Large-scale ocean front identification method

Abstract

The invention discloses a large-scale ocean front identification method, and relates to the field of Bayesian statistics and ocean mesoscale phenomenon intelligent identification, and the method comprises the steps: carrying out the blocking processing of a large-scale sea area according to ocean knowledge, and determining a plurality of blocking regions; based on a Bayesian algorithm, extracting a sharp region in each block region; the sharp area is an area containing a sharp surface; extracting a skeleton of the sharp area from the sharp area by using a mathematical morphology algorithm; trimming the skeleton by using a discrete skeleton evolution algorithm, and determining a trimmed skeleton; on the basis of the trimmed skeleton, connecting adjacent frontal surfaces in each block area by using a sharp weaving ring elimination algorithm, removing a ring structure connected with the rear frontal surfaces, and extracting the frontal surface from which the ring structure is removed in each block area; and combining the frontal surfaces after the ring structures are removed in all the block regions, and determining the global frontal surface of the large-scale sea area.

Description

Technical Field

[0001] The present application relates to the field of Bayesian statistics and intelligent identification of mesoscale phenomena in the ocean, and in particular to a method for identifying large-scale ocean fronts. Background Art

[0002] In recent years, the huge energy carried by mesoscale phenomena in the ocean has been confirmed as the main form of energy in the ocean, and there has been a huge demand for the identification of one of its main manifestations, the ocean front. Previous ocean front intelligent identification algorithms, such as variance analysis, edge detection, Laplace index and machine learning, have many problems. Among them, the truncation nature of variance analysis makes it difficult to handle multi-peak phenomena; edge detection only uses gradient (first-order derivative) information, resulting in a small amount of information used and the need to manually set threshold parameters. Manually setting threshold parameters will also lead to poor continuity and cumbersome elimination and merging operations, and uneven gradient distribution limits the research scale; Laplace index basically identifies the front next to the vortex, which cannot reflect large-scale ocean peaks; supervised learning is usually considered to be only an approximation of the solution of traditional algorithms; unsupervised learning requires setting the number of cluster centers, and manually setting the number of cluster centers will affect the front identification results. In summary, the existing ocean peak intelligent identification algorithms cannot accurately identify large-scale ocean peaks. Summary of the invention

[0003] The purpose of this application is to provide a large-scale ocean front identification method to solve the problem of being unable to accurately identify large-scale ocean peaks.

[0004] To achieve the above objectives, this application provides the following solutions.

[0005] In a first aspect, the present application provides a large-scale ocean front identification method, comprising the following steps.

[0006] Based on oceanographic knowledge, large-scale sea areas are divided into blocks and multiple block areas are determined.

[0007] Based on the Bayesian algorithm, the frontal zone in each block area is extracted; the frontal zone is a region containing a front.

[0008] The skeleton of the frontal zone is extracted from the frontal zone using a mathematical morphology algorithm.

[0009] The skeleton is pruned using a discrete skeleton evolution algorithm to determine a pruned skeleton.

[0010] Based on the pruned skeleton, the adjacent fronts in each block area are connected using the front weaving ring elimination algorithm, and the ring structure of the connected fronts is removed, so as to extract the front in each block area after removing the ring structure.

[0011] The fronts after removing the ring structures in all block areas are merged to determine the global front of the large-scale sea area.

[0012] According to the specific embodiments provided by the present application, the present application discloses the following technical effects: based on oceanographic knowledge, Bayesian algorithm, mathematical morphology algorithm, discrete skeleton evolution algorithm, and front weaving ring elimination algorithm, the present application performs front zone extraction, skeleton extraction, skeleton pruning, front connection, and ring structure removal on the divided sea area, and finally obtains the global front of the large-scale sea area. Different from the existing ocean peak intelligent identification algorithm, the present application does not involve variance analysis, artificially set threshold parameters, Laplace indicators, and machine learning, and therefore does not have many problems existing in these methods. The present application adopts a new large-scale ocean front identification method, which can accurately identify large-scale ocean peaks. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. 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 paying creative work.

[0014] Figure 1 Flowchart of the large-scale ocean front identification method provided in this application.

[0015] Figure 2 This is a schematic diagram of neighborhood points provided by this application.

[0016] Figure 3 This is a schematic diagram of the front connection provided for this application.

[0017] Figure 4 This is a schematic diagram of the ring structure removal provided in this application.

[0018] Figure 5 This is a schematic diagram of the dual threshold provided in this application.

[0019] Figure 6 This is a rendering of the frontal zone identified by the present invention provided in this application.

[0020] Figure 7 Schematic diagram of extracting the skeleton of a triangle region using the maximum disk method provided in this application.

[0021] Figure 8 Schematic diagram of extracting the skeleton of a square region using the maximum disk method provided in this application.

[0022] Fig. 9 This is a schematic diagram of the DSE algorithm provided in this application; wherein, Fig. 9 (a) is the original skeleton image; Fig. 9 (b) is the pruned skeleton image with the weight threshold set to 0.0001; Fig. 9 (c) is the pruned skeleton image with the weight threshold set to 0.0002; Fig. 9 (d) is the pruned skeleton image with the weight threshold set to 0.0005; Fig. 9 (e) is the pruned skeleton image with the weight threshold set to 0.001; Fig. 9 Middle (f) is the pruned skeleton image with the weight threshold set to 0.005.

[0023] Fig.10 This is the front effect diagram identified by this application.

[0024] Fig.11 A schematic flow chart of another large-scale ocean front identification method provided in this application. DETAILED DESCRIPTION

[0025] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0026] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0027] The embodiment of the present application provides a large-scale ocean front identification method, which is executed by a computer device, and can be specifically executed by a computer device such as a terminal or a server alone, or can be executed by a terminal and a server together. In the embodiment of the present application, Figure 1 As shown, the method includes the following steps.

[0028] S1: Divide the large-scale sea area into blocks based on oceanographic knowledge and determine multiple block areas.

[0029] S2: Based on the Bayesian algorithm, the frontal zone in each block area is extracted; the frontal zone is the area containing the front.

[0030] S3: Using a mathematical morphology algorithm, extract the skeleton of the frontal zone from the frontal zone.

[0031] S4: Pruning the skeleton using a discrete skeleton evolution algorithm to determine a pruned skeleton.

[0032] S5: Based on the pruned skeleton, the adjacent fronts in each block area are connected using the front weaving ring elimination algorithm, and the ring structure of the connected fronts is removed, and the front after the ring structure is removed is extracted in each block area.

[0033] S6: Merge the fronts in all the block areas after removing the ring structures to determine the global front of the large-scale sea area.

[0034] In an exemplary embodiment, S1 needs to divide the sea area into different sea areas, namely, block areas, according to basic oceanographic knowledge.

[0035] Taking the Western Pacific as an example, it is divided into four sea areas: the South China Sea, the Bohai Sea, the East China Sea, the Kuroshio Current and the tropical Pacific according to its offshore distance, latitude, river estuaries and tidal characteristics.

[0036] In an exemplary embodiment, S2 may be replaced by the following steps.

[0037] S21: For each block area, the Sobel operator is used to calculate the gradient and gradient direction of the ocean elements; the ocean elements include sea surface temperature, salinity and density.

[0038] Furthermore, the Sobel operator is as follows.

[0039] ; ; Among them, Fx and Fy are two convolution kernels. For each point, weighted operations are performed on the 9 points around it and its own points. The gradients in the meridian direction can be calculated separately. T y With the latitudinal gradient T x .

[0040] : ;in, is the ocean element matrix, which is convolved with the above two operators to obtain the meridional gradient of each point T y and the latitudinal gradient T x .

[0041] ; ;in, is the gradient direction; is the gradient size.

[0042] The above two gradients can be used to obtain the gradient size and direction of each point. The gradient direction is determined by the arc tangent function of the longitude and latitude directions, and the range is .

[0043] S22: determining a quantile according to the gradient, and determining a frontal zone type according to the quantile; the frontal zone type includes a frontal zone, a non-frontal zone, and a to-be-determined frontal zone.

[0044] S23: Determine the prior probability of each pixel point in the undetermined frontal zone according to the gradient.

[0045] S24: Calculate the local degree of edge (Local Degree of Edge, LDE) and block dispersion (Block Deviation, BD) of each pixel point in the undetermined frontal zone using the ocean elements.

[0046] S25: Calculate the likelihood function of each pixel point in the to-be-determined front area according to the local edge degree and the block discreteness.

[0047] S26: Perform Bayesian decision making according to the likelihood function and the prior probability, determine the frontal zone and the non-frontal zone in the pending frontal zone, and extract all the frontal zones in each block area.

[0048] In an exemplary embodiment, S22 may be replaced by the following steps.

[0049] The gradient sequence is generated by sorting the gradient values ​​from large to small.

[0050] The first 10% of the gradient positions in the gradient sequence are taken as the first quantile, and the first 10% of the gradient positions are taken as the upper threshold.

[0051] The first 20% of the gradient positions in the gradient sequence are taken as the second quantile, and the first 20% of the gradient positions are taken as the lower threshold.

[0052] The gradient corresponding to a value greater than the upper threshold is regarded as the frontal area.

[0053] The gradient corresponding to a value less than the lower threshold is regarded as a non-frontal area.

[0054] The gradient corresponding to the upper threshold and the lower threshold is used as the to-be-determined frontal zone.

[0055] In an exemplary embodiment, S23 may be replaced by the following steps.

[0056] use and Determine the prior probability of each pixel point in the undetermined frontal zone; wherein, is the prior probability of each pixel point when the undetermined front point E is the front point, i It is the frontal zone; is the prior probability of each pixel point when the pending frontal point E is a non-frontal point, j It is a non-frontal area; is the upper threshold; is the lower threshold; is the gradient of the undetermined frontal point E.

[0057] In an exemplary embodiment, S24 may be replaced by the following steps.

[0058] Let any undetermined front point in the undetermined front area be located at the center of the 3×3 area, and a set of vectors consisting of the undetermined front point E and the neighboring points of the undetermined front point E .

[0059] use Calculate the local edge degree between two neighboring points in the pending frontal zone; wherein, is the local edge degree; A and I are any two opposite points in the 3×3 area; For vector The maximum value of an element, is a vector The minimum value of an element, is a vector Mean value of elements; vector It is a set of vectors consisting of the undetermined frontal point E and its neighboring points.

[0060] The local edge degree of the undetermined frontal point E is calculated based on the local edge degree between the four pairs of neighboring points. .

[0061] use Calculating the block discreteness between two neighboring points in the pending frontal zone; is the block discreteness.

[0062] The block discreteness of the undetermined frontal point E is calculated based on the block discreteness between the four pairs of neighboring points. .

[0063] Furthermore, the ocean elements are used to calculate the LDE and BD values ​​of each point. , which is located in the center The 8 neighborhood points in the area form a vector .

[0064] These 8 points are about Can form 4 pairs of symmetrical points, For example, Figure 2 As shown, the calculation formulas for LDE and BD are as follows.

[0065] ; ;in, is a vector The maximum value of an element, is a vector The minimum value of an element, is a vector The average value of the elements.

[0066] Average the 4 pairs of LDE and BD values ​​to get the point LDE and BD values: ; .

[0067] In an exemplary embodiment, S25 may be replaced by the following steps.

[0068] use Calculating the likelihood function that the undetermined frontal point E belongs to the frontal zone i; is the likelihood function that the undetermined frontal point E belongs to the frontal zone i; The gradient is greater than The number of points; For The number of differences less than 0.1; For The number of differences less than 0.1.

[0069] use Calculating the likelihood function that the undetermined frontal point E belongs to the non-frontal zone j; is the likelihood function that the undetermined frontal point E belongs to the non-frontal zone j; The gradient is less than The number of points; For The number of differences less than 0.1; For The number of differences less than 0.1.

[0070] In an exemplary embodiment, S26 may be replaced by the following steps.

[0071] when , the pending front zone point E is determined to be the front zone.

[0072] when When the frontal zone point to be determined is Determined to be a non-frontal area.

[0073] Furthermore, ; ;in, is the posterior probability that the point is a frontal point, The fact that there is also a relationship between the two operators of the front and non-front, is the posterior probability that the point is a non-frontal point.

[0074] In an exemplary embodiment, S3 may be replaced by the following steps.

[0075] The frontal zone is extracted using the maximum disk method in the mathematical morphology algorithm, which is a combination of erosion and opening operations. Skeleton ;in, , ; For the identified frontal zone Continuous Second corrosion, is the convolution kernel, For the opening operation, For corrosion operations; for is eroded to an empty set The last iteration step before is: .

[0076] In an exemplary embodiment, S4 may be replaced by the following steps.

[0077] In practical applications, in order to simplify the formula, Can S Indicated by.

[0078] The skeleton All vertices on are divided into endpoints and intersections, and As All endpoints on As distance The nearest intersection will As a line segment between two points.

[0079] according to Determine the frontal zone after skeleton reconstruction; among them, The frontal zone after skeleton reconstruction; For The largest disk centered in the frontal zone The radius of the skeleton The point on.

[0080] use For each endpoint Assign weights; among them, is the weight; Based on the skeleton Reconstructed frontal zone; is the area function.

[0081] All endpoint weights are initialized, and the endpoint weights are iterated, the endpoint weights of the iteration are determined, and the minimum endpoint weight of the iteration during the iteration is determined.

[0082] A pruned skeleton is determined according to the minimum endpoint weights.

[0083] Furthermore, The closer it is to 0, the more similar the frontal zone reconstructed after removing the branch it represents is to the original one, so it can be removed, and a greedy algorithm can be obtained: Step 1. Initialize the initial skeleton All endpoint weights on : .

[0084] Step 2. In the iteration step: .

[0085] Step 3. Choose the smallest weight .if If it is less than the given threshold, then jump to step 4; otherwise, stop the iteration and output As the final result.

[0086] Step 4. Remove the branch , get the new skeleton, that is, the pruned skeleton: ;in, is the endpoint corresponding to the minimum weight branch; is the intersection point corresponding to the minimum weight branch.

[0087] In an exemplary embodiment, S5 may be replaced by the following steps.

[0088] Based on the pruned skeleton, a search radius is set.

[0089] For any front, determine whether the head and tail of an adjacent front fall within the search radius of the head and tail of the front.

[0090] If there are multiple adjacent fronts, they are connected according to the priority of the distance between the head and the tail and the absolute value of the difference in the gradient direction of the head and the tail to determine the connected rear front.

[0091] If the tail of any front in each connected rear front is within the 3×3 area of ​​the head, the front is judged as a ring structure and is removed.

[0092] If the tail of each connected front is not within the 3×3 area of ​​the head, determine whether each point in the front is within the 3×3 area of ​​the tail starting from the head. If so, the front is judged as a ring structure and removed.

[0093] Until the data of all fronts are no longer updated, extract the fronts in each block area after removing the ring structure.

[0094] Furthermore, we first set a search radius, set the search radius to 3, and examine a certain front. If the head / tail of another front falls within the search radius of the head / tail of this front, then they need to be connected.

[0095] If there are multiple front heads / tails that fall within the search radius, they are connected based on 1) the distance between the heads / tails; 2) the absolute value of the difference in the gradient direction between the heads / tails. The connection process is as follows: Figure 3 As shown in the figure, the pixel points filled in the diagonal direction are used as new front points, and the two fronts are merged to update the front data information. In order to prevent the connection after the update, it is necessary to iterate in this way until the data of all fronts no longer changes.

[0096] Furthermore, the head and tail of each front are first determined. If the tail is near the head (in the nine-square grid), the entire front is determined to be a ring structure and removed. If not, each point is determined from the beginning to determine whether it is near the tail, and this section of the ring structure is removed. Then the same process is repeated from the end. Finally, the front data information is updated. In order to prevent the generation of new ring structures after the update, it is necessary to iterate in this way until the data of all fronts no longer changes. Figure 4 As shown, the filled part is the determined ring structure.

[0097] In an exemplary embodiment, the fronts identified in each block are merged according to the latitude and longitude information. If there is an overlapping area when the blocks are divided, the union of the fronts identified in the overlapping parts is taken as the final output result.

[0098] Based on the large-scale ocean front identification method shown above in this application, this application is further explained with actual data below.

[0099] like Figure 5 As shown in the figure, the cumulative function of the sea surface temperature gradient in the Kuroshio Sea area on January 1, 2022 is shown. The first 10% quantile is 0.27, and the first 20% quantile is 0.18. Therefore, points with a gradient greater than 0.27 are directly determined as frontal areas, points with a gradient less than 0.18 are determined as non-frontal areas, and points between 0.18-0.27 are pending frontal areas, and their prior probabilities are calculated. Figure 6 As shown in Figure 1, the frontal zone identified by the sea surface temperature in the Kuroshio Sea area on January 1, 2022; Figure 7-Figure 8 As shown in , the skeleton of a simple shape is extracted using the maximum disk method; Fig. 9 As shown, the process of pruning the skeleton using the DSE algorithm; Fig.10 Shown is the front identified by the sea surface temperature in the Kuroshio area on January 1, 2022.

[0100] like Fig.11As shown, the main process of this application includes: (1) dividing the large-scale sea area into blocks; (2) using the Bayesian algorithm to make Bayesian decisions for each sea area and identify the frontal zone; (3) using the mathematical morphology algorithm to extract the frontal zone as the front; (4) using the front weaving ring elimination algorithm to achieve the connection and de-ringing of the front; (5) merging the front results.

[0101] Therefore, the present application has the following beneficial effects: (1) the analysis scale is large, combined with oceanographic knowledge, and has a wider range of applications; (2) more information is integrated through the likelihood function, and the positioning is more accurate; (3) the research object is changed to the frontal zone, which enhances the continuity of frontal identification; (4) there is no need to set a threshold, and the robustness is strong; (5) the use of mathematical morphology algorithms removes unimportant branches, and the front has better smoothness; (6) the connection and de-ringing algorithms further enhance the continuity and eliminate unreasonable ring structures.

[0102] This application changes the research object to the frontal zone, and then extracts the front from the frontal zone using a mathematical morphology algorithm to ensure continuity; it also proposes a front merging algorithm to further improve continuity; and invents a de-ringing algorithm to remove unreasonable fronts. At the same time, the algorithm based on block Bayesian decision-making can obtain better front recognition results.

[0103] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0104] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application; at the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A large-scale ocean front identification method, characterized in that: The large-scale ocean front identification method comprises: Based on oceanographic knowledge, large-scale sea areas are divided into blocks to determine multiple block areas; Based on the Bayesian algorithm, the frontal zone in each block area is extracted; the frontal zone is an area containing a front; Extracting the skeleton of the frontal zone from the frontal zone by using a mathematical morphology algorithm; Using a discrete skeleton evolution algorithm, pruning the skeleton to determine a pruned skeleton; Based on the pruned skeleton, the adjacent fronts in each block area are connected using the front weaving ring elimination algorithm, and the ring structure of the connected fronts is removed, so as to extract the fronts in each block area after the ring structure is removed; The fronts after removing the ring structures in all block areas are merged to determine the global front of the large-scale sea area.

2. The large-scale ocean front identification method according to claim 1, characterized in that: Based on the Bayesian algorithm, the frontal zone in each block area is extracted, including: For each block area, the Sobel operator is used to calculate the gradient and gradient direction of the ocean elements; the ocean elements include sea surface temperature, salinity and density; Determine a quantile according to the gradient, and determine a frontal zone type according to the quantile; the frontal zone type includes a frontal zone, a non-frontal zone, and a to-be-determined frontal zone; Determine the prior probability of each pixel point in the undetermined frontal zone according to the gradient; Calculating the local edge degree and block discreteness of each pixel point in the undetermined frontal zone by using the ocean elements; Calculating the likelihood function of each pixel point in the to-be-determined frontal zone according to the local edge degree and the block discreteness; Bayesian decision making is performed according to the likelihood function and the prior probability to determine the frontal area and the non-frontal area in the pending frontal area, and all the frontal areas in each block area are extracted.

3. The large-scale ocean front identification method according to claim 2, characterized in that: Determining a quantile according to the gradient, and determining a frontal zone type according to the quantile, specifically includes: Generate a gradient sequence according to the order of the gradient values ​​from large to small; The first 10% of the gradient position in the gradient sequence is taken as the first quantile, and the first 10% of the gradient is taken as the upper threshold; The first 20% of the gradient positions in the gradient sequence are used as the second quantile, and the first 20% of the gradient positions are used as the lower threshold; Taking the gradient corresponding to a value greater than the upper threshold as the frontal zone; The gradient corresponding to a value less than the lower threshold is regarded as a non-frontal area; The gradient corresponding to the upper threshold and the lower threshold is used as the to-be-determined frontal zone.

4. The large-scale ocean front identification method according to claim 3 is characterized in that: Determining the prior probability of each pixel point in the undetermined frontal zone according to the gradient specifically includes: use and Determine the prior probability of each pixel point in the undetermined frontal zone; wherein, is the prior probability of each pixel point when the undetermined front point E is the front point, i It is the frontal zone; is the prior probability of each pixel point when the pending frontal point E is a non-frontal point, j It is a non-frontal area; is the upper threshold; is the lower threshold; is the gradient of the undetermined frontal point E.

5. The large-scale ocean front identification method according to claim 4, characterized in that: The local edge degree and block discreteness of each pixel point in the undetermined frontal zone are calculated by using the ocean elements, specifically including: Let any undetermined front point in the undetermined front area be located at the center of the 3×3 area, and a set of vectors consisting of the undetermined front point E and the neighboring points of the undetermined front point E ; use Calculate the local edge degree between two neighboring points in the pending frontal zone; wherein, is the local edge degree; A and I are any two opposite points in the 3×3 area; For vector The maximum value of an element, is a vector The minimum value of an element, is a vector Mean value of elements; vector is a set of vectors consisting of the undetermined frontal point E and its neighboring points; The local edge degree of the undetermined frontal point E is calculated based on the local edge degree between the four pairs of neighboring points. ; use Calculating the block discreteness between two neighboring points in the pending frontal zone; is the block discreteness; The block discreteness of the undetermined frontal point E is calculated based on the block discreteness between the four pairs of neighboring points. .

6. The large-scale ocean front identification method according to claim 5, characterized in that: Calculating the likelihood function of each pixel point in the to-be-determined frontal area according to the local edge degree and the block discreteness specifically includes: use Calculating the likelihood function that the undetermined frontal point E belongs to the frontal zone i; is the likelihood function that the undetermined frontal point E belongs to the frontal zone i; The gradient is greater than The number of points; For The number of differences less than 0.1; For The number of differences less than 0.1; use Calculating the likelihood function that the undetermined frontal point E belongs to the non-frontal zone j; is the likelihood function that the undetermined frontal point E belongs to the non-frontal zone j; The gradient is less than The number of points; For The number of differences less than 0.1; For The number of differences less than 0.

1.

7. The large-scale ocean front identification method according to claim 6, characterized in that: Performing Bayesian decision making according to the likelihood function and the prior probability to determine the frontal zone and the non-frontal zone in the undetermined frontal zone specifically includes: when When , the undetermined frontal zone point E is determined as the frontal zone; when When the frontal zone point to be determined is Determined to be a non-frontal area.

8. The large-scale ocean front identification method according to claim 1, characterized in that: The skeleton of the frontal zone is extracted from the frontal zone by using a mathematical morphology algorithm, specifically including: Use the maximum disk method in mathematical morphology algorithm to extract the skeleton of the frontal zone ;in, , ; For the identified frontal zone Continuous Second corrosion, is the convolution kernel, For the opening operation, For corrosion operations, for is eroded to an empty set The last iteration step before is: .

9. The large-scale ocean front identification method according to claim 8, characterized in that: The skeleton is pruned using a discrete skeleton evolution algorithm to determine a pruned skeleton, specifically including: The skeleton All vertices on are divided into endpoints and intersections, and As All endpoints on As distance The nearest intersection will as a line segment between two points; S for ; according to Determine the frontal zone after skeleton reconstruction; among them, The frontal zone after skeleton reconstruction; For The largest disk centered in the frontal zone The radius of the skeleton The point above; use For each endpoint Assign weights; among them, is the weight; Based on the skeleton Reconstructed frontal zone; is the area function; Initializing all endpoint weights, iterating the endpoint weights, determining the iterated endpoint weights, and determining the minimum iterative endpoint weight during the iterative process; A pruned skeleton is determined according to the minimum endpoint weights.

10. The large-scale ocean front identification method according to claim 1, characterized in that: Based on the pruned skeleton, the adjacent fronts in each block area are connected using the front weaving ring elimination algorithm, and the ring structure of the connected fronts is removed, and the fronts after the ring structure is removed in each block area are extracted, specifically including: Based on the pruned skeleton, setting a search radius; For any front, determine whether the head and tail of the adjacent front fall within the search radius of the head and tail of the front; If there are multiple adjacent fronts, they are connected according to the priority of the distance between the head and the tail and the absolute value of the difference in the gradient direction between the head and the tail to determine the connected rear front; If the tail of any front in each connected rear front is within the 3×3 area of ​​the head, the front is judged as a ring structure and removed; If the tail of each connected front is not within the 3×3 area of ​​the head, determine whether each point in the front is within the 3×3 area of ​​the tail from the head. If so, determine the front as a ring structure and remove it. Until the data of all fronts are no longer updated, extract the fronts in each block area after removing the ring structure.

Images