A method and device for identifying harmful algal bloom regions based on remote sensing data

Abstract

The application provides a harmful algal bloom area identification method and device based on remote sensing data, and belongs to the field of marine remote sensing and environmental monitoring. The method comprises the following steps: determining a plurality of sensitive bands of harmful algal blooms; acquiring remote sensing data of a to-be-identified area in each sensitive band; determining an identification interference factor of the to-be-identified area; combining each sensitive band according to the relationship between the identification interference factor and each sensitive band; calculating an identification interference factor index corresponding to each combination; correcting the remote sensing data corresponding to the first sensitive band combination and the second sensitive band combination according to the identification interference factor index; calculating a harmful algal bloom index according to the remote sensing data corresponding to the third sensitive band combination after correction; and judging whether the to-be-identified area is a harmful algal bloom area according to the harmful algal bloom index and a dynamic threshold corresponding to the to-be-identified area. The harmful algal bloom area identification method and device based on remote sensing data provided by the application can realize high-precision and high-stability identification of harmful algal bloom areas.

Description

A method and apparatus for identifying harmful algal bloom areas based on remote sensing data Technical Field

[0001] This application relates to the field of marine remote sensing and environmental monitoring technology, and in particular to a method and apparatus for identifying harmful algal bloom areas based on remote sensing data. Background Technology

[0002] Harmful algal blooms, commonly known as "red tides," refer to an ecological disaster caused by the rapid proliferation or high aggregation of algae (such as diatoms and dinoflagellates) in water bodies under specific environmental conditions, reaching abnormally high densities and resulting in water discoloration. Some algal bloom species can produce toxins or consume large amounts of oxygen in the water during their decay, posing a serious threat to aquatic ecosystems, aquaculture, drinking water safety, and public health. Therefore, large-scale, high-frequency, and high-precision dynamic monitoring of harmful algal blooms is an urgent need for marine and inland water environment management and disaster early warning.

[0003] In recent years, with the development of satellite remote sensing technology, using optical remote sensing data to invert water composition and monitor algal blooms has become a mainstream method. Current techniques mainly rely on the spectral characteristics of algae in certain specific wavelength bands, such as the absorption valleys of chlorophyll a in the blue and red light bands and the reflection peaks in the near-infrared band, to construct various empirical, semi-empirical, or bio-optical models to estimate chlorophyll concentration and thus infer whether algal blooms have occurred.

[0004] However, existing technologies still have significant limitations in practical applications, especially in nearshore waters with complex optical characteristics. First, optical remote sensing cannot penetrate cloud layers; thick clouds lead to data loss, while thin clouds or cloud shadows severely pollute the water's spectral signal, rendering algal bloom identification algorithms based on specific spectral features ineffective, and potentially causing false alarms due to the high reflectivity of clouds. Second, in turbid waters such as nearshore areas and estuaries, suspended sediments significantly enhance the water's reflectivity in the blue and green bands through Mie scattering. This strong scattering signal masks the weak absorption characteristics of algal pigments, leading to spectral confusion and making it difficult for traditional methods to accurately distinguish between highly turbid water and algal bloom water, resulting in misjudgments or missed detections. Furthermore, many existing algorithms are based on the optical assumptions of clean water bodies or rely on fixed empirical thresholds. When applied to water bodies in different seasons, geographical regions, or hydrological conditions, their recognition accuracy and stability significantly decrease due to changes in background optical characteristics, lacking robustness. In complex nearshore environments, pixels often contain multiple types of information, such as algae and suspended matter. Existing methods mostly perform inversion for single targets, lacking mechanisms for collaborative analysis and hierarchical processing of multiple interfering factors, resulting in limited accuracy in extracting algal bloom information from mixed pixels. Therefore, there is an urgent need for a remote sensing identification method for harmful algal blooms that can effectively overcome interference from clouds and suspended sediments and adapt to different environmental conditions, in order to achieve high-precision and high-stability automated monitoring of harmful algal bloom areas. Summary of the Invention

[0005] In view of this, this application provides a method and apparatus for identifying harmful algal bloom areas based on remote sensing data, which can overcome the severe interference of clouds and suspended sediments in nearshore waters with complex optical characteristics, and achieve high-precision and high-stability automated identification of harmful algal bloom areas.

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

[0007] The first aspect of this application provides a method and apparatus for identifying harmful algal bloom areas based on remote sensing data, the method comprising:

[0008] Identify multiple sensitive bands for harmful algal blooms;

[0009] Acquire remote sensing data of the area to be identified in various sensitive bands;

[0010] Identify the identification interference factors in the region to be identified;

[0011] Based on the relationship between the identified interference factors and each sensitive band, the sensitive bands are combined to obtain multiple sensitive band combinations. Each sensitive band combination plays a different role in the method for identifying harmful algal bloom areas.

[0012] At least the identification interference factor index corresponding to each combination should be calculated based on the remote sensing data corresponding to the first sensitive band combination and the second sensitive band combination, respectively.

[0013] The remote sensing data corresponding to the first sensitive band combination and the second sensitive band combination are corrected based on the index of identified interference factors.

[0014] The harmful algal bloom index is calculated based on the corrected remote sensing data corresponding to the third sensitive band combination.

[0015] The determination of whether the region to be identified is a harmful algal bloom region is based on the harmful algal bloom index and the dynamic threshold corresponding to the region to be identified.

[0016] The second aspect of this application provides a device for identifying harmful algal bloom areas based on remote sensing data. The device includes a determination module, an acquisition module, a combination module, a calculation module, a correction module, and a judgment module.

[0017] The determining module is used to determine multiple sensitive bands of harmful algal blooms;

[0018] The acquisition module is used to acquire remote sensing data of the area to be identified in various sensitive bands;

[0019] The determining module is also used to determine the identification interference factors of the area to be identified;

[0020] The combination module is used to combine the various sensitive bands according to the relationship between the identified interference factors and each sensitive band to obtain multiple sensitive band combinations. Each sensitive band combination plays a different role in the method for identifying harmful algal bloom areas.

[0021] The calculation module is used to calculate the identification interference factor index corresponding to each combination based on the remote sensing data corresponding to the first sensitive band combination and the second sensitive band combination, respectively.

[0022] The correction module is used to correct the remote sensing data corresponding to the first sensitive band combination and the second sensitive band combination according to the identified interference factor index.

[0023] The calculation module is also used to calculate the harmful algal bloom index based on the corrected remote sensing data corresponding to the third sensitive band combination.

[0024] The judgment module is used to determine whether the region to be identified is a harmful algal bloom region based on the harmful algal bloom index and the dynamic threshold corresponding to the region to be identified.

[0025] The method and apparatus for identifying harmful algal bloom areas based on remote sensing data provided in this application, taking advantage of the different sensitivities of each sensitive band to different features, selects multiple sensitive bands to form different combinations, and uses different combinations to remove different interference factors. Based on the removed data, a harmful algal bloom index is calculated, enabling targeted use of remote sensing data from the same source for interference removal and identification index calculation. First, by comparing the spectral reflectance curves of harmful algae and background water, bands sensitive to both algal characteristics and major interference factors (clouds, suspended sediments) are accurately selected. Then, based on the spectral response characteristics of the interference factors, the aforementioned sensitive bands are functionally combined, with the first and second sensitive band combinations used to calculate the identification interference factor indices for suppressing cloud interference and suspended sediment interference, respectively. Next, an effective pixel mask is generated based on the identification interference factor indices and a dynamic threshold to correct the remote sensing data and remove invalid pixels. Finally, the harmful algal bloom index is calculated using the corrected remote sensing data corresponding to the third sensitive band combination, and combined with a dynamic threshold adapted to real-time environmental conditions, a final determination is made as to whether the area to be identified is a harmful algal bloom area. To address different interference factors and identification targets, a limited number of sensitive bands are functionally combined and synergistically processed, achieving accurate and stable identification of harmful algal bloom areas in complex aquatic environments. This approach ensures the systematic nature of the identification logic while enhancing the method's adaptability to different observation conditions and aquatic environments, effectively overcoming the problems of insufficient interference suppression and limited identification accuracy in traditional identification methods. Specifically, by identifying the relationship between interference factors and sensitive bands, signal confusion caused by single bands or indiscriminate combinations is avoided, improving the targeting of interference suppression and algal bloom signal extraction. By calculating the interference factor index and correcting remote sensing data, invalid interference information can be extracted from the raw data, providing a high-quality data foundation for subsequent algal bloom index calculation, reducing the impact of interference factors on the identification results, and improving the reliability of the identification results. The method of calculating the algal bloom index and making judgments using corrected data, with dynamic thresholds adapted to the actual situation of the area to be identified, avoids the problems of misjudgment and omission due to fixed thresholds in different environments, further enhancing the universality and identification accuracy of the method. Attached Figure Description

[0026] Figure 1 is a flowchart of an embodiment of the harmful algal bloom region identification method based on remote sensing data provided in this application;

[0027] Figure 2 shows the spectral reflectance curves of harmful algae and various background water bodies as illustrated in this application.

[0028] Figure 3 is a schematic diagram of the second embodiment of the harmful algal bloom area identification device based on remote sensing data provided in this application. Detailed Implementation

[0029] 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.

[0030] 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.

[0031] 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."

[0032] The following specific embodiments are given to illustrate the technical solution of this application in detail.

[0033] Example 1

[0034] Figure 1 is a flowchart of an embodiment of the harmful algal bloom region identification method based on remote sensing data provided in this application. Referring to Figure 1, the method provided in this embodiment may include:

[0035] S101. Identify multiple sensitive bands for harmful algal blooms.

[0036] It should be noted that harmful algal blooms refer to the phenomenon in specific marine, lake, and other aquatic environments where harmful algae proliferate in large numbers due to nutrient enrichment and suitable water temperatures, leading to an imbalance in the aquatic ecosystem. These blooms cause problems such as oxygen depletion and toxin production, endangering not only aquatic life but also shipping safety and nearshore ecological environments, making them a marine ecological disaster requiring precise monitoring. Remote sensing identification of harmful algal blooms utilizes the differences in reflectance between algae and background water at different spectral bands. Due to their physiological structures (such as algal cells, chlorophyll a, and carotenoids), algae exhibit unique spectral responses within specific wavelength ranges, while background water lacks these characteristics. These significant differences in spectral reflectance between harmful algae and background water, and the specific wavelength ranges that can effectively distinguish between the two, are known as sensitive bands.

[0037] Specifically, several sensitive bands for identifying harmful algal blooms were determined, including:

[0038] (1) Obtain spectral reflectance data of various harmful algae and various background water bodies.

[0039] It should be noted that harmful algae refer to algae that can cause algal blooms and harm the ecological environment and biological health. For example, in this application, they may specifically include diatoms, dinoflagellates, and dinoflagellates, which are common dominant algal species in nearshore waters. Background water refers to various water bodies that have not experienced harmful algal blooms. Similarly, in this application, it mainly includes nearshore clear water (such as clean nearshore waters with low suspended solids and low nutrients), turbid water (such as waters containing large amounts of suspended sediments), and open seawater (such as offshore waters far from the coastline). Spectral reflectance data refers to the ratio (unit: dimensionless) of the reflected electromagnetic wave energy to the incident energy after electromagnetic waves of different wavelengths irradiate the surface of a water body (or algae). It is a type of data used in remote sensing technology to characterize the optical properties of water bodies and substances in water, usually expressed as R. rs (λ) represents (λ is a specific wavelength, such as R) rs (560) refers to the remote sensing reflectance of the 560nm band.

[0040] It should be noted that due to the different molecular structures and physical forms of various substances (algae, background water), their absorption and scattering capabilities for electromagnetic waves of different wavelengths will also vary. This difference is quantified through reflectance data. Specifically, spectral reflectance data can be obtained by using a field spectrometer to measure the reflectance of laboratory-cultured samples of diatoms, dinoflagellates, etc., in the area to be studied, as well as the in-situ reflectance of the corresponding background water. Alternatively, it can be obtained directly from satellite remote sensing data, such as L2A-level surface reflectance data acquired by the coastal zone imager aboard the HY-1C / D satellite. This data has undergone atmospheric correction and can be directly used for spectral analysis. The acquired spectral reflectance data should cover the visible-near-infrared band of 400-750 nm to ensure that the core response wavelengths of algal pigment absorption and cell scattering are included.

[0041] (2) Compare and analyze the spectral reflectance curves of the harmful algae with the spectral reflectance curves of the background water bodies of various types.

[0042] It should be noted that after obtaining spectral reflectance data, these data can be used to plot the spectral reflectance characteristic curves of harmful algae and background water. By intuitively comparing the curve shapes and conducting quantitative analysis, we can find the unique optical characteristics of harmful algae, while excluding the optical characteristics shared by harmful algae and background water, thereby determining the wavelength range that only harmful algae possess or that is significantly different from the background water.

[0043] In the spectral reflectance characteristic curve, the X-axis (horizontal axis) represents wavelength in nanometers (nm), characterizing different bands of electromagnetic waves; the Y-axis (vertical axis) represents remote sensing reflectance (R0). rs ), a dimensionless quantity, characterizes the ability of a water body to reflect incident light of a specific wavelength.

[0044] Figure 2 shows the spectral reflectance curves of harmful algae and various background water bodies as shown in this application. Referring to Figure 2, the three reflectance curves of diatoms, dinoflagellates, and algae are superimposed with the three curves of clear nearshore water, turbid water, and open seawater to observe the peak and valley positions, reflectance value differences, and trends of the curves.

[0045] Specifically, by quantitatively observing the differences in reflectance values ​​between harmful algae and various background water bodies at the same wavelength, a decision-making basis for band selection is provided. For example, at a wavelength of 560 nm, if algae show a reflectance peak due to cell scattering, their reflectance is significantly higher than that of background water bodies; while at a wavelength of 650 nm, algae show an absorption valley due to pigment absorption, while background water bodies have a flat reflectance in this range without absorption valley characteristics. This stable and significant reflectance difference existing in a specific band can be used to determine whether the band is a sensitive band. The larger the reflectance difference and the more stable it is in comparison with various algae and various background water bodies, the stronger the ability of this band to distinguish algal bloom water bodies from background water bodies, and the more likely it is to be selected as a sensitive band for constructing an identification index.

[0046] Furthermore, quantitative analysis methods can be used to calculate the difference and percentage of the reflectance between algae and background water at the same wavelength. For example, if the reflectance of dinoflagellates is 0.015 and the reflectance of turbid water is 0.008 at a wavelength of 560nm, the difference is 0.007, and the percentage of the difference is 87.5%, indicating that the difference between the two is significant at this wavelength.

[0047] (3) Based on the comparison analysis results, select multiple characteristic bands with a difference between the reflectance of harmful algae and background water that is greater than or equal to a preset threshold as the sensitive bands.

[0048] Referring to Figure 2, based on step (2), we have determined the approximate wavelength range of differences. Further, we need to select the specific bands that have the most significant differences, strong anti-interference ability, and can stably characterize the algal features. Among them, significant differences can be judged by the proportion of the reflectance difference between algae and at least two types of background water bodies at the same wavelength being greater than a preset threshold, such as greater than or equal to 50%; stability means that the difference exists in all three types of harmful algae (rather than being unique to a single algal species); anti-interference ability means that the band should be able to respond to the interference factors to be suppressed later. For example, suspended sediments have strong scattering in the blue light band, so the blue light band is selected for subsequent suspension interference suppression.

[0049] Referring to Figure 2, in the blue light band (440-490nm), algae exhibit a significant absorption valley (low reflectance) due to the strong absorption of chlorophyll a, while background water (especially turbid water) shows no significant absorption valley in this range, and reflectance increases with wavelength. In the green light band (560-580nm), algae exhibit a significant reflection peak (high reflectance) due to the scattering effect of algal cells, and the peak value is significantly higher than that of the three types of background water. In the red light band (650-670nm), algae show a further decrease in reflectance due to the absorption of auxiliary pigments such as carotenoids, forming a secondary absorption valley, while background water shows a flat reflectance in this range without absorption valley characteristics.

[0050] Based on the above analysis, three characteristic bands were ultimately selected as sensitive bands: the 460nm blue light band (corresponding to the blue light band difference range), the 560nm green light band (corresponding to the green light band reflection peak range), and the 650nm red light band (corresponding to the red light band absorption valley range). In other words, multiple sensitive bands include the blue light band, which is sensitive to scattering by suspended sediments; the green light band, which is sensitive to scattering by algal cells; and the red light band, which is sensitive to absorption by algal pigments.

[0051] The data for the sensitive bands comes from the Coastal Zone Imager (CZI) carried by the HY-1C / D satellite. The band configuration of this sensor fully covers the identified sensitive bands. Specifically, the sensor characteristics of the HY-1C / D CZI are shown in Table 1.

[0052]

[0053] Please refer to Table 1. The 460nm blue light band corresponds to band 1 (421-500nm) of the CZI, the 560nm green light band corresponds to band 2 (517-598nm) of the CZI, and the 650nm red light band corresponds to band 3 (608-690nm) of the CZI. Therefore, the selected sensitive bands directly match the channel settings of the HY-1C / D CZI sensor, ensuring that this method has a reliable data source and is feasible for operational implementation.

[0054] Furthermore, it should be noted that by selecting multiple rather than a single sensitive band, the unique spectral characteristics of harmful algae can be accurately captured, and the interference factors such as suspended sediments and clouds that need to be suppressed can be targeted, laying the foundation for subsequent band combination, interference suppression, and algal bloom signal extraction.

[0055] S102. Obtain remote sensing data of the area to be identified in each sensitive band.

[0056] It should be noted that after determining the sensitive wavelengths (i.e., blue, green, and red wavelengths) used to identify harmful algal blooms, it is necessary to extract the observation data of the water area to be monitored in these specific wavelengths from the remote sensing data source to provide direct data input for subsequent spectral analysis and feature extraction.

[0057] Remote sensing data refers to raw observation data acquired from satellite or airborne sensors, encompassing multiple spectral bands. The remote sensing data used in this method refers to the remote sensing reflectance (R0) extracted from the aforementioned raw data within the sensitive band range and obtained after preprocessing such as radiometric calibration and atmospheric correction. rs The purpose of this method is to convert the raw radiation signals received by satellite sensors into optical quantities that are only related to the internal composition of water bodies (such as algae and suspended matter), thereby removing the influence of factors such as the atmosphere and the angle of sunlight, and making data acquired at different times and locations comparable.

[0058] Specifically, remote sensing data can originate from satellites or airborne platforms equipped with multispectral imagers. To achieve continuous and stable monitoring of large areas of water, ocean color satellites with operational capabilities are preferred. For example, a satellite constellation consisting of my country's independently launched Haiyang-1C (HY-1C) and Haiyang-1D (HY-1D) satellites can be used. The Coastal Zone Imager (CZI) on this constellation has multiple spectral channels covering the visible to near-infrared range, with its band settings fully covering the identified sensitive bands. It also boasts a high signal-to-noise ratio and spatial resolution, making it particularly suitable for detailed monitoring of complex nearshore waters.

[0059] It should be noted that the raw satellite observation data (Level 0) requires a series of processing steps before it can be used for quantitative analysis. This application mainly utilizes L1B and L2A level data products. L1B level data has undergone radiometric calibration, converting the sensor's raw digital quantization (DN) values ​​into physically meaningful radiance values. This data can be used for preliminary radiometric information analysis. L2A level data, based on L1B, undergoes further atmospheric correction to eliminate the contribution of atmospheric molecules and aerosols to the signal, ultimately yielding surface reflectance or remotely sensed reflectance (R0). rs Image data products, which most directly represent the inherent optical properties of water bodies, include strict geographic coordinate information (usually latitude and longitude) for L1B and L2A levels. By matching these geographic coordinates with the predefined spatial range of the area to be identified (such as vector boundaries or latitude and longitude ranges), image data belonging to the area to be identified can be accurately filtered out, thus ensuring that all subsequent analysis and processing are performed on the target area.

[0060] Based on the center wavelength of the sensitive band determined in step S101, remote sensing reflectance data of the water area to be identified in each corresponding band are extracted from selected satellite data products (such as the L2A product of HY-1C / DCZI). These data are spatially organized in pixels, with each pixel containing its Rreflectance in each sensitive band. rs These values ​​together form the data foundation for subsequent calculations and analyses.

[0061] This step obtains and processes high-quality remote sensing reflectance data of the area to be identified in specific sensitive bands from operational satellites, providing reliable data support for the subsequent accurate identification of harmful algal blooms.

[0062] S103. Determine the identification interference factors of the area to be identified.

[0063] It should be noted that interference factors refer to non-target ground features or environmental conditions in optical remote sensing images whose spectral signals are mixed with harmful algal bloom signals and are not easily separated by simple methods, thus affecting the automatic identification of algal bloom areas.

[0064] Specifically, the interference factors mainly include the following two categories. First, clouds. Clouds are one of the most common interference factors in optical remote sensing. Thick clouds can completely block the sensor's observation of the land surface, leading to the loss of water signals and creating invalid information areas in the image. Thin clouds or cloud shadows can alter the spectral composition of the signals reaching the sensor. Clouds typically exhibit high and flat reflectivity in the visible light range, which can cover the relatively weak spectral characteristics of water bodies, causing algal bloom identification algorithms based on specific absorption and scattering peaks to fail. Therefore, the high reflectivity of clouds may resemble that of highly scattering algal bloom water bodies in certain wavelengths. If not distinguished, cloud areas can easily be misjudged as algal bloom areas, leading to incorrect identification. Second, suspended sediments. Suspended sediments are the most significant endogenous interference factor in nearshore turbid water bodies. Specifically, suspended particles (such as silt) significantly enhance the reflectivity of water bodies through Mie scattering, especially in the blue and green light bands. This strong scattering signal dilutes and masks the weak absorption valleys formed by algal pigment absorption. Furthermore, the reflectance signals contributed by suspended matter scattering and algal cell scattering may have similar spectral shapes, causing spectral confusion and making it difficult to accurately distinguish between highly turbid water bodies and algal bloom water bodies using reflectance thresholds alone. Therefore, the presence of suspended sediments can alter the overall spectral profile of a water body, for example, significantly increasing its reflectance in the blue light band and exhibiting specific trends, which differs significantly from the algal-dominated spectral profile of clean water bodies.

[0065] It should be further clarified that the identification of these interfering factors is based on the understanding of nearshore water remote sensing monitoring practices. This application does not list all possible interferences, but focuses on the two factors that have the greatest impact on harmful algal bloom signals in optical remote sensing images.

[0066] S104. Based on the relationship between the identified interference factors and each sensitive band, the sensitive bands are combined to obtain multiple sensitive band combinations. Each sensitive band combination plays a different role in the method for identifying harmful algal bloom areas.

[0067] It should be noted that the combination here is not a simple band aggregation. The index calculated afterward will be used specifically to identify or suppress specific interference factors, as well as to extract the core algal bloom signal.

[0068] Specifically, based on the relationship between the identified interference factors and each sensitive band, the sensitive bands are combined to obtain multiple sensitive band combinations, including:

[0069] (1) For each identified interference factor, analyze the spectral response characteristics of the current interference factor under different sensitive bands.

[0070] It should be noted that clouds typically exhibit high reflectivity in the visible light range, and the difference in reflectivity between bands such as green and red light is relatively small. In contrast, suspended particulate matter exhibits strong Mie scattering in the blue light band, resulting in a significant increase in reflectivity in this band. The difference in scattering intensity between this particulate matter and the green light band constitutes an effective feature for identifying turbidity.

[0071] (2) Based on the spectral response characteristics, at least two sensitive bands are selected from the sensitive bands to form a sensitive band combination; wherein, each sensitive band in the same sensitive band combination has a different spectral response from the current interfering factor.

[0072] It should be noted that, regarding cloud interference, based on the characteristic that clouds exhibit high reflectivity and minimal difference in both the green and red light bands, the green and red light bands are selected to form the first sensitive band combination (specifically, the 560nm green light band and the 650nm red light band). This combination is used to maximize the separability of clouds and background water in this two-dimensional spectral space. Regarding suspended sediment interference, based on the characteristic that suspended sediments scatter strongly in the blue light band and relatively weakly in the green light band, the blue and green light bands are selected to form the second sensitive band combination (specifically, the 460nm blue light band and the 560nm green light band). This combination can effectively capture changes in the spectral morphology of water caused by suspended matter.

[0073] (3) Multiple sensitive band combinations are formed for different identification interference factors, and the multiple sensitive band combinations share at least one sensitive band.

[0074] Based on the preceding description, the green band is simultaneously included in both the first sensitive band combination (for cloud detection) and the second sensitive band combination (for sediment detection). This band reuse strategy allows the three sensitive bands to form multiple functionally distinct combinations through sharing, greatly improving data utilization efficiency and ensuring the inherent consistency of information generated by different combinations. Furthermore, for the identification of harmful algal blooms, a third sensitive band combination (also including both green and red bands) is constructed based on the spectral characteristics of algae—a scattering peak in the green band and an absorption valley in the red band. This combination is specifically designed to accurately extract algal bloom signals after interference has been eliminated.

[0075] Thus, the three sensitive band combinations constitute a well-defined collaborative identification system. The first sensitive band combination is responsible for identifying and suppressing cloud interference; the second sensitive band combination is responsible for identifying and suppressing suspended sediment interference; and the third sensitive band combination is responsible for accurately extracting harmful algal bloom signals based on the effective data constructed by the first two combinations.

[0076] S105. Calculate the identification interference factor index corresponding to each combination based on the remote sensing data corresponding to the first sensitive band combination and the second sensitive band combination, respectively.

[0077] It should be noted that the interference factor identification index here is an index constructed by the ratio of the difference and sum of reflectance in a specific band. This index, through the difference / sum operation, can effectively suppress the influence of common-mode noise such as changes in lighting conditions and sensor observation angle, thereby extracting the spectral characteristics of the target ground object (cloud or suspended sediment) more stably and purely.

[0078] Specifically, at least based on the remote sensing data corresponding to the first sensitive band combination and the second sensitive band combination, the identification interference factor index corresponding to each combination is calculated, including: obtaining the remote sensing reflectance data corresponding to the green band and the red band in the first sensitive band combination; calculating the difference between the reflectance of the green band and the reflectance of the red band, and the sum of the reflectance of the green band and the reflectance of the red band; dividing the difference by the sum to obtain the first identification interference factor index; the first identification interference factor index is used to identify and quantify cloud interference.

[0079] It should be noted that the green band (R) is extracted from the first sensitive band combination from the preprocessed remote sensing image. rs (560) and red light band (R rs(650)) The remote sensing reflectance data is used to calculate the difference between the green band reflectance and the red band reflectance at the same pixel point. This is to amplify the spectral contrast between the two. For healthy vegetation or certain ground features, this value is positive; for clouds, since their reflectance in both the green and red bands is high and similar, the difference is close to zero. Further, the sum of the green band reflectance and the red band reflectance at the same pixel point is calculated, and this value is used as a normalization factor. Finally, the calculated difference is divided by the sum, and the final result is the first identification interference factor index (Cloud index, CI). Specifically, the formula is as follows: The index ranges from -1 to 1, and cloud-covered areas typically exhibit specific low-value characteristics, making them easily identifiable.

[0080] Furthermore, it should be noted that calculating the identification interference factor index corresponding to each combination based on the remote sensing data corresponding to the first sensitive band combination and the second sensitive band combination, respectively, also includes: obtaining the remote sensing reflectance data corresponding to the blue light band and the green light band in the second sensitive band combination; calculating the difference between the reflectance of the blue light band and the reflectance of the green light band, and the sum of the reflectance of the blue light band and the reflectance of the green light band; dividing the difference by the sum to obtain the second identification interference factor index; the second identification interference factor index is used to identify and quantify suspended sediment interference.

[0081] Similarly, from the same remote sensing image, the blue band (R) is extracted from the second sensitive band combination. rs (460) and green band (R) rs (560) Based on the remote sensing reflectance data, calculate the difference between the blue band reflectance and the green band reflectance at the same pixel. Since the scattering enhancement effect of suspended sediments is usually stronger in the blue band than in the green band, high-turbidity water bodies will show differences from other water bodies in this dimension. Furthermore, calculate the sum of the blue band reflectance and the green band reflectance at the same pixel, divide the calculated difference by the sum, and the final result is the second identification interference factor index (Sand index, SI). Specifically, the formula is as follows: This index effectively characterizes the relative turbidity of water bodies; a higher value generally indicates a greater contribution from suspended sediment scattering.

[0082] In this step, by performing two sets of parallel normalized difference calculations, the original reflectivity data is transformed into two interference factor indices (CI and SI) with clear physical meaning, providing a basis for the next step of precise data correction and thus for generating an effective pixel mask.

[0083] S106. Calibrate the remote sensing data corresponding to the first sensitive band combination and the second sensitive band combination according to the identified interference factor index.

[0084] It should be noted that the calibration here does not refer to the radiation or spectral correction of the original remote sensing reflectance value, but rather a data screening or masking process based on logical judgment. Its purpose is to generate a spatial filter to exclude severely interfered invalid pixels from subsequent analyses.

[0085] Specifically, the calibration of the remote sensing data corresponding to the first sensitive band combination and the second sensitive band combination according to the identified interference factor index includes:

[0086] (1) Compare the first identified interference factor index with the first dynamic threshold, and compare the second identified interference factor index with the second dynamic threshold.

[0087] It should be noted that when comparing the first identified interference factor index (CI) with the first dynamic threshold (denoted as a), this first dynamic threshold is used to distinguish cloud-covered pixels from clear-sky pixels. Since clouds cause the CI value to be low, the rule is set as CI > a. In this way, only pixels with a CI value higher than the first dynamic threshold a are considered to be basically not covered by clouds and are valid candidate pixels.

[0088] The second identified interference factor index (SI) is compared with the second dynamic threshold (denoted as b), and this second dynamic threshold is used to distinguish highly turbid water bodies from relatively clear water bodies. Since suspended sediments cause the SI value to be high, the rule is set as SI < b. In this way, only pixels with an SI value lower than the threshold b are considered to be less interfered by the scattering of suspended sediments.

[0089] Specifically, a represents a constant used to eliminate pixels affected by cloud contamination, usually set between 0 and 0.5; b represents a constant used to identify and exclude water bodies with a high content of suspended particles, usually set between 0 and 0.3. These two parameters will be dynamically adjusted according to the on-site environmental conditions to enhance the adaptability and expandability of the model. Specifically, when the regional cloud cover is high or the solar altitude angle is low, resulting in insufficient light, the first dynamic threshold (a) will be moderately relaxed to avoid misjudging clear-sky water bodies with insufficient light as clouds and over-eliminating them; for waters with a high background turbidity (such as estuary areas), the second dynamic threshold (b) will be correspondingly reduced to more strictly exclude the interference of suspended sediments. Among them, in one way, the thresholds can be set based on the statistical distribution (such as mean, quantile) of the CI and SI indices in the current image.

[0090] (2) Generate a binary effective pixel mask based on the comparison results; wherein, when a pixel point simultaneously satisfies that the first recognition interference factor index is greater than the first dynamic threshold and the second recognition interference factor index is less than the second dynamic threshold, the pixel point is marked as valid in the mask.

[0091] It should be noted that for a pixel point, only when it simultaneously satisfies the two conditions of CI > a and SI < b, it is considered an effective pixel applicable to algal bloom recognition. Based on this, the generated effective pixel mask (V mask ) is a binary image with exactly the same spatial resolution as the original remote sensing image. Specifically, the expression of the effective pixel mask is . At all pixel positions that meet the above conditions, V mask is marked as 1 (or True), indicating that the pixel is valid. At all pixel positions that do not meet the conditions (i.e., affected by clouds or high turbidity), V mask is marked as 0 (or False), indicating that the pixel is invalid.

[0092] In this step, by applying dynamic thresholds to perform logical judgments on the two interference factor indices and generating a unified binary effective pixel mask, the removal of irrelevant features from the original remote sensing data is achieved, and this mask will be used as input and directly applied to the next step of calculating the harmful algal bloom index, ensuring that the final recognition result is only based on high-quality and low-interference observation data.

[0093] S107. Calculate the harmful algal bloom index based on the corrected remote sensing data corresponding to the third sensitive band combination.

[0094] It should be noted that the data used here is the data environment that has been screened by the effective pixel mask, so the quality of the input data is already guaranteed.

[0095] It should be further noted that the data correction step S106 is a synchronous processing of the original remote sensing reflectance data of all sensitive bands (blue, green, red). Although the red band (650nm) in the third sensitive band combination (green and red bands) is not used to calculate the aforementioned recognition interference factor indices (CI, SI), as a key band for calculating the final harmful algal bloom index S107, it must be screened by the effective pixel mask (V mask ) together with the blue and green bands in the correction step. This synchronous correction operation ensures that when calculating the harmful algal bloom index subsequently, the green and red band data called are all from the same set of pixels that have undergone quality control and are free from cloud and high turbidity interference, thus ensuring the consistency and reliability of the final recognition result in the spectral and spatial dimensions.

[0096] Specifically, the step of calculating the harmful algal bloom index based on the corrected remote sensing data corresponding to the third sensitive band combination includes: obtaining the remote sensing reflectance data corresponding to the green band and the red band in the corrected third sensitive band combination; calculating the difference between the reflectance of the green band and the reflectance of the red band; and multiplying the difference by the effective pixel mask to obtain the harmful algal bloom index.

[0097] It should be noted that, from V mask In the processed data scenario, the green band (R) contained in the third sensitive band combination is extracted. rs (560) and red light band (R rs (650)) remote sensing reflectance data. At this point, the acquired data theoretically no longer includes the data affected by V. mask The reflectance values ​​of invalid pixels (clouds, high-turbidity water bodies) are filtered out, or these pixels have been marked and will be ignored in the calculation. Then, the difference between the green and red reflectance at the same valid pixel is calculated. This difference directly quantifies the contrast between the scattering peak of harmful algal blooms in the green band and the absorption valley in the red band. The higher the algal density, the more significant this peak-valley difference, and the larger the calculated difference. Finally, the generated valid pixel masks are spatially multiplied pixel by pixel to obtain the harmful algal bloom index. The multiplication operation achieves the final fusion of spectral feature extraction and spatial validity screening, ensuring that the final harmful algal bloom index map not only numerically characterizes the algal bloom intensity but also accurately identifies which algal bloom signals are reliable (i.e., based on valid water body pixels) and which should be ignored (clouds, flares, high-turbidity water body areas).

[0098] Specifically, the expression for the harmful algal bloom index is: .

[0099] This step calculates the spectral characteristics of algal blooms using clean data and uses a mask for spatial constraint, ultimately generating a highly reliable harmful algal bloom index that can directly and reliably determine the final algal bloom region.

[0100] S108. Determine whether the region to be identified is a harmful algal bloom region based on the harmful algal bloom index and the dynamic threshold corresponding to the region to be identified.

[0101] It should be noted that, due to the differences in lighting conditions and water background optical characteristics in different seasons and water bodies, using a fixed threshold for judgment can easily lead to misjudgment or missed judgment. Therefore, by setting a dynamic threshold, the self-adjustment of the dynamic threshold can be used to overcome this limitation of the fixed threshold.

[0102] Specifically, the dynamic threshold value is set based on the real-time environmental conditions of the area to be identified. These environmental conditions may include season and solar altitude angle, background optical characteristics of the regional water body, and real-time image statistical features. Season and solar altitude angle affect light intensity, indirectly affecting the absolute value of reflectance; different water bodies have different SHDI background values ​​for non-algal bloom water due to their inherent optical properties (such as background turbidity and concentration of yellow substances). Real-time image statistical features can adaptively determine the threshold for distinguishing between algal blooms and non-algal blooms based on the overall distribution of harmful algal bloom indices in the current image. By considering these factors, the dynamic threshold can better match the actual situation of the current data and improve the accuracy of the judgment. For example, the dynamic range of SHDI is appropriately adjusted according to the specific environmental conditions on the day of algal bloom detection, usually set between 0 and 0.005; values ​​outside this range are classified as noise or non-algal features.

[0103] In practice, determining the dynamic threshold includes:

[0104] (1) On the true-color remote sensing image covering the area to be identified on the same day, reference sample pixels are selected based on the color and texture features of the image to construct a reference sample set, which includes pure water samples and harmful algal bloom samples.

[0105] It should be noted that true-color remote sensing imagery refers to an image synthesized by assigning data acquired by satellite sensors in the red, green, and blue visible light bands to the red, green, and blue channels respectively, resulting in colors that closely resemble human visual perception. This imagery is used to visually identify the types of land cover. Specifically, the color and texture characteristics of the imagery refer to the fact that in true-color imagery, pure water bodies typically appear deep blue or bluish-black with a uniform color; while harmful algal blooms, due to algal pigments (such as chlorophyll and carotenoids) and cell scattering, may appear as abnormal colors such as reddish-brown, yellowish-green, or brown, and spatially may exhibit non-uniform texture structures such as patches or stripes.

[0106] It should be noted that the reference sample pixels were selected interactively. Interactive selection refers to operators with professional knowledge using remote sensing image processing software to manually draw or click on pixels on the image based on the aforementioned color and texture features through visual interpretation. The reason for using interactive selection instead of a fully automated algorithm is that in complex nearshore environments, spectral ambiguity of ground features is severe, making it difficult for automatic classification algorithms to reliably and initially distinguish between pure water samples and harmful algal blooms. Manual interactive selection ensures the purity and representativeness of the samples.

[0107] Among them, a pure water sample refers to a set of pixels that exhibits typical clean water color (deep blue / blue-black) on a true-color image, is spatially uniform, and has been confirmed to be unaffected by thin clouds, solar flares, ship wakes, or high concentrations of suspended sediment. A harmful algal bloom sample refers to a set of pixels that exhibits typical abnormal color of the target algal bloom (such as reddish-brown), has relatively uniform texture, is spatially continuous, and has been confirmed as an algal bloom area through historical records or simultaneous on-site observations.

[0108] (2) For the pure water sample and the harmful algal bloom sample respectively, calculate their respective index values ​​according to the harmful algal bloom index calculation formula, and analyze the statistical characteristics of the index of each type of sample.

[0109] Substitute all pixels of each sample selected in step (1) into the calculation formula of the harmful algal bloom index (SHDI) defined in S107 for calculation.

[0110] Descriptive statistical analysis was performed on all SHDI values ​​calculated for both pure water samples and samples with harmful algal blooms. Specifically, statistical characteristics included the mean, standard deviation, and numerical distribution range. The mean reflects the central tendency of the SHDI for this type of sample; the standard deviation reflects the dispersion of SHDI within the same sample group, used to assess the consistency of sample selection; an excessively large standard deviation suggests that atypical pixels may have been included in the sample; and the numerical distribution range, i.e., the minimum and maximum values, reflects the upper and lower limits of the SHDI values ​​for this type of sample.

[0111] (3) Determine target parameters based on the statistical characteristics, including the maximum value of the harmful algal bloom index corresponding to the pure water sample and the minimum value of the harmful algal bloom index corresponding to the harmful algal bloom sample.

[0112] It should be noted that the target parameters refer to the two key values ​​extracted from the statistical results of step (2) and used to directly calculate the initial dynamic threshold. The maximum value of SHDI for the pure water sample is denoted as... This represents the upper limit of SHDI for clean water bodies free from algal blooms; the minimum SHDI for samples with harmful algal blooms is denoted as . , representing the lower limit of SHDI for a typical algal bloom region under high-concentration algal aggregation. These two parameters are directly derived from the numerical distribution range obtained in step (2). The role of analyzing the mean and standard deviation is to help determine the reliability of these two extreme values. For example, in one possible case, if Values ​​significantly higher than the mean plus two standard deviations of pure water samples are likely outliers; similarly, if... If the sample purity is significantly lower than the mean of harmful algal blooms minus two standard deviations, the sample purity needs to be rechecked.

[0113] (4) Calculate the initial dynamic threshold based on the maximum and minimum values.

[0114] Specifically, using step (3) to determine and Calculate the initial dynamic threshold. Specifically, the calculation formula is as follows:

[0115] ;

[0116] Where T is the initial dynamic threshold. This represents the maximum SHDI value for a pure water sample. This represents the minimum SHDI value for harmful algal bloom samples.

[0117] It should be noted that the actual meaning of this calculation formula is to take the midpoint between the boundaries of the two types of land cover feature ranges on the SHDI one-dimensional feature axis as the initial segmentation point. If the distribution ranges of the two types of samples do not overlap, the threshold can clearly segment them at this time; if there is a small amount of overlap, it also provides a reasonable starting point for the next step of optimization.

[0118] (5) Apply the initial dynamic threshold to the region to be identified to generate a preliminary algal bloom extraction map and compare it with the true color image for spatial consistency. Based on the comparison results, fine-tune the initial dynamic threshold to obtain the final dynamic threshold.

[0119] Specifically, the obtained initial dynamic threshold is used as the discrimination threshold and applied to all effective pixels in the entire area to be identified to generate a binary preliminary algal bloom extraction map. Then, the extraction map is spatially overlaid and visually compared with the true color remote sensing image used in step (1).

[0120] Furthermore, the initial dynamic threshold is fine-tuned based on the comparison results, with the aim of achieving the highest spatial consistency. Specifically, operators observe the comparison results to determine whether the extracted algal bloom patch outlines highly match the abnormally colored areas on the true-color image; whether obvious clouds, flares, or turbid water masses are misidentified as algal blooms (false positives); and whether faint, visible suspected algal bloom areas on the true-color image are missed (false negatives). Based on these comparison results, the initial dynamic threshold is slightly increased or decreased, using the best visually judged spatial consistency as the criterion. For example, if there are too many false positives, the threshold is appropriately increased; if there are too many false negatives, the threshold is appropriately decreased. After several iterative adjustments, until the spatial pattern of the extracted results achieves an optimal match with the visually interpreted algal bloom distribution, the threshold used at this point is determined as the final dynamic threshold.

[0121] In addition, it should be noted that the dynamic threshold setting is regional; the same threshold can be used for the same research area.

[0122] Further, the harmful algal bloom index (SHDI) calculated for each pixel is compared with the final dynamic threshold determined above. If the harmful algal bloom index of a certain pixel is greater than or equal to the dynamic threshold, it is determined that the area represented by the pixel belongs to the harmful algal bloom area. If the harmful algal bloom index of a certain pixel is less than the dynamic threshold, it is determined that the area represented by the pixel does not belong to the harmful algal bloom area. Preferably, based on the above pixel-by-pixel judgment results, a binary harmful algal bloom distribution map can be generated. In this map, pixels marked as algal bloom areas can be assigned specific colors or values, so as to intuitively and clearly display the occurrence range and spatial pattern of harmful algal blooms.

[0123] In this step, by introducing a dynamic threshold determination method that integrates interactive sample selection, statistical feature analysis, extreme value parameter extraction, and spatial consistency optimization verification, the traditional fixed threshold is replaced, enabling the threshold to adapt to the real-time environmental conditions and ground object distributions of specific images, thereby intelligently discriminating high-confidence harmful algal bloom indices and ultimately achieving precise, reliable, and highly adaptable automated monitoring of harmful algal bloom areas.

[0124] The method provided in this embodiment screens out three sensitive bands of 460 nm blue light, 560 nm green light, and 650 nm red light by comparing the spectral reflectance curves of various harmful algae and background water bodies. It not only captures the spectral characteristics of the absorption valley of algal chlorophyll a, the scattering peak of cells, and the absorption valley of accessory pigments, but also can specifically respond to the optical characteristics of interference factors such as suspended sediments and clouds, providing a data carrier for subsequent precise identification and avoiding signal confusion caused by single bands or non-feature bands. Priority is given to using L1B-level (after radiometric calibration) and L2A-level (after atmospheric correction) data of the CZI sensor on HY-1C / D satellites. Such data has removed interference such as atmospheric scattering and sensor noise and directly provides precise remote sensing reflectance (R rs ) data; at the same time, effective pixels with CI > a and SI < b are screened through dynamic threshold masking to completely exclude invalid signals from clouds and high-turbidity water bodies, ensuring that subsequent index calculations are only based on high-quality and low-interference data, greatly reducing the risk of misjudgment. In addition, functionalized band combinations are designed based on the spectral response characteristics of interference factors to calculate CI to suppress cloud interference and calculate SI to suppress suspended sediment interference, and the SHDI is extracted by multiplexing bands to extract algal bloom signals. This not only improves the data utilization efficiency through band multiplexing but also effectively solves the pain point that interference signals in coastal waters mask algal bloom signals. It should also be noted that, in view of the water optical characteristics and seasonal light differences in different areas to be identified, the CI threshold, SI threshold, and SHDI determination threshold are dynamically adjusted to avoid missed or misjudged problems of traditional fixed thresholds in different environments, enabling the method to stably adapt to the complex and changeable observation conditions in coastal areas and having broad potential for operational applications.

[0125] Embodiment 2

[0126] Corresponding to the aforementioned embodiment of a method for identifying harmful algal bloom areas based on remote sensing data, this application also provides an embodiment of a device for identifying harmful algal bloom areas based on remote sensing data.

[0127] Figure 3 is a schematic diagram of the structure of a second embodiment of the harmful algal bloom area identification device based on remote sensing data provided in this application. Referring to Figure 3, the device provided in this embodiment includes a determining module 310, an acquiring module 320, a combining module 330, a calculating module 340, a correcting module 350, and a judging module 360;

[0128] The determining module 310 is used to determine multiple sensitive bands of harmful algal blooms;

[0129] The acquisition module 320 is used to acquire remote sensing data of the area to be identified in various sensitive bands;

[0130] The determining module 310 is further configured to determine the identification interference factors of the area to be identified;

[0131] The combination module 330 is used to combine the various sensitive bands according to the relationship between the identified interference factors and each sensitive band to obtain multiple sensitive band combinations. Each sensitive band combination plays a different role in the method for identifying harmful algal bloom areas.

[0132] The calculation module 340 is used to calculate the identification interference factor index corresponding to each combination based on the remote sensing data corresponding to the first sensitive band combination and the second sensitive band combination, respectively.

[0133] The correction module 350 is used to correct the remote sensing data corresponding to the first sensitive band combination and the second sensitive band combination according to the identified interference factor index.

[0134] The calculation module 340 is also used to calculate the harmful algal bloom index based on the corrected remote sensing data corresponding to the third sensitive band combination.

[0135] The judgment module 360 ​​is used to determine whether the region to be identified is a harmful algal bloom region based on the harmful algal bloom index and the dynamic threshold corresponding to the region to be identified.

[0136] The apparatus in this embodiment can be used to execute the steps of the method embodiment shown in FIG1. ​​The specific implementation principle and process are similar and will not be described again here.

[0137] 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.

[0138] 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.

[0139] 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 and apparatus for identifying harmful algal bloom areas based on remote sensing data, characterized in that, The method includes: determining multiple sensitive bands for harmful algal blooms; acquiring remote sensing data of the area to be identified in each sensitive band; determining the identification interference factors of the area to be identified; combining the sensitive bands according to the relationship between the identification interference factors and each sensitive band to obtain multiple sensitive band combinations, each sensitive band combination having a different role in the method for identifying harmful algal blooms; calculating the identification interference factor index corresponding to each combination based at least on the remote sensing data corresponding to the first and second sensitive band combinations; correcting the remote sensing data corresponding to the first and second sensitive band combinations based on the identification interference factor index; calculating the harmful algal bloom index based on the corrected remote sensing data corresponding to the third sensitive band combination; determining whether the area to be identified is a harmful algal bloom area based on the harmful algal bloom index and a dynamic threshold corresponding to the area to be identified; the dynamic threshold is based on the real-time environmental conditions of the area to be identified. The determination of the dynamic threshold includes: on the true-color remote sensing image covering the area to be identified on the same day, selecting reference sample pixels based on the image color and texture features to construct a reference sample set, which includes pure water samples and harmful algal bloom samples; calculating the index values ​​of the pure water samples and the harmful algal bloom samples according to the harmful algal bloom index calculation formula, and analyzing the statistical characteristics of the index of each type of sample; determining target parameters based on the statistical characteristics, which include the maximum value of the harmful algal bloom index corresponding to the pure water samples and the minimum value of the harmful algal bloom index corresponding to the harmful algal bloom samples; calculating an initial dynamic threshold based on the maximum and minimum values; applying the initial dynamic threshold to the area to be identified to generate a preliminary algal bloom extraction map, and performing a spatial consistency comparison with the true-color remote sensing image; fine-tuning the initial dynamic threshold based on the comparison results to obtain the final dynamic threshold.

2. The method according to claim 1, characterized in that, The determination of multiple sensitive bands for harmful algal blooms includes: acquiring spectral reflectance data of various harmful algae and various background water bodies; comparing and analyzing the spectral reflectance curves of the harmful algae with the spectral reflectance curves of various background water bodies; and selecting multiple characteristic bands with a difference greater than or equal to a preset threshold between the reflectance of the harmful algae and the background water bodies as the sensitive bands based on the comparison and analysis results.

3. The method according to claim 1, characterized in that, The multiple sensitive bands include the blue light band, which is sensitive to scattering by suspended sediments; the green light band, which is sensitive to scattering by algal cells; and the red light band, which is sensitive to absorption by algal pigments. The interfering factors include cloud cover and suspended sediment interference.

4. The method according to claim 1, characterized in that, The step of combining the sensitive bands according to the relationship between the identified interference factors and each sensitive band to obtain multiple sensitive band combinations includes: for each identified interference factor, analyzing the spectral response characteristics of the current interference factor under different sensitive bands; based on the spectral response characteristics, selecting at least two sensitive bands from the sensitive bands to form a sensitive band combination; wherein each sensitive band in the same sensitive band combination has a spectral response difference with the current interference factor; forming multiple sensitive band combinations for different identified interference factors, and the multiple sensitive band combinations sharing at least one sensitive band.

5. The method according to claim 1, characterized in that, The step of calculating the identification interference factor index corresponding to each combination based on the remote sensing data corresponding to the first sensitive band combination and the second sensitive band combination includes: obtaining the remote sensing reflectance data corresponding to the green band and the red band in the first sensitive band combination; calculating the difference between the reflectance of the green band and the reflectance of the red band, and the sum of the reflectance of the green band and the reflectance of the red band; dividing the difference by the sum to obtain the first identification interference factor index.

6. The method according to claim 1, characterized in that, The step of calculating the identification interference factor index corresponding to each combination based on the remote sensing data corresponding to the first sensitive band combination and the second sensitive band combination, respectively, further includes: obtaining the remote sensing reflectance data corresponding to the blue light band and the green light band in the second sensitive band combination; calculating the difference between the reflectance of the blue light band and the reflectance of the green light band, and the sum of the reflectance of the blue light band and the reflectance of the green light band; dividing the difference by the sum to obtain the second identification interference factor index.

7. The method according to claim 1, characterized in that, The step of correcting the remote sensing data corresponding to the first sensitive band combination and the second sensitive band combination based on the identified interference factor index includes: comparing the first identified interference factor index with a first dynamic threshold, and comparing the second identified interference factor index with a second dynamic threshold; generating a binarized effective pixel mask based on the comparison result; wherein, when a pixel simultaneously satisfies that the first identified interference factor index is greater than the first dynamic threshold and the second identified interference factor index is less than the second dynamic threshold, the pixel is marked as effective in the mask.

8. The method according to claim 7, characterized in that, The step of calculating the harmful algal bloom index based on the corrected remote sensing data corresponding to the third sensitive band combination includes: acquiring the remote sensing reflectance data corresponding to the green band and the red band in the corrected third sensitive band combination; calculating the difference between the reflectance of the green band and the reflectance of the red band; and multiplying the difference by the effective pixel mask to obtain the harmful algal bloom index.

9. A device for identifying harmful algal bloom areas based on remote sensing data, characterized in that, The device includes a determining module, an acquiring module, a combining module, a calculating module, a correcting module, and a judging module. The determining module is used to determine multiple sensitive bands for harmful algal blooms. The acquiring module is used to acquire remote sensing data of the area to be identified in each sensitive band. The determining module is also used to determine the identification interference factors of the area to be identified. The combining module is used to combine the sensitive bands according to the relationship between the identification interference factors and each sensitive band, obtaining multiple sensitive band combinations, each sensitive band combination having a different role in the harmful algal bloom area identification method. The calculating module is used to calculate the identification interference factor index corresponding to each combination, at least based on the remote sensing data corresponding to the first and second sensitive band combinations. The correcting module is used to correct the remote sensing data corresponding to the first and second sensitive band combinations based on the identification interference factor index. The calculating module is also used to calculate a harmful algal bloom index based on the corrected remote sensing data corresponding to the third sensitive band combination. The judging module is used to judge whether the area to be identified is a harmful algal bloom area based on the harmful algal bloom index and the dynamic threshold corresponding to the area to be identified. The dynamic threshold is set based on the real-time environmental conditions of the area to be identified. Determining the dynamic threshold includes: selecting reference sample pixels on the true-color remote sensing image covering the area to be identified on the same day, based on the image's color and texture features, to construct a reference sample set, which includes pure water samples and harmful algal bloom samples; calculating the index values ​​of the pure water samples and the harmful algal bloom samples according to the harmful algal bloom index calculation formula, and analyzing the statistical characteristics of each sample index; determining target parameters based on the statistical characteristics, including the maximum value of the harmful algal bloom index corresponding to the pure water samples and the minimum value of the harmful algal bloom index corresponding to the harmful algal bloom samples; calculating an initial dynamic threshold based on the maximum and minimum values; applying the initial dynamic threshold to the area to be identified to generate a preliminary algal bloom extraction map, and performing a spatial consistency comparison with the true-color remote sensing image; fine-tuning the initial dynamic threshold based on the comparison results to obtain the final dynamic threshold.

Images