CN-115639555-B - Semi-automatic prediction method for under-ice melting and freezing distribution based on ice radar data

CN115639555BCN 115639555 BCN115639555 BCN 115639555BCN-115639555-B

Abstract

The invention provides a semi-automatic prediction method suitable for under-ice melting and freezing distribution of ice radar data, and belongs to the field of ice radar data processing. The method comprises the steps of generating a substrate reflectivity change curve, correcting a melting and freezing screening threshold value, automatically detecting a melting and freezing transition zone and predicting melting and freezing distribution under ice. Aiming at the possible error problem of echo power attenuation items, the method adopts the water body under ice in the area as a reference, corrects the detection threshold value of melting and freezing of the area, defines and extracts corresponding characteristics according to the characteristics of the melting and freezing transition area, realizes automatic detection of the melting and freezing transition area, and can obtain more accurate melting and freezing distribution in the melting and freezing transition area. And finally, more accurate melting and freezing distribution under ice can be generated, and the method is suitable for different areas and a large amount of ice radar data.

Inventors

LANG SHINAN
YANG MINGZHU
Cai Diehang
LI GUIQIANG

Assignees

北京工业大学

Dates

Publication Date: 20260512
Application Date: 20220907

Claims (7)

1. A semi-automatic prediction method for a sub-ice melt freeze distribution of ice radar data, comprising the steps of: a substrate reflectivity change curve generation stage: (1) Calculating echo power of the gas-ice interface according to the input echo amplitudes of the gas-ice interface and the ice substrate interface Echo power interfacing with ice substrates And calculating the geometric propagation loss term to the air-ice interface And a geometric propagation loss term to ice base interface Acquiring single-pass transmission loss of other known echo power attenuation item air-ice interface Reflection loss at gas-ice interface Reflection loss at interface with ice bedrock Is a theoretical value of (2); (2) According to the obtained 、、、、、 And Generating an ice absorption loss term Depth and depth of Is subjected to linear fitting by using a RANSAC method, and the slope of the fitting is the average ice absorption loss rate ; (3) According to the obtained Substituted into And Calculating the linear model of the ice base interface corresponding to all the azimuth samples And according to the change of the reflectivity of the substrate Is generated along the azimuth direction A curve; melting and freezing screening threshold correction stage: (4) Using the presence of a body of water under ice in the area to which the method is to be applied as a reference, correcting the threshold value used to screen the melting and freezing states in the base reflectivity profile And ; Automatic detection stage of melting and freezing transition zone: (5) According to Curve, find along azimuth to make each azimuth sample correspondent reflectivity produce the size as Drawing a statistical distribution map by analyzing the minimum number of azimuth samples required by the variation of the first peak value in the statistical distribution map As a feature calculation window Of (2), wherein Is the difference between the melting and freezing thresholds; (6) According to the characteristics of the thawing and freezing transition zone, five characteristics are defined to be used for representing respectively, namely 1) whether transition between thawing and freezing states occurs in a characteristic calculation window, 2) whether reflectivity in the characteristic calculation window has a significant change trend, 3) whether the terrain in the characteristic calculation window has an opposite change trend with the reflectivity, 4) whether the reflectivity in the characteristic calculation window has a large span, 5) whether the terrain trend in the characteristic calculation window changes, and finally, all the defined characteristics are calculated along the azimuth directions of all the obtained measuring line segments to complete characteristic extraction so as to obtain an original data set; (7) The method comprises the steps of adopting a few-over-sampling technology for synthesizing an original data set and an SMOTE+ENN (nearest neighbor) editing algorithm to realize equalization processing of the data set, and only carrying out standardization processing on the equalized data set to obtain an input data set of a classification model; (8) Inputting the data set into a Support Vector Machine (SVM) classification model of a Radial Basis Function (RBF) core to finish classification of a thawing frozen transition region and other regions, and realizing automatic detection of the thawing frozen transition region; Stage of prediction of melting and freezing distribution under ice: (9) The method comprises the steps of detecting melting and freezing distribution in a melting transition zone, generating the melting and freezing distribution in the melting and freezing transition zone by using a reference freezing position and a reference melting position in the melting and freezing transition zone according to the detected melting and freezing transition zone, generating the melting and freezing distribution of other zones according to corrected melting and freezing screening thresholds for the melting and freezing distribution of other zones, and combining the two results to generate the under-ice melting and freezing distribution of the zone.
2. A method of semi-automatic prediction of the ice melt freeze distribution for ice radar data as claimed in claim 1, wherein: At the end of the chain And Is obtained by linear fitting of the scatter diagram of (2) In this case, a robust sample random consistency fitting method RANSAC is adopted to effectively shield the influence of outliers and outliers, so as to obtain a more accurate linear fitting slope.
3. A method of semi-automatic prediction of the ice melt freeze distribution for ice radar data as claimed in claim 1, wherein: correcting a threshold value for detecting melting and freezing by taking a water body under ice contained in the measuring line in the area as a reference, and calculating the first Within the range of the water body under ice The average value of the curve is recorded as the melting screening threshold value of the water body under ice after correction (dB): (Equation 1) Wherein the method comprises the steps of And The range of the water body under the ice is determined; Represent the first Position of water under ice Where (a) A value; final corrected for multiple water under ice conditions in the area (DB) is expressed as: (equation 2) Wherein the method comprises the steps of Representing the number of sub-ice bodies of water present in the application area, each sub-ice body of water being given the same weight to obtain a more universal ; The freeze screening threshold is modified to Wherein Representing the known reflectivity of the frozen bedrock interface, Indicating the reflectivity of an ideal ice water interface.
4. A method of semi-automatic prediction of the ice melt freeze distribution for ice radar data as claimed in claim 1, wherein: According to the characteristics of the thawing and freezing transition zone, five characteristics are defined to realize automatic detection of the thawing and freezing transition zone in the opposite measuring line; For any azimuth sampling point : Defining features Characterizing whether a transition between the melted and frozen states occurs within the feature calculation window, Is represented as follows: (equation 3) Wherein, the And Windows respectively Inner part Maximum and minimum of (2); defining features To characterize whether the reflectivity within the feature computation window has a significant trend of change, Is represented as follows: (equation 4) Wherein, the Is fitted in a window according to least square method Is a slope of (2); defining features To characterize whether the topography within the feature calculation window has an opposite trend of change than the reflectivity, Is represented as follows: (equation 5) Wherein the method comprises the steps of Is the slope of the terrain within the window fitted according to the least squares method, The number of the distance samples corresponding to the distance of one kilometer in the radar chart; defining features To characterize whether the reflectivity within the feature computation window has a large span, Is represented as follows: (equation 6) Wherein the method comprises the steps of Representative of the sample points in azimuth A kind of electronic device The value of the sum of the values, Representative of At the position of The average value within the corresponding calculation window is calculated, , ; Defining features To characterize whether the topographical trend within the feature calculation window has changed, Is represented as follows: (equation 7) Representative of the sample points in azimuth Is provided with an elevation value of the ice base interface, Representative of the topography of the ground Direction sample points of fitting straight lines in corresponding calculation windows Elevation values of (2); the final feature vector is represented as follows: (equation 8).
5. A semi-automatic prediction method for the distribution of melting and freezing under ice of ice radar data according to claim 1, wherein the input data of a classification model for realizing the automatic detection of a melting and freezing transition area is an unbalanced data set, and the section of the melting and freezing transition area and other areas except the melting and freezing transition area are selected in a test line for carrying out subsequent feature extraction work in order to achieve the integrity of positive and negative sample proportion, data set size and inclusion.
6. The method for semi-automatically predicting the ice melting and freezing distribution of ice radar data according to claim 1, wherein the method is characterized in that for a data set of a training and testing classification model, a synthetic minority oversampling technology and an edit nearest neighbor algorithm SMOTE+ENN are used for carrying out certain equalization processing on the data, adaptation on an unbalanced data set is realized by combining punishment parameters of an SVM classification model, and only the processed data set is subjected to standardization processing so that the classification model has universality.
7. A semi-automatic prediction method for the distribution of ice melting and freezing of ice radar data according to claim 1, wherein the prediction phase of the distribution of ice melting and freezing is as follows: When the final under-ice thawing and freezing distribution of the region is generated according to the detected thawing and freezing transition region and thawing and freezing screening threshold value, firstly, for the reflectivity change curve, the thawing and freezing transition region is arranged in the window corresponding to the thawing and freezing transition region The lowest point of the curve is taken as a reference freezing position and is matched with the reference freezing position Is greater than Is determined as a reference thawing position and the reference freezing position and the reference thawing position To determine a melting-freezing distribution in a melting-freezing transition zone, above which it is considered to be likely to be melting, and below which it is considered to be likely to be freezing, and then detecting the freezing and melting positions for the other zone using a freezing threshold and a melting threshold, respectively, i.e Indicating the frozen position of the container, Indicating the melting position, thereby creating a complete distribution of the melt frozen state under ice.

Description

Semi-automatic prediction method for under-ice melting and freezing distribution based on ice radar data Technical Field The invention relates to the field of ice radar data processing, in particular to a semi-automatic prediction method suitable for under-ice melting and freezing distribution of ice radar data. Background With the development of a material balance research of the ice cover and a research of a water system under ice, the melting and freezing distribution under the ice cover can be used as information capable of describing the environment under ice to give out the positions of melting and freezing substrates in an area, and the method has very high research and application values. At present, the technology is mainly realized by two modes of thermodynamic model solving and radio echo detection (RES) data processing. The method for obtaining the temperature distribution under the ice and the melting rate distribution of the substrate by solving the thermodynamic model so as to further predict the melting and freezing distribution under the ice can quantitatively explain the melting and freezing under the ice in the current and even past. But this approach typically has regional limitations and with existing large error geothermal fluxes as boundary conditions, it can lead to large uncertainties in the predicted sub-ice melt freeze distribution. The method for predicting the melting and freezing distribution under ice by calculating the reflectivity of the substrate according to the echo power of the substrate or other layers through RES data processing can be easily popularized and applied to different radar systems and areas, the geothermal flux with larger error is not needed to be used as a boundary condition, and the steps are simple and the calculated amount is small. However, errors that may exist in the estimation of echo power decay terms will reduce the accuracy of the results, and there are problems of insufficient attention and inaccurate melt freeze estimation in areas where dry and wet conditions of significant research significance are transitional, resulting in a reduced accuracy of the melt freeze profile under ice in the predicted areas. With the development of antarctic ice detection and the continuous enhancement of the demand for high-resolution antarctic ice topography, a large amount of RES data will be generated in the future, and the prediction of ice melting and freezing distribution based on RES data processing becomes a research hotspot in view of the advantages of the RES data processing method in processing a large amount of data. In order to enable the RES data processing-based method to obtain more accurate under-ice melting and freezing distribution, on one hand, the method solves the error problem in echo power attenuation item estimation by correcting screening thresholds of under-ice melting and freezing states to a certain extent, and on the other hand, the method analyzes characteristics of a special area of a melting and freezing transition area and achieves automatic extraction of the melting and freezing transition area in the area, and solves the problems of insufficient attention and inaccurate melting and freezing estimation of the melting and freezing transition area. Thereby more accurately predicting the under-ice melting and freezing distribution of the region. Disclosure of Invention In order to solve the problem that the accuracy of a result is reduced due to the estimation error of an echo power attenuation item and insufficient attention to a thawing freezing transition region in the existing method for predicting thawing freezing under ice based on RES data processing, the invention provides a semi-automatic prediction method for thawing freezing distribution under ice based on ice radar data, which has the advantages that firstly, a thawing and freezing detection threshold value of an area is corrected by taking a water body under ice in the area as a reference, and the problem of errors in the echo power attenuation item is solved to a great extent; secondly, according to the characteristics of the melting and freezing transition zone, the characteristics are defined and extracted, and the automatic detection of the melting and freezing transition zone is realized, so that more accurate melting and freezing distribution is obtained in the melting and freezing transition zone. According to the method, the under-ice thawing and freezing distribution of the area can be obtained efficiently and accurately according to the corrected threshold value and the detected thawing and freezing transition area. The total flow chart of the method is shown in figure 1, and is mainly divided into four modules, namely a substrate reflectivity change curve generation module, a thawing freezing screening threshold correction module, a thawing freezing transition zone automatic detection module and an under-ice thawing freezing distribution prediction module.