Sea ice classification and thickness detection integrated inversion method and system based on high-resolution polarimetric SAR

Abstract

The invention provides a sea ice classification and thickness detection integrated inversion method and system based on a high-resolution polarimetric SAR. The method comprises an integrated inversion method combining sea ice type and thickness information, integrated inversion based on type support, integrated inversion based on thickness support and integrated inversion based on thickness and type combined support. The method is suitable for simultaneously performing sea ice classification and thickness detection inversion by using high-resolution polarimetric SAR data, an integrated inversion method of sea ice classification and thickness detection is constructed, and the precision of sea ice classification and thickness detection is further improved.

Description

Technical Field

[0001] The present invention relates to the technical field of satellite product SAR payload technology, and in particular to an integrated inversion method and system for sea ice classification and thickness detection based on high-resolution polarimetric SAR. Background Art

[0002] In SAR sea ice detection, existing research methods have always performed sea ice type identification and sea ice thickness inversion separately. When sea ice conditions are particularly complex, such as when the ice is covered with snow, the accuracy of SAR sea ice classification will be reduced. However, sea ice thickness detection is not affected by snow accumulation and can still achieve good inversion accuracy. For SAR sea ice thickness inversion, current inversion methods require necessary control points (measured ice thickness data) to determine model parameters. When control points are missing, the accuracy of polarimetric SAR sea ice thickness inversion results will be significantly inaccurate. Therefore, for SAR sea ice detection, there are certain limitations when performing either type identification or thickness inversion separately.

[0003] In reality, sea ice type and thickness are closely related. In particular, sea ice types based on their growth process correspond to a specific thickness range. The thickness ranges for primary ice, early-stage ice, and thin first-year ice are relatively narrow, and the correspondence between sea ice type and thickness is relatively precise. Therefore, this patent considers integrating both sea ice type and thickness information for integrated detection to further enhance SAR sea ice detection capabilities.

[0004] Currently, many spaceborne SAR systems, such as RADARSAT-2 and ALOS-2, have achieved high-resolution polarimetric imaging, combining both high-resolution and polarimetric characteristics. Due to the correlation between sea ice type and thickness, the SAR detection results of sea ice type and thickness can serve as constraints on each other. Therefore, the present invention utilizes high-resolution polarimetric SAR data to simultaneously perform sea ice classification and thickness inversion, constructing an integrated inversion method for sea ice classification and thickness detection, further improving the accuracy of sea ice classification and thickness detection.

[0005] Patent application CN107516317A discloses a SAR image sea ice classification method based on a deep convolutional neural network, comprising the following steps: S01: Segmenting existing sea ice SAR images; S02: Data preprocessing; S03: Model training and model establishment; S04: Processing the sea ice SAR images to be classified; and S05: Merging the classification results. However, this patent fails to consider the relationship between sea ice thickness and sea ice classification, and does not incorporate sea ice thickness into the sea ice classification process. This inevitably affects the accuracy of SAR sea ice classification when the sea ice conditions are particularly complex, such as when the ice is covered with snow.

[0006] Patent application document CNIO6842205A discloses a method for automatic sea ice and seawater identification using synthetic aperture radar (SAR), including: Step S1: preprocessing a quad-polarized SAR image; removing quad-polarized SAR image noise using the Lee filter method; Step S2: calculating the directional drift angle and its standard deviation; Step S3: providing the directional drift angle formula in Step 2; Step S4: calculating the polarization ratio of each pixel in the SAR image; Step S5: calculating the theoretical polarization average of the SAR image; Step S6: calculating the difference between the actual SAR image polarization ratio and the theoretical polarization average. If the SAR image polarization ratio is greater than the theoretical average polarization ratio, it is determined to be seawater; otherwise, it is determined to be sea ice. However, this patent only identifies sea ice and seawater and cannot invert sea ice type and thickness.

[0007] Patent application CN107092933A discloses a method for classifying sea ice using synthetic aperture radar (SAR) scanning mode images. The method comprises: an image preprocessing step for preprocessing the SAR scanning mode image to obtain an optimized SAR image to be classified; a texture information extraction and optimization step for obtaining SAR texture information that meets sea ice classification requirements; a sea ice classification step for preliminary classification of sea ice based on the SAR texture information; and a sea ice classification optimization step for optimizing the preliminary classification results in combination with sea ice geometry information. However, this patent only processes sea ice in SAR scanning mode images and does not consider the relationship between sea ice thickness and sea ice classification, failing to incorporate sea ice thickness into the sea ice classification process.

[0008] Patent document CN201754185U discloses a sea ice microwave remote sensing monitoring system, comprising: a GNSS signal source, a RADARSAT signal source, a GNSS-R receiver, a mobile platform, a platform control system, and a data receiving and processing system. The GNSS signal source provides an L-band signal, the RADARSAT signal source provides a backscattered signal, the GNSS-R receiver receives the direct signal from the GNSS signal source and the forward scattered signal incident on the sea ice surface and scattered, the data receiving and processing system receives the forward scattered signal and the backscattered signal, processes and analyzes the received signals, and the platform control system controls the movement of the mobile platform. However, this patent primarily utilizes RADARSAT satellite signals and GNSS signals as signal sources for monitoring sea ice, does not consider the relationship between sea ice thickness and sea ice classification, and does not clarify whether sea ice thickness monitoring or sea ice classification can be performed.

[0009] Patent application document CN107679476A discloses a remote sensing classification method for sea ice types, including: the first step: preparing training data and data to be classified, and reading information from the data; the second step: extracting waveform features, based on the radar echo waveforms of all measurement points in the training data and the data to be classified, extracting corresponding waveform features as training features of the classifier for subsequent sea ice classification; the third step: spatially matching the radar waveform features extracted from the training data and the data to be classified with the corresponding longitude and latitude coordinates, and converting the matched data into vector point data with longitude and latitude coordinates; the fourth step: pre-classification processing, only using Using waveforms with leading edge widths less than or equal to 14, waveforms with pulse widths less than 0.3 and stack standard deviations greater than 4 are eliminated. Step 5: Generate training samples, spatially match vector points with latitude and longitude coordinates and waveform features within a certain period with sea ice type information data within the same period, and then obtain sea ice type information for each vector point, using this data as training samples. Step 6: Use the training samples to train a random forest classifier. After training, use the random forest classifier to classify the vector point data to be classified. Step 7: Obtain classified sea ice type raster data with uniform spatial resolution in the same coordinate system. However, this patent primarily utilizes Cryosat satellite data for sea ice monitoring, which is satellite radar altimeter data. This method does not consider the relationship between sea ice thickness and sea ice classification, but only performs sea ice classification.

[0010] Patent application CN106871877A discloses a method and apparatus for determining sea ice identification, including: first establishing a grid of equal latitude and longitude on the Earth's surface; then obtaining sea ice density values ​​for each grid point in the grid of equal latitude and longitude within a predetermined time period; obtaining sea surface temperature values ​​for each grid point in the grid of equal latitude and longitude; and finally, determining the sea ice identification of the grid of equal latitude and longitude based on the sea ice density values ​​and the sea surface temperature values. However, this patent primarily uses sea ice density and sea surface temperature values ​​for sea ice identification, and does not consider the relationship between sea ice thickness and sea ice classification. The method only performs sea ice identification, and does not clarify whether sea ice thickness monitoring or sea ice classification can be performed.

[0011] Patent application CN107730549A discloses a method for calculating sea ice age, including: converting a daily sea ice density image from year X#N to year X into a binary image of sea ice and seawater based on a preset sea ice density threshold, where X is the year and N is the sea ice age; determining the maximum sea ice area corresponding to year X and the minimum sea ice area corresponding to year X#1 based on the number of pixels in the binary image; determining sea ice areas with an ice age of 1 and a total ice age of 2 to N based on the maximum sea ice area corresponding to year X and the minimum sea ice area corresponding to year X#1; and extracting sea ice areas with an ice age of two and three years, and so on, from the total sea ice age of 2 to N. However, this patent primarily uses sea ice density to calculate sea ice age, and the method does not consider the relationship between sea ice thickness and sea ice classification. It only calculates sea ice age and does not clarify whether sea ice thickness monitoring can be performed.

[0012] Patent application CN107170006A discloses a method and apparatus for separating sea ice and seawater information from synthetic aperture radar images, comprising: calculating at least two texture features of an HV or VH cross-polarization image in the synthetic aperture radar image to be separated; determining a gradient matrix for at least one target texture feature among the at least two texture features; determining a set of points in the gradient matrix including gradient maxima and minima; determining multiple maximum points near each point in the point set, and defining the closed area formed by the multiple maximum points as a patch; determining patches with an average energy value greater than a preset energy threshold as seawater samples, and determining patches with an average entropy value greater than a preset entropy threshold as sea ice samples; and separating sea ice and seawater information from the synthetic aperture radar image based on the sea ice and seawater samples. However, while this patent automatically identifies sea ice and seawater samples while ensuring their accuracy, it does not consider the relationship between sea ice thickness and sea ice classification and only performs sea ice and seawater identification. It also does not clarify whether sea ice thickness monitoring or sea ice classification can be performed. Summary of the Invention

[0013] In view of the defects in the prior art, the purpose of the present invention is to provide an integrated inversion method and system for sea ice classification and thickness detection based on high-resolution polarimetric SAR.

[0014] The integrated inversion method for sea ice classification and thickness detection based on high-resolution polarimetric SAR provided by the present invention includes:

[0015] Step S1: Acquire high-resolution polarimetric SAR images;

[0016] Step S2: for high-resolution polarimetric SAR images, sea ice classification is performed using SAR high-resolution characteristics and category attributes are assigned;

[0017] Step S3: for high-resolution polarimetric SAR images, the SAR polarimetric characteristics are used to detect the sea ice thickness and assign thickness attributes;

[0018] Step S4: merging sea ice category attributes and thickness attributes;

[0019] Step S5: Calculate the confidence of sea ice classification;

[0020] Step S6: Obtaining measured sea ice thickness data;

[0021] Step S7: Processing the detection results containing the measured data of sea ice thickness, and correcting and improving the classification results based on the measured data of thickness;

[0022] Step S8: Processing detection results that do not contain measured sea ice thickness data but whose sea ice classification accuracy meets the preset requirements, and correcting and improving the thickness based on the classification results and classification confidence;

[0023] Step S9: When processing detection results containing measured sea ice thickness data and whose sea ice classification accuracy meets preset requirements, first improve the sea ice classification results based on the integrated detection supported by thickness, including correcting the sea ice type attributes and category confidence of the sea ice containing measured thickness data, and correcting the sea ice type and thickness information of the sea ice without measured thickness data; then adopt the integrated inversion method based on type support and use the improved sea ice classification results to correct the sea ice thickness detection results.

[0024] Preferably, step S7 includes:

[0025] Step S7.1: Matching sea ice with measured thickness data to determine whether sea ice contains measured thickness data;

[0026] Step S7.2: If the sea ice contains measured thickness data, process the sea ice containing the measured thickness data and modify the category attribute of the sea ice;

[0027] Step S7.3: If the sea ice does not contain measured thickness data, process the sea ice without measured thickness data and modify the sea ice category attributes and thickness data;

[0028] Step S7.4: Output the results of sea ice type and thickness detection.

[0029] Preferably, the step S7.2 includes:

[0030] Assume that the sea ice classification result contains N patches, M pixels, and the Sth i The pixels in the patch are represented by P ij Represents; i=1,2,…,N; j=1,2,…,m i ;m iis the total number of patch pixels, i.e. Σm i =M; the confidence of the category to which the patch belongs is determined by C i The thickness range corresponding to the plaque classification result is expressed as (h imin ,h imax ), the center value is h ia The thickness range corresponding to the classification results of each pixel in the patch is the same as the patch thickness range. The mean patch thickness calculated from the measured thickness data is H ia , with variance σ ia , then the thickness range of this patch is set to (H ia -σ ia ,H ia +σ ia ), by (H imin ,H imax ) indicates that the pixel thickness range obtained by the actual thickness measurement data is (H ijmin ,H ijmax ), the mean is H ija , the confidence threshold is determined by C t If , then the improvement strategy based on the premise that the plaque thickness attribute remains unchanged is:

[0031] like Then the patch category attribute remains unchanged, and the category confidence C i =C t ;

[0032] If (h imin ,h imax )∩(H imin ,H imax )≠0, and h ia ∈(H imin ,H imax ), the patch category attribute remains unchanged, and the category confidence C i =C t ×f1(S i );

[0033] If (h imin ,h imax )∩(H imin ,H imax )≠0, and Then find h ia The ice type corresponding to the thickness range, let the center value of this thickness range be H a , modify the patch category attributes, category confidence C i =C t ×f2(S i );

[0034] If (h imin ,h imax )∩(Himin ,H imax )=0, then find h ia The ice type corresponding to the thickness range, let the center value of this thickness range be H a , modify the patch category attributes, category confidence C i =C t ×f3(S i );

[0035] Among them, f1(S i )、f2(S i ) and f3(S i ) is the plaque category confidence correction function, and the plaque category confidence analysis is calculated using the standard Gaussian model;

[0036] For f1(S i ), let plaque S i In the classification results, the thickness range corresponding to the measured pixel thickness and the classification results satisfies H ija ∈(h imin ,h imax ) has a pixel number of m i1 ,but:

[0037]

[0038] For f2(S i ):

[0039] If h ia <H a ,but

[0040] If h ia >H a ,but

[0041] For f3(S i ):

[0042] If h ia <H a ,but

[0043] If h ia >H a ,but

[0044] This results in corrections and improvements to the sea ice classification patch category confidence and classification results.

[0045] Preferably, the step S7.3 includes:

[0046] The average thickness of all pixels obtained by the sea ice thickness detection method is given by H iIndicates; H ij is the thickness value of the j-th pixel of the patch extracted by the thickness model, then the improvement strategy is:

[0047] For patches without measured thickness data, the classification confidence of the sea ice patches without measured thickness data is corrected based on the measured thickness data. The expression is:

[0048]

[0049] Among them, C′ i represents the confidence of the original category of the patch, C i Represents the modified category confidence, R i It represents the ratio of the number of pixels with correct thickness to the total number of pixels in the thickness measurement data of the same sea ice type;

[0050] If H i ∈(h imin ,h imax ), the plaque category attribute remains unchanged, and the thickness attribute remains unchanged;

[0051] like And C i <C t , then the thickness attribute remains unchanged, find h ia Ice type corresponding to the thickness range, modify the patch category attributes, category confidence C i =C t ;

[0052] like And C i ≥C t , then the plaque category attribute remains unchanged, and the thickness attribute is determined as:

[0053] The pixels in the patch satisfy H ij ∈(h imin ,h imax ), then the pixel thickness value H ij No modification is required;

[0054] if Then the pixel thickness value H ij =(h imin +h imax ) / 2;

[0055] The patch thickness is the average thickness of all pixels in the patch.

[0056] Preferably, the step S8 includes:

[0057] Step S8.1: Compare the category and thickness attributes of the sea ice patch to determine whether the thickness of the sea ice patch is within the thickness range corresponding to its category; if the thickness of the sea ice patch is within the thickness range corresponding to its category, directly output the sea ice type and thickness detection results; if the thickness of the sea ice patch is not within the thickness range corresponding to its category, calculate the sea ice patch category confidence C i , and read the sea ice classification confidence threshold C t ;

[0058] Assume that the sea ice classification results contain N patches, and the Sth i The category confidence of each patch is given by C i Indicates that the confidence threshold is represented by C t Indicates, i=1,2,…,N; when the sea ice classification accuracy meets the preset requirements, the confidence threshold C t = 0.5, that is, the sea ice classification result is trusted, and the thickness range corresponding to the patch classification result is expressed as (h imin ,h imax ); plaque S i It is composed of multiple pixels, and its ice thickness is regarded as the average thickness of all pixels in the patch obtained by the sea ice thickness detection method, which is represented by H i express;

[0059] Step S8.2: Compare sea ice patch class confidence C i and the confidence threshold C for sea ice classification t If the sea ice patch category confidence level C i >Sea ice classification confidence threshold C t , then modify the sea ice patch thickness attribute, and the sea ice patch category attribute remains unchanged; if the sea ice patch category confidence C i <Sea ice classification confidence threshold C t , then the sea ice patch category attribute is modified, and the sea ice patch thickness attribute remains unchanged;

[0060] Find the sea ice type corresponding to the thickness of the sea ice patch and modify the sea ice patch type attribute; according to the positive correlation between the backscatter coefficient and the sea ice thickness, i The backscatter coefficients of all pixels in the image are analyzed by histogram, and the patch S i The minimum backscatter coefficient of the middle pixel corresponds to the patch S i The minimum thickness of the category, the maximum backscatter coefficient corresponds to the patch S i The maximum thickness of the category to which it belongs is calculated, and the corresponding relationship between the quantitative pixel backscatter coefficient and the sea ice thickness is established. i In the backscatter coefficient histogram of all pixels, the thickness corresponding to the backscatter coefficient with the highest frequency is regarded as the ice thickness of the patch, thereby realizing the ice thickness of the patch S. iCorrection of sea ice thickness attributes; using the correspondence between sea ice type and thickness, find H i The sea ice type corresponding to the thickness range is used as the category attribute of the patch;

[0061] Step S8.3: If H i ∈(h imin ,h imax ), that is, plaque S i The average thickness H i Within the thickness range corresponding to the ice type to which the patch belongs, the category attribute and thickness attribute of the patch remain unchanged, and the sea ice type and thickness detection results are directly output.

[0062] The integrated inversion system for sea ice classification and thickness detection based on high-resolution polarimetric SAR provided by the present invention comprises:

[0063] Module M1: Acquire high-resolution polarimetric SAR images;

[0064] Module M2: Based on high-resolution polarimetric SAR images, classify sea ice using SAR high-resolution characteristics and assign category attributes;

[0065] Module M3: Detect sea ice thickness using SAR polarization characteristics and assign thickness attributes based on high-resolution polarimetric SAR images.

[0066] Module M4: Merge sea ice category attributes and thickness attributes;

[0067] Module M5: Calculating sea ice classification confidence;

[0068] Module M6: Obtaining measured data on sea ice thickness;

[0069] Module M7: Processing detection results containing measured sea ice thickness data, and correcting and improving classification results based on the measured thickness data;

[0070] Module M8: Processes detection results that do not contain measured sea ice thickness data but whose sea ice classification accuracy meets the preset requirements, and corrects and improves the thickness based on the classification results and classification confidence;

[0071] Module M9: When processing detection results containing measured sea ice thickness data and whose sea ice classification accuracy meets the preset requirements, the sea ice classification results are first improved based on the integrated detection supported by thickness, including the correction of sea ice type attributes and category confidence for sea ice containing measured thickness data, as well as the correction of sea ice type and thickness information for sea ice without measured thickness data; then, the integrated inversion method based on type support is adopted to use the improved sea ice classification results to correct the sea ice thickness detection results.

[0072] Preferably, the module M7 includes:

[0073] Module M7.1: Match sea ice and thickness measured data to determine whether sea ice contains thickness measured data;

[0074] Module M7.2: If the sea ice contains measured thickness data, process the sea ice containing measured thickness data and modify the sea ice category attributes;

[0075] Module M7.3: If the sea ice does not contain measured thickness data, process the sea ice without measured thickness data and modify the sea ice category attributes and thickness data;

[0076] Module M7.4: Outputs the results of sea ice type and thickness detection.

[0077] Preferably, the module M7.2 includes:

[0078] Assume that the sea ice classification result contains N patches, M pixels, and the Sth i The pixels in the patch are represented by P ij Represents; i=1,2,…,N; j=1,2,…,m i ;m i is the total number of patch pixels, i.e. Σm i =M; the confidence of the category to which the patch belongs is determined by C i The thickness range corresponding to the plaque classification result is expressed as (h imin ,h imax ), the center value is h ia The thickness range corresponding to the classification results of each pixel in the patch is the same as the patch thickness range. The mean patch thickness calculated from the measured thickness data is H ia , with variance σ ia , then the thickness range of this patch is set to (H ia -σ ia ,H ia +σ ia ), by (H imin ,H imax ) indicates that the pixel thickness range obtained by the actual thickness measurement data is (H ijmin ,H ijmax ), the mean is H ija , the confidence threshold is determined by C t If , then the improvement strategy based on the premise that the plaque thickness attribute remains unchanged is:

[0079] like Then the patch category attribute remains unchanged, and the category confidence C i =C t ;

[0080] If (h imin ,h imax )∩(H imin ,Himax )≠0, and h ia ∈(H imin ,H imax ), the patch category attribute remains unchanged, and the category confidence C i =C t ×f1(S i );

[0081] If (h imin ,h imax )∩(H imin ,H imax )≠0, and Then find h ia The ice type corresponding to the thickness range, let the center value of this thickness range be H a , modify the patch category attributes, category confidence C i =C t ×f2(S i );

[0082] If (h imin ,h imax )∩(H imin ,H imax )=0, then find h ia The ice type corresponding to the thickness range, let the center value of this thickness range be H a , modify the patch category attributes, category confidence C i =C t ×f3(S i );

[0083] Among them, f1(S i )、f2(S i ) and f3(S i ) is the plaque category confidence correction function, and the plaque category confidence analysis is calculated using the standard Gaussian model;

[0084] For f1(S i ), let plaque S i In the classification results, the thickness range corresponding to the measured pixel thickness and the classification results satisfies H ija ∈(h imin ,h imax ) has a pixel number of m i1 ,but:

[0085]

[0086] For f2(S i ):

[0087] If h ia <H a ,but

[0088] If h ia >H a ,but

[0089] For f3(S i ):

[0090] If h ia <H a ,but

[0091] If h ia >H a ,but

[0092] This results in corrections and improvements to the sea ice classification patch category confidence and classification results.

[0093] Preferably, the module M7.3 includes:

[0094] The average thickness of all pixels obtained by the sea ice thickness detection method is given by H i Indicates; H ij is the thickness value of the j-th pixel of the patch extracted by the thickness model, then the improvement strategy is:

[0095] For patches without measured thickness data, the classification confidence of the sea ice patches without measured thickness data is corrected based on the measured thickness data. The expression is:

[0096]

[0097] Among them, C′ i represents the confidence of the original category of the patch, C i Represents the modified category confidence, R i It represents the ratio of the number of pixels with correct thickness to the total number of pixels in the thickness measurement data of the same sea ice type;

[0098] If H i ∈(h imin ,h imax ), the plaque category attribute remains unchanged, and the thickness attribute remains unchanged;

[0099] like And C i <C t , then the thickness attribute remains unchanged, find h ia Ice type corresponding to the thickness range, modify the patch category attributes, category confidence C i =C t ;

[0100] like And Ci ≥C t , then the plaque category attribute remains unchanged, and the thickness attribute is determined as:

[0101] The pixels in the patch satisfy H ij ∈(h imin ,h imax ), then the pixel thickness value H ij No modification is required;

[0102] if Then the pixel thickness value H ij =(h imin +h imax ) / 2;

[0103] The patch thickness is the average thickness of all pixels in the patch.

[0104] Preferably, the module M8 includes:

[0105] Module M8.1: Compare the category and thickness attributes of the sea ice patch to determine whether the thickness of the sea ice patch is within the thickness range corresponding to its category; if the thickness of the sea ice patch is within the thickness range corresponding to its category, directly output the sea ice type and thickness detection results; if the thickness of the sea ice patch is not within the thickness range corresponding to its category, calculate the confidence level C of the sea ice patch category. i , and read the sea ice classification confidence threshold C t ;

[0106] Assume that the sea ice classification results contain N patches, and the Sth i The category confidence of each patch is given by C i Indicates that the confidence threshold is represented by C t Indicates, i=1,2,…,N; when the sea ice classification accuracy meets the preset requirements, the confidence threshold C t = 0.5, that is, the sea ice classification result is trusted, and the thickness range corresponding to the patch classification result is expressed as (h imin ,h imax ); plaque S i It is composed of multiple pixels, and its ice thickness is regarded as the average thickness of all pixels in the patch obtained by the sea ice thickness detection method, which is represented by H i express;

[0107] Module M8.2: Comparing confidence levels of sea ice patch categories C i and the confidence threshold C for sea ice classification t If the sea ice patch category confidence level C i >Sea ice classification confidence threshold C t , then modify the sea ice patch thickness attribute, and the sea ice patch category attribute remains unchanged; if the sea ice patch category confidence C i <Sea ice classification confidence threshold Ct , then the sea ice patch category attribute is modified, and the sea ice patch thickness attribute remains unchanged;

[0108] Find the sea ice type corresponding to the thickness of the sea ice patch and modify the sea ice patch type attribute; according to the positive correlation between the backscatter coefficient and the sea ice thickness, i The backscatter coefficients of all pixels in the image are analyzed by histogram, and the patch S i The minimum backscatter coefficient of the middle pixel corresponds to the patch S i The minimum thickness of the category, the maximum backscatter coefficient corresponds to the patch S i The maximum thickness of the category to which it belongs is calculated, and the corresponding relationship between the quantitative pixel backscatter coefficient and the sea ice thickness is established. i In the backscatter coefficient histogram of all pixels, the thickness corresponding to the backscatter coefficient with the highest frequency is regarded as the ice thickness of the patch, thereby realizing the ice thickness of the patch S. i Correction of sea ice thickness attributes; using the correspondence between sea ice type and thickness, find H i The sea ice type corresponding to the thickness range is used as the category attribute of the patch;

[0109] Module M8.3: If H i ∈(h imin ,h imax ), that is, plaque S i The average thickness H i Within the thickness range corresponding to the ice type to which the patch belongs, the category attribute and thickness attribute of the patch remain unchanged, and the sea ice type and thickness detection results are directly output.

[0110] Compared with the prior art, the present invention has the following beneficial effects:

[0111] 1. This invention adopts an improved strategy based on type support for integrated detection, combining sea ice thickness detection and sea ice classification results for integrated detection. This can improve the accuracy of sea ice thickness detection and provide feedback to the correction of sea ice classification results, ultimately obtaining a result map that includes both sea ice type and thickness.

[0112] 2. This invention uses an improved strategy based on thickness support integrated detection to correct the sea ice thickness detection and sea ice classification results, and feeds back into the sea ice classification and thickness detection results, ultimately obtaining a fused result map that includes both sea ice type and thickness.

[0113] 3. The present invention uses high-resolution polarimetric SAR data to simultaneously perform sea ice classification and thickness detection inversion, constructing an integrated inversion method for sea ice classification and thickness detection, further improving the accuracy of sea ice classification and thickness detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0114] Other features, objects and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments with reference to the following drawings:

[0115] Figure 1 It is a schematic flow chart of the working method of the present invention;

[0116] Figure 2 A detailed flow chart of the integrated detection method based on type support in the present invention;

[0117] Figure 3 This is a detailed flow chart of the integrated detection method based on thickness support in the present invention. DETAILED DESCRIPTION

[0118] The present invention will be described in detail below with reference to specific embodiments. The following examples will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those skilled in the art, several changes and improvements can be made without departing from the scope of the present invention. These all fall within the scope of protection of the present invention.

[0119] Example

[0120] The present invention provides an integrated inversion method for sea ice classification and thickness detection based on high-resolution polarimetric SAR. Figure 1 Shown, including:

[0121] Step S1: input high-resolution polarimetric SAR image;

[0122] Step S2: Classify sea ice using SAR high-resolution features and assign category attributes;

[0123] Step S3: Detecting sea ice thickness using SAR polarization characteristics and assigning thickness attributes;

[0124] Step S4: Merge sea ice patch category and thickness attributes;

[0125] Step S5: Calculate the confidence of sea ice classification;

[0126] Step S6: inputting measured sea ice thickness data;

[0127] Step S7: Processing the detection results with measured sea ice thickness data, and correcting and improving the classification results based on the thickness data;

[0128] Step S8: Processing detection results without measured sea ice thickness data but with high sea ice classification accuracy, and correcting and improving the thickness based on the classification results;

[0129] Step S9: When processing detection results with measured sea ice thickness data and the sea ice classification accuracy is high, first use the integrated detection based on thickness support to improve the overall classification accuracy, and then use the integrated detection based on type support to improve the sea ice thickness detection results.

[0130] like Figure 2 If the processed SAR image contains the detection result of the measured data of sea ice thickness, step S7 is executed, including:

[0131] Step S7.1: Matching sea ice patches with measured ice thickness data;

[0132] Step S7.2: Determine whether the sea ice patch contains measured ice thickness data;

[0133] Step S7.3: If the sea ice patch contains measured ice thickness data, process the sea ice patch containing the measured data and modify the category attribute of the sea ice patch;

[0134] Step S7.4: If the sea ice patch does not contain measured ice thickness data, process the sea ice patch without measured data to modify the category attributes and thickness data of the sea ice patch;

[0135] Step S7.5: Output the results of sea ice type and thickness detection.

[0136] If the processed SAR image has the detection result of the measured sea ice thickness data and the sea ice classification accuracy is low, then executing step S7.3 includes:

[0137] Assume that the sea ice classification result contains N patches, M pixels, and the Sth i The pixels in a patch (i=1,2,…,N) are represented by P ij Indicates that j = 1, 2, ..., m i , m i is the total number of patch pixels, i.e. Σm i =M, the confidence of the category to which the patch belongs is given by C i The thickness range corresponding to the plaque classification result is expressed as (h imin ,h imax ), the center value is h ia The thickness range corresponding to the classification results of each pixel in the patch is the same as the patch thickness range. The mean patch thickness calculated from the measured data is H ia , with variance σ ia , then the thickness range of this patch is set to (H ia -σ ia ,H ia +σ ia ), by (H imin ,H imax ) represents; the pixel thickness range obtained by the measured data is (H ijmin ,Hijmax ), the mean is H ija , the confidence threshold is determined by C t If , then the improvement strategy based on the premise that the plaque thickness attribute remains unchanged is:

[0138] (i)If Then the patch category attribute remains unchanged, and the category confidence C i =C t .

[0139] (ii) If (h imin ,h imax )∩(H imin ,H imax )≠0, and h ia ∈(H imin ,H imax ), the patch category attribute remains unchanged, and the category confidence C i =C t ×f1(S i ).

[0140] (iii) If (h imin ,h imax )∩(H imin ,H imax )≠0, and Then find h ia Which ice type falls into the corresponding thickness range? Let the center value of this thickness range be H a , modify the patch category attributes, category confidence C i =C t ×f2(S i ).

[0141] (iv) If (h imin ,h imax )∩(H imin ,H imax )=0, then find h ia Which ice type falls into the corresponding thickness range? Let the center value of this thickness range be H a , modify the patch category attributes, category confidence C i =C t ×f3(S i ).

[0142] Among them, f1(S i )、f2(S i ) and f3(S i ) is the plaque category confidence correction function. This patent still uses the standard Gaussian model to calculate the plaque category confidence analysis.

[0143] For f1(S i ), let plaque Si In the classification results, the thickness range corresponding to the measured pixel thickness and the classification results satisfies H ija ∈(h imin ,h imax ) has a pixel number of m i1 ,but:

[0144]

[0145] For f2(S i ):

[0146]

[0147] For f3(S i ):

[0148]

[0149] This can be used to correct and improve the confidence level of sea ice classification patch categories and the classification results.

[0150] If the processed SAR image does not contain any sea ice thickness measurement data, then execute step S7.4:

[0151] Similar to the case of patches with measured data, patch S i The thickness is the average thickness of all pixels obtained by the sea ice thickness detection method, which is given by H i Indicates that H ij is the thickness value of the j-th pixel of the patch extracted by the thickness model, then the improvement strategy is:

[0152] (i) Modify the confidence level of patches that do not contain measured sea ice thickness data.

[0153] Since sea ice thickness is the result extracted using measured data and models, it cannot be guaranteed to be accurate. Therefore, for patches that do not contain measured sea ice thickness data, the classification confidence of the sea ice patches without measured data needs to be corrected based on the measured ice thickness data. The correction method is shown in formula (4).

[0154]

[0155] In formula (4), C′ i represents the confidence of the original category of the patch, C i Represents the modified category confidence, R i It represents the ratio of the number of pixels with correct thickness to the total number of pixels in the measured thickness data for the same sea ice type.

[0156] (ii) If H i ∈(h imin ,h imax), the plaque category attribute remains unchanged, and the thickness attribute remains unchanged.

[0157] (iii) If And C i <C t , then the thickness attribute remains unchanged, find h ia According to the thickness range of ice type, the patch category attributes are modified, and the category confidence C is determined. i =C t .

[0158] (iv) If And C i ≥C t , then the plaque category attribute remains unchanged, and the determination of the thickness attribute is divided into two cases:

[0159] The pixels in the patch satisfy H ij ∈(h imin ,h imax ), then the pixel thickness value H ij No modification is required;

[0160] if Then the pixel thickness value H ij =(h imin +h imax ) / 2.

[0161] The patch thickness is the average thickness of all pixels in the patch.

[0162] like Figure 3 If the processed SAR image contains the detection result of the measured data of sea ice thickness, then step S8 is executed:

[0163] Step S8.1: Compare the sea ice patch categories and thickness attributes;

[0164] Step S8.2: Determine whether the thickness of the sea ice patch is within the thickness range corresponding to its category;

[0165] Step S8.3: If the thickness of the sea ice patch is within the thickness range corresponding to its category, directly output the sea ice type and thickness detection results;

[0166] Step S8.4: If the thickness of the sea ice patch is not within the thickness range corresponding to its category, calculate the confidence level C of the sea ice patch category. i , and read the sea ice classification confidence threshold C t ;

[0167] Step S8.5: Compare sea ice patch class confidence C i and the confidence threshold C for sea ice classification t size;

[0168] Step S8.6: Sea ice patch category confidence C i >Sea ice classification confidence threshold C t , modify the sea ice patch thickness attribute, and the sea ice patch category attribute remains unchanged;

[0169] Step S8.7: Sea ice patch category confidence C i <Sea ice classification confidence threshold C t , modify the sea ice patch category attribute, and the sea ice patch thickness attribute remains unchanged;

[0170] Step S8.8: Output the results of sea ice type and thickness detection.

[0171] If the thickness of the sea ice patch is not within the thickness range corresponding to its category, execute step S8.4:

[0172] Step S8.4.1 Assume that the sea ice classification results contain N patches, and the Sth i The category confidence of each patch (i=1,2,…,N) is given by C i Indicates that the confidence threshold is represented by C t express.

[0173] Step S8.4.2 When the sea ice classification accuracy is relatively high, the confidence threshold can be set to a relatively small value, such as C t = 0.5, that is, trust the sea ice classification results. The thickness range corresponding to the patch classification results is expressed as (h imin ,h imax ).

[0174] Step S8.4.3 Plaque S i It is composed of many pixels, and its ice thickness is regarded as the average thickness of all pixels in the patch obtained by the sea ice thickness detection method, which is represented by H i express.

[0175] If the confidence level of sea ice patch category C i ≥Sea ice classification confidence threshold C t , modify the sea ice patch thickness attribute and execute step S8.6:

[0176] Step S8.6.1: Find the sea ice type corresponding to the sea ice patch thickness;

[0177] Step S8.6.2: Modify the sea ice patch type attribute. When the patch type is determined, only the range of sea ice thickness for that patch can be given. In order to be able to give a definite value of the patch thickness, we use the property of positive correlation between backscatter coefficient and sea ice thickness to modify the patch S. i The backscatter coefficients of all pixels in the image are analyzed by histogram. Assuming that the patch S i The minimum backscatter coefficient of the middle pixel corresponds to the patch S iThe minimum thickness of the category, the maximum backscatter coefficient corresponds to the patch S i The maximum thickness of the category, the backscatter coefficient and the sea ice thickness satisfy the linear relationship, so that the quantitative correspondence between the pixel backscatter coefficient and the sea ice thickness can be established. i In the backscatter coefficient histogram of all pixels, the thickness corresponding to the backscatter coefficient with the highest frequency can be regarded as the ice thickness of the patch. i Correction to sea ice thickness property.

[0178] If the confidence level of sea ice patch category C i <Sea ice classification confidence threshold C t , modify the sea ice patch thickness attribute, and execute step S8.7 including:

[0179] Step S8.7.1: Establish a quantitative relationship between the backscatter coefficient of the pixels in the patch and the sea ice thickness;

[0180] Step S8.7.2: Calculate the backscatter coefficient histogram for all pixels in the sea ice patch;

[0181] Step S8.7.3: Select the thickness corresponding to the highest frequency in the backscatter coefficient histogram;

[0182] Step S8.7.4: Modify the sea ice patch thickness attribute. Using the correspondence between sea ice type and thickness, find H i If a patch falls within the thickness range corresponding to a certain sea ice type, the category attribute of the patch will be changed to that ice type.

[0183] When processing detection results with measured sea ice thickness data and the sea ice classification accuracy is high, execute step S9:

[0184] Step S9.1: Improve the sea ice classification results based on the integrated detection of thickness support, including revising the sea ice type attributes and category confidence for sea ice with measured ice thickness data, and revising the sea ice type and thickness information for sea ice without measured ice thickness data;

[0185] Step S9.2: Use the integrated inversion method based on type support and the improved sea ice classification results to correct the sea ice thickness detection results.

[0186] Example 2

[0187] The present invention provides an integrated inversion system for sea ice classification and thickness detection based on high-resolution polarimetric SAR, comprising:

[0188] Module M1: Acquire high-resolution polarimetric SAR images;

[0189] Module M2: Based on high-resolution polarimetric SAR images, classify sea ice using SAR high-resolution characteristics and assign category attributes;

[0190] Module M3: Detect sea ice thickness using SAR polarization characteristics and assign thickness attributes based on high-resolution polarimetric SAR images.

[0191] Module M4: Merge sea ice category attributes and thickness attributes;

[0192] Module M5: Calculating sea ice classification confidence;

[0193] Module M6: Obtaining measured data on sea ice thickness;

[0194] Module M7: Processing detection results containing measured sea ice thickness data, and correcting and improving classification results based on the measured thickness data;

[0195] Module M8: Processes detection results that do not contain measured sea ice thickness data but whose sea ice classification accuracy meets the preset requirements, and corrects and improves the thickness based on the classification results and classification confidence;

[0196] Module M9: When processing detection results containing measured sea ice thickness data and whose sea ice classification accuracy meets the preset requirements, the sea ice classification results are first improved based on the integrated detection supported by thickness, including the correction of sea ice type attributes and category confidence for sea ice containing measured thickness data, as well as the correction of sea ice type and thickness information for sea ice without measured thickness data; then, the integrated inversion method based on type support is adopted to use the improved sea ice classification results to correct the sea ice thickness detection results.

[0197] The module M7 includes:

[0198] Module M7.1: Match sea ice and thickness measured data to determine whether sea ice contains thickness measured data;

[0199] Module M7.2: If the sea ice contains measured thickness data, process the sea ice containing measured thickness data and modify the sea ice category attributes;

[0200] Module M7.3: If the sea ice does not contain measured thickness data, process the sea ice without measured thickness data and modify the sea ice category attributes and thickness data;

[0201] Module M7.4: Outputs the results of sea ice type and thickness detection.

[0202] The module M7.2 includes:

[0203] Assume that the sea ice classification result contains N patches, M pixels, and the Sth i The pixels in the patch are represented by P ijRepresents; i=1,2,…,N; j=1,2,…,m i ;m i is the total number of patch pixels, i.e. Σm i =M; the confidence of the category to which the patch belongs is determined by C i The thickness range corresponding to the plaque classification result is expressed as (h imin ,h imax ), the center value is h ia The thickness range corresponding to the classification results of each pixel in the patch is the same as the patch thickness range. The mean patch thickness calculated from the measured thickness data is H ia , with variance σ ia , then the thickness range of this patch is set to (H ia -σ ia ,H ia +σ ia ), by (H imin ,H imax ) indicates that the pixel thickness range obtained by the actual thickness measurement data is (H ijmin ,H ijmax ), the mean is H ija , the confidence threshold is determined by C t If , then the improvement strategy based on the premise that the plaque thickness attribute remains unchanged is:

[0204] like Then the patch category attribute remains unchanged, and the category confidence C i =C t ;

[0205] If (h imin ,h imax )∩(H imin ,H imax )≠0, and h ia ∈(H imin ,H imax ), the patch category attribute remains unchanged, and the category confidence C i =C t ×f1(S i );

[0206] If (h imin ,h imax )∩(H imin ,H imax )≠0, and Then find h ia The ice type corresponding to the thickness range, let the center value of this thickness range be H a , modify the patch category attributes, category confidence C i =C t ×f2(S i );

[0207] If (h imin ,h imax )∩(H imin ,H imax )=0, then find h ia The ice type corresponding to the thickness range, let the center value of this thickness range be H a , modify the patch category attributes, category confidence C i =C t ×f3(S i );

[0208] Among them, f1(S i )、f2(S i ) and f3(S i ) is the plaque category confidence correction function, and the plaque category confidence analysis is calculated using the standard Gaussian model;

[0209] For f1(S i ), let plaque S i In the classification results, the thickness range corresponding to the measured pixel thickness and the classification results satisfies H ija ∈(h imin ,h imax ) has a pixel number of m i1 ,but:

[0210]

[0211] For f2(S i ):

[0212] If h ia <H a ,but

[0213] If h ia >H a ,but

[0214] For f3(S i ):

[0215] If h ia <H a ,but

[0216] If h ia >H a ,but

[0217] This results in corrections and improvements to the sea ice classification patch category confidence and classification results.

[0218] The module M7.3 includes:

[0219] The average thickness of all pixels obtained by the sea ice thickness detection method is given by H i Indicates; H ij is the thickness value of the j-th pixel of the patch extracted by the thickness model, then the improvement strategy is:

[0220] For patches without measured thickness data, the classification confidence of the sea ice patches without measured thickness data is corrected based on the measured thickness data. The expression is:

[0221]

[0222] Among them, C′ i represents the confidence of the original category of the patch, C i Represents the modified category confidence, R i It represents the ratio of the number of pixels with correct thickness to the total number of pixels in the thickness measurement data of the same sea ice type;

[0223] If H i ∈(h imin ,h imax ), the plaque category attribute remains unchanged, and the thickness attribute remains unchanged;

[0224] like And C i <C t , then the thickness attribute remains unchanged, find h ia Ice type corresponding to the thickness range, modify the patch category attributes, category confidence C i =C t ;

[0225] like And C i ≥C t , then the plaque category attribute remains unchanged, and the thickness attribute is determined as:

[0226] The pixels in the patch satisfy H ij ∈(h imin ,h imax ), then the pixel thickness value H ij No modification is required;

[0227] if Then the pixel thickness value H ij =(h imin +h imax ) / 2;

[0228] The patch thickness is the average thickness of all pixels in the patch.

[0229] The module M8 includes:

[0230] Module M8.1: Compare the category and thickness attributes of the sea ice patch to determine whether the thickness of the sea ice patch is within the thickness range corresponding to its category; if the thickness of the sea ice patch is within the thickness range corresponding to its category, directly output the sea ice type and thickness detection results; if the thickness of the sea ice patch is not within the thickness range corresponding to its category, calculate the confidence level C of the sea ice patch category. i , and read the sea ice classification confidence threshold C t ;

[0231] Assume that the sea ice classification results contain N patches, and the Sth i The category confidence of each patch is given by C i Indicates that the confidence threshold is represented by C t Indicates, i=1,2,…,N; when the sea ice classification accuracy meets the preset requirements, the confidence threshold C t = 0.5, that is, the sea ice classification result is trusted, and the thickness range corresponding to the patch classification result is expressed as (h imin ,h imax ); plaque S i It is composed of multiple pixels, and its ice thickness is regarded as the average thickness of all pixels in the patch obtained by the sea ice thickness detection method, which is represented by H i express;

[0232] Module M8.2: Comparing confidence levels of sea ice patch categories C i and the confidence threshold C for sea ice classification t If the sea ice patch category confidence level C i >Sea ice classification confidence threshold C t , then modify the sea ice patch thickness attribute, and the sea ice patch category attribute remains unchanged; if the sea ice patch category confidence C i <Sea ice classification confidence threshold C t , then the sea ice patch category attribute is modified, and the sea ice patch thickness attribute remains unchanged;

[0233] Find the sea ice type corresponding to the thickness of the sea ice patch and modify the sea ice patch type attribute; according to the positive correlation between the backscatter coefficient and the sea ice thickness, i The backscatter coefficients of all pixels in the image are analyzed by histogram, and the patch S i The minimum backscatter coefficient of the middle pixel corresponds to the patch S i The minimum thickness of the category, the maximum backscatter coefficient corresponds to the patch S i The maximum thickness of the category to which it belongs is calculated, and the corresponding relationship between the quantitative pixel backscatter coefficient and the sea ice thickness is established. i In the backscatter coefficient histogram of all pixels, the thickness corresponding to the backscatter coefficient with the highest frequency is regarded as the ice thickness of the patch, thereby realizing the ice thickness of the patch S. iCorrection of sea ice thickness attributes; using the correspondence between sea ice type and thickness, find H i The sea ice type corresponding to the thickness range is used as the category attribute of the patch;

[0234] Module M8.3: If H i ∈(h imin ,h imax ), that is, plaque S i The average thickness H i Within the thickness range corresponding to the ice type to which the patch belongs, the category attribute and thickness attribute of the patch remain unchanged, and the sea ice type and thickness detection results are directly output.

[0235] Those skilled in the art will appreciate that, in addition to implementing the system, device, and various modules provided by the present invention in purely computer-readable program code, it is entirely possible to implement the same program in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, embedded microcontrollers, and the like by logically programming the method steps. Therefore, the system, device, and various modules provided by the present invention can be considered a hardware component, and the modules included therein for implementing various programs can also be considered structures within the hardware component; the modules for implementing various functions can also be considered both software programs for implementing the method and structures within the hardware component.

[0236] The above describes specific embodiments of the present invention. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art may make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. The embodiments of this application and the features in the embodiments may be combined with each other in any manner unless there is a conflict.

Claims

1. An integrated inversion method for sea ice classification and thickness detection based on high-resolution polarimetric SAR, characterized in that: include: Step S1: Acquire high-resolution polarimetric SAR images; Step S2: for high-resolution polarimetric SAR images, sea ice classification is performed using SAR high-resolution characteristics and category attributes are assigned; Step S3: for high-resolution polarimetric SAR images, the SAR polarimetric characteristics are used to detect the sea ice thickness and assign thickness attributes; Step S4: merging sea ice category attributes and thickness attributes; Step S5: Calculate the confidence of sea ice classification; Step S6: Obtaining measured sea ice thickness data; Step S7: Processing the detection results containing the measured data of sea ice thickness, and correcting and improving the classification results based on the measured data of thickness; Step S8: Processing detection results that do not contain measured sea ice thickness data but whose sea ice classification accuracy meets the preset requirements, and correcting and improving the thickness based on the classification results and classification confidence; Step S9: When processing detection results containing measured sea ice thickness data and whose sea ice classification accuracy meets preset requirements, first improve the sea ice classification results based on the integrated detection supported by thickness, including correcting the sea ice type attributes and category confidence of the sea ice containing measured thickness data, and correcting the sea ice type and thickness information of the sea ice without measured thickness data; then adopt the integrated inversion method based on type support and use the improved sea ice classification results to correct the sea ice thickness detection results.

2. The integrated inversion method for sea ice classification and thickness detection based on high-resolution polarimetric SAR according to claim 1 is characterized in that: The step S7 comprises: Step S7.1: Matching sea ice with measured thickness data to determine whether sea ice contains measured thickness data; Step S7.2: If the sea ice contains measured thickness data, process the sea ice containing the measured thickness data and modify the category attribute of the sea ice; Step S7.3: If the sea ice does not contain measured thickness data, process the sea ice without measured thickness data and modify the sea ice category attributes and thickness data; Step S7.4: Output the results of sea ice type and thickness detection.

3. The integrated inversion method for sea ice classification and thickness detection based on high-resolution polarimetric SAR according to claim 2 is characterized in that: The step S7.2 includes: Assume that the sea ice classification result contains N patches, M pixels, and the Sth i The pixels in the patch are represented by P ij Represents; i=1,2,…,N; j=1,2,…,m i ;m i is the total number of patch pixels, i.e. Σm i =M; the confidence of the category to which the patch belongs is determined by C i The thickness range corresponding to the plaque classification result is expressed as (h imin ,h imax ), the center value is h ia The thickness range corresponding to the classification results of each pixel in the patch is the same as the patch thickness range. The mean patch thickness calculated from the measured thickness data is H ia , with variance σ ia , then the thickness range of this patch is set to (H ia -σ ia ,H ia +σ ia ), by (H imin ,H imax ) indicates that the pixel thickness range obtained by the actual thickness measurement data is (H ijmin ,H ijmax ), the mean is H ija , the confidence threshold is determined by C t If , then the improvement strategy based on the premise that the plaque thickness attribute remains unchanged is: like Then the patch category attribute remains unchanged, and the category confidence C i =C t ; If (h imin ,h imax )∩(H imin ,H imax )≠0, and h ia ∈(H imin ,H imax ), the patch category attribute remains unchanged, and the category confidence C i =C t ×f1(S i ); If (h imin ,h imax )∩(H imin ,H imax )≠0, and Then find h ia The ice type corresponding to the thickness range, let the center value of this thickness range be H a , modify the patch category attributes, category confidence C i =C t ×f2(S i ); If (h imin ,h imax )∩(H imin ,H imax )=0, then find h ia The ice type corresponding to the thickness range, let the center value of this thickness range be H a , modify the patch category attributes, category confidence C i =C t ×f3(S i ); Among them, f1(S i )、f2(S i ) and f3(S i ) is the plaque category confidence correction function, and the plaque category confidence analysis is calculated using the standard Gaussian model; For f1(S i ), let plaque S i In the classification results, the thickness range corresponding to the measured pixel thickness and the classification results satisfies H ija ∈(h imin ,h imax ) has a pixel number of m i1 ,but: For f2(S i ): If h ia <H a ,but If h ia >H a ,but For f3(S i ): If h ia <H a ,but If h ia >H a ,but This results in corrections and improvements to the sea ice classification patch category confidence and classification results.

4. The integrated inversion method for sea ice classification and thickness detection based on high-resolution polarimetric SAR according to claim 3 is characterized in that: The step S7.3 includes: The average thickness of all pixels obtained by the sea ice thickness detection method is given by H i Indicates; H ij is the thickness value of the j-th pixel of the patch extracted by the thickness model, then the improvement strategy is: For patches without measured thickness data, the classification confidence of the sea ice patches without measured thickness data is corrected based on the measured thickness data. The expression is: Among them, C′ i represents the confidence of the original category of the patch, C i Represents the modified category confidence, R i It represents the ratio of the number of pixels with correct thickness to the total number of pixels in the thickness measurement data of the same sea ice type; If H i ∈(h imin ,h imax ), the plaque category attribute remains unchanged, and the thickness attribute remains unchanged; like And C i <C t , then the thickness attribute remains unchanged, find h ia Ice type corresponding to the thickness range, modify the patch category attributes, category confidence C i =C t ; like And C i ≥C t , then the plaque category attribute remains unchanged, and the thickness attribute is determined as: The pixels in the patch satisfy H ij ∈(h imin ,h imax ), then the pixel thickness value H ij No modification is required; if Then the pixel thickness value H ij =(h imin +h imax ) / 2; The patch thickness is the average thickness of all pixels in the patch.

5. The integrated inversion method for sea ice classification and thickness detection based on high-resolution polarimetric SAR according to claim 4 is characterized in that: The step S8 comprises: Step S8.1: Compare the category and thickness attributes of the sea ice patch to determine whether the thickness of the sea ice patch is within the thickness range corresponding to its category; if the thickness of the sea ice patch is within the thickness range corresponding to its category, directly output the sea ice type and thickness detection results; if the thickness of the sea ice patch is not within the thickness range corresponding to its category, calculate the sea ice patch category confidence C i , and read the sea ice classification confidence threshold C t ; Assume that the sea ice classification results contain N patches, and the Sth i The category confidence of each patch is given by C i Indicates that the confidence threshold is represented by C t Indicates, i=1,2,…,N; when the sea ice classification accuracy meets the preset requirements, the confidence threshold C t = 0.5, that is, the sea ice classification result is trusted, and the thickness range corresponding to the patch classification result is expressed as (h imin ,h imax ); plaque S i It is composed of multiple pixels, and its ice thickness is regarded as the average thickness of all pixels in the patch obtained by the sea ice thickness detection method, which is represented by H i express; Step S8.2: Compare sea ice patch class confidence C i and the confidence threshold C for sea ice classification t If the sea ice patch category confidence level C i >Sea ice classification confidence threshold C t , then modify the sea ice patch thickness attribute, and the sea ice patch category attribute remains unchanged; if the sea ice patch category confidence C i <Sea ice classification confidence threshold C t , then the sea ice patch category attribute is modified, and the sea ice patch thickness attribute remains unchanged; Find the sea ice type corresponding to the thickness of the sea ice patch and modify the sea ice patch type attribute; according to the positive correlation between the backscatter coefficient and the sea ice thickness, i The backscatter coefficients of all pixels in the image are analyzed by histogram, and the patch S i The minimum backscatter coefficient of the middle pixel corresponds to the patch S i The minimum thickness of the category, the maximum backscatter coefficient corresponds to the patch S i The maximum thickness of the category to which it belongs is calculated, and the corresponding relationship between the quantitative pixel backscatter coefficient and the sea ice thickness is established. i In the backscatter coefficient histogram of all pixels, the thickness corresponding to the backscatter coefficient with the highest frequency is regarded as the ice thickness of the patch, thereby realizing the ice thickness of the patch S. i Correction of sea ice thickness attributes; using the correspondence between sea ice type and thickness, find H i The sea ice type corresponding to the thickness range is used as the category attribute of the patch; Step S8.3: If H i ∈(h imin ,h imax ), that is, plaque S i The average thickness H i Within the thickness range corresponding to the ice type to which the patch belongs, the category attribute and thickness attribute of the patch remain unchanged, and the sea ice type and thickness detection results are directly output.

6. An integrated inversion system for sea ice classification and thickness detection based on high-resolution polarimetric SAR, characterized in that: include: Module M1: Acquire high-resolution polarimetric SAR images; Module M2: Based on high-resolution polarimetric SAR images, classify sea ice using SAR high-resolution characteristics and assign category attributes; Module M3: Detect sea ice thickness using SAR polarization characteristics and assign thickness attributes based on high-resolution polarimetric SAR images. Module M4: Merge sea ice category attributes and thickness attributes; Module M5: Calculating sea ice classification confidence; Module M6: Obtaining measured data on sea ice thickness; Module M7: Processing detection results containing measured sea ice thickness data, and correcting and improving classification results based on the measured thickness data; Module M8: Processes detection results that do not contain measured sea ice thickness data but whose sea ice classification accuracy meets the preset requirements, and corrects and improves the thickness based on the classification results and classification confidence; Module M9: When processing detection results containing measured sea ice thickness data and whose sea ice classification accuracy meets the preset requirements, the sea ice classification results are first improved based on the integrated detection supported by thickness, including the correction of sea ice type attributes and category confidence for sea ice containing measured thickness data, as well as the correction of sea ice type and thickness information for sea ice without measured thickness data; then, the integrated inversion method based on type support is adopted to use the improved sea ice classification results to correct the sea ice thickness detection results.

7. The integrated inversion system for sea ice classification and thickness detection based on high-resolution polarimetric SAR according to claim 6 is characterized in that: The module M7 includes: Module M7.1: Match sea ice and thickness measured data to determine whether sea ice contains thickness measured data; Module M7.2: If the sea ice contains measured thickness data, process the sea ice containing measured thickness data and modify the sea ice category attributes; Module M7.3: If the sea ice does not contain measured thickness data, process the sea ice without measured thickness data and modify the sea ice category attributes and thickness data; Module M7.4: Outputs the results of sea ice type and thickness detection.

8. The integrated inversion system for sea ice classification and thickness detection based on high-resolution polarimetric SAR according to claim 7 is characterized in that: The module M7.2 includes: Assume that the sea ice classification result contains N patches, M pixels, and the Sth i The pixels in the patch are represented by P ij Represents; i=1,2,…,N; j=1,2,…,m i ;m i is the total number of patch pixels, i.e. Σm i =M; the confidence of the category to which the patch belongs is determined by C i The thickness range corresponding to the plaque classification result is expressed as (h imin ,h imax ), the center value is h ia The thickness range corresponding to the classification results of each pixel in the patch is the same as the patch thickness range. The mean patch thickness calculated from the measured thickness data is H ia , with variance σ ia , then the thickness range of this patch is set to (H ia -σ ia ,H ia +σ ia ), by (H imin ,H imax ) indicates that the pixel thickness range obtained by the actual thickness measurement data is (H ijmin ,H ijmax ), the mean is H ija , the confidence threshold is determined by C t If , then the improvement strategy based on the premise that the plaque thickness attribute remains unchanged is: like Then the patch category attribute remains unchanged, and the category confidence C i =C t ; If (h imin ,h imax )∩(H imin ,H imax )≠0, and h ia ∈(H imin ,H imax ), the patch category attribute remains unchanged, and the category confidence C i =C t ×f1(S i ); If (h imin ,h imax )∩(H imin ,H imax )≠0, and Then find h ia The ice type corresponding to the thickness range, let the center value of this thickness range be H a , modify the patch category attributes, category confidence C i =C t ×f2(S i ); If (h imin ,h imax )∩(H imin ,H imax )=0, then find h ia The ice type corresponding to the thickness range, let the center value of this thickness range be H a , modify the patch category attributes, category confidence C i =C t ×f3(S i ); Among them, f1(S i )、f2(S i ) and f3(S i ) is the plaque category confidence correction function, and the plaque category confidence analysis is calculated using the standard Gaussian model; For f1(S i ), let plaque S i In the classification results, the thickness range corresponding to the measured pixel thickness and the classification results satisfies H ija ∈(h imin ,h imax ) has a pixel number of m i1 ,but: For f2(S i ): If h ia <H a ,but If h ia >H a ,but For f3(S i ): If h ia <H a ,but If h ia >H a ,but This results in corrections and improvements to the sea ice classification patch category confidence and classification results.

9. The integrated inversion system for sea ice classification and thickness detection based on high-resolution polarimetric SAR according to claim 8 is characterized in that: The module M7.3 includes: The average thickness of all pixels obtained by the sea ice thickness detection method is given by H i Indicates; H ij is the thickness value of the j-th pixel of the patch extracted by the thickness model, then the improvement strategy is: For patches without measured thickness data, the classification confidence of the sea ice patches without measured thickness data is corrected based on the measured thickness data. The expression is: Among them, C′ i represents the confidence of the original category of the patch, C i Represents the modified category confidence, R i It represents the ratio of the number of pixels with correct thickness to the total number of pixels in the thickness measurement data of the same sea ice type; If H i ∈(h imin ,h imax ), the plaque category attribute remains unchanged, and the thickness attribute remains unchanged; like And C i <C t , then the thickness attribute remains unchanged, find h ia Ice type corresponding to the thickness range, modify the patch category attributes, category confidence C i =C t ; like And C i ≥C t , then the plaque category attribute remains unchanged, and the thickness attribute is determined as: The pixels in the patch satisfy H ij ∈(h imin ,h imax ), then the pixel thickness value H ij No modification is required; if Then the pixel thickness value H ij =(h imin +h imax ) / 2; The patch thickness is the average thickness of all pixels in the patch.

10. The integrated inversion system for sea ice classification and thickness detection based on high-resolution polarimetric SAR according to claim 9 is characterized in that: The module M8 includes: Module M8.1: Compare the category and thickness attributes of the sea ice patch to determine whether the thickness of the sea ice patch is within the thickness range corresponding to its category; if the thickness of the sea ice patch is within the thickness range corresponding to its category, directly output the sea ice type and thickness detection results; if the thickness of the sea ice patch is not within the thickness range corresponding to its category, calculate the confidence level C of the sea ice patch category. i , and read the sea ice classification confidence threshold C t ; Assume that the sea ice classification results contain N patches, and the Sth i The category confidence of each patch is given by C i Indicates that the confidence threshold is represented by C t Indicates, i=1,2,…,N; when the sea ice classification accuracy meets the preset requirements, the confidence threshold C t = 0.5, that is, the sea ice classification result is trusted, and the thickness range corresponding to the patch classification result is expressed as (h imin ,h imax ); plaque S i It is composed of multiple pixels, and its ice thickness is regarded as the average thickness of all pixels in the patch obtained by the sea ice thickness detection method, which is represented by H i express; Module M8.2: Comparing confidence levels of sea ice patch categories C i and the confidence threshold C for sea ice classification t If the sea ice patch category confidence level C i >Sea ice classification confidence threshold C t , then modify the sea ice patch thickness attribute, and the sea ice patch category attribute remains unchanged; if the sea ice patch category confidence C i <Sea ice classification confidence threshold C t , then the sea ice patch category attribute is modified, and the sea ice patch thickness attribute remains unchanged; Find the sea ice type corresponding to the thickness of the sea ice patch and modify the sea ice patch type attribute; according to the positive correlation between the backscatter coefficient and the sea ice thickness, i The backscatter coefficients of all pixels in the image are analyzed by histogram, and the patch S i The minimum backscatter coefficient of the middle pixel corresponds to the patch S i The minimum thickness of the category, the maximum backscatter coefficient corresponds to the patch S i The maximum thickness of the category to which it belongs is calculated, and the corresponding relationship between the quantitative pixel backscatter coefficient and the sea ice thickness is established. i In the backscatter coefficient histogram of all pixels, the thickness corresponding to the backscatter coefficient with the highest frequency is regarded as the ice thickness of the patch, thereby realizing the ice thickness of the patch S. i Correction of sea ice thickness attributes; using the correspondence between sea ice type and thickness, find H i The sea ice type corresponding to the thickness range is used as the category attribute of the patch; Module M8.3: If H i ∈(h imin ,h imax ), that is, plaque S i The average thickness H i Within the thickness range corresponding to the ice type to which the patch belongs, the category attribute and thickness attribute of the patch remain unchanged, and the sea ice type and thickness detection results are directly output.

Images