A method for calculating sea ice drift velocity vector

Through polarization ratio feature image and local outlier factor algorithm, the problem of large spatial resolution and low computational efficiency in sea ice drift velocity calculation is solved, and high-resolution and high-precision sea ice drift velocity monitoring is realized, supporting polar regions Efficiency and accuracy of drift speed observation.

CN117876422BActive Publication Date: 2024-11-19NANJING UNIV OF INFORMATION SCI & TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410053835.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-01-15
Publication Date
2024-11-19
Estimated Expiration
2044-01-15

AI Technical Summary

Technical Problem

The existing sea ice drift velocity vector calculation methods have problems with large spatial resolution and low computational efficiency, and it is difficult to effectively monitor rotating and deformed sea ice.

Method used

The polarization ratio feature image is used to calculate the sea ice drift velocity vector based on feature tracking method and local outlier factor algorithm, and improve spatial resolution and calculation efficiency through SAR image preprocessing, feature point matching and local outlier factor refinement.

Benefits of technology

It realizes high-resolution and high-precision sea ice drift velocity vector monitoring, with an error ratio of less than 6%, supporting the efficiency and accuracy of polar drift velocity observation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117876422B_ABST
    Figure CN117876422B_ABST
Patent Text Reader

Abstract

The present invention discloses a method for calculating sea ice drift velocity vector, comprising the following steps: (1) acquiring raw data; (2) preprocessing and decibelizing HHHV polarized SAR images based on two-scene SAR images; (3) respectively optimizing HV polarized backscatter coefficients and optimized HH polarized backscatter coefficients corresponding to each pixel; (4) detecting the grayscale image pairs by a feature tracking method to obtain respective feature points; (5) performing a nearest neighbor distance ratio test on the feature points to filter out mismatched point pairs to correct the matching point pairs, and preliminarily screening and deleting the mismatched point pairs; (6) calculating the sea ice drift velocity vector for the retained matching point pairs, and then performing maximum threshold filtering on the sea ice drift velocity vector to obtain a preliminary sea ice drift velocity vector result; (7) refining the gridded drift velocity vector based on a local outlier factor algorithm to obtain a sea ice drift velocity vector field; the present invention can obtain a high-precision and high-resolution sea ice drift velocity vector.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of remote sensing geoscience application, and in particular to a method for calculating sea ice drift velocity vector. Background Art

[0002] According to the 2022 Arctic Climate Report, the frequency of extreme warming events in the Arctic is increasing, accompanied by the accelerated melting of the Greenland ice sheet and an overall reduction in the extent of Arctic sea ice. The sea ice drift speed is an indicator of Arctic sea ice changes and an important factor affecting the Arctic sea ice balance. In addition, for the newly opened routes in the Arctic ice area caused by global warming in recent years, it is crucial to strengthen the real-time monitoring and numerical forecasting capabilities of sea ice drift speed. Therefore, sea ice drift research plays an important role in Arctic climate and environmental research, Arctic shipping safety, and marine economic activities.

[0003] Previously, many sets of sea ice drift velocity remote sensing products were released abroad, and the data sources included satellite-borne radiometers, scatterometers, and synthetic aperture radar data, but the spatial resolution was generally low (6.25-25km). There are currently two main methods for inverting sea ice drift velocity vectors applied to remote sensing data: feature tracking and pattern matching methods. The pattern matching method is widely used in sea ice drift monitoring due to its simple algorithm principle and good sea ice motion velocity vector tracking effect; however, the algorithm has defects such as strong image dependence, inability to match rotated and deformed sea ice, and very large amount of calculation. Compared with the pattern matching method, the biggest advantage of the feature tracking algorithm is its high computational efficiency, and it can overcome the defects of rotating sea ice matching; however, its matching accuracy is low. Summary of the invention

[0004] Purpose of the invention: The purpose of the present invention is to provide a method for calculating sea ice drift velocity vector. In view of the problems commonly existing in existing sea ice drift velocity vector products, such as large spatial resolution and low computational efficiency, a method for monitoring sea ice drift velocity vector based on feature tracking method and local outlier factor algorithm using polarization ratio feature image is provided. This method can obtain richer grayscale information and more efficiently extract high-resolution and high-precision sea ice drift velocity vector.

[0005] Technical solution: A method for calculating sea ice drift velocity vector according to the present invention comprises the following steps:

[0006] (1) Obtaining raw data;

[0007] (2) The scene with the earlier observation time in the SAR image pair is taken as the first image, and the other scene is taken as the second image. The HHHV polarization SAR image is preprocessed and decibelized based on the two SAR images; the HV polarization backscatter coefficient and HH polarization backscatter coefficient corresponding to each pixel are obtained;

[0008] (3) Obtain the polarization ratio corresponding to each pixel based on the optimized HV polarization backscattering coefficient and the optimized HH polarization backscattering coefficient, and normalize the polarization ratio image into a grayscale image;

[0009] (4) Perform feature tracking on the grayscale image pairs to obtain their respective feature points;

[0010] (5) Perform a nearest neighbor distance ratio test on the feature points to filter out the wrong matching point pairs to correct the matching point pairs, and preliminarily screen and delete the wrong matching point pairs;

[0011] (6) Combined with the projected geographic coordinates, the sea ice drift velocity vector is calculated for the retained matching point pairs, and then the sea ice drift velocity vector is filtered out with the maximum threshold to obtain the preliminary sea ice drift velocity vector result;

[0012] (7) The gridded drift velocity vector is refined based on the local outlier factor algorithm to obtain the sea ice drift velocity vector field.

[0013] Furthermore, the step (1) is specifically as follows: downloading SAR satellite HHHV polarimetric data pairs with a time interval of 1 day, using two pairs of SAR HHHV data pairs of one meter C-SAR / 01 and Sentinel-1, a total of four SAR images, and each SAR data pair covers most of the same area.

[0014] Furthermore, the specific process of step (2) is as follows: for each HHHV polarized SAR image pair, radiation calibration, geometric correction, and multi-view processing are performed to obtain the HV polarized backscattering coefficient corresponding to each pixel. and HH polarization backscattering coefficient ; The spatial resolution of the image is 250m×250m.

[0015] Furthermore, the step (3) is specifically as follows: for each pixel corresponding to the HV polarization backscattering coefficient and HH polarization backscattering coefficient , the polarization ratio image is calculated:

[0016] ;

[0017] Normalize the polarization ratio image to obtain a 0-255 grayscale image:

[0018] ;

[0019] in, and are the 0.5th and 100th percentiles respectively.

[0020] Furthermore, the step (4) is specifically as follows: using the ORB algorithm to construct an image pyramid to extract feature points from each image pair, and the maximum number of retained points is set to 30,000.

[0021] Furthermore, the step (5) is specifically as follows: using a brute force matching method to match the feature points of the image pair to obtain a matching point pair, setting a nearest neighbor distance ratio test threshold of the matching point pair to 0.75, and screening the distance ratio greater than the nearest neighbor distance ratio threshold as an erroneous matching point pair.

[0022] Furthermore, the step (6) is specifically as follows: according to the positions of the retained matching point pairs and the acquisition interval time of the two images, the sea ice drift velocity vector is calculated; wherein the maximum threshold of the sea ice drift velocity is set to 50 cm / s, and the results of the sea ice drift velocity vector greater than 50 cm / s are defaulted to erroneous results and then deleted.

[0023] Furthermore, the step (7) is specifically as follows: the gridded drift velocity vector is refined based on a local outlier factor algorithm to obtain a final sea ice drift vector.

[0024] Furthermore, the local outlier factor algorithm is used to refine the gridded drift velocity vector as follows: first, the local reachable density of each data point is calculated, and then the local reachable density of the point is compared with the local reachable density of its neighboring points to obtain the local outlier factor of the point; wherein the local outlier factor represents the degree of abnormality of the point relative to its neighboring points, and the larger the value, the more abnormal it is; thereby identifying outliers in the entire data set; wherein the number of neighboring points in the local reachable density is set to 160.

[0025] A device described in the present invention includes a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, any one of the methods for calculating the sea ice drift velocity vector is implemented.

[0026] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: by using the improved ORB feature tracking algorithm to utilize the SAR polarization ratio characteristics to invert the high-resolution, high-precision and efficient sea ice drift velocity method, support is provided for polar drift velocity observations, and a high-precision, high-resolution sea ice drift velocity vector is obtained. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is a flow chart of the SAR polarization ratio characteristic sea ice drift velocity vector inversion of the present invention;

[0028] Figure 2 This is a case study of the inversion of sea ice drift velocity vector from a one-meter C-SAR / 01 SAR image of the present invention;

[0029] Figure 3 This is a case study of the inversion of sea ice drift velocity vector from Sentinel-1 SAR images of the present invention;

[0030] Figure 4 This is a comparative analysis diagram of the sea ice drift velocity vector extracted from the Sentinel-1 SAR image based on the method of the present invention and the sea ice drift velocity vector calculated by the IABP buoy.

[0031] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

[0032] like Figure 1 As shown, an embodiment of the present invention provides a method for calculating sea ice drift velocity vector, including downloading SAR satellite HHHV polarization data pairs with a time interval of 1 day, and performing preprocessing operations such as radiation calibration, Lee filtering and multi-viewing to obtain HH polarization and HV polarization backscattering coefficient image pairs respectively; the HH and HV polarization backscattering coefficients are calculated to obtain polarization ratio images, and normalized and calculated as grayscale images. The grayscale image pairs are used to perform feature tracking algorithms to extract feature points and match them, and the matching point pairs are filtered out based on the nearest neighbor distance ratio test method to filter out false matches, and the preliminary sea ice drift velocity vector field is calculated for the filtered matching point pairs; finally, the sea ice drift velocity vector is refined using the local outlier factor method to obtain the final sea ice drift velocity vector result, which specifically includes the following steps:

[0033] S1, the scene with the earlier observation time in the SAR image pair is taken as image 1, and the other scene is taken as image 2. Based on the preprocessed and decibelized HHHV polarized SAR images of the two SAR images, the HV polarized backscatter coefficients corresponding to each pixel are obtained. and HH polarization backscattering coefficient ; The spatial resolution of the image is 250m×250m;

[0034] S2, respectively obtain the polarization ratio corresponding to each pixel for the optimized HV polarization backscattering coefficient and the optimized HH polarization backscattering coefficient corresponding to each pixel, and normalize the polarization ratio image into a grayscale image.

[0035] Polarization ratio calculation formula:

[0036] ;

[0037] Normalize the polarization ratio image to obtain a 0-255 grayscale image:

[0038] ;

[0039] in, and are the 0.5th and 100th percentiles respectively.

[0040] S3, using the ORB algorithm to build an image pyramid to extract feature points from image pairs, and the maximum number of retained points is set to 30,000.

[0041] S4, using the brute force matching method to match the feature points of the image pair to obtain the matching point pair, by setting the nearest neighbor distance ratio test threshold of the matching point pair to 0.75, screening the distance ratio greater than the nearest neighbor distance ratio threshold as the wrong matching point pair.

[0042] S5, calculate the sea ice drift velocity vector for the retained matching point pairs in combination with their geographic coordinates, set the maximum threshold of the sea ice drift velocity to 50 cm / s, and default all results of the sea ice drift velocity vector greater than 50 cm / s as erroneous results and delete them, to obtain preliminary sea ice drift velocity vector results.

[0043] S6, the gridded drift velocity vector is further refined based on the local outlier factor algorithm. First, a local outlier factor model is initialized, and the local outlier factor model is trained with the x vector of the sea ice drift velocity vector, and the outlier value of the x vector is predicted, and the data classified as normal values ​​in the prediction result is retained; the same operation is performed on the y vector to obtain the final sea ice drift velocity vector result. The local outlier factor algorithm first calculates the local accessible density of each data point, and then compares the local accessible density of the point with the local accessible density of its neighboring points to obtain the local anomaly factor of the point. This local anomaly factor represents the degree of abnormality of the point relative to its neighboring points, and the larger the value, the more abnormal it is. By calculating the local anomaly factor of each data point, the local outlier factor algorithm can find data points with abnormal density relative to the surrounding data points, thereby identifying outliers in the entire data set. Among them, the number of neighboring points to be considered for the local accessible density is set to 160, which is more accurate for density estimation.

[0044] The calculation results of the case of the present invention are as follows Figure 2 , Figure 3 and Figure 4 shown.

[0045] in, Figure 2 This is an example of sea ice drift velocity vector inversion from C-SAR / 01 SAR image: (a) is the result of confidence interval threshold filtering; (b) is the result of local outlier factor algorithm. Figure 2 (a) and Figure 2 From the comparison of (b), it can be seen that when the confidence interval threshold method is used, an obvious gap appears in the lower right corner of the image, while the local interest factor algorithm retains this part of the result.

[0046] in, Figure 3 This is an example of inversion of sea ice drift velocity vector from Sentinel-1 image: (a) is the result of filtering using the confidence interval threshold method; (b) is the result of the local outlier factor algorithm. Figure 3 Results and Figure 2 Similarly, the local benefit factor algorithm results ( Figure 3 (b) can retain the correct sea ice drift velocity vector in the lower right corner of the image while filtering out other erroneous results.

[0047] in, Figure 4 This is a comparative analysis case of the sea ice drift velocity vector extracted from Sentinel-1 SAR images based on the method of the present invention and the sea ice drift velocity vector calculated by the IABP buoy. The results show that the sea ice drift velocity vector result obtained by this method can achieve high accuracy, with the RMSE of size and direction being 0.36cm / s and 1.71° respectively, and the error ratio is less than 6%.

[0048] The present invention designs a method for calculating sea ice drift velocity vector. To improve ice-water contrast, the HHHV polarization SAR backscatter coefficient is first logarithmized for decibelization, and the decibelized polarization ratio is obtained on this basis, and the next step of calculation is performed based on the decibelized polarization ratio. Richer grayscale information is obtained based on the polarization ratio feature image pair, and the matching point pair is quickly obtained using the feature tracking method. The preliminary sea ice drift velocity vector is calculated using a specific nearest neighbor distance ratio test and a maximum retention number threshold combined with the geographic location. Finally, the local outlier factor algorithm is used to retain the correct sea ice drift vector to obtain the final sea ice drift velocity vector field. The calculation method of the present invention provides a physical identification method for sea ice drift velocity monitoring technology. This method is based on the difference in dielectric constants between sea ice and seawater. In order to address the common problems of large spatial resolution and low computational efficiency in existing sea ice drift velocity vector products, a method for monitoring sea ice drift velocity vector is used based on feature tracking method and local outlier factor algorithm using polarization ratio feature image. This method can obtain richer grayscale information and more efficiently extract high-resolution and high-precision sea ice drift velocity vectors. The sea ice drift velocity vector results of Sentinel-1 image pairs are analyzed for accuracy with the sea ice drift velocity vector results extracted based on the International Arctic Buoy Program (IABP) as a reference. The results show that the sea ice drift velocity vector results obtained by this method can achieve the highest accuracy, with RMSE of size and direction of 0.36cm / s and 1.71° respectively, and the error ratio is less than 6%.

[0049] An embodiment of the present invention also provides a device, including a memory, a processor, and a program stored in the memory and executable on the processor, wherein when the processor executes the program, any one of the methods for calculating the sea ice drift velocity vector is implemented.

Claims

1. A method for calculating sea ice drift velocity vector, It is characterized in that The following steps are involved: (1) Obtaining raw data; (2) The scene with the earlier observation time in the SAR image pair is taken as the first image, and the other scene is taken as the second image. The HHHV polarized SAR image is preprocessed and decibelized based on the two SAR images; the HV polarized backscatter coefficient and HH polarized backscatter coefficient corresponding to each pixel are obtained; (3) Obtain the polarization ratio corresponding to each pixel based on the optimized HV polarization backscattering coefficient and the optimized HH polarization backscattering coefficient, and normalize the polarization ratio image into a grayscale image; (4) Perform feature tracking on the grayscale image pairs to obtain their respective feature points; (5) Perform a nearest neighbor distance ratio test on the feature points to filter out the wrong matching point pairs to correct the matching point pairs, and preliminarily screen and delete the wrong matching point pairs; (6) Combined with the projected geographic coordinates, the sea ice drift velocity vector is calculated for the retained matching point pairs, and then the sea ice drift velocity vector is filtered out with the maximum threshold to obtain the preliminary sea ice drift velocity vector result; (7) The gridded drift velocity vector is refined based on the local outlier factor algorithm to obtain the sea ice drift velocity vector field. The implementation process of refining the gridded drift velocity vector based on the local outlier factor algorithm is as follows: First, the local reachable density of each data point is calculated. Then, the local reachable density of the point is compared with the local reachable density of its neighboring points to obtain the local anomaly factor of the point. The local anomaly factor indicates the degree of abnormality of the point relative to its neighboring points. The larger the value, the more abnormal it is. In this way, outliers in the entire data set are identified. The number of neighboring points in the local reachable density is set to 160.

2. A method for calculating sea ice drift velocity vector according to claim 1, It is characterized in that The specific steps of step (1) are as follows: downloading SAR satellite HHHV polarization data pairs with a time interval of 1 day, using two pairs of SAR HHHV data pairs of one-meter C-SAR / 01 and Sentinel-1, and a total of four SAR images.

3. A method for calculating sea ice drift velocity vector according to claim 1, It is characterized in that The specific process of step (2) is as follows: for each HHHV polarimetric SAR image pair, radiation calibration, geometric correction, and multi-view processing are performed to obtain the HV polarimetric backscattering coefficient corresponding to each pixel. and HH polarization backscattering coefficient ; The spatial resolution of the image is 250m×250m.

4. A method for calculating sea ice drift velocity vector according to claim 1, It is characterized in that The step (3) is specifically as follows: for each pixel corresponding to the HV polarization backscattering coefficient and HH polarization backscattering coefficient , the polarization ratio image is calculated: ; Normalize the polarization ratio image to obtain a 0-255 grayscale image: ; in, and are the 0.5th and 100th percentiles respectively.

5. A method for calculating sea ice drift velocity vector according to claim 1, It is characterized in that The step (4) specifically includes: using the ORB algorithm to construct an image pyramid to extract feature points from each image pair, and the maximum number of retained points is set to 30,000.

6. A method for calculating sea ice drift velocity vector according to claim 1, It is characterized in that The step (5) specifically includes: using a brute force matching method to match the feature points of the image pair to obtain a matching point pair, setting a nearest neighbor distance ratio test threshold of the matching point pair to 0.75, and screening the distance ratio greater than the nearest neighbor distance ratio threshold as an erroneous matching point pair.

7. A method for calculating sea ice drift velocity vector according to claim 1, It is characterized in that The step (6) is specifically as follows: according to the positions of the retained matching point pairs and the acquisition interval time of the two images, the sea ice drift velocity vector is calculated; wherein the maximum threshold of the sea ice drift velocity is set to 50 cm / s, and the results of the sea ice drift velocity vector greater than 50 cm / s are defaulted to erroneous results and then deleted.

8. A sea ice drift velocity vector calculation device, comprising a memory, a processor, and a program stored in the memory and executable on the processor, It is characterized in that When the processor executes the program, a method for calculating a sea ice drift velocity vector is implemented as described in any one of claims 1 to 7.