Adaptive cross-validation radar data fusion method and system

Abstract

The present invention discloses an adaptive cross-validation radar and video data fusion method and system, which relates to the field of hydrological monitoring. The method includes: S1, collecting radar data R and video data V within a time window, respectively calculating the median absolute deviation (MAD) of the radar data R and the video data V; S2, constructing a weight calculation model for the proportion of outliers to obtain the reliability weight W of the data. outlier , used to evaluate the reliability of the data; S3, build a weight calculation model for instantaneous consistency and obtain the instantaneous consistency weight W of the data instant , used to evaluate whether the data is abnormal; S4, build the information entropy weight calculation model to obtain the data stability weight W entropy , used to evaluate the stability of data within the time window; S5, the obtained reliability weight W outlier , instantaneous consistency weight W instant , stability weight W entropy Perform comprehensive weight calculation to obtain fusion data results. Through the above method, the impact of data mutation, instability and abnormal data on data fusion results can be solved.

Description

Technical Field

[0001] The present invention relates to the technical field of hydrological monitoring, and in particular to an adaptive cross-validation radar-visual data fusion method and system. Background Art

[0002] Currently, non-contact hydrological monitoring usually uses radar or video monitoring technology. However, radar monitoring will produce measurement errors when the water surface fluctuates violently or there are complex reflectors. Video monitoring, on the other hand, will cause measurement errors due to unclear video images in complex environments such as heavy rain and fog. Therefore, it is proposed to use a hydrological monitoring device based on radar and video fusion.

[0003] Traditional weighted averaging methods use fixed weights and are unable to adapt to dynamic environmental changes (such as sudden changes in water levels). They also lack a dynamic assessment of hydrological data reliability and struggle to suppress outlier interference. Data fusion lacks a theoretical basis, and weight assignment is highly subjective, resulting in low credibility in the fusion results.

[0004] The Chinese patent (publication number: CN114911789A) "Water Level Data Fusion Method and Device Based on Radar Multi-Sensor Data" provides a data fusion method that takes into account the discreteness and correlation between various data sources and automatically detects and eliminates abnormal data. However, the method requires constructing theoretical functions, solving the Lagrange equation, and performing complex calculations of the relative entropy matrix and the correlation matrix. The calculation amount is large and difficult to process in real time. The weight distribution of this method is based on static relative entropy and correlation matrices, and it cannot cope with rapid mutations in water level and flow rate during high flood periods. The complex method steps limit its application in actual dynamic environments. Summary of the Invention

[0005] Aiming at the problem that the existing technology of radar data fusion calculation cannot cope with the rapid mutation and outlier interference of hydrological data, resulting in large errors in calculation results, the purpose of the present invention is to provide an adaptive cross-validation radar data fusion method.

[0006] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0007] S1, collect data from t0 to t n The radar data R and video data V in the time window are calculated to obtain the median absolute deviation MAD of the radar data R and video data V respectively. R and MAD V ;

[0008] S2. Build a weight calculation model based on the proportion of outliers to obtain the reliability weight W of the data outlier , which is used to evaluate the reliability of data;

[0009] S3. Build a weight calculation model based on instantaneous consistency to obtain the instantaneous consistency weight W of the data instant , which is used to evaluate whether the data has anomalies;

[0010] S4. Build a weight calculation model based on information entropy to obtain the stability weight W of the data entropy , used to evaluate the time from t0 to t n Stability of data within the time window;

[0011] S5. Obtain the reliability weight W outlier , instantaneous consistency weight W instant , stability weight W entropy Perform comprehensive weight calculation to obtain fused data results.

[0012] Preferably, step S1 further includes:

[0013] S101. Calculate the median absolute deviation (MAD) based on the radar data R and the video data V. The MAD calculation formula is as follows:

[0014] MAD R =median(|R-median(R)|), MAD V =median(|V--median(V)|);

[0015] Among them, median represents the median function.

[0016] Preferably, step S2 further includes:

[0017] S201. Use MAD R and MAD V Outlier judgment is performed on the radar data R and the video data V respectively. The judgment formula is as follows:

[0018] |R t -μ R |>k·MAD R ,|V t -μ V |>k·MAD V ;

[0019] R t is the radar data at time t, V t is the video data at time t, μ R is the median of radar monitoring data, μ V is the median of the video monitoring data, and k is the empirical coefficient.

[0020] Preferably, step S2 further includes:

[0021] S202, through MADR and MAD V Perform cross-outlier ratio detection on radar data R and video data V, and score based on the cross-outlier ratio:

[0022]

[0023] and Represent the scores of radar data and video data through cross outlier detection, Q R→R : Radar data R in its own MAD R The outlier ratio under the standard, Q V→R : Radar data R in video MAD V The outlier ratio under the standard, Q R→V :Video data V in radar MAD R The outlier ratio under the standard, Q V→V : Video data V in its own MAD V The outlier ratio under the standard, γ represents the trust coefficient;

[0024] right and Normalize and obtain their respective weights:

[0025]

[0026] is the reliability weight of radar data R, is the reliability weight of the video data V.

[0027] Preferably, step S3 further includes:

[0028] S301, radar data R and video data V are respectively from t0 to t n The degree of deviation of the median of the data within the time window is used to express its instantaneous consistency. The formula is as follows:

[0029]

[0030] Among them, r t is the radar water level corresponding to the current time t, v t is the video water level corresponding to the current time t, median(R) is the median of the radar water level sequence in the window, median(V) is the median of the video water level sequence in the window, The MAD of radar data R relative to itself R The instantaneous deviation from the standard, The video data V is relative to its own MAD V the degree of instantaneous deviation from the standard; R is the radar data relative to the video MADV The instantaneous deviation from the standard, is the video data V relative to the radar MAD R The instantaneous degree of deviation from the standard.

[0031] Preferably, step S3 further includes:

[0032] S302: Perform instantaneous consistency weight synthesis on the cross instantaneous deviation degree of the radar data R and the video data V. The synthesis formula is as follows:

[0033]

[0034] Among them, C R represents the instantaneous consistency of the radar data R at the current moment, C V represents the instantaneous consistency of the video data V at the current moment, and λ is the deviation penalty coefficient;

[0035] Normalize the above formula to obtain the instantaneous consistency weight:

[0036]

[0037] is the instantaneous consistency weight of radar data R, is the instantaneous consistency weight of the video data V.

[0038] Preferably, step S4 further includes:

[0039] S401, for t0 to t n Each historical data point in the time window is assigned an exponential decay weight w i =α T-i , where α∈(0,1) controls the attenuation rate of the influence of historical data, and then calculates the dynamic variance. The formula is as follows:

[0040]

[0041] in and are the variances of radar data R and video data V, r i and v i Represent the radar water level and video water level data at time i respectively, and are the weighted averages of radar data R and video data V respectively;

[0042] By assigning an exponential decay weight w i =α T-i , reduce the impact of recent data on the current state and effectively avoid the interference of mutation noise.

[0043] Preferably, step S4 further includes:

[0044] S402: Calculate the entropy of the radar data R and the video data V using the following formula:

[0045]

[0046] Among them H R is the entropy value of radar data R, H V is the entropy value of the video data V;

[0047] Normalize the above formula:

[0048]

[0049] is the stability weight of radar data R, is the stability weight of the video data V.

[0050] Preferably, step S5 further includes:

[0051] The obtained reliability weight W outlier , instantaneous consistency weight W instant , stability weight W entropy Perform comprehensive weight calculation, the calculation formula is as follows:

[0052]

[0053] Weight R Weight is the comprehensive weight of radar data R. V is the comprehensive weight of the video data V, and the fusion calculation is performed according to the following formula to obtain the fusion data result:

[0054] Fused t =Weight R ·R t +Weight V ·V t ;

[0055] Among them, Fused t is the fusion data result at time t.

[0056] The present invention also provides an adaptive cross-validation radar data fusion system, comprising the following modules:

[0057] Data acquisition module, used to collect data from t0 to t n Radar data R and video data V within the time window;

[0058] Calculation module, used to calculate the median absolute deviation MAD of radar data R and video data V R and MADV ;

[0059] Model building module, used to build a weight calculation model based on the proportion of outliers and obtain the reliability weight W of the data outlier , which is used to evaluate the reliability of data; construct a weight calculation model based on instantaneous consistency to obtain the instantaneous consistency weight W of the data instant , which is used to evaluate whether the data is abnormal; construct a weight calculation model based on information entropy to obtain the stability weight W of the data entropy , used to evaluate the time from t0 to t n Stability of data within the time window;

[0060] Data fusion module is used to obtain the reliability weight W outlier , instantaneous consistency weight W instant , stability weight W entropy Perform comprehensive weight calculation to obtain fused data results.

[0061] Compared with the prior art, the advantages of the present invention are:

[0062] (1) Cross-validation and bidirectional outlier detection models are used to address the problem of single sensor detection being prone to misjudgment. Outliers are identified using dual standards (self-distribution and cross-modal distribution), the cross-modal outlier ratio is calculated, and a dynamic trust coefficient is introduced. Reliability is assessed from the dual dimensions of data distribution difference and modal consistency, effectively suppressing misjudgments caused by single device errors and improving the accuracy of multi-source data consistency judgments.

[0063] (2) Design a dynamic time window and exponential decay weight model to address the response lag problem of static time windows. The window length is dynamically adjusted according to the scenario and integrity constraints are set. The real-time performance and trend analysis are balanced through semi-overlapping sliding. The historical data is given an exponential decay weight, which automatically enhances the impact of recent data, significantly improving the response speed to dynamic environments such as water level changes and shortening the system reaction time.

[0064] (3) The median absolute deviation (MAD) is combined with a multi-dimensional weighted synthesis model to address the problem of insufficient robustness of traditional standard deviation detection. MAD has a stronger anti-interference ability for non-normally distributed data. Combining the three-dimensional weights of reliability, instantaneous consistency, and stability, it comprehensively evaluates data from multiple dimensions such as long-term quality, real-time deviation, and historical fluctuation, forming a multi-level suppression of outliers and improving the detection robustness in complex noise environments.

[0065] (4) The information entropy quantification and instantaneous consistency test model are introduced to solve the problem of weight allocation relying on empirical parameters. The information entropy is used to transform data stability into a computable quantity, and an instantaneous anomaly evaluation mechanism is established based on the two-way deviation. All weights are derived based on the statistical characteristics of the data (median, MAD, dynamic variance), forming an axiomatic system that meets mathematical constraints, providing an explainable theoretical basis for weight allocation and avoiding subjective setting defects. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0067] Figure 1 This is a flow chart of the adaptive cross-validation radar data fusion method according to an embodiment of the present invention;

[0068] Figure 2 Schematic diagram of the structure of the adaptive cross-validation radar data fusion system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0069] To make the purpose, technical solutions, and advantages of the embodiments of this application more clear, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0070] The embodiments of the present invention are further described in detail below with reference to the accompanying drawings.

[0071] S1, collect data from t0 to t n The radar data R and video data V in the time window are calculated to obtain the median absolute deviation MAD of the radar data R and video data V respectively. R and MAD V ;

[0072] The time window sequence is required to be basically coherent, with no more than three consecutive interval units missing and no more than 20% of the data volume missing. The window length T needs to be dynamically adjusted according to the actual scenario (such as T = 20 minutes in flood monitoring). The purpose of setting the window length is to balance the needs of real-time and historical trend analysis and avoid single-point noise interference.

[0073] S101. Calculate the median absolute deviation (MAD) based on the radar data R and the video data V. The MAD calculation formula is as follows:

[0074] MAD R =median(|R-median(R)|), MAD V =median(|V-median(V)|)

[0075] Here, median represents the median function. Since it is calculated based on the median, it is more robust to outliers and is therefore superior to the standard deviation calculated based on the mean.

[0076] S2. Build a weight calculation model based on the proportion of outliers to obtain the reliability weight W of the data outlier , which is used to evaluate the reliability of data;

[0077] In order to reduce the misjudgment caused by the standard deviation of a single monitoring device, MAD is used to judge the short-term reliability of radar data and video data. The differences and accuracy of radar data and video data are explained through cross-validation to evaluate the reliability of the data in the short term.

[0078] S201. Use MAD R and MAD V Outlier detection is performed on radar data R and video data V respectively, and the detection conditions are as follows:

[0079] |R t -μ R |>k·MAD R ,|V t -μ V |>k·MAD V

[0080] R t is the radar data at time t, V t is the video data at time t, μ R is the median of radar monitoring data, μ V is the median of the video monitoring data, and k is the empirical coefficient. Using the 3σ principle close to the normal distribution, we set k = 5 and MAD = 0.674σ.

[0081] S202, through MAD R and MAD V Perform cross-outlier ratio detection on radar data R and video data V, and score based on the cross-outlier ratio:

[0082]

[0083] and Represent the scores of radar data and video data through cross outlier detection, Q R→R : Radar data R in its own MAD R The outlier ratio under the standard, Q V→R : Radar data R in video MAD V The outlier ratio under the standard, Q R→V :Video data V in radar MAD R The outlier ratio under the standard, Q V→V : Video data V in its own MAD V The outlier ratio under the standard, γ represents the trust coefficient.

[0084] The outlier ratio represents the percentage of outliers in the total data volume, such as Q R→R It is the number of outliers in the radar data under its own MAD standard divided by the total amount of radar data. The smaller the Q value, the fewer outliers there are, and the more reliable the recent measurement is unilaterally. This reliability is represented by 1-Q.

[0085] In addition to calculating the outlier ratio under its own MAD standard, it also calculates the outlier ratio under the other party's MAD standard, where Q R→R and Q V→R Working together on radar data, Q V→V and Q R→V For example, when the radar water level is close to the median of the video water level, if the distribution of the video data is more discrete, then Q R→R ≤Q V→V , Q V→R ≥Q R→R , Q V→R ≥Q V→V , the radar data will eventually get a greater weight. If the distribution of radar data is more discrete, then Q R→R ≥Q V→V , Q V→R ≥Q R→R , Q V→R ≥Q V→V , and eventually the video data will get a greater weight.

[0086] The trust coefficient γ is introduced when synthesizing outlier weights. The value of γ ranges from 0 to 1. When γ is greater than 0.5, it means that radar data is more trusted, and when γ is less than 0.5, it means that video data is more trusted.

[0087] right and Normalize and obtain their respective weights:

[0088]

[0089] in, is the reliability weight of radar data R, is the reliability weight of the video data V.

[0090] S3. Build a weight calculation model based on instantaneous consistency to obtain the instantaneous consistency weight W of the data instant , which is used to evaluate whether the data at time t is a significant outlier.

[0091] Combined with the numerical distribution and relative changes of the data at the current moment, cross-check whether the data at the current moment is abnormal to avoid sudden noise affecting the final fusion result.

[0092] S301, radar data R and video data V are respectively from t0 to t n The degree of deviation of the median of the data within the time window is used to express its instantaneous consistency. The formula is as follows:

[0093]

[0094] Among them, r t is the radar water level corresponding to the current time t, v t is the video water level corresponding to the current time t, median(R) is the median of the radar water level sequence in the window, median(V) is the median of the video water level sequence in the window, The MAD of radar data R relative to itself R The instantaneous deviation from the standard, The video data V is relative to its own MAD V the degree of instantaneous deviation from the standard; R is the radar data relative to the video MAD V The instantaneous deviation from the standard, is the video data V relative to the radar MAD R The instantaneous degree of deviation from the standard.

[0095] S302: Perform instantaneous consistency weight synthesis on the cross instantaneous deviation degree of the radar data R and the video data V. The synthesis formula is as follows:

[0096]

[0097] Among them, C R represents the instantaneous consistency of the radar data R at the current moment, C V represents the instantaneous consistency of the video data V at the current moment, λ is the deviation penalty coefficient, and increasing λ strengthens the suppression of anomalies;

[0098] Normalize the above formula to obtain the instantaneous consistency weight:

[0099]

[0100] in, is the instantaneous consistency weight of radar data R, is the instantaneous consistency weight of the video data V.

[0101] S4. Build a weight calculation model based on information entropy to obtain the stability weight W of the data entropy , used to evaluate the selection of t0 to t n The stability of the data within the time window.

[0102] Information entropy is used to measure the uncertainty of data probability distributions and reconcile it with instantaneous consistency. We engineered improvements to this by using a weight calculation model based on information entropy. This model quantifies data volatility through dynamic variance, maps the variance to entropy through a logarithmic transformation, and uses the inverse of entropy to assign weights. This prioritizes recently stable data during fusion.

[0103] S401. Construct a weight calculation model based on information entropy to obtain the stability weight W of the data entropy , used to evaluate the time from t0 to t n Stability of data within the time window:

[0104]

[0105] in and are the variances of radar data R and video data V, r i and v i Represent the radar water level and video water level data at time i respectively, and are the weighted averages of radar data R and video data V respectively.

[0106] By assigning an exponential decay weight w i =α T-1 , reduce the impact of recent data on the current state and effectively avoid the interference of mutation noise.

[0107] S402: Calculate the entropy of the radar data R and the video data V using the following formula:

[0108]

[0109] Among them H R is the entropy value of radar data R, H V is the entropy value of the video data V; the lower the entropy, the more stable the data;

[0110] Normalize the above formula:

[0111]

[0112] in, is the stability weight of radar data R, is the stability weight of the video data V.

[0113] S5. Obtain the reliability weight W outlier , instantaneous consistency weight W instant , stability weight W entropy Perform comprehensive weight calculation, the calculation formula is as follows:

[0114]

[0115] Among them, Weight R Weight is the comprehensive weight of radar data R. V is the comprehensive weight of the video data V, and the fusion calculation is performed according to the following formula to obtain the fusion data result:

[0116] Fused t =Weight R ·R t +Weight V ·V t

[0117] Among them, Fused t is the fusion data result at time t.

[0118] The embodiment of the present invention adopts cross-validation and two-way outlier detection mechanism, constructs a dual standard based on the median absolute deviation (MAD), evaluates reliability from the data's own distribution and cross-modal consistency, suppresses single device misjudgment, and solves the problem of inconsistency in multi-source data; designs a dynamic time window combined with an exponential decay weight, adaptively adjusts the window length according to the scenario and strengthens the influence of recent data, balances real-time response and trend analysis, and improves the ability to quickly respond to dynamic environments such as water level mutations; constructs a three-dimensional weight synthesis system that includes reliability, instantaneous consistency, and stability, replaces the traditional standard deviation with MAD to enhance the robustness of non-normal data, quantifies data stability through information entropy and introduces an instantaneous deviation penalty mechanism to form an interpretable weight distribution model, and improves the accuracy of outlier identification, environmental change response speed, and multi-source data collaboration reliability in the process of hydrological data fusion from the aspects of noise suppression, anomaly detection, and weight optimization, effectively avoiding the problem of being unable to quickly respond to hydrological data mutations based on static relative entropy and correlation matrix.

[0119] In addition, the adaptive cross-validation radar data fusion system provided by the present invention is also described. The adaptive cross-validation radar data fusion system described below and the adaptive cross-validation radar data fusion method described above can be compared with each other.

[0120] Data acquisition module, used to collect data from t0 to t n Radar data R and video data V within the time window;

[0121] Calculation module, used to calculate the median absolute deviation MAD of radar data R and video data V R and MAD V ;

[0122] Model building module, used to build a weight calculation model based on the proportion of outliers and obtain the reliability weight W of the data outlier , which is used to evaluate the reliability of data; construct a weight calculation model based on instantaneous consistency to obtain the instantaneous consistency weight W of the data instant , which is used to evaluate whether the data is abnormal; construct a weight calculation model based on information entropy to obtain the stability weight W of the data entropy , used to evaluate the time from t0 to t n Stability of data within the time window;

[0123] Data fusion module is used to obtain the reliability weight W outlier , instantaneous consistency weight W instant , stability weight W entropy Perform comprehensive weight calculation to obtain fused data results.

[0124] In the description of this application, it should be noted that the terms "upper" and "lower" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on this application. Unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or an indirect connection through an intermediate medium, or it can be internal communication between two elements. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to the specific circumstances.

[0125] It should be noted that, in this application, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any 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 a process, method, article or device comprising 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, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element.

[0126] The foregoing is merely a list of specific embodiments of the present application, intended to enable those skilled in the art to understand or implement the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the broadest scope consistent with the principles and novel features of the present application.

Claims

1. An adaptive cross-validation radar data fusion method, characterized in that: include: S1, collect data from t0 to t n The radar data R and video data V in the time window are calculated to obtain the median absolute deviation MAD of the radar data R and video data V respectively. R and MAD V ; S2. Build a weight calculation model based on the proportion of outliers to obtain the reliability weight W of the data outlier , which is used to evaluate the reliability of data; The outlier ratio includes the radar data R in its own MAD R Standard and Video MAD V The outlier ratio Q under the standard R→R and Q V→R , and the video data V are respectively in their own MAD V Standard and Radar MAD R The outlier ratio Q under the standard V→V and Q R→V ; S3. Build a weight calculation model based on instantaneous consistency to obtain the instantaneous consistency weight W of the data instant , which is used to evaluate whether the data has anomalies; S4. Build a weight calculation model based on information entropy to obtain the stability weight W of the data entropy , used to evaluate the time from t0 to t n Stability of data within the time window; S5. Obtain the reliability weight W outlier , instantaneous consistency weight W instant , stability weight W entropy Perform comprehensive weight calculation to obtain fused data results.

2. The adaptive cross-validation radar data fusion method according to claim 1, wherein: Step S1 further includes: S101. Calculate the median absolute deviation (MAD) based on the radar data R and the video data V. The MAD calculation formula is as follows: MAD R =median(|R-median(R)|),MAD V =median(|V-median(V)|); Among them, median represents the median function.

3. The adaptive cross-validation radar data fusion method according to claim 1, wherein: Step S2 further includes: S201. Use MAD R and MAD V Outlier judgment is performed on the radar data R and the video data V respectively. The judgment formula is as follows: |R t -m R |>k·MAD R ,|V t -m V |>k·MAD V ; R t is the radar data at time t, V t is the video data at time t, μ R is the median of radar monitoring data, μ V is the median of the video monitoring data, and k is the empirical coefficient.

4. The adaptive cross-validation radar data fusion method according to claim 3, wherein: Step S2 further includes: S202, through MAD R and MAD V Perform cross-outlier ratio detection on radar data R and video data V, and score based on the cross-outlier ratio: and Represent the scores of radar data and video data through cross outlier detection, Q R→R : Radar data R in its own MAD R The outlier ratio under the standard, Q V→R : Radar data R in video MAD V The outlier ratio under the standard, Q R→V :Video data V in radar MAD R The outlier ratio under the standard, Q V→V : Video data V in its own MAD V The outlier ratio under the standard, γ represents the trust coefficient; right and Normalize and obtain their respective weights: is the reliability weight of radar data R, is the reliability weight of the video data V.

5. The adaptive cross-validation radar data fusion method according to claim 1, wherein: Step S3 further includes: S301, radar data R and video data V are respectively from t0 to t n The degree of deviation of the median of the data within the time window is used to express its instantaneous consistency. The formula is as follows: Among them, r t is the radar water level corresponding to the current time t, v t is the video water level corresponding to the current time t, median(R) is the median of the radar water level sequence in the window, median(V) is the median of the video water level sequence in the window, The MAD of the radar data relative to itself R The instantaneous deviation from the standard, The MAD of the video data relative to itself V the degree of instantaneous deviation from the standard; Radar data relative to video MAD V The instantaneous deviation from the standard, is the video data relative to radar MAD R The instantaneous degree of deviation from the standard.

6. The adaptive cross-validation radar data fusion method according to claim 5, characterized in that: Step S3 further includes: S302: Perform instantaneous consistency weight synthesis on the cross instantaneous deviation degree of the radar data R and the video data V. The synthesis formula is as follows: Among them, C R represents the instantaneous consistency of the radar data R at the current moment, C V represents the instantaneous consistency of the video data V at the current moment, and λ is the deviation penalty coefficient; Normalize the above formula to obtain the instantaneous consistency weight: is the instantaneous consistency weight of radar data R, is the instantaneous consistency weight of the video data V.

7. The adaptive cross-validation radar data fusion method according to claim 1, wherein: Step S4 further includes: S401, for t0 to t n Each historical data point in the time window is assigned an exponential decay weight w i =α T-i , where α∈(0,1) controls the attenuation rate of the influence of historical data, and then calculates the dynamic variance. The formula is as follows: in and are the variances of radar data R and video data V, r i and v i Represent the radar water level and video water level data at time i respectively, and are the weighted averages of radar data R and video data V respectively, and T is the window length; By assigning an exponential decay weight w i =α T-i , reduce the impact of recent data on the current state and effectively avoid the interference of mutation noise.

8. The adaptive cross-validation radar data fusion method according to claim 7, wherein: Step S4 further includes: S402: Calculate the entropy of the radar data R and the video data V using the following formula: Among them H R is the entropy value of radar data R, H V is the entropy value of the video data V; Normalize the above formula: is the stability weight of radar data R, is the stability weight of the video data V.

9. The adaptive cross-validation radar data fusion method according to claim 1, wherein: Step S5 further includes: The obtained reliability weight W outlier , instantaneous consistency weight W instant , stability weight W entropy Perform comprehensive weight calculation, the calculation formula is as follows: Weight R Weight is the comprehensive weight of radar data R. V is the comprehensive weight of the video data V, and the fusion calculation is performed according to the following formula to obtain the fusion data result: Fused t =Weight R ·R t +Weight V ·V t ; Among them, Fused t is the fusion data result at time t.

10. Adaptive cross-validation radar data fusion system, characterized by: Includes the following modules: Data acquisition module, used to collect data from t0 to t n Radar data R and video data V within the time window; Calculation module, used to calculate the median absolute deviation MAD of radar data R and video data V R and MAD V ; Model building module, used to build a weight calculation model based on the proportion of outliers and obtain the reliability weight W of the data outlier , which is used to evaluate the reliability of the data; the outlier ratio includes the radar data R in its own MAD R Standard and Video MAD V The outlier ratio Q under the standard R→R and Q V→R , and the video data V are respectively in their own MAD V Standard and Radar MAD R The outlier ratio Q under the standard V→V and Q R→V ; Build a weight calculation model based on instantaneous consistency to obtain the instantaneous consistency weight W of the data instant , which is used to evaluate whether the data has anomalies; Construct a weight calculation model based on information entropy to obtain the stability weight W of the data entropy , used to evaluate the time from t0 to t n Stability of data within the time window; Data fusion module is used to obtain the reliability weight W outlier , instantaneous consistency weight W instant , stability weight W entropy Perform comprehensive weight calculation to obtain fused data results.

Images