A method for visualizing marine environment remote sensing based on numerical feature statistics

Abstract

The present invention provides a method for visual expression of marine environment remote sensing based on numerical feature statistics, including: a data preprocessing module, which performs noise reduction processing on data to improve data resolution and weaken edge jaggedness; a data grouping and feature statistics module, which obtains data grouping and performs feature statistics through cluster analysis and feature statistics for the clustering phenomenon existing in the numerical distribution of marine environment remote sensing inversion raster data; a group spacing optimization and color mapping module, which establishes an index relationship between a set color and a numerical interval through a set numerical interval to perform color mapping and optimize the visual display effect. Beneficial effects of the present invention: a method for visual expression of marine environment remote sensing based on numerical feature statistics, which uses a data preprocessing module to improve data resolution and weaken edge jaggedness for the problem that the inverted raster data inevitably contains noise and errors, and the data pattern and its features are easily hidden.

Description

Technical Field

[0001] The invention belongs to the field of marine environment remote sensing, and in particular relates to a marine environment remote sensing visualization expression method based on numerical feature statistics. Background Art

[0002] Marine remote sensing inversion data is a quantitative inversion of marine environment and phenomenon information. Due to the obvious differences in the properties of marine and terrestrial environments and the large scale of the areas covered by marine satellites, the information expression of satellite remote sensing inversion data is different from that of land. At present, marine environmental data is mainly stored in the form of raster data. Due to the complex distribution of marine environmental information in large-scale areas and uneven information distribution, the commonly used marine element visualization rendering methods include single value rendering, graded rendering, and stretched rendering. These rendering methods all face the problem of reasonable grading of color bands. The traditional method of using fixed levels to visualize marine time series elements often causes pixel values ​​to be overly concentrated in individual small intervals, seriously affecting the resolution effect of areas with high color concentration, resulting in low visual contrast during rendering, large discreteness of individual data, and serious edge jaggedness, which in turn affects the ability to express map information and visual aesthetics.

[0003] At present, the analysis of ocean inversion data visualization has not received enough attention. ESRI provides a method to increase the visual contrast of raster display by stretching, but its principle is just to grade the color bands by uniform density segmentation, without analyzing or specially processing the data. The method results also fail to substantially solve the problem of poor visualization of the above-mentioned marine environment remote sensing inversion information. Summary of the invention

[0004] In view of this, the present invention aims to propose a method for visualizing remote sensing of the marine environment based on numerical feature statistics, so as to at least solve at least one problem in the background technology.

[0005] To achieve the above object, the technical solution of the present invention is achieved as follows:

[0006] A method for visualizing marine environment remote sensing based on numerical feature statistics, including:

[0007] The data preprocessing module performs noise reduction on the data to improve data resolution and weaken edge jaggedness;

[0008] The data grouping and feature statistics module aims at the clustering phenomenon in the numerical distribution of the marine environment remote sensing inversion raster data. Through cluster analysis and feature statistics, the data is grouped and feature statistics are performed.

[0009] The group spacing optimization and color mapping module establishes an index relationship between the set color and the numerical interval by setting the numerical interval to perform color mapping and optimize the visual display effect.

[0010] Furthermore, in the data preprocessing module, the marine environment remote sensing inversion raster data is preprocessed to remove abnormal values ​​and mutation values, and the special abundance distribution type data is converted, as follows:

[0011] A1. Outlier elimination: Based on the attribute, region and climatological characteristics of the elements, the data is tested for range and landing, and outliers are eliminated;

[0012] A2. Remove sudden changes: Find isolated peaks in the data and remove them;

[0013] A3. Data preprocessing of special abundance distribution types: For the visualization of raster data of large gradient change process elements facing estuary / nearshore-basin / ocean mixed areas, logarithmic preprocessing is first performed to initially even out their abundance distribution.

[0014] Furthermore, in step A2, the following steps are specifically included:

[0015] (1) Traverse and calculate the average value D of the mutation between the grid point and the neighboring grid points ij :

[0016]

[0017] Among them, g ij is the target grid value, g pq is the value of the neighborhood grid point, and n is the number of valid values ​​of the neighborhood grid points;

[0018] (2) Calculate the spatial standard deviation σ of the mutation mean D :

[0019]

[0020] Among them, N is the number of valid values ​​of raster data, r is the total number of rows of raster data, and r is the total number of columns of raster data. is the average value of the grid data mutation;

[0021] (3) Traversing the grid point mutation average value D ij and spatial standard deviation σ D Relationship, determination D ij >3σ D The grid points are mutation outliers, and the invalid value noDataValue is assigned to them to complete the elimination process.

[0022] Furthermore, in step A3, the following steps are specifically included:

[0023] (1) Calculate the range of raster data;

[0024] (2) Performing histogram analysis on the raster data with no less than 10 groups, when the cumulative group interval of the first n maximum frequency groups with cumulative frequencies greater than 90% is less than 10% of the range, logarithmic processing is performed on the raster data;

[0025] (3) Store the logarithmic processing mapping relationship and use the inverse mapping to determine the color band coordinates after completing the subsequent hierarchical optimization process.

[0026] Furthermore, in the data preprocessing module, interpolation algorithms are used to improve the density of spatially discontinuous or low-resolution data, supplement missing data, eliminate and weaken internal and boundary mosaic phenomena, and more smoothly and meticulously depict the true distribution of marine environmental remote sensing elements, as follows:

[0027] B1. Mask construction: superimpose the land vector on the raster data, set the noDataValue position of the land part to -1, set the noDataValue position of the other sea parts to 0, and set the other valid value positions to 1 to obtain the inverted raster data mask;

[0028] B2. Hole filling: selectively fill the missing values ​​in a small range where the mask gray value is 0;

[0029] B3. Internal interpolation: To prevent data distortion, the interpolation ratio of internal data is set to 0.5-3. According to the abundance of the spatiotemporal sequence of data, DINEOF or bilinear interpolation is used to process the internal continuous data.

[0030] B4. Edge interpolation: Bilinear interpolation cannot solve the problem of jagged edges. For edge data, a natural transition with the internal effective value is required. The present invention adopts the following steps to smooth the edges and weaken the mosaic phenomenon.

[0031] Furthermore, in step B2, the following steps are specifically included:

[0032] (1) For the internal missing values ​​whose grayscale value of the mask is 0, extract the connected domain S i , i = 1, 2, 3, ..., m, m is the total number of connected domains, which are divided into small and large categories according to the size of the connected domains. The connected domains with an area of ​​≤ 5 pixels are small connected domains, and the rest are large connected domains;

[0033] (2) For each small connected domain, the filling method is divided into time series regularity filling and adjacent value filling according to whether it is time series data. For the inversion information containing time series data, it is reconstructed by the empirical orthogonal function interpolation method; for the single inversion information without time series data, the method of statistical adjacent value characteristics is used to assign the missing value, and the filling is performed in sequence from the outside to the inside. The filling value is the average of all adjacent values. Among them, no filling processing is performed on each large connected domain.

[0034] Furthermore, in step B4, the following steps are used to smooth the edges and weaken the mosaic phenomenon:

[0035] (1) Edge search: traverse the four neighborhoods of the pixel with a mask value of 1. If any pixel in the four neighborhoods is -1, the pixel is an edge pixel.

[0036] (2) Reconstruct the best edge polyline based on edge shape fitting. For each edge, divide it into long edge and short edge combinations according to the number of pixels. Use the MLAA algorithm to divide the edge into "Z-type", "U-type", and "L-type", and reconstruct the best edge line;

[0037] (3) Assign a value to each noDataValue pixel that the reconstructed edge line passes through. The value of each pixel that is crossed by the reconstructed edge line is the mixed result of the length ratio of the reconstructed edge line passing through the adjacent pixels. The value of the pixel to be assigned is:

[0038]

[0039] Among them, T is the number of adjacent edge pixels of the pixel to be assigned, g i is the value of the i-th adjacent pixel, r i is the length of the reconstructed edge line passing through the i-th edge pixel, r is the pixel length, g a The value of the pixel to be assigned.

[0040] Furthermore, in the data grouping and feature statistics module, the following steps are specifically included:

[0041] C1. Determine the number of groups n, group the continuous raster data, and the data within the group should have obvious clustering characteristics;

[0042] C2, performing intra-group feature statistics according to the grouping determined in step C1;

[0043] C3. Adjust the group spacing for steps C1 and C2;

[0044] In step C1, the number of groups should be determined according to the following method:

[0045] (1) Perform histogram statistics on the raster data to obtain the pixel value range [g min, g max ]The number of pixels corresponding to each value;

[0046] (3) extract the histogram contour and smooth it;

[0047] (3) Use the findpeaks function in Matlab to find the histogram peaks, set the minimum interval between two peaks to 5, and the total number of peaks that meet the conditions is the number of groups S;

[0048] Wherein, in step C2, the specific steps are as follows:

[0049] The value range [g min , g max ] are divided into S groups, and the number of corresponding pixels n in each group is counted k (k∈[1,S]), the value range of each group is [str k ,end k ];

[0050] Wherein, in step C3, the specific steps are as follows:

[0051] (1) According to the data histogram, calculate [str k ,end k The sum of the probability densities corresponding to all values ​​in the interval is used to obtain the probability density P of the values ​​within the group. k :

[0052]

[0053] Among them, n is the total number of pixels, n k For the value in [str k ,end k ]The number of all pixels in the interval;

[0054] (2) The degree of discreteness of the numerical space distribution is measured by the proportion of the covariance matrix norm of all numerical space coordinates in the group. Find [str k ,end k ] The spatial coordinates corresponding to all values ​​in the interval Calculate the dispersion D of the value distribution within the group k , the formula is as follows:

[0055] COV(X k ,Y k )=E[(X k -E(X k ))(Y k -E(Y k ))]

[0056]

[0057]

[0058]

[0059] in, E(X k ) is X k Expectation, E(Y k ) is Y k Expectation, C is the covariance matrix; is the singular value of C; F k is the covariance norm of the kth group;

[0060] (3) Get each interval [str k ,end k ]The number of pixel values ​​and the next adjacent interval [str k+1 ,end k+1 ] pixel value change gradient V k :

[0061]

[0062]

[0063] Among them, n k+1 For the value in [str k+1 ,end k+1 ]The number of all pixels in the interval;

[0064] Wherein, in step C4, the specific steps are as follows:

[0065] (1) Based on the numerical feature parameters of each group counted in step C3, the feature weight A in each group is formed k :

[0066]

[0067] (2) By adjusting the feature weights, the intervals between groups are optimized, and the range [g min , g max ] Readjust according to the feature weights within the group, then the distance between the kth group is:

[0068] F k =(g max -g min )×A k .

[0069] Furthermore, in the group spacing optimization and color mapping module, the following steps are specifically included:

[0070] Establish a value-color band mapping table: select a color model for the optimal S group value after determination, and perform data mapping. The data mapping formula is as follows:

[0071] IMG(x,y)→{[str1,end1],[str2,end2]……[str s ,end s ]}→ColorBar

[0072] Among them, IMG is raster data, [x, y] is the spatial horizontal and vertical coordinates, and ColorBar is the color band.

[0073] Furthermore, the present solution discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, implements a method for visualizing the remote sensing of the marine environment based on numerical feature statistics.

[0074] Compared with the prior art, the method for visualizing the remote sensing of the marine environment based on numerical feature statistics described in the present invention has the following beneficial effects:

[0075] (1) The method for visualizing remote sensing of marine environment based on numerical feature statistics of the present invention uses a data preprocessing module to improve data resolution and weaken edge jaggedness in order to solve the problem that the inverted raster data inevitably contains noise and errors and the data pattern and its features are easily hidden;

[0076] (2) The method for visualizing the marine environment remote sensing based on numerical feature statistics described in the present invention utilizes group spacing optimization and color mapping modules to ensure that the numerical probability density distribution of each group interval is uniform and reasonable, and the differences between groups are obvious, thereby optimizing the visualization display effect, highlighting the expression of marine environmental phenomena, and solving the problem of unsatisfactory visualization effect of the above-mentioned marine remote sensing inversion information raster data. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:

[0078] Figure 1 It is a schematic diagram of the overall processing flow according to an embodiment of the present invention;

[0079] Figure 2 A schematic diagram of a process of assigning and filling values ​​in sequence from outside to inside in an interpolation calculation process according to an embodiment of the present invention;

[0080] Figure 3 A schematic diagram of an edge interpolation process according to an embodiment of the present invention;

[0081] Figure 4 The present invention is a schematic diagram of establishing a value-color band mapping method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0082] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0083] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0084] The present invention is aimed at the marine environment remote sensing inversion raster data, and mainly includes the following modules:

[0085] (1) Data preprocessing. Since the inverted raster data inevitably contains noise and errors, the data pattern and its characteristics are easily hidden, so it is necessary to preprocess the data. First, remove outliers and invalid data and make necessary conversions; then interpolate the data, mainly including mask construction, hole filling, internal interpolation, and edge interpolation. This will eliminate holes caused by missing values, improve data resolution, and weaken edge jaggedness;

[0086] (2) Data grouping and feature statistics. The distribution of the values ​​of the marine environment remote sensing inversion raster data is clustered. Through cluster analysis and feature statistics, reasonable grouping can be obtained. First, the histogram is statistically analyzed to determine the cluster center of the raster data and the number of groups. Then, the feature statistical analysis of the data within the group is performed, mainly including the probability density of the values, the distribution dispersion, and the change gradient.

[0087] (3) Group spacing optimization and color mapping. The marine environment remote sensing inversion raster data is color mapped by setting a reasonable numerical interval and establishing an index relationship between the set color and the numerical interval. First, the weight of each group is determined through the characteristic statistical value within the group, and the group spacing is optimized. Then, the grouped data is grouped and rendered according to the designed color band to ensure that the numerical probability density distribution of each group interval is uniform and reasonable, and the difference between groups is obvious. The visualization display effect is optimized, the expression of marine environmental phenomena is highlighted, and the problem of unsatisfactory visualization effect of the above-mentioned marine remote sensing inversion information raster data is solved. The overall process is attached. Figure 1 .

[0088] step:

[0089] 1. Data preprocessing: Data preprocessing is performed on the marine environment remote sensing inversion raster data to remove abnormal values ​​and mutation values, and to convert data of special abundance distribution types.

[0090] 1-1: Outlier removal: Based on feature attributes, regional and climatological characteristics, the data is tested for range and landing, and outliers are removed.

[0091] 1-2: Mutation value removal: Find isolated peaks in the data and remove them

[0092] (1) Traverse and calculate the average value D of the mutation between the grid point and the neighboring grid points ij :

[0093]

[0094] Among them, g ij is the target grid value, g pq is the value of the neighborhood grid point, and n is the number of valid values ​​of the neighborhood grid points. (2) Calculate the spatial standard deviation σ of the mutation mean D :

[0095]

[0096] Among them, N is the number of valid values ​​of raster data, r is the total number of rows of raster data, and r is the total number of columns of raster data. is the average value of the grid data mutation.

[0097] (3) Traversing the grid point mutation average value D ij and spatial standard deviation σ D Relationship, determination D ij >3σ D The grid points are mutation outliers, and the invalid value noDataValue is assigned to them to complete the elimination process.

[0098] 1-3: Data preprocessing of special abundance distribution types: For the visualization of raster data of large gradient variation process elements (such as chlorophyll, suspended matter, etc.) facing estuaries / nearshore-basin / ocean mixed areas, logarithmic preprocessing is first performed to initially even out their abundance distribution.

[0099] (1) Calculate the range of raster data;

[0100] (2) Performing histogram analysis on the raster data with no less than 10 groups, when the cumulative group interval of the first n maximum frequency groups with cumulative frequencies greater than 90% is less than 10% of the range, logarithmic processing is performed on the raster data;

[0101] (3) Store the logarithmic processing mapping relationship and use the inverse mapping to determine the color band coordinates after completing the subsequent hierarchical optimization process.

[0102] 2. Interpolation calculation. Interpolation algorithms are used for spatially discontinuous or low-resolution data to improve their density, supplement missing data, eliminate and weaken internal and boundary mosaic phenomena, and more smoothly and meticulously depict the true distribution of marine environmental remote sensing elements. Due to the large amount of marine inversion remote sensing raster data, traditional interpolation algorithms are time-consuming and cannot take into account internal and edge data. The present invention uses the following method to interpolate the data:

[0103] 2-1: Mask construction. Overlay the raster data with the land vector and set the noDataValue position of the land part to

[0104] -1, the noDataValue position of other sea areas is set to 0, and the remaining valid value positions are set to 1 to obtain the inversion raster data mask.

[0105] 2-2: Hole filling: Selectively fill the missing values ​​in a small range where the grayscale value of the mask is 0.

[0106] (1) For the internal missing values ​​whose grayscale value of the mask is 0, extract the connected domain S i , i = 1, 2, 3, ..., m, m is the total number of connected domains. According to the size of the connected domain, it is divided into small and large categories (the area of ​​the connected domain is ≤ 5 pixels, which is a small connected domain, and the rest are large connected domains).

[0107] (2) For each small connected domain, the filling is divided into time series regularity filling and adjacent value filling according to whether it is time series data. For the inversion information containing time series data, it is reconstructed by the empirical orthogonal function interpolation method (DINEOF); ​​for the single inversion information without time series data, the method of statistical adjacent value characteristics is used to assign values ​​to the missing values, and the filling is performed in sequence from the outside to the inside (see Appendix). Figure 2 ), the filling value is the average of all adjacent values. No filling is performed for each large interconnection domain.

[0108] 2-3: Internal interpolation. For internal data, in order to prevent data distortion, the interpolation ratio is set to 0.5-3, and DINEOF or bilinear interpolation is used to process internal continuous data according to the abundance of the spatiotemporal sequence of the data.

[0109] 2-4: Edge interpolation. Bilinear interpolation cannot solve the edge jagged problem. For edge data, a natural transition with the internal effective value is required. The present invention adopts the following steps to smooth the edge and weaken the mosaic phenomenon.

[0110] (1) Edge search: traverse the four neighborhoods of the pixel with a mask value of 1. If any pixel in the four neighborhoods is -1, the pixel is an edge pixel.

[0111] (2) Reconstruct the best edge polyline based on edge shape fitting. For each edge, divide it into long edge and short edge combinations according to the number of pixels. According to the MLAA algorithm, divide the edge into "Z-type", "U-type" and "L-type", and reconstruct the best edge line (see Appendix Figure 3 , where the white line is the optimal edge line for reconstruction);

[0112] (3) Assign a value to each noDataValue pixel that the reconstructed edge line passes through. The value of each pixel that the reconstructed edge line passes through is the mixed result of the length ratio of the adjacent pixels that the reconstructed edge line passes through (see Appendix Figure 3 ), the value of the pixel to be assigned is:

[0113]

[0114] Among them, T is the number of adjacent edge pixels of the pixel to be assigned, g i is the value of the i-th adjacent pixel, r i is the length of the reconstructed edge line passing through the i-th edge pixel, r is the pixel length, g a The value of the pixel to be assigned.

[0115] 3. Raster data grouping.

[0116] 3-1: Determine the number of groups n. When grouping continuous raster data, the data within the group should have obvious clustering characteristics, and the number of groups should be determined according to the following method.

[0117] (1) Perform histogram statistics on the raster data to obtain the pixel value range [g min , g max ]The number of pixels corresponding to each value.

[0118] (2) extracting the histogram contour and smoothing it;

[0119] (3) Use the findpeaks function in Matlab to find the histogram peaks, set the minimum interval between two peaks to 5, and the total number of peaks that meet the conditions is the number of groups S.

[0120] 3-2: Intra-group feature statistics.

[0121] (1) Set the value range [g min , g max ] are divided into S groups, and the number of corresponding pixels n in each group is counted k (k∈[1,S]), the value range of each group is [str k ,end k ];

[0122] 3-3: Perform characteristic statistical analysis on the values ​​within the group to obtain the probability density P of the values ​​within each group k , distribution dispersion D k , change gradient V k .

[0123] (1) According to the data histogram, calculate [str k ,end kThe sum of the probability densities corresponding to all values ​​in the interval is used to obtain the probability density P of the values ​​within the group. k .

[0124]

[0125] Among them, n is the total number of pixels, n k For values ​​in [str k ,end k ]The number of all pixels in the interval.

[0126] (2) The degree of discreteness of the numerical space distribution is measured by the proportion of the covariance matrix norm of all numerical space coordinates in the group. Find [str k ,end k ] The spatial coordinates corresponding to all values ​​in the interval Calculate the dispersion D of the value distribution within the group k , the formula is as follows:

[0127] COV(X k ,Y k )=E[(X k -E(X k ))(Y k -E(Y k ))]

[0128]

[0129]

[0130]

[0131] in, E(X k ) is X k Expectation, E(Y k ) is Y k Expectation, C is the covariance matrix; is the singular value of C; F k is the covariance norm of the kth group.

[0132] (3) Get each interval [str k ,end k ]The number of pixel values ​​and the next adjacent interval [str k+1 ,end k+1 ] pixel value change gradient V k .

[0133]

[0134]

[0135] Among them, n k+1 For the value in [str k+1 ,end k+1 ]The number of all pixels in the interval.

[0136] 3-4: Adjust the distance between groups.

[0137] (1) Based on the numerical feature parameters of each group counted in step 3-3, the feature weight A within each group is formed k .

[0138]

[0139] (2) Optimize the distance between groups by adjusting the feature weights. min , g max ] Readjust according to the feature weights within the group, then the distance between the kth group is:

[0140] F k =(g max -g min )×A k

[0141] 4. Establish a value-color band mapping table. Select a color model for the optimal S group value after determination and perform data mapping. See attached Figure 4 The data mapping formula is as follows:

[0142] IMG(x,y)→{[str1,end1],[str2,end2]……[str s ,end s ]}→ColorBar

[0143] Among them, IMG is raster data, [x, y] is the spatial horizontal and vertical coordinates, and ColorBar is the color band.

[0144] Those of ordinary skill in the art will appreciate that the units and method steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0145] In the several embodiments provided in the present application, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the division of the units described above is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. The above-mentioned units may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiments of the present invention.

[0146] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

[0147] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for visualizing the remote sensing of the marine environment based on numerical feature statistics, characterized in that: include: The data preprocessing module performs noise reduction on the data to improve data resolution and weaken edge jaggedness; The data grouping and feature statistics module aims at the clustering phenomenon in the numerical distribution of the marine environment remote sensing inversion raster data. Through cluster analysis and feature statistics, the data is grouped and feature statistics are performed. The group spacing optimization and color mapping module establishes the index relationship between the set color and the numerical interval through the set numerical interval to perform color mapping and optimize the visual display effect; In the data preprocessing module, the marine environment remote sensing inversion raster data is preprocessed to remove abnormal values ​​and mutation values, and the special abundance distribution type data is converted as follows: A1. Outlier elimination: Based on the attribute, region and climatological characteristics of the elements, the data is tested for range and landing, and outliers are eliminated; A2. Remove sudden changes: Find isolated peaks in the data and remove them; A3. Preprocessing of data of special abundance distribution types: For the visualization of raster data of large gradient change process elements facing estuary / nearshore-sea basin / ocean mixed areas, logarithmic preprocessing is first performed to initially even out their abundance distribution; In step A2, the following steps are specifically included: (1) Traverse and calculate the average value D of the mutation between the grid point and the neighboring grid points ij : Among them, g ij is the target grid value, g pq is the value of the neighborhood grid point, and n is the number of valid values ​​of the neighborhood grid points; (2) Calculate the spatial standard deviation σ of the mutation mean D : Among them, N is the number of valid values ​​of raster data, r is the total number of rows of raster data, and c is the total number of columns of raster data. is the average value of the grid data mutation; (3) Traversing the grid point mutation average value D ij and spatial standard deviation σ D Relationship, determination D ij >3σ D The grid points are mutation outliers, and the invalid value noDataValue is assigned to them to complete the elimination process.

2. The method for visualizing the remote sensing of the marine environment based on numerical feature statistics according to claim 1, characterized in that: In step A3, the following steps are specifically included: (1) Calculate the range of raster data; (2) Performing histogram analysis on the raster data with no less than 10 groups, when the cumulative group interval of the first n maximum frequency groups with cumulative frequencies greater than 90% is less than 10% of the range, logarithmic processing is performed on the raster data; (3) Store the logarithmic processing mapping relationship and use the inverse mapping to determine the color band coordinates after completing the subsequent hierarchical optimization process.

3. The method for visualizing the remote sensing of the marine environment based on numerical feature statistics according to claim 1, characterized in that: In the data preprocessing module, interpolation algorithms are used to improve the density of spatially discontinuous or low-resolution data, supplement missing data, eliminate and weaken internal and boundary mosaic phenomena, and more smoothly and meticulously depict the true distribution of marine environmental remote sensing elements. The details are as follows: B1. Mask construction: superimpose the land vector on the raster data, set the noDataValue position of the land part to -1, set the noDataValue position of the other sea parts to 0, and set the other valid value positions to 1 to obtain the inverted raster data mask; B2. Hole filling: selectively fill the missing values ​​in a small range where the mask gray value is 0; B3. Internal interpolation: To prevent data distortion, the interpolation ratio of internal data is set to 0.5-3. According to the abundance of the spatiotemporal sequence of data, DINEOF or bilinear interpolation is used to process the internal continuous data. B4. Edge interpolation: Bilinear interpolation cannot solve the edge jagged problem. For edge data, a natural transition with the internal effective value is required; In step B2, the following steps are specifically included: (1) For the internal missing values ​​whose grayscale value of the mask is 0, extract the connected domain S b , b=1,2,3,……,m, m is the total number of connected domains, which are divided into small and large categories according to the size of the connected domains. The connected domains with an area of ​​≤5 pixels are small connected domains, and the rest are large connected domains; (2) For each small connected domain, the filling method is divided into time series regularity filling and adjacent value filling according to whether it is time series data. For the inversion information containing time series data, it is reconstructed by the empirical orthogonal function interpolation method; for the single inversion information without time series data, the method of statistical adjacent value characteristics is used to assign the missing value, and the filling is performed in sequence from the outside to the inside. The filling value is the average of all adjacent values. Among them, no filling processing is performed on each large connected domain.

4. The method for visualizing the remote sensing of the marine environment based on numerical feature statistics according to claim 3 is characterized in that: In step B4, the method of smoothing the edge and weakening the mosaic phenomenon includes: (1) Edge search: traverse the four neighborhoods of the pixel with a mask value of 1. If any pixel in the four neighborhoods is -1, the pixel is an edge pixel. (2) Reconstruct the best edge polyline based on edge shape fitting. For each edge, divide it into long edge and short edge combinations according to the number of pixels. Use the MLAA algorithm to divide the edge into "Z-type", "U-type", and "L-type", and reconstruct the best edge line. (3) Assign a value to each noDataValue pixel that the reconstructed edge line passes through. The value of each pixel that is crossed by the reconstructed edge line is the mixed result of the length ratio of the reconstructed edge line passing through the adjacent pixels. The value of the pixel to be assigned is: Among them, T is the number of adjacent edge pixels of the pixel to be assigned, g e is the value of the e-th adjacent pixel, Q e is the length of the reconstructed edge line passing through the e-th edge pixel, Q is the pixel length, g a The value of the pixel to be assigned.

5. The method for visualizing the remote sensing of the marine environment based on numerical feature statistics according to claim 1 is characterized in that: In the data grouping and feature statistics module, it includes: C1. Determine the number of groups n, group the continuous raster data, and the data within the group should have obvious clustering characteristics; C2, performing intra-group feature statistics according to the grouping determined in step C1; C3. Adjust the group spacing for steps C1 and C2; In step C1, the number of groups should be determined according to the following method: (1) Perform histogram statistics on the raster data to obtain the pixel value range [g min , g max ]The number of pixels corresponding to each value; (2) extracting the histogram contour and smoothing it; (3) Use the findpeaks function in Matlab to find the histogram peaks, set the minimum interval between two peaks to 5, and the total number of peaks that meet the conditions is the number of groups S; Wherein, in step C2, the specific steps are as follows: The value range [g min , g max ] are divided into S groups, and the number of corresponding pixels n in each group is counted k (k∈[1,S]), the value range of each group is [str k ,end k ]; Wherein, in step C3, the specific steps are as follows: (1) According to the data histogram, calculate [str k ,end k The sum of the probability densities corresponding to all values ​​in the interval is used to obtain the probability density P of the values ​​within the group. k : Among them, n is the total number of pixels, n k For the value in [str k ,end k ]The number of all pixels in the interval; (2) The degree of discreteness of the numerical space distribution is measured by the proportion of the covariance matrix norm of all numerical space coordinates in the group. Find [str k ,end k ] The spatial coordinates corresponding to all values ​​in the interval Calculate the dispersion D of the value distribution within the group k , the formula is as follows: COV(X k ,Y k )=E[(X k -E(X k ))(Y k -E(Y k ))] in, E(X k ) is X k Expectation, E(Y k ) is Y k Expectation, C is the covariance matrix; is the singular value of C; F k is the covariance norm of the kth group; (3) Get each interval [str k ,end k ]The number of pixel values ​​and the next adjacent interval [str k+1 ,end k+1 ] pixel value change gradient V k : Among them, n k+1 For the value in [str k+1 ,end k+1 ]The number of all pixels in the interval; Wherein, in step C4, the specific steps are as follows: (1) Based on the numerical feature parameters of each group counted in step C3, the feature weight A in each group is formed k : (2) By adjusting the feature weights, the intervals between groups are optimized, and the range [g min , g max ] Readjust according to the feature weights within the group, then the distance between the kth group is: U k =(g max -g min )×A k 。 6. The method for visualizing the remote sensing of the marine environment based on numerical feature statistics according to claim 5 is characterized in that: In the group spacing optimization and color mapping module, including: Establish a value-color band mapping table: select a color model for the optimal S group value after determination, and perform data mapping. The data mapping formula is as follows: IMG(x,y)→{[str1,end1],[str2,end2]……[str s ,end s ]}→ColorBar Among them, IMG is raster data, [x, y] is the spatial horizontal and vertical coordinates, and ColorBar is the color band.

7. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for visualizing and expressing marine environment remote sensing based on numerical feature statistics as described in any one of claims 1 to 6 is implemented.

Images