a gross error marking result of the RTK positioning result and a value of at least one gross error related feature, wherein the gross error related feature includes: one or more of a GDOP value, a number of satellites fixedly eliminated for partial ambiguity, a number of fixed satellites, a residual of an ionosphere-free combination, a residual of a wide-lane combination, a positioning difference between a first-frequency positioning result and a second-frequency positioning result, a positioning difference between a positioning result of an ionosphere-free combination and a RTK positioning result, and a positioning difference between a positioning result of a wide-lane combination and a RTK positioning result; Using a classification algorithm, as well as the gross error marking result of each RTK positioning result and the gross error related features, a gross error detection model is trained to determine a reference bias value of the gross error detection model and a reference weight coefficient corresponding to each gross error related feature; The gross error related features corresponding to the RTK positioning result to be tested are substituted into the trained gross error test model to perform a gross error test on the RTK positioning result to be tested.

2. The method according to claim 1, characterized in that The gross error related features also include ratio values ​​and / or residuals of RTK positioning.

3. The method according to claim 2, characterized in that The RTK positioning result is a single-frequency RTK positioning result, and the gross error related features include: the residual of RTK positioning, the GDOP value, the number of satellites fixedly eliminated for partial ambiguity, and the number of fixed satellites.

4. The method according to claim 2, characterized in that The RTK positioning result is a dual-frequency RTK positioning result, and the gross error related features include: the residual of RTK positioning, the GDOP value, the number of satellites fixedly eliminated for partial ambiguity, the number of fixed satellites, the positioning difference between the positioning result of the ionosphere-free combination and the RTK positioning result, and the residual of the ionosphere-free combination.

5. The method according to claim 1, characterized in that In the step of training the gross error detection model by using the classification algorithm, the gross error marking result of each RTK positioning result and the gross error related features to determine the reference bias value of the gross error detection model and the reference weight coefficient corresponding to each gross error related feature, the step further includes: Performing nonlinear mapping on the values ​​of the one or more gross error related features; The gross error detection model is trained using the gross error related features after nonlinear mapping.

6. The method according to claim 5, characterized in that When the gross error related feature includes a fixed number of satellites, in the step of performing nonlinear mapping on the value of the gross error related feature, the fixed number of satellites is nonlinearly mapped in the following manner: N fix =10lg(n fix )+1 Among them, n fix is the number of fixed satellites; N fix is the mapped sample feature corresponding to the fixed number of satellites.

7. The method according to claim 5, characterized in that When the gross error related feature includes the number of satellites that are fixedly excluded with partial ambiguity, in the step of performing nonlinear mapping on the value of the gross error related feature, the number of satellites that are fixedly excluded with partial ambiguity is nonlinearly mapped in the following manner: N kick =n kick / (n kick +n fix ) Among them, n kick N is the number of satellites removed when the ambiguity of the RTK positioning algorithm is fixed; kick is the corresponding sample feature after mapping, n fix is a fixed number of satellites.

8. The method according to claim 5, characterized in that When the gross error related feature includes the residual of the ionosphere-free combination or the residual of the wide lane combination, in the step of performing non-linear mapping on the value of each gross error related feature, the residual of the ionosphere-free combination or the residual of the wide lane combination is non-linearly mapped in the following manner: in, is the residual of the wide-lane combination or the residual of the ionosphere-free combination of the i-th fixed satellite; is the mean of the corresponding residual; Std * is the sample feature after mapping corresponding to the residual, n fix is a fixed number of satellites.

9. The method according to claim 5, characterized in that When the gross error related feature includes a positioning difference between a first frequency positioning result and a second frequency positioning result, in the step of performing nonlinear mapping on the value of each gross error related feature, the positioning difference between the first frequency positioning result and the second frequency positioning result is nonlinearly mapped in the following manner: Among them, (x1, y1, z1) is the first frequency positioning result, and (x2, y2, z2) is the second frequency positioning result; is the corresponding sample feature after mapping.

10. The method according to claim 5, characterized in that When the gross error related feature includes the positioning difference between the positioning result of the ionosphere-free combination and the RTK positioning result, in the step of performing non-linear mapping on the value of each gross error related feature, the positioning difference between the positioning result of the ionosphere-free combination and the RTK positioning result is non-linearly mapped in the following manner: Among them, (x iono-free ,y iono-free , z iono-free ) is the positioning result of the ionosphere-free combination, (x, y, z) is the RTK positioning result to be tested; dP iono-free It is the sample feature after mapping of the positioning difference between the positioning result of the ionosphere-free combination and the positioning result of the RTK positioning; and / or When the gross error related feature includes the positioning difference between the positioning result of the wide lane combination and the RTK positioning result, in the step of performing nonlinear mapping on the value of each gross error related feature, the positioning difference between the positioning result of the wide lane combination and the RTK positioning result is nonlinearly mapped in the following manner: Among them, (x wl ,y wl , z wl ) is the positioning result of the wide lane combination, (x, y, z) is the RTK positioning result to be tested; dP wl It is the sample feature after mapping the positioning difference between the positioning result of the wide lane combination and the positioning result of the RTK positioning.

11. The method according to claim 2, characterized in that In the step of training the gross error detection model by using the classification algorithm, the gross error marking result of each RTK positioning result and the gross error related features to determine the reference bias value of the gross error detection model and the reference weight coefficient corresponding to each gross error related feature, the step further includes: Performing nonlinear mapping on the values ​​of one or more of the gross error related features; The gross error detection model is trained using the gross error related features after nonlinear mapping.

12. The method according to claim 11, characterized in that When the gross error related feature includes a ratio value, in the step of performing nonlinear mapping on the value of the gross error related feature, the ratio value is nonlinearly mapped in the following manner: Ratio=10lg(ratio) Among them, ratio is the ratio value of the ambiguity fixation of the RTK positioning algorithm; Ratio is the corresponding sample feature after mapping.

13. The method according to claim 11, characterized in that When the gross error related feature includes the residual of RTK positioning, in the step of performing nonlinear mapping on the value of each gross error related feature, the residual of RTK positioning is nonlinearly mapped in the following manner: in, is the residual error of the RTK positioning of the i-th fixed satellite; The mean of the residuals of RTK positioning; Std rtk The sample features after mapping corresponding to the residual of RTK positioning, n fix is a fixed number of satellites.

14. The method according to claim 1, wherein: The method further comprises: cleaning the values ​​of the gross error related features to remove abnormal data, wherein the abnormal data comprises at least one of erroneous observation data, erroneous ephemeris data, erroneous reference coordinate data and gross error feature missing data.

15. A gross error detection system for RTK positioning results, characterized in that: Include: an acquisition unit, configured to acquire a gross error marking result of an RTK positioning result and a value of at least one gross error related feature, wherein the gross error related feature comprises one or more of a GDOP value, a number of satellites fixedly eliminated for partial ambiguity, a number of fixed satellites, a residual of an ionosphere-free combination, a residual of a wide-lane combination, a positioning difference between a first-frequency positioning result and a second-frequency positioning result, a positioning difference between a positioning result of an ionosphere-free combination and a RTK positioning result, and a positioning difference between a positioning result of a wide-lane combination and a RTK positioning result; A training unit is used to train a gross error detection model by using a classification algorithm, a gross error marking result of each RTK positioning result and the gross error related features, so as to determine a reference bias value of the gross error detection model and a reference weight coefficient corresponding to each gross error related feature; The gross error detection unit is used to substitute the gross error related features corresponding to the RTK positioning result to be checked into the trained gross error detection model to perform gross error detection on the RTK positioning result to be checked.

16. A gross error inspection device for RTK positioning results, characterized in that: include: A memory for storing computer executable instructions; as well as, A processor, configured to implement the steps in the method according to any one of claims 1 to 12 when executing the computer executable instructions.

Images