A tidal flat extraction method and device based on satellite remote sensing data

Abstract

The application provides a tidal flat extraction method and device based on satellite remote sensing data, and belongs to the technical field of marine remote sensing monitoring. The method comprises the following steps: acquiring satellite remote sensing data; subtracting the green light band reflectivity data from the red light band reflectivity data to obtain a first difference value, adding the green light band reflectivity data and the red light band reflectivity data to obtain a first sum value, and dividing the first difference value by the first sum value to obtain a first ratio value; subtracting the red light band reflectivity data from the near-infrared band reflectivity data to obtain a second difference value, adding the red light band reflectivity data and the near-infrared band reflectivity data to obtain a second sum value, and dividing the second difference value by the second sum value to obtain a second ratio value; subtracting the second ratio value from the first ratio value to obtain a tidal flat extraction index value; and comparing the tidal flat extraction index value with a preset tidal flat identification threshold range. The method and device can effectively distinguish the tidal flat from spectral similar features, and have cross-satellite sensor and regional application adaptability.

Description

Technical Field

[0001] This application relates to the field of marine remote sensing monitoring technology, and in particular to a method and apparatus for extracting tidal flats based on satellite remote sensing data. Background Technology

[0002] As an important ecological and land resource in the nearshore coastal zone, tidal flats possess multiple values, including ecological protection, resource development, and coastal defense. Accurate monitoring of their spatial distribution and area changes is crucial for marine ecological protection, coastal zone planning and management, and marine economic development. Satellite remote sensing technology, with its advantages of wide coverage, short revisit cycles, and convenient data acquisition, has become a core technology for tidal flat extraction and dynamic monitoring. By analyzing the spectral characteristics of satellite remote sensing data to classify land features, it can quickly obtain large-scale tidal flat distribution information. Compared to traditional field surveys, this significantly improves the efficiency and timeliness of tidal flat monitoring, making it the mainstream technology in the field of coastal zone remote sensing monitoring.

[0003] Currently, methods for extracting tidal flats based on satellite remote sensing data mostly rely on traditional remote sensing indices such as the Normalized Difference Vegetation Index (NDVI) and the Normalized Difference Water Index (NDWI), or are achieved through single-band threshold segmentation and multi-band combination classification. Specifically, the NDVI primarily highlights vegetation features through a combination of near-infrared and red bands, achieving a preliminary distinction between tidal flats and vegetation; the NDWI amplifies the spectral differences between water and land through a combination of green and near-infrared bands, enabling the identification of tidal flats and water bodies; single-band threshold segmentation often selects the near-infrared band, utilizing the reflectance differences between tidal flats and water bodies for simple classification. Furthermore, existing methods are mostly developed based on remote sensing data from specific satellite sensors, such as the Sentinel-2 MSI sensor and the Landsat OLI sensor. Extraction schemes need to be designed separately for the spectral response characteristics of different sensors, and some methods also adjust band combinations and threshold parameters based on the land cover characteristics of the study area.

[0004] However, existing tidal flat extraction technologies still face many unresolved problems, making it difficult to meet the demands for high-precision and universal tidal flat extraction. Firstly, traditional remote sensing indices lack specificity. NDVI is only advantageous for vegetation identification, with weak ability to distinguish between tidal flats and rocks or bare land. While NDWI can distinguish between water and land, its accuracy in identifying spectral differences between tidal flats and shallow water areas or wetland vegetation is low, easily leading to misidentification or omission of tidal flats and similar spectral features, making it difficult to guarantee the accuracy of the extraction results. Secondly, existing methods have poor cross-satellite sensor adaptability. Different satellite sensors have different spectral response ranges, resolutions, and radiometric accuracy, resulting in deviations in reflectance acquisition results for the same feature. Existing extraction methods are often designed for single-sensor calibration. Directly applying these methods to other sensors leads to a significant drop in extraction accuracy, requiring repeated parameter adjustments, which is cumbersome and lacks versatility. Thirdly, their regional adaptability is limited. Existing methods often determine band combinations and threshold parameters based on the land cover characteristics of specific study areas. Different geographical regions exhibit significant differences in climate, topography, and land cover composition, resulting in variations in the spectral characteristics of tidal flats and confounding features. Directly applying existing methods leads to extraction errors due to insufficient band discrimination and poor threshold adaptability, making accurate tidal flat extraction across regions difficult. Therefore, there is an urgent need to develop a tidal flat extraction method that can effectively distinguish tidal flats from spectrally similar features and is adaptable to cross-satellite sensor and cross-regional applications to improve the accuracy, versatility, and stability of tidal flat extraction. Summary of the Invention

[0005] In view of this, this application provides a method and apparatus for tidal flat extraction based on satellite remote sensing data, which can effectively distinguish tidal flats from spectrally similar land features, and has cross-satellite sensor adaptability and cross-regional application adaptability, thereby improving the versatility and identification accuracy of tidal flat extraction.

[0006] Specifically, this application is implemented through the following technical solution:

[0007] The first aspect of this application provides a method for extracting tidal flats based on satellite remote sensing data, the method comprising:

[0008] Acquire satellite remote sensing data of the target area, and extract green light band reflectance data, red light band reflectance data and near-infrared band reflectance data from the satellite remote sensing data;

[0009] The first difference is obtained by subtracting the green band reflectance data from the red band reflectance data, the first sum is obtained by adding the green band reflectance data to the red band reflectance data, and the first ratio is obtained by dividing the first difference by the first sum.

[0010] The second difference is obtained by subtracting the red band reflectance data from the near-infrared band reflectance data, the second sum is obtained by adding the red band reflectance data to the near-infrared band reflectance data, and the second ratio is obtained by dividing the second difference by the second sum.

[0011] Subtracting the second ratio from the first ratio yields the tidal flat extraction index value;

[0012] The mudflat extraction index value is compared with a preset mudflat identification threshold range. If the mudflat extraction index value is within the threshold range, the corresponding pixel is determined to be a mudflat; if the mudflat extraction index value is outside the threshold range, the corresponding pixel is determined to be a non-mudflat.

[0013] A second aspect of this application provides a device for extracting tidal flats based on satellite remote sensing data, the device comprising an acquisition module, a processing module, and a judgment module;

[0014] The acquisition module is used to acquire satellite remote sensing data of the target area and extract green light band reflectance data, red light band reflectance data and near-infrared band reflectance data from the satellite remote sensing data;

[0015] The processing module is used to subtract the green light band reflectance data from the red light band reflectance data to obtain a first difference, add the green light band reflectance data to the red light band reflectance data to obtain a first sum, and divide the first difference by the first sum to obtain a first ratio.

[0016] The processing module is further configured to subtract the red band reflectance data from the near-infrared band reflectance data to obtain a second difference, add the red band reflectance data to the near-infrared band reflectance data to obtain a second sum, and divide the second difference by the second sum to obtain a second ratio.

[0017] The processing module is further configured to subtract the second ratio from the first ratio to obtain the tidal flat extraction index value;

[0018] The judgment module is used to compare the mudflat extraction index value with a preset mudflat identification threshold range. If the mudflat extraction index value is within the threshold range, the corresponding pixel is determined to be a mudflat; if the mudflat extraction index value is outside the threshold range, the corresponding pixel is determined to be a non-mudflat.

[0019] The method and apparatus for tidal flat extraction based on satellite remote sensing data provided in this application extract reflectance data of specific bands and construct a tidal flat extraction index through a series of normalization operations. Then, it combines threshold range comparison to realize tidal flat identification. Compared with the existing technology, it effectively distinguishes tidal flats from spectrally similar land features, and greatly improves the accuracy and stability of tidal flat extraction. Specifically, by extracting reflectance data from the green, red, and near-infrared bands as the basic data for tidal flat extraction, core spectral information capable of distinguishing tidal flats from easily confused features such as water, vegetation, and rocks was determined, avoiding interference introduced by invalid bands. Secondly, the first ratio was obtained by calculating the difference and sum of the green and red band reflectances. Normalization was then used to eliminate the influence of external environmental factors such as illumination and topography on the reflectance data, while simultaneously amplifying the spectral differences between tidal flats and vegetation / rocks, highlighting their spectral distinguishing characteristics and making the spectral feature differences between tidal flats and these types of confusing land features more significant. The second ratio was obtained by performing the same normalization operation on the red and near-infrared band reflectances, further offsetting external interference such as atmospheric residues, amplifying the spectral differences between tidal flats and water, and highlighting the differences between tidal flats and these types of confusing land features. The distinguishing characteristics of confounding features in water areas compensate for the inadequacy of single bands or ratios in differentiating water bodies. Then, the first ratio is subtracted from the second ratio to obtain the tidal flat extraction index value, integrating the advantages of both ratios. This achieves comprehensive differentiation of tidal flats from vegetation, rocks, and all major confounding features in water bodies. Simultaneously, it amplifies the overall spectral differences between tidal flats and non-tidal flat features, creating an independent and concentrated numerical range for the index values ​​corresponding to tidal flats. This effectively avoids misidentification and missed identification problems caused by the limitations of single ratios. Finally, the tidal flat extraction index value is compared with a preset tidal flat identification threshold range to determine the pixel type. This transforms abstract spectral differences into quantifiable identification standards, enabling rapid and accurate classification of every satellite remote sensing pixel within the target area, thus achieving clear differentiation of tidal flats in the target area. Attached Figure Description

[0020] Figure 1 A flowchart of an embodiment of the method for extracting tidal flats based on satellite remote sensing data provided in this application;

[0021] Figure 2 This is a schematic diagram of the second embodiment of the tidal flat extraction device based on satellite remote sensing data provided in this application. Detailed Implementation

[0022] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application.

[0023] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used herein are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0024] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0025] The following specific embodiments are given to illustrate the technical solution of this application in detail.

[0026] Example 1

[0027] Figure 1 This is a flowchart of an embodiment of the method for extracting tidal flats based on satellite remote sensing data provided in this application. Please refer to... Figure 1 The method provided in this embodiment may include:

[0028] S101. Obtain satellite remote sensing data of the target area, and extract green light band reflectance data, red light band reflectance data and near-infrared band reflectance data from the satellite remote sensing data.

[0029] It should be noted that satellite remote sensing data contains spectral information of ground objects within the target area. Different ground objects have different abilities to reflect and absorb electromagnetic waves of different wavelengths due to differences in their material composition and structural characteristics.

[0030] Specifically, the satellite remote sensing data used in this application are L1B and L2A level data products from the CZI sensor of the HY-1C / D satellite. The HY-1C / D satellite is part of my country's ocean color series satellites. Its onboard Coastal Zone Imager (CZI) can acquire remote sensing data in four bands: blue, green, red, and near-infrared, with a spatial resolution of 50 meters and a revisit period of 3 days, suitable for dynamic monitoring of nearshore coastal areas. L1B level data requires radiometric calibration and Rayleigh scattering correction. L2A level data products are standard data products that have already undergone radiometric calibration and Rayleigh scattering correction, and users can directly obtain and use them without additional preprocessing. Furthermore, to reduce the impact of aerosol scattering on nearshore marine remote sensing data, L1B and L2A level data products obtained under clear or dry weather conditions can be selected as satellite remote sensing data for tidal flat extraction calculations.

[0031] After acquiring the aforementioned satellite remote sensing data, reflectance data in the green band, red band, and near-infrared band were extracted. Specifically, the green band corresponds to band 2 (wavelength range 0.517-0.598 μm) of the HY-1C / D satellite's CZI sensor, the red band corresponds to band 3 (wavelength range 0.608-0.690 μm), and the near-infrared band corresponds to band 4 (wavelength range 0.761-0.891 μm).

[0032] Specifically, acquiring satellite remote sensing data of the target area and extracting green band reflectance data, red band reflectance data, and near-infrared band reflectance data from the satellite remote sensing data includes:

[0033] (1) Collect spectral reflectance data of tidal flats, water bodies, vegetation and rock features in the target area.

[0034] Spectral reflectance data refers to the ratio of electromagnetic wave energy reflected by a ground object to the incident electromagnetic wave energy within a specific wavelength range. The value typically ranges from 0 to 1; a higher value indicates a stronger reflectivity of the ground object for that wavelength. Specifically, the spectral reflectance data collected in this step includes ground object spectral reflectance data obtained through inversion from L1B and L2A level satellite remote sensing data. This data covers the entire target area, ensuring data comprehensiveness. It also includes spectral reflectance data obtained through ground-based measurements. Sampling points are set up in different typical areas within the target area, such as high-coverage mudflats, mudflat-water interfaces, and mudflat-vegetation interfaces. Spectrometer measurements are performed in the field to ensure data accuracy. This cross-validation of two types of data avoids biases caused by a single data source.

[0035] Because tidal flats are often adjacent to or mixed with water bodies, vegetation, rocks, and other landforms in the natural environment, and the above four types of landforms are the most common and most easily confused with the spectral characteristics of tidal flats, spectral reflectance data of these four types of landforms are collected. During the collection process, it is necessary to ensure the representativeness of the sampling points, taking into account landforms under different terrains, coverage, and humidity conditions within the target area. The number of sampling points for each landform type should not be less than a specified number, such as 30, to avoid spectral data deviations caused by single sampling points or single environmental conditions.

[0036] (2) Based on the spectral reflectance data of various land features, draw corresponding spectral curves and compare and analyze the spectral differences between tidal flats and other land features.

[0037] A spectral curve is a curve that visualizes the spectral reflectance data of various land features, with wavelength as the x-axis and spectral reflectance as the y-axis. It reflects the spectral response characteristics of land features within different wavelength ranges and can quickly identify the spectral differences between tidal flats and other mixed land features. Specifically, when plotting spectral curves, professional spectral analysis software (such as ENVI or Origin) can be used. The collected satellite-derived and ground-measured spectral reflectance data can be imported and plotted according to land feature type to ensure the accuracy and readability of the curves. The wavelength range of each band should also be labeled.

[0038] After plotting the spectral curves, a detailed comparative analysis was conducted on the spectral curves of the tidal flats, water bodies, vegetation, and rocks to clarify the spectral differences among the four types of land cover in different wavelength bands. Specifically, water bodies exhibit extremely low reflectance in the near-infrared band, with the curve showing a clear downward trend, while the reflectance of the tidal flats in the near-infrared band is significantly higher than that of water bodies, showing a clear difference. Vegetation exhibits a weak reflection peak in the green band and an absorption trough in the red band, which is significantly different from the reflectance fluctuation characteristics of the tidal flats in the green and red bands. The spectral reflectance of rocks is generally high, and the reflectance changes relatively smoothly in the green and red bands, unlike the spectral fluctuation characteristics of the tidal flats. Through this comparative analysis, the wavelength range that can effectively distinguish the tidal flats from other confused land cover can be preliminarily determined.

[0039] (3) Based on the comparison results of spectral differences, the validity of each band of the satellite remote sensing data is judged, and the green band, red band and near-infrared band are identified as the effective bands for extraction from the tidal flats.

[0040] Band validity determination refers to assessing, based on the preceding spectral difference analysis, whether each band of the satellite remote sensing data can effectively distinguish mudflats from other confounding features, and selecting bands with high contribution to mudflat identification and low interference as valid bands. Combining the above spectral difference comparison results, the green band can effectively distinguish mudflats from vegetation, the red band can effectively distinguish mudflats from rocks, and the near-infrared band can quickly distinguish mudflats from water bodies. Therefore, the green, red, and near-infrared bands were ultimately determined as valid bands for mudflat extraction.

[0041] S102. Subtract the green light band reflectance data from the red light band reflectance data to obtain a first difference value, add the green light band reflectance data to the red light band reflectance data to obtain a first sum value, and divide the first difference value by the first sum value to obtain a first ratio value.

[0042] It should be noted that in coastal areas, vegetation exhibits reflective properties in the green light band and absorptive properties in the red light band, resulting in a significant difference between the reflectance of green and red light. Rocks have relatively high reflectance in both the green and red light bands, and the variation is gradual, with the difference between the two being relatively stable. In contrast, the reflectance distribution patterns of tidal flats in the green and red light bands are distinguishable from those of the two types of land cover mentioned above.

[0043] By first performing a difference operation and then a sum operation, and finally quotienting the two to obtain the first ratio, the influence of environmental factors such as light and topography can be eliminated while amplifying the spectral differences between mudflats and vegetation / rocks, reducing the interference of external conditions on the stability of the index. Specifically, the first ratio can be based on the difference in reflectance between the green and red light bands, highlighting the spectral distinguishing features between mudflats and vegetation / rocks. Through this normalized calculation, the calculation results can be kept within a relatively stable numerical range, avoiding the problem of difficulty in determining the subsequent identification threshold due to excessively large differences in absolute reflectance values.

[0044] The first difference is used to quantify the degree of difference in reflectance between the green and red light bands, reflecting the difference in spectral response of different land features in these two bands. The first difference for vegetation is positive and relatively large (high green light reflectance and strong red light absorption), the first difference for rocks is positive and relatively flat (high reflectance in both bands with small differences), and the first difference for tidal flats is between the two and relatively stable. The spectral characteristics of the three types of land features can be preliminarily distinguished by the first difference.

[0045] The first sum is used to offset interference from external environmental factors such as light intensity and terrain. Since there are differences in light intensity and terrain slope in different areas, the absolute value of reflectance of ground objects will fluctuate. For example, the absolute value of reflectance of the same ground object on sunny days and cloudy days is quite different. By adding the reflectance of the two bands to obtain the first sum, this fluctuation of absolute value can be normalized and offset, ensuring that the subsequent ratio calculation results are not affected by the external environment, and improving the stability and universality of the first ratio.

[0046] Specifically, the first difference = green band reflectance data - red band reflectance data; the first sum = green band reflectance data + red band reflectance data; the first ratio = first difference ÷ first sum.

[0047] S103. Subtract the red band reflectance data from the near-infrared band reflectance data to obtain a second difference value, add the red band reflectance data to the near-infrared band reflectance data to obtain a second sum value, and divide the second difference value by the second sum value to obtain a second ratio value.

[0048] Similarly, in coastal areas, water exhibits strong absorption characteristics and extremely low reflectivity in the near-infrared band, while showing weak reflection in the red band, resulting in a significant difference between the reflectivity of the red and near-infrared bands. Tidal flats exhibit stable reflectivity in both the red and near-infrared bands, with a moderate and gradual difference in reflectivity. Rocks have relatively high reflectivity in both the red and near-infrared bands, with a relatively stable difference, and the difference range is clearly distinguishable from that of tidal flats.

[0049] By first performing difference calculations and then summing them, the second ratio is obtained by quotienting the two. This ratio can offset the effects of sunlight and atmospheric residues while amplifying the spectral differences between tidal flats, water bodies, and rocks. It compensates for the insufficient ability of the first ratio to distinguish between tidal flats and water bodies, further enhancing the distinguishability of spectral features of ground features. Specifically, the second ratio can highlight the spectral distinguishing features between tidal flats and water bodies based on the difference in reflectance between the red and near-infrared bands. At the same time, it helps to distinguish between tidal flats and rocks. Through normalized calculations, the results are kept within a stable numerical range, complementing the first ratio and providing support for the subsequent construction of a complete tidal flat extraction index.

[0050] The second difference is used to quantify the degree of reflectance difference between the red light band and the near-infrared band, reflecting the difference in spectral response of different land features in these two bands. Specifically, the second difference for water bodies is positive and significantly larger (slightly higher red light reflectance and extremely strong near-infrared absorption), the second difference for rocks is positive and relatively flat (high reflectance in both bands with small differences), and the second difference for tidal flats is between the two and stable. The second difference can be used to initially distinguish tidal flats from water bodies and rocks, and can especially quickly identify significant differences between water bodies and tidal flats.

[0051] The second sum is used to offset the effects of external interference factors such as light intensity and atmospheric residue, and to avoid fluctuations in the absolute value of ground reflectance due to changes in light intensity at different times and near-shore aerosol residue. The second sum is obtained by adding the reflectance of the two bands, which normalizes and offsets the fluctuations in the absolute value of reflectance, ensuring the stability and universality of the second ratio, so that the second ratio of the same type of ground feature remains relatively consistent under different environmental conditions.

[0052] Specifically, the second difference = red band reflectance data - near-infrared band reflectance data; the second sum = red band reflectance data + near-infrared band reflectance data; the second ratio = second difference ÷ second sum.

[0053] S104. Subtract the second ratio from the first ratio to obtain the tidal flat extraction index value.

[0054] It should be noted that the first ratio calculated by S102 primarily highlights the spectral distinguishing features between tidal flats and vegetation / rocks, effectively differentiating tidal flats from the two main types of terrestrial features, but its ability to distinguish tidal flats from water is relatively weak. The second ratio calculated by S103 primarily highlights the spectral distinguishing features between tidal flats and water, effectively differentiating tidal flats from water—the main body of terrestrial features—and also assisting in distinguishing tidal flats from rocks, but its effect on distinguishing tidal flats from vegetation is not as good as the first ratio. Therefore, subtracting the second ratio from the first ratio to obtain the tidal flat extraction index value integrates the advantages of both ratios, achieving comprehensive differentiation of various types of features, while further amplifying the spectral differences between tidal flats and all features, thus improving the accuracy and stability of tidal flat identification.

[0055] Specifically, the difference calculation between the first and second ratios superimposes the distinguishing features between tidal flats and vegetation / rocks, as well as the distinguishing features between tidal flats and water bodies. This results in the index values ​​corresponding to tidal flats forming a concentrated numerical range, while the index values ​​corresponding to vegetation, rocks, and water bodies form numerical ranges significantly separated from those of tidal flats. This effectively avoids misidentification and omissions caused by the limited distinguishing ability of a single ratio. Furthermore, this difference calculation can further offset the influence of external interference factors such as sunlight, topography, and atmospheric residues, ensuring the stability and universality of the extracted index values ​​from tidal flats.

[0056] The formula for calculating the mudflat extraction index value is based on the optimization and improvement of existing mature indices. Through scientific comparison and screening, it ensures that the accuracy of the index in identifying mudflats is superior to existing conventional indices. Specifically, it includes:

[0057] (1) Obtain the normalized water index and normalized vegetation index, and calculate the identification accuracy of the normalized water index and normalized vegetation index at the sampling point of the tidal flat, respectively, as the first control accuracy and the second control accuracy.

[0058] The Normalized Difference Water Index (NDWI) is a mature existing index for identifying aquatic features. Its advantage lies in distinguishing between water and land features, but its ability to distinguish between mudflats and vegetation / rock is relatively weak. The Normalized Difference Vegetation Index (NDVI) is a mature existing index for vegetation identification. Its advantage lies in distinguishing between vegetation and non-vegetation features, but its effectiveness in distinguishing between mudflats and water / rock is poor. This step uses these two existing conventional indices as benchmarks. By calculating their identification accuracy at mudflat sampling points, the shortcomings of existing indices in mudflat identification are identified. At the same time, clear performance standards are set for the selection of subsequent candidate indices, ensuring that the final constructed mudflat extraction index outperforms existing conventional indices in terms of identification accuracy. The identification accuracy is calculated by statistically analyzing the ratio of the number of sampling points correctly identified as mudflats by the existing index to the total number of all mudflat sampling points. A higher ratio indicates a better identification effect of the index on mudflats.

[0059] (2) Based on the formula configuration of the normalized water index and the normalized vegetation index, the band combination used in them is replaced with different combinations of green light band, red light band and near-infrared band to generate multiple candidate indices.

[0060] The core formula configuration of both the Normalized Difference in Reflectance (NDR) and the Normalized Difference in Vegetation (NDV) is a normalized form of the difference in reflectance between two bands divided by the sum of reflectance in the two bands. This configuration can effectively offset external interference and amplify the spectral differences between the target feature and the confused features. However, the band combinations used in the two existing indices (the NDR mostly uses green and near-infrared bands, and the NDR mostly uses near-infrared and red bands) have not been optimized for the spectral characteristics of tidal flats and the three types of confused features, resulting in limited accuracy in tidal flat identification. Therefore, this step, based on this configuration, replaces the band combinations with different combinations of the already determined effective bands (green, red, and near-infrared bands) to generate multiple candidate indices. This retains the advantages of the normalized configuration while also being specifically optimized based on the spectral characteristics of tidal flats, ensuring that all candidate indices have the potential for tidal flat identification.

[0061] (3) Calculate the recognition accuracy of each candidate index at the sampling point on the tidal flat.

[0062] Specifically, all candidate indices are applied to both mudflat sampling points and non-mudflat sampling points. The number of mudflat sampling points that each candidate index can correctly identify and the number of non-mudflat sampling points that it can correctly exclude are calculated. The identification accuracy of each candidate index is then calculated, and the calculation logic is consistent with that of the accuracy calculation logic of the control index in step (1). In this way, the performance of all candidate indices is initially quantitatively evaluated, and candidate indices with good mudflat identification effects can be screened out, while indices with low identification accuracy or that cannot meet the requirements for mudflat extraction are eliminated.

[0063] (4) Compare the recognition accuracy of each candidate index with the first control accuracy and the second control accuracy to select the candidate index with the highest recognition accuracy.

[0064] Specifically, by comparing each candidate index pairwise, the recognition accuracy is compared with the first control accuracy (normalized water index accuracy) and the second control accuracy (normalized vegetation index accuracy). Candidate indices with lower accuracy than the control indices are eliminated, and the one with the highest accuracy is selected from the remaining candidate indices. This ensures that the tidal flat recognition performance of the obtained candidate indices reaches the optimal level, which can effectively solve the problems of low accuracy and easy misidentification of existing tidal flat recognition indices.

[0065] (5) The calculation formula of the candidate index with the highest identification accuracy is determined as the mudflat extraction index.

[0066] After comparison and screening through the above steps, the candidate index with the highest accuracy rate has been identified. It possesses the advantages of integrating two types of ratios, comprehensively distinguishing tidal flats from all confusing features, and offsetting external interference, thus meeting the needs of tidal flat extraction in nearshore coastal areas. Its calculation formula has been established as the calculation formula for the tidal flat extraction index, providing a foundation for subsequent tidal flat identification.

[0067] S105. The mudflat extraction index value is compared with a preset mudflat identification threshold range. If the mudflat extraction index value is within the threshold range, the corresponding pixel is determined to be a mudflat; if the mudflat extraction index value is outside the threshold range, the corresponding pixel is determined to be a non-mudflat.

[0068] It should be noted that the obtained tidal flat extraction index value integrates the advantages of the first and second ratios, amplifying the spectral differences between tidal flats and the three types of mixed features: vegetation, rocks, and water. This results in a significantly separated numerical distribution characteristic between the index values ​​corresponding to tidal flat and non-tidal flat features. Therefore, by setting a clear threshold range for tidal flat identification, this spectral difference is transformed into a quantifiable identification standard, enabling accurate classification of each satellite remote sensing pixel within the target area, clearly distinguishing between tidal flat and non-tidal flat features, and completing the tidal flat extraction of the target area.

[0069] Specifically, satellite remote sensing data records ground feature information in pixels. Each pixel corresponds to a certain area of ​​the land surface within the target region. By comparing the mudflat extraction index value of each pixel with a preset mudflat identification threshold range, it is possible to quickly determine whether the land surface corresponding to that pixel is a mudflat.

[0070] Before comparing the extracted tidal flat index value with a preset tidal flat identification threshold range, the preset tidal flat identification threshold range needs to be calibrated to ensure its rationality and adaptability. Specifically, this includes:

[0071] (1) Select sampling points on the mudflats and non-mudflats within the target area.

[0072] Specifically, following the sampling specifications established in S101, tidal flat sampling points must cover tidal flat areas with different coverage, humidity, and topography within the target area. Non-tidal flat sampling points should focus on vegetation, rock, and water areas whose spectral characteristics are easily confused with those of tidal flats, ensuring a reasonable number of sampling points for each land cover type. This selection method ensures that the sampling points comprehensively reflect the spectral characteristics of both tidal flats and non-tidal flats within the target area, avoiding threshold calibration deviations caused by single sampling points or incomplete coverage.

[0073] (2) Substitute the tidal flat sampling points and non-tidal flat sampling points into the tidal flat extraction index to calculate the tidal flat extraction index value corresponding to all sampling points.

[0074] Specifically, the green, red, and near-infrared reflectance data of all mudflat and non-mudflat sampling points selected in step (1) are substituted into the mudflat extraction index calculation formula determined in S104, and the mudflat extraction index value corresponding to each sampling point is calculated one by one. At the same time, the calculation results are sorted and verified, and outliers are removed, such as index values ​​that deviate from the normal distribution range due to sampling errors or calculation errors, so as to ensure the accuracy and effectiveness of the index values ​​of all sampling points.

[0075] (3) Determine the candidate threshold search interval based on the distribution range of the index values ​​extracted from the mudflats at all sampling points.

[0076] The purpose of determining the candidate threshold search interval is to define a reasonable numerical range and narrow down the subsequent threshold screening range to avoid inefficiency or threshold deviation caused by blind searching. Specifically, statistical analysis is performed on all the index values ​​of the sampling points after step (2) to obtain the overall distribution range of the index values ​​of the tidal flat sampling points and the overall distribution range of the index values ​​of the non-tidal flat sampling points. Based on the intersection or adjacent intervals of the two types of land cover index values, and combined with the dispersion of the data distribution, the candidate threshold search interval is determined. Usually, the minimum value of all sampling point index values ​​can be used as the lower endpoint of the search interval, and the maximum value can be used as the upper endpoint of the search interval to ensure that the candidate threshold search interval can completely cover the distribution range of the index values ​​of tidal flat and non-tidal flat land cover.

[0077] (4) Divide the candidate threshold search interval into multiple consecutive sub-intervals.

[0078] Based on the numerical span of the candidate threshold search interval and the actual extraction accuracy requirements, the candidate threshold search interval is divided into multiple continuous and non-overlapping sub-intervals using an equidistant division method. The number of sub-intervals can be set according to the actual situation, and the numerical span of each sub-interval remains consistent.

[0079] (5) Using the endpoint value of each sub-interval as a candidate threshold, pixels with a mudflat extraction index value greater than or equal to the candidate threshold are identified as mudflats, and pixels with a mudflat extraction index value less than the candidate threshold are identified as non-mudflats. The number of correctly identified mudflat sampling points and the number of incorrectly identified non-mudflat sampling points under each candidate threshold are counted.

[0080] Specifically, the upper and lower endpoints of each sub-interval are used as candidate thresholds, and each is substituted into a preset judgment rule (index value ≥ candidate threshold is mudflat, < candidate threshold is non-mudflat) to identify and judge all sampling points. At the same time, the number of mudflat sampling points that are correctly identified as mudflats (correct identification number) and the number of non-mudflat sampling points that are incorrectly identified as mudflats (incorrect identification number) are counted. By analyzing these two data, the recognition accuracy of the candidate threshold can be reflected. The more correct identifications and the fewer incorrect identifications, the better the recognition effect of the candidate threshold.

[0081] (6) Calculate the recognition accuracy corresponding to each candidate threshold based on the number of correct recognitions and the number of incorrect recognitions.

[0082] Specifically, the formula for calculating the recognition accuracy is: Recognition accuracy = Number of correctly identified points ÷ (Total number of mudflat sampling points + Total number of non-mudflat sampling points - Number of incorrectly identified points) × 100%. This formula comprehensively considers the correct identification of mudflat sampling points and the incorrect identification of non-mudflat sampling points, and can fully reflect the overall recognition performance of the candidate threshold. The higher the accuracy value, the better the candidate threshold is at distinguishing between mudflat and non-mudflat features, and the more suitable it is as a threshold for mudflat identification.

[0083] (7) The two endpoint values ​​of the sub-interval where the candidate threshold with the highest recognition accuracy is located are determined as the upper and lower endpoint values ​​of the preset mudflat recognition threshold range.

[0084] Among all the candidate threshold recognition accuracy rates calculated in step (6), the candidate threshold with the highest recognition accuracy is selected, and the sub-interval containing this candidate threshold is found. The two endpoints of this sub-interval are respectively determined as the upper and lower endpoints of the preset tidal flat recognition threshold range. In this way, using the endpoints of the sub-interval as the threshold range can both take into account the recognition accuracy of the optimal threshold and provide a tolerance space for small fluctuations in the tidal flat extraction index value, ensuring that even if there are slight fluctuations in the index value, the pixel type can be accurately determined, further improving the reliability and stability of tidal flat extraction.

[0085] It should be noted that the aforementioned threshold range for tidal flat identification is based on L2A level data from the HY-1C / D satellite CZI sensor and is adapted to the spectral response characteristics of this sensor model. Since different satellite sensor models have differences in spectral response range, spectral resolution, and radiometric accuracy, slight deviations may occur in the acquisition of reflectance data for the same ground feature. Directly applying the same threshold range for tidal flat identification would lead to a decrease in the accuracy of tidal flat extraction and an increase in the probability of misidentification. Therefore, the preset threshold range for tidal flat identification needs to correspond one-to-one with the satellite sensor model. Different satellite sensor models correspond to different threshold ranges for tidal flat identification to ensure accurate extraction of tidal flats regardless of the satellite sensor model used to acquire remote sensing data. Specifically, this includes:

[0086] (1) Obtain the satellite sensor model information corresponding to the satellite remote sensing data to be processed.

[0087] Specifically, satellite remote sensing data to be processed usually carries metadata information such as sensor model and satellite platform. The corresponding satellite sensor model information can be directly extracted from the metadata of the satellite remote sensing data. If the metadata does not explicitly indicate the sensor model, the sensor model corresponding to the remote sensing data can be queried through data acquisition channels, such as satellite data publishing platforms, to ensure that the obtained model information is accurate.

[0088] (2) Based on the model information, retrieve the pre-calibrated tidal flat identification threshold range of the corresponding model sensor as the preset tidal flat identification threshold range.

[0089] The threshold range for identifying mudflats, predetermined for the corresponding sensor model, is based on the threshold range of a benchmark satellite sensor and corrected by considering the difference in spectral response between the target sensor and the benchmark sensor. The specific process includes:

[0090] (i) Obtain the spectral response curves of the reference satellite sensor in the green, red and near-infrared bands.

[0091] Specifically, this application selects the HY-1C / D satellite CZI sensor as the reference satellite sensor. The spectral response curve characterizes the sensor's sensitivity to electromagnetic waves at different wavelengths, reflecting its acquisition characteristics across three effective wavelength bands. In practice, standard spectral response curves of the reference satellite sensor in the green, red, and near-infrared bands can be obtained from the official technical manual and spectral database of the satellite sensor.

[0092] (ii) Obtain the spectral response curves of the target satellite sensor in the green, red and near-infrared bands.

[0093] The target satellite sensor is the satellite sensor corresponding to the current remote sensing data to be processed (such as the Sentinel-2 satellite MSI sensor). This step adopts the same acquisition method as step (1) to acquire the spectral response curves of the target satellite sensor in the three effective bands of green light, red light and near infrared, and ensure that the band range and data format of the two curves are consistent.

[0094] (iii) Compare the spectral response curves of the target satellite sensor in each band with the spectral response curves of the reference satellite sensor in the corresponding band, and calculate the spectral response difference within the wavelength overlap range.

[0095] Specifically, since there may be slight differences in the wavelength range of different sensor models, it is first necessary to determine the wavelength overlap range of the target sensor and the reference sensor for the corresponding wavelength bands (green light to green light, red light to red light, near-infrared to near-infrared). Comparison is only performed within the overlap range to avoid calculation errors caused by inconsistent wavelength ranges. During the comparison process, the two spectral response curves are compared band by band to analyze the difference in responsivity between the target sensor and the reference sensor at the same wavelength. Then, the spectral response difference of each band within the wavelength overlap range is calculated. For example, this can be achieved by calculating the root mean square error of the two curves within the wavelength overlap range. The smaller the root mean square error, the smaller the difference in spectral response between the two types of sensors. The spectral response difference reflects the difference in the response sensitivity of the two types of sensors to electromagnetic waves of the same wavelength and band.

[0096] (iiii) Based on the spectral response difference, determine the mudflat identification threshold correction coefficient of the target satellite sensor relative to the reference satellite sensor.

[0097] The difference in spectral response can cause a deviation between the reflectance data of ground features collected by the target sensor and that collected by the reference sensor, thus affecting the calculation results of the tidal flat extraction index. Directly using the threshold range of the reference sensor will lead to identification errors. Therefore, it is necessary to determine a corresponding threshold correction coefficient based on the difference in spectral response to correct the reference threshold range and adapt it to the characteristics of the target sensor. The specific determination process includes:

[0098] ① Align the spectral response curves of the target satellite sensor and the reference satellite sensor in the corresponding bands according to the wavelength horizontal axis to obtain a set of matching points where the vertical axis response rate values ​​of the two curves correspond one-to-one.

[0099] Specifically, with wavelength as the abscissa, the spectral response curves of the target sensor and the reference sensor in the corresponding bands are aligned to ensure that the abscissas (wavelengths) of the two curves are completely consistent. At this time, for each wavelength point on the abscissa, the response rate values ​​of the target sensor and the reference sensor can be obtained, forming a one-to-one matching point. Each matching point contains a wavelength value, a target response rate value, and a reference response rate value. All matching points form a matching point set.

[0100] ② Within the wavelength overlap range, for each matching point, calculate the difference between the target satellite sensor response rate value and the reference satellite sensor response rate value, and use it as the ordinate difference value of the matching point.

[0101] Specifically, for each matching point within the wavelength overlap range, the difference in the vertical coordinate of each matching point is calculated by subtracting the reference response rate from the target response rate value. This value can be positive or negative. A positive value indicates that the response sensitivity of the target sensor at that wavelength point is higher than that of the reference sensor, while a negative value indicates that the response sensitivity of the target sensor is lower than that of the reference sensor. The larger the absolute value of the vertical coordinate difference, the more significant the difference in response between the two types of sensors at that wavelength point.

[0102] ③ Divide the difference in the ordinate of each matching point by the response rate value of the reference satellite sensor at the matching point to obtain the response rate correction percentage of the matching point.

[0103] Specifically, the formula for calculating the response rate correction percentage is: Response rate correction percentage = (vertical axis difference value ÷ reference sensor response rate value) × 100%. This percentage reflects the degree of response rate deviation of the target sensor relative to the reference sensor at this wavelength point. Compared with the simple vertical axis difference value, the correction percentage can eliminate the influence of the reference response rate value itself and more objectively characterize the response difference between the two types of sensors. For example, for the same difference value, the deviation is greater when the reference response rate value is smaller.

[0104] ④ Calculate the arithmetic mean of the response rate correction percentage for all matching points within the wavelength overlap range, and determine the arithmetic mean as the mudflat identification threshold correction coefficient of the target satellite sensor relative to the reference satellite sensor.

[0105] Specifically, the percentage correction for the response rates of all matching points within the wavelength overlap range is summed and divided by the total number of matching points to obtain the arithmetic mean. This mean is the threshold correction coefficient. Using the arithmetic mean as the correction coefficient comprehensively considers the response differences of all wavelength points, avoids correction deviations caused by extreme differences at a single wavelength point, and ensures the rationality and stability of the correction coefficient.

[0106] (iiiii) Using the tidal flat identification threshold range corresponding to the reference satellite sensor as the reference threshold, the threshold correction coefficient is superimposed on the endpoint value of the reference threshold to obtain the tidal flat identification threshold range pre-calibrated by the target satellite sensor.

[0107] Specifically, the threshold range for tidal flat identification corresponding to the benchmark satellite sensor, i.e., the threshold range calibrated in S105, is used. The determined threshold correction coefficient is then superimposed onto the upper and lower endpoints of this benchmark threshold range to obtain the pre-calibrated threshold range for the target satellite sensor. The correction formula is: Target threshold endpoint value = Benchmark threshold endpoint value × (1 + threshold correction coefficient). This correction process allows the benchmark threshold range to be adapted to the spectral response characteristics of the target sensor, offsetting the extraction deviation caused by the response differences between the two types of sensors. This ensures that the remote sensing data acquired by the target sensor, when used for tidal flat identification with the corresponding threshold range, achieves the same extraction accuracy as the benchmark sensor, thereby expanding the application range of the method.

[0108] It should be noted that after achieving cross-satellite sensor adaptation for tidal flat extraction, the application adaptability to different geographical regions must also be considered. Due to differences in climate conditions and topography across different geographical regions, the composition and distribution characteristics of land cover types, as well as the acquisition quality and distribution patterns of satellite remote sensing data, will vary. For example, some coastal areas have high vegetation cover, while others are dominated by rocks and water; this difference in the proportion of land cover area will affect the differentiation effect of the original effective bands. Simultaneously, the number of effective pixels and the spatial uniformity of satellite data will also differ across regions, thus affecting the stability of band effectiveness. Therefore, when applying the described method to tidal flat extraction in different geographical regions, it is necessary to determine whether it is necessary to re-determine the effective bands based on the land cover type composition and satellite data distribution characteristics of the target geographical region. This ensures that the effective bands always adapt to the actual situation of the target region, guaranteeing the accuracy of tidal flat extraction. Specifically, this includes:

[0109] (1) Obtain the distribution data of land cover types in the target geographical area and count the area proportion of the four types of land cover: mudflats, water bodies, vegetation and rocks.

[0110] Specifically, the distribution data of land cover types in the target geographic area can be obtained through existing land use data and high-resolution remote sensing image interpretation results, or supplemented and corrected by combining ground survey data. The area proportions of the four land cover types—tidal flats, water bodies, vegetation, and rocks—are statistically analyzed. The area proportion of a certain land cover type is calculated as: (Total area of ​​that land cover type in the target area ÷ Total area of ​​the target area) × 100%. By clarifying the land cover composition structure of the target area, the distribution proportions of each land cover type can be determined. This provides basic data for subsequent comparative analysis with the original area (the target area used when pre-determining effective bands), and further determines whether differences in land cover composition affect the applicability of the original effective bands.

[0111] (2) Obtain satellite remote sensing data covering the target geographic area and count the number of effective pixels and the uniformity of spatial distribution of each band data.

[0112] Satellite remote sensing data covering the target geographic area will be used, employing the L2A-level atmospheric correction product data described earlier (corresponding to the satellite sensor model to be adapted), and acquiring remote sensing data in the three original effective bands: green, red, and near-infrared. Effective pixels refer to pixels that are unaffected by clouds, fog, shadows, etc., and can accurately reflect the spectral characteristics of ground objects. Satellite data will be preprocessed using professional remote sensing software (such as ENVI) to remove invalid pixels, and the number and spatial uniformity of effective pixels in each band will be statistically analyzed. A higher number of effective pixels and a more uniform spatial distribution indicate better satellite data quality and more stable differentiation of the original effective bands; conversely, insufficient or unevenly distributed effective pixels may affect the effectiveness of the original effective bands.

[0113] (3) Compare the area ratio of land cover types in the target geographic area with the area ratio of the corresponding land cover types in the target area used when the effective band is determined in advance, and calculate the difference value of the area ratio of each type of land cover.

[0114] The target area (original area) used when determining the effective bands in advance, i.e., the area corresponding to the determination of green, red, and near-infrared bands as effective bands in S101, has its land cover type area ratio as the basis for the adaptability of the original effective bands. This step compares the area ratio of the four types of land cover in the target geographical area with the area ratio of the corresponding land cover in the original area one by one. The difference in area ratio of each type of land cover is calculated by using the absolute value of |the percentage of a certain type of land cover in the target area - the percentage of a certain type of land cover in the original area|. This difference reflects the degree of difference in land cover composition between the target area and the original area. The larger the difference, the more significant the difference in land cover composition between the two areas, and the weaker the adaptability of the original effective bands may be.

[0115] (4) Compare the satellite data distribution characteristics of the target geographic area with the satellite data distribution characteristics of the target area used when the effective band is determined in advance, and calculate the distribution similarity.

[0116] Specifically, the satellite data distribution characteristics mainly refer to the number of effective pixels in each effective band and the uniformity of spatial distribution. This step adopts the same comparison logic as step (3) to compare the distribution characteristics of the satellite data in the target area with those in the original area. The similarity between the two is calculated by similarity algorithms, such as cosine similarity and Pearson correlation coefficient. The similarity value is usually between 0 and 1. The closer the value is to 1, the closer the satellite data distribution characteristics of the target area and the original area are, and the stronger the applicability of the original effective bands. The closer the value is to 0, the greater the difference in the distribution characteristics between the two, and the original effective bands may not be able to adapt to the satellite data quality of the target area.

[0117] (5) If the area ratio difference of any land cover type exceeds the preset ratio difference threshold, it is determined that the effective band needs to be re-determined; if the distribution similarity is lower than the preset similarity threshold, it is determined that the effective band needs to be re-determined.

[0118] The preset percentage difference threshold and similarity threshold can be set and fine-tuned according to the accuracy requirements of the actual application scenario. Specifically, if the area percentage difference value of any land cover type exceeds the percentage difference threshold, it indicates that the land cover composition of the target area and the original area are significantly different, and the original effective bands will have a reduced effect on distinguishing land cover in this area. It is necessary to reselect effective bands that are suitable for the land cover characteristics of this area. If the distribution similarity is lower than the similarity threshold, it indicates that the satellite data quality and distribution pattern of the target area are significantly different from those of the original area. The original effective bands cannot be adapted to the satellite data of this area, and it is also necessary to redetermine the effective bands.

[0119] (6) If the area ratio difference of all land cover types does not exceed the preset ratio difference threshold, and the distribution similarity is not lower than the preset similarity threshold, then it is determined that the determined effective bands will continue to be used.

[0120] If this criterion is met, it indicates that the differences in land cover composition and satellite data distribution characteristics between the target area and the original area are small. The existing effective bands can effectively distinguish the mudflats from various mixed land features in the target area and are compatible with the satellite data quality of the target area, eliminating the need to re-determine the effective bands. This criterion logic avoids unnecessary band re-selection, saving computational costs and improving extraction efficiency. It also ensures the stability and accuracy of mudflat extraction, enabling flexible adaptation of this method to different geographical regions and further expanding its application scope.

[0121] It should be noted that while continuing to use the established effective bands indicates that the differences in land cover composition and satellite data distribution characteristics between the target geographic area and the original area are small, and the discriminative power of the effective bands is suitable for the target area, subtle differences in the microenvironment of different geographic areas, such as local humidity, soil composition, and lighting conditions, can lead to a slight shift in the distribution range of the tidal flat extraction index values. Directly using the tidal flat identification threshold calibrated for the original area may result in slight misidentification or omission, affecting extraction accuracy. Therefore, after deciding to continue using the established effective bands, it is necessary to specifically optimize the tidal flat identification threshold based on the actual situation of the target geographic area to ensure that the threshold adapts to the microenvironmental characteristics of the target area, further guaranteeing the accuracy of tidal flat extraction. Specifically, this includes:

[0122] (1) Based on satellite remote sensing data of the target geographic area, select sampling points on tidal flats and sampling points on non-tidal flats.

[0123] Specifically, the sampling in this step follows the sampling requirements outlined above. Furthermore, the tidal flat sampling points must cover tidal flat areas with different microenvironments within the target region. Non-tidal flat sampling points should be selected from vegetation, rock, and water areas whose spectral characteristics are easily confused with those of the tidal flats. All sampling data are derived from satellite remote sensing data covering the target geographical area (corresponding to the compatible sensor model, L1B and L2A level atmospheric correction products).

[0124] (2) Construct a candidate threshold set based on the mudflat extraction index values ​​corresponding to the mudflat sampling points and non-mudflat sampling points.

[0125] Specifically, firstly, the green, red, and near-infrared reflectance data of all mudflat and non-mudflat sampling points selected in step (1) are substituted into the mudflat extraction index calculation formula determined in S104 above, and the mudflat extraction index value corresponding to each sampling point is calculated one by one. At the same time, the calculation results are sorted and verified, and outliers caused by sampling errors and calculation errors are removed. Subsequently, based on the distribution range of mudflat extraction index values ​​of all valid sampling points, possible threshold values ​​are selected to construct a candidate threshold set. The candidate threshold set needs to cover the transition range of index values ​​between mudflat and non-mudflat sampling points, ensuring that it includes the optimal threshold that can accurately distinguish between the two types of land features, without omitting potential optimal values ​​or including obviously unreasonable values.

[0126] (3) Calculate the recognition accuracy of each candidate threshold for the mudflat sampling point and the non-mudflat sampling point.

[0127] This step uses the same calculation logic and formula as in S105 for the recognition accuracy, to quantitatively evaluate each threshold in the candidate threshold set and clarify the recognition performance of each candidate threshold. Specifically, each threshold in the candidate threshold set is substituted into the preset judgment rule one by one to identify and judge all sampling points; at the same time, the number of correctly identified mudflat sampling points and the number of incorrectly identified non-mudflat sampling points corresponding to each candidate threshold are counted. Substituting these numbers into the formula Recognition Accuracy = Number of Correctly Identified Points ÷ (Total Number of Mudflat Sampling Points + Total Number of Non-Mudflat Sampling Points - Number of Incorrectly Identified Points) × 100%, the recognition accuracy of each candidate threshold is calculated. The higher the accuracy value, the better the candidate threshold's ability to distinguish between mudflat and non-mudflat features in the target area, and the stronger its adaptability.

[0128] (4) The candidate threshold with the highest recognition accuracy is determined as the optimized mudflat recognition threshold for the target geographical area.

[0129] Among all the candidate threshold recognition accuracies calculated in step (3), the candidate threshold with the highest recognition accuracy is selected and determined as the optimized tidal flat recognition threshold for the target geographical area. If multiple candidate thresholds have the same recognition accuracy and are all the highest, the threshold located in the middle of the transition range of the index values ​​between tidal flat and non-tidal flat sampling points can be selected to further improve the fault tolerance of the threshold. Through this optimization process, the advantages of the original effective bands are preserved, and the impact of the micro-environmental differences in the target area is offset by targeted threshold adjustments. This ensures that the method can still achieve high tidal flat extraction accuracy when the original effective bands are used in different geographical areas.

[0130] The method provided in this embodiment is based on L2A level data from the HY-1C / D satellite CZI sensor. It collects spectral reflectance data from easily confused land features such as tidal flats, water bodies, vegetation, and rocks, plots spectral curves, and compares and analyzes the spectral differences between these features. This identifies three effective wavelength bands: green, red, and near-infrared, thus establishing a spectral characteristic basis for accurately distinguishing tidal flats from confused land features and avoiding interference from ineffective wavelength bands. Subsequently, a first ratio is obtained by normalizing the green and red wavelengths, and a second ratio is obtained by normalizing the red and near-infrared wavelengths. The difference between the two ratios is then used to construct the tidal flat extraction index. This index integrates the advantages of the first ratio in distinguishing tidal flats from vegetation and rocks, and the advantages of the second ratio in distinguishing tidal flats from water bodies. This achieves comprehensive differentiation of various confusing land features. At the same time, the normalization operation cancels out the influence of external environmental factors such as sunlight, topography, and atmospheric residues, amplifies the spectral differences between tidal flat and non-tidal flat land features, and makes the index values ​​corresponding to tidal flats form a concentrated numerical range that is significantly separated from non-tidal flat land features. This effectively solves the problems of insufficient distinguishing ability of a single index and easy misidentification and omission in the existing technology. Building upon this foundation, this method calibrates the threshold range for tidal flat identification by selecting representative sampling points, determining candidate threshold search intervals, dividing sub-intervals, and statistically analyzing the recognition accuracy of each candidate threshold. This avoids relying on subjective experience, making threshold calibration more objective. Furthermore, the threshold range provides tolerance for minor fluctuations in the index value, further enhancing the reliability of tidal flat identification. Simultaneously, addressing the differences in spectral response characteristics among different satellite sensors, this method compares spectral response curves, calculates spectral response differences, determines threshold correction coefficients, and adjusts the baseline threshold. This achieves a one-to-one correspondence between the tidal flat identification threshold and the satellite sensor model, giving the method good cross-satellite sensor adaptability and solving the problem of existing methods only adapting to a single sensor and experiencing a significant decrease in accuracy when used across different sensors. In addition, this method considers the differences in land cover composition and satellite data distribution characteristics across different geographical regions. By statistically analyzing the area ratio of land cover and calculating distribution similarity, it determines whether to re-determine the effective band. When deciding to continue using the original effective band, it can also optimize the tidal flat identification threshold based on the micro-environmental characteristics of the target area, offsetting the index value distribution shift caused by regional micro-environmental differences, and achieving cross-regional application adaptability of the method.

[0131] Example 2

[0132] Corresponding to the aforementioned embodiment of a method for extracting tidal flats based on satellite remote sensing data, this application also provides an embodiment of a device for extracting tidal flats based on satellite remote sensing data.

[0133] Figure 2 This is a schematic diagram of the second embodiment of the tidal flat extraction device based on satellite remote sensing data provided in this application. Please refer to... Figure 2The device provided in this embodiment includes an acquisition module 210, a processing module 220, and a judgment module 230;

[0134] The acquisition module 210 is used to acquire satellite remote sensing data of the target area and extract green light band reflectance data, red light band reflectance data and near-infrared band reflectance data from the satellite remote sensing data.

[0135] The processing module 220 is used to subtract the green light band reflectance data from the red light band reflectance data to obtain a first difference, add the green light band reflectance data to the red light band reflectance data to obtain a first sum, and divide the first difference by the first sum to obtain a first ratio.

[0136] The processing module 220 is further configured to subtract the red band reflectance data from the near-infrared band reflectance data to obtain a second difference, add the red band reflectance data to the near-infrared band reflectance data to obtain a second sum, and divide the second difference by the second sum to obtain a second ratio.

[0137] The processing module 220 is further configured to subtract the second ratio from the first ratio to obtain the tidal flat extraction index value;

[0138] The judgment module 230 is used to compare the mudflat extraction index value with a preset mudflat identification threshold range. If the mudflat extraction index value is within the threshold range, the corresponding pixel is determined to be a mudflat; if the mudflat extraction index value is outside the threshold range, the corresponding pixel is determined to be a non-mudflat.

[0139] The apparatus of this embodiment can be used to perform... Figure 1 The steps of the method embodiment shown are similar in principle and process, and will not be repeated here.

[0140] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0141] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0142] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for extracting tidal flats based on satellite remote sensing data, characterized in that, The method includes: Acquire satellite remote sensing data of the target area, and extract green light band reflectance data, red light band reflectance data and near-infrared band reflectance data from the satellite remote sensing data; The first difference is obtained by subtracting the green band reflectance data from the red band reflectance data, the first sum is obtained by adding the green band reflectance data to the red band reflectance data, and the first ratio is obtained by dividing the first difference by the first sum. The second difference is obtained by subtracting the red band reflectance data from the near-infrared band reflectance data, the second sum is obtained by adding the red band reflectance data to the near-infrared band reflectance data, and the second ratio is obtained by dividing the second difference by the second sum. Subtracting the second ratio from the first ratio yields the tidal flat extraction index value; The mudflat extraction index value is compared with a preset mudflat identification threshold range. If the mudflat extraction index value is within the threshold range, the corresponding pixel is determined to be a mudflat; if the mudflat extraction index value is outside the threshold range, the corresponding pixel is determined to be a non-mudflat.

2. The method according to claim 1, characterized in that, The preset threshold range for tidal flat identification corresponds one-to-one with the model of the satellite sensor. Different models of satellite sensors correspond to different threshold ranges for tidal flat identification, including: Obtain the satellite sensor model information corresponding to the satellite remote sensing data to be processed; Based on the model information, the pre-calibrated tidal flat identification threshold range of the corresponding model sensor is retrieved as the preset tidal flat identification threshold range.

3. The method according to claim 2, characterized in that, The step of retrieving the pre-calibrated tidal flat identification threshold range for the corresponding sensor model based on the model information includes: Obtain the spectral response curves of the benchmark satellite sensor in the green, red, and near-infrared bands; Obtain the spectral response curves of the target satellite sensor in the green, red, and near-infrared bands; The spectral response curves of the target satellite sensor in each band are compared with the spectral response curves of the reference satellite sensor in the corresponding band, and the spectral response difference within the wavelength overlap range is calculated. Based on the spectral response difference, determine the mudflat identification threshold correction coefficient of the target satellite sensor relative to the reference satellite sensor; Using the tidal flat identification threshold range corresponding to the reference satellite sensor as the reference threshold, the threshold correction coefficient is superimposed on the endpoint value of the reference threshold to obtain the tidal flat identification threshold range pre-calibrated by the target satellite sensor.

4. The method according to claim 3, characterized in that, The step of determining the mudflat identification threshold correction coefficient of the target satellite sensor relative to the reference satellite sensor based on the spectral response difference includes: Align the spectral response curves of the target satellite sensor and the reference satellite sensor in the corresponding bands according to the wavelength horizontal axis to obtain a set of matching points where the vertical axis response rate values ​​of the two curves correspond one-to-one. Within the wavelength overlap range, for each matching point, the difference between the target satellite sensor responsivity value and the reference satellite sensor responsivity value is calculated and used as the ordinate difference value of the matching point. Divide the ordinate difference value of each matching point by the response rate value of the reference satellite sensor at the matching point to obtain the response rate correction percentage of the matching point. The arithmetic mean of the response rate correction percentages for all matching points within the wavelength overlap range is calculated, and the arithmetic mean is determined as the mudflat identification threshold correction coefficient of the target satellite sensor relative to the reference satellite sensor.

5. The method according to claim 1, characterized in that, Before comparing the extracted tidal flat index value with a preset tidal flat identification threshold range, the following steps are included: Select sampling points on the mudflats and outside the mudflats within the target area; Substitute the tidal flat sampling points and non-tidal flat sampling points into the tidal flat extraction index to calculate the tidal flat extraction index value corresponding to all sampling points. Based on the distribution range of the index values ​​extracted from the mudflats at all sampling points, the candidate threshold search interval is determined; The candidate threshold search interval is divided into multiple consecutive sub-intervals; Using the endpoint value of each sub-interval as a candidate threshold, pixels with a mudflat extraction index value greater than or equal to the candidate threshold are identified as mudflats, and pixels with a mudflat extraction index value less than the candidate threshold are identified as non-mudflats. The number of correctly identified mudflat sampling points and the number of incorrectly identified non-mudflat sampling points under each candidate threshold are counted. Based on the number of correctly identified and incorrectly identified items, calculate the recognition accuracy corresponding to each candidate threshold; The two endpoints of the sub-interval containing the candidate threshold with the highest recognition accuracy are determined as the upper and lower endpoints of the preset tidal flat recognition threshold range.

6. The method according to claim 1, characterized in that, The establishment of the calculation formula for the tidal flat extraction index value includes: Obtain the normalized water index and normalized vegetation index, and calculate the identification accuracy of the normalized water index and normalized vegetation index at the sampling point on the tidal flat, respectively, as the first control accuracy and the second control accuracy. Based on the formula configuration of the normalized water index and the normalized vegetation index, the band combinations used in them are replaced with different combinations of green light band, red light band and near-infrared band to generate multiple candidate indices. Calculate the recognition accuracy of each candidate index at the sampling points on the mudflats; The identification accuracy of each candidate index is compared with the first control accuracy and the second control accuracy to select the candidate index with the highest identification accuracy. The formula for calculating the candidate index with the highest identification accuracy is determined as the mudflat extraction index.

7. The method according to claim 1, characterized in that, The acquisition of satellite remote sensing data of the target area, and the extraction of green band reflectance data, red band reflectance data, and near-infrared band reflectance data from the satellite remote sensing data, includes: Collect spectral reflectance data of tidal flats, water bodies, vegetation, and rock features within the target area; Based on the spectral reflectance data of various land features, corresponding spectral curves were plotted, and the spectral differences between tidal flats and other land features were compared and analyzed. Based on the comparison of spectral differences, the validity of each band of the satellite remote sensing data was determined, and the green light band, red light band, and near-infrared band were identified as the effective bands for extraction from the tidal flats.

8. The method according to claim 7, characterized in that, When the method is applied to tidal flat extraction in different geographical regions, it is determined whether it is necessary to re-determine the effective bands based on the land cover composition and satellite data distribution characteristics of the target geographical region, including: Obtain data on the distribution of land cover types in the target geographic area, and statistically analyze the area proportion of four types of land cover: mudflats, water bodies, vegetation, and rocks. Acquire satellite remote sensing data covering the target geographic area, and statistically analyze the number of effective pixels and the spatial distribution uniformity of data in each band; The area ratio of land cover types in the target geographic area is compared with the area ratio of the corresponding land cover types in the target area used when the effective bands are determined in advance, and the difference value of the area ratio of each type of land cover is calculated. The distribution characteristics of satellite data in the target geographic area are compared with the distribution characteristics of satellite data in the target area used when the effective bands are determined in advance, and the distribution similarity is calculated. If the area ratio difference of any land cover type exceeds the preset ratio difference threshold, it is determined that the effective band needs to be re-determined; if the distribution similarity is lower than the preset similarity threshold, it is determined that the effective band needs to be re-determined. If the area ratio difference of all land cover types does not exceed the preset ratio difference threshold, and the distribution similarity is not lower than the preset similarity threshold, then it is determined that the determined effective bands will continue to be used.

9. The method according to claim 8, characterized in that, After determining that the already identified valid bands should continue to be used, this includes: Based on satellite remote sensing data of the target geographic area, sampling points on tidal flats and non-tidal flats were selected; Based on the mudflat extraction index values ​​corresponding to the mudflat sampling points and non-mudflat sampling points, a candidate threshold set is constructed. Calculate the recognition accuracy of each candidate threshold for the mudflat sampling points and non-mudflat sampling points respectively; The candidate threshold with the highest recognition accuracy is determined as the optimized mudflat recognition threshold for the target geographical area.

10. A device for extracting tidal flats based on satellite remote sensing data, characterized in that, The device includes an acquisition module, a processing module, and a judgment module; The acquisition module is used to acquire satellite remote sensing data of the target area and extract green light band reflectance data, red light band reflectance data and near-infrared band reflectance data from the satellite remote sensing data; The processing module is used to subtract the green light band reflectance data from the red light band reflectance data to obtain a first difference, add the green light band reflectance data to the red light band reflectance data to obtain a first sum, and divide the first difference by the first sum to obtain a first ratio. The processing module is further configured to subtract the red band reflectance data from the near-infrared band reflectance data to obtain a second difference, add the red band reflectance data to the near-infrared band reflectance data to obtain a second sum, and divide the second difference by the second sum to obtain a second ratio. The processing module is further configured to subtract the second ratio from the first ratio to obtain the tidal flat extraction index value; The judgment module is used to compare the mudflat extraction index value with a preset mudflat identification threshold range. If the mudflat extraction index value is within the threshold range, the corresponding pixel is determined to be a mudflat; if the mudflat extraction index value is outside the threshold range, the corresponding pixel is determined to be a non-mudflat.

Images