Method and system for detecting lake cyanobacterial bloom by fusing remote sensing images

Abstract

The application relates to the technical field of remote sensing monitoring, and discloses a lake cyanobacterial bloom detection method and system fusing remote sensing images. The method comprises the following steps: acquiring multispectral remote sensing images and synthetic aperture radar images of a target lake; calculating phycocyanobilin characteristic indexes and chlorophyll fluorescence peak indexes according to characteristic band reflectivity, screening pixels meeting a discrimination condition to obtain an optical water bloom distribution mask; calculating a radar-retrieved phycocyanobilin index according to a dual-polarization backscattering coefficient through a nonlinear regression model to obtain a radar water bloom distribution mask; and fusing the two masks to obtain a cyanobacterial bloom monitoring result. The application solves the problems that the existing cyanobacterial bloom detection method cannot realize all-weather monitoring, lacks cyanobacterial specificity recognition capability, and has insufficient reliability of a single data source, and improves the timeliness, accuracy and reliability of cyanobacterial bloom detection.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of remote sensing monitoring, and in particular to a lake cyanobacterial bloom detection method and system fusing remote sensing images. BACKGROUND

[0002] Lake cyanobacterial bloom is an ecological phenomenon formed by the rapid reproduction and aggregation of cyanobacteria on the surface layer of water under suitable conditions, which can seriously affect water quality safety and ecosystem health. Remote sensing technology has become an important means for monitoring cyanobacterial bloom due to its advantages of wide range and high frequency. The existing technology mainly uses multispectral remote sensing images to detect cyanobacterial bloom, and identifies the bloom area by calculating optical characteristic parameters such as normalized vegetation index and chlorophyll concentration. This method can obtain good monitoring results under clear sky conditions.

[0003] However, the existing cyanobacterial bloom detection method based on optical remote sensing has obvious deficiencies: on the one hand, optical remote sensing is severely dependent on weather conditions and cannot obtain effective data under cloud and fog obstruction, resulting in poor timeliness of monitoring; on the other hand, the use of a single normalized vegetation index and other general optical indicators lacks specific recognition ability for cyanobacterial pigments, and other algae or aquatic plants may be misjudged as cyanobacterial bloom, so the detection accuracy needs to be improved. SUMMARY

[0004] The present application provides a lake cyanobacterial bloom detection method and system fusing remote sensing images, which solves the problems of existing cyanobacterial bloom detection methods that cannot achieve all-weather monitoring, lack specific recognition ability for cyanobacteria, and single data source has insufficient reliability, and improves the timeliness, accuracy and reliability of cyanobacterial bloom detection.

[0005] In a first aspect, the present application provides a lake cyanobacterial bloom detection method fusing remote sensing images, which comprises:

[0006] Step S1, obtaining multispectral remote sensing images and synthetic aperture radar images of a target lake, wherein the multispectral remote sensing images comprise a plurality of characteristic band reflectances, and the synthetic aperture radar images comprise dual-polarization backscatter coefficients;

[0007] Step S2, calculating phycocyanobilin characteristic index and chlorophyll fluorescence peak index according to the plurality of characteristic band reflectances;

[0008] Step S3, screening pixels that simultaneously satisfy the phycocyanobilin characteristic index and the chlorophyll fluorescence peak index discrimination conditions to obtain an optical bloom distribution mask;

[0009] Step S4, calculating the radar inversion phycocyanin index according to the dual-polarized backscattering coefficient through a nonlinear regression model, screening the pixels satisfying the discrimination condition to obtain a radar water bloom distribution mask, and fusing the optical water bloom distribution mask and the radar water bloom distribution mask to obtain a blue-green algae water bloom monitoring result.

[0010] In a second aspect, the present application provides a lake blue-green algae water bloom detection system fusing remote sensing images, which comprises:

[0011] An extraction module is configured to acquire a multispectral remote sensing image and a synthetic aperture radar image of a target lake, wherein the multispectral remote sensing image comprises a plurality of characteristic band reflectances, and the synthetic aperture radar image comprises a dual-polarized backscattering coefficient.

[0012] A calculation module is configured to calculate a phycocyanin characteristic index and a chlorophyll fluorescence peak index according to the plurality of characteristic band reflectances.

[0013] A screening module is configured to screen pixels satisfying both the discrimination condition of the phycocyanin characteristic index and the discrimination condition of the chlorophyll fluorescence peak index to obtain an optical water bloom distribution mask.

[0014] A fusion module is configured to calculate a radar inversion phycocyanin index according to the dual-polarized backscattering coefficient through a nonlinear regression model, screen pixels satisfying the discrimination condition to obtain a radar water bloom distribution mask, and fuse the optical water bloom distribution mask and the radar water bloom distribution mask to obtain a blue-green algae water bloom monitoring result.

[0015] In a third aspect, a lake blue-green algae water bloom detection device fusing remote sensing images is provided, which comprises a memory and at least one processor, wherein the memory stores instructions, and the at least one processor invokes the instructions in the memory to enable the lake blue-green algae water bloom detection device fusing remote sensing images to perform the lake blue-green algae water bloom detection method fusing remote sensing images.

[0016] In a fourth aspect, a computer readable storage medium is provided, which stores instructions, and when the instructions are run on a computer, the computer is enabled to perform the lake blue-green algae water bloom detection method fusing remote sensing images.

[0017] The technical scheme provided in the application overcomes the limitations of the prior art by fusing multispectral remote sensing images and synthetic aperture radar images for lake cyanobacterial bloom detection. The dual-index discrimination mechanism of phycocyanin characteristic index and chlorophyll fluorescence peak index is used to accurately identify the phycocyanin pigment absorption characteristics and chlorophyll fluorescence peak characteristics specific to cyanobacteria. Compared with traditional general indicators such as normalized vegetation index, the dual-index discrimination mechanism can effectively distinguish cyanobacterial blooms from other algae or aquatic plants, and significantly improves the specificity and accuracy of detection. A nonlinear regression model based on dual-polarized backscatter coefficient is constructed to calculate the radar inversion phycocyanin index, and a quantitative relationship between the radar backscatter characteristics and the cyanobacterial biomass is established, so that the synthetic aperture radar image can continue to provide effective monitoring data under adverse weather conditions such as clouds and fog, solving the problem of severe dependence of optical remote sensing on weather conditions and realizing all-weather monitoring capability. By fusing the optical water bloom distribution mask and the radar water bloom distribution mask, the advantages of high spectral resolution of optical remote sensing under clear sky conditions and the all-day and all-weather observation advantages of radar remote sensing are fully utilized, forming a complementary dual verification mechanism that improves the reliability of the detection results and enhances the timeliness and continuity of the monitoring, providing a more stable and reliable technical means for operational monitoring of lake cyanobacterial blooms.

[0018] The introduction of the nonlinear regression model establishes a mathematical mapping relationship between the radar polarized backscatter ratio and the phycocyanin concentration. The model describes the nonlinear response law between the water surface roughness change caused by the aggregation of cyanobacterial blooms and the radar echo through a logarithmic function form, which is more consistent with the physical process of cyanobacterial bloom radar scattering than a simple linear relationship, and improves the inversion accuracy of radar data. The dual-mask fusion strategy realizes the synergy of heterogeneous data sources through logical discrimination and spatial superposition, and takes into account the advantages and limitations of optical and radar in algorithm design, forming a more robust comprehensive discrimination system that can maintain high spatial resolution and ensure the integrity of the time series, significantly improving the practical value and operational application level of cyanobacterial bloom monitoring as a whole. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0020] Figure 1 An embodiment of the lake cyanobacterial bloom detection method of the application is shown in the following figure.

[0021] Figure 2An embodiment of a lake cyanobacterial bloom detection system fusing remote sensing images in the embodiment of the application is shown in the figure;

[0022] Figure 3 An embodiment of a lake cyanobacterial bloom detection system fusing remote sensing images in the embodiment of the application is shown in the figure; DETAILED DESCRIPTION

[0023] The embodiment of the application provides a lake cyanobacterial bloom detection method and system fusing remote sensing images. The terms "first", "second", "third", "fourth" and the like (if any) in the specification and claims of the application and the above-mentioned drawings are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" or "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to only those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to the process, method, product or device.

[0024] For the convenience of understanding, the specific process of the embodiment of the application is described below. Please refer to Figure 1 An embodiment of a lake cyanobacterial bloom detection method fusing remote sensing images in the embodiment of the application includes the following steps.

[0025] Step S1, obtaining multispectral remote sensing images and synthetic aperture radar images of a target lake, the multispectral remote sensing images including a plurality of characteristic band reflectivities, and the synthetic aperture radar images including dual-polarization backscatter coefficients;

[0026] Step S2, calculating phycocyanobilin characteristic index and chlorophyll fluorescence peak index according to the plurality of characteristic band reflectivities;

[0027] Step S3, screening pixels that satisfy the discrimination conditions of the phycocyanobilin characteristic index and the chlorophyll fluorescence peak index to obtain an optical water bloom distribution mask;

[0028] Step S4, calculating a radar-inverted phycocyanobilin index through a nonlinear regression model according to the dual-polarization backscatter coefficients, screening pixels that satisfy the discrimination conditions of the radar-inverted phycocyanobilin index to obtain a radar water bloom distribution mask, and fusing the optical water bloom distribution mask and the radar water bloom distribution mask to obtain a cyanobacterial bloom monitoring result.

[0029] Specifically, when acquiring the multispectral remote sensing image and the synthetic aperture radar image of the target lake, the Sentinel-3 / OLCI multispectral remote sensing image original data and the Sentinel-1 synthetic aperture radar image original data are acquired through a satellite platform. The multispectral remote sensing image original data records the original digital quantization value received by the sensor, and needs to be subjected to a radiometric calibration process to convert the digital quantization value into a physical quantity. The radiometric calibration process calculates the top-of-atmosphere reflectance data according to the sensor calibration coefficient. The reflectance data contains the influence of atmospheric scattering and absorption, and needs to be further subjected to an atmospheric correction process. The atmospheric correction process simulates the transmission process of light in the atmosphere by using a radiative transfer model, inputs the solar zenith angle, observation zenith angle, relative azimuth angle and aerosol optical depth and other parameters at the imaging time, and calculates to eliminate the influence of Rayleigh scattering and aerosol scattering to obtain water surface ex-water reflectance data. The data truly reflects the spectral reflectance characteristics of the water surface. The synthetic aperture radar image original data also records the echo signal strength received by the radar, and is subjected to a radiometric calibration process to convert the signal strength into backscatter intensity data. The data is affected by the terrain undulation and will produce geometric distortion, and needs to be subjected to a terrain correction process. The terrain correction process uses a range-Doppler terrain correction algorithm to calculate the range and Doppler offset caused by the terrain by using a digital elevation model, performs geometric correction on the image, and obtains backscatter coefficient data. When extracting the phycocyanin absorption characteristic band from the water surface ex-water reflectance data, a band with a center wavelength of 620 nanometers is selected as the phycocyanin absorption characteristic band because the phycocyanin pigment in the blue-green algae cells has a strong absorption peak at this band. A band with a center wavelength of 665 nanometers is selected as the chlorophyll absorption characteristic band, a band with a center wavelength of 681 nanometers is selected as the chlorophyll fluorescence characteristic band, and a band with a center wavelength of 709 nanometers is selected as the near-infrared characteristic band. The reflectance data of the four bands constitutes multiple characteristic band reflectances. When extracting the vertical polarization component and the cross-polarization component from the backscatter coefficient data, the vertical polarization component corresponds to the backscatter coefficient of vertical emission and vertical reception, and the cross-polarization component corresponds to the backscatter coefficient of vertical emission and horizontal reception. The backscatter coefficients of the two polarization components constitute dual-polarization backscatter coefficients.

[0030] When calculating the phycocyanin characteristic index from the multiple characteristic band reflectances, the phycocyanin absorption characteristic band reflectance and the near-infrared characteristic band reflectance are extracted from the multiple characteristic band reflectances, and the two band reflectances are subjected to difference normalization processing. The difference normalization processing first calculates the difference between the phycocyanin absorption characteristic band reflectance and the near-infrared characteristic band reflectance, which reflects the absolute difference between the two band reflectances. Then, the sum of the phycocyanin absorption characteristic band reflectance and the near-infrared characteristic band reflectance is calculated. The normalized phycocyanin characteristic index is obtained by dividing the difference by the sum. The index value ranges from negative one to positive one. When the concentration of blue-green algae in the water body increases, the absorption of light at the 620 nm band by the phycocyanin pigment is enhanced, resulting in a decrease in the reflectance at this band. The near-infrared band reflectance is less affected by blue-green algae, so the difference becomes larger, and the phycocyanin characteristic index value increases. When calculating the chlorophyll fluorescence peak index, the chlorophyll absorption characteristic band reflectance, the chlorophyll fluorescence characteristic band reflectance, and the near-infrared characteristic band reflectance are extracted from the multiple characteristic band reflectances. The chlorophyll absorption characteristic band reflectance and the near-infrared characteristic band reflectance are subjected to linear baseline construction processing. The linear baseline construction processing takes the two band reflectances as the two endpoints of a linear function, calculates the function value of the linear function at the position of the chlorophyll fluorescence characteristic band as the baseline reflectance value, and represents the reflectance at this band without considering the emission of chlorophyll fluorescence. Then, the chlorophyll fluorescence characteristic band reflectance and the baseline reflectance value are subjected to difference calculation processing. The difference calculation processing directly calculates the chlorophyll fluorescence characteristic band reflectance minus the baseline reflectance value to obtain the chlorophyll fluorescence peak index. The index reflects the emission intensity of chlorophyll fluorescence at the 681 nm band. When the concentration of chlorophyll a in the water body increases, chlorophyll a emits fluorescence at the 681 nm band after being excited by sunlight, resulting in a higher actual reflectance at this band than the baseline reflectance value, and the chlorophyll fluorescence peak index value increases. When performing per-pixel mapping processing on the phycocyanin characteristic index and the chlorophyll fluorescence peak index, each pixel in the image is traversed, and the phycocyanin characteristic index and the chlorophyll fluorescence peak index of the pixel are calculated according to the corresponding values of the pixel in the multiple characteristic band reflectances. The calculated index values are assigned to the pixel. After the calculation of all pixels is completed, the phycocyanin characteristic index image and the chlorophyll fluorescence peak index image are obtained. The index values of each pixel in the two index images correspond to the phycocyanin spectral characteristics and the chlorophyll fluorescence spectral characteristics at the same position, respectively.

[0031] When screening the pixels satisfying the discrimination conditions of phycocyanin characteristic index and chlorophyll fluorescence peak index to obtain the optical water bloom distribution mask, the phycocyanin characteristic index image and the chlorophyll fluorescence peak index image are subjected to double-threshold discrimination processing. The double-threshold discrimination processing needs to determine the phycocyanin characteristic index threshold and the chlorophyll fluorescence peak index threshold. When determining the two thresholds, the phycocyanin characteristic index values and the chlorophyll fluorescence peak index values of the sample pixels in the known cyanobacterial water bloom region of the target lake are counted. The sample pixels in the known cyanobacterial water bloom region are determined by artificial investigation or historical monitoring records. The regions in which the cyanobacterial water bloom has occurred in the lake are determined, and the pixels in these regions are extracted as sample pixels. When counting the phycocyanin characteristic index values of the sample pixels, the arithmetic mean of the phycocyanin characteristic index values of all sample pixels is calculated to obtain the mean value of the phycocyanin characteristic index values. The square of the difference between the phycocyanin characteristic index value of each sample pixel and the mean value is calculated, and the arithmetic mean of the square values is calculated to obtain the standard deviation of the phycocyanin characteristic index values. The mean value and the standard deviation of the chlorophyll fluorescence peak index values are calculated in the same way. When determining the threshold according to the mean value and the standard deviation of the phycocyanin characteristic index values, the threshold is obtained by subtracting twice the standard deviation from the mean value. This threshold corresponds to the lower limit of the distribution covering about 95 percent of the samples. When determining the threshold according to the mean value and the standard deviation of the chlorophyll fluorescence peak index values, the threshold is obtained by subtracting 1.5 times the standard deviation from the mean value. When traversing all pixels of the phycocyanin characteristic index image and the chlorophyll fluorescence peak index image, the value of each pixel in the phycocyanin characteristic index image and the value of each pixel in the chlorophyll fluorescence peak index image are read in sequence. It is judged whether the phycocyanin characteristic index value of the pixel is greater than the phycocyanin characteristic index threshold and whether the chlorophyll fluorescence peak index value of the pixel is greater than the chlorophyll fluorescence peak index threshold. If both conditions are met, the pixel is screened as a suspected water bloom pixel. After the judgment of all pixels is completed, a set of suspected water bloom pixels is obtained. When performing eight-neighbor connected component labeling processing on the set of suspected water bloom pixels, the eight-neighbor connected component labeling processing is a method of image connectivity analysis. For each pixel in the set of suspected water bloom pixels, it is checked whether the adjacent pixels in eight directions around the pixel also belong to the set of suspected water bloom pixels. If the adjacent pixels are also suspected water bloom pixels, the two pixels are labeled as belonging to the same connected domain. Through iterative scanning of all suspected water bloom pixels, the spatially connected suspected water bloom pixels are aggregated into a connected domain. Each connected domain constitutes a suspected water bloom pixel cluster, and multiple suspected water bloom pixel clusters are obtained. When calculating the number of pixels in each suspected water bloom pixel cluster, the number of pixels contained in each suspected water bloom pixel cluster is counted.When the area screening processing is performed on the multiple suspected water bloom pixel clusters according to the number of pixels, the number of pixels of each suspected water bloom pixel cluster is multiplied by the area of a single pixel to obtain the total area of the suspected water bloom pixel cluster, the area of a single pixel is determined by the spatial resolution of the image, whether the total area of each suspected water bloom pixel cluster is greater than an area threshold is judged, the area threshold is set to be 0.5 square kilometers, the suspected water bloom pixel cluster with the total area greater than the area threshold is retained, the suspected water bloom pixel cluster with the total area less than the area threshold is removed, and the retained suspected water bloom pixel cluster constitutes the optical water bloom distribution mask.

[0032] When calculating the phycocyanin index by the nonlinear regression model according to the dual-polarized backscatter coefficient, first, the vertical polarization backscatter coefficient and the cross polarization backscatter coefficient are extracted from the dual-polarized backscatter coefficient, and the polarization backscatter ratio of the cross polarization backscatter coefficient to the vertical polarization backscatter coefficient is calculated, which reflects the difference in scattering characteristics of radar waves on the water surface. When there is a blue-green algae biofilm on the water surface, the biofilm increases the roughness of the water surface, resulting in an increase in cross polarization backscatter and polarization backscatter ratio. When the polarization backscatter ratio is substituted into the nonlinear regression model for logarithmic regression calculation and processing, the construction of the nonlinear regression model requires a training data set. The historical multispectral remote sensing image and the historical synthetic aperture radar image obtained simultaneously under clear sky conditions are selected as the training data set. Training sample pixels are extracted from the training data set in the confirmed blue-green algae bloom area. The blue-green algae bloom area is determined by the method of steps two and three in the historical multispectral remote sensing image. The pixels corresponding to the position of the historical synthetic aperture radar image in these areas are extracted as training sample pixels. The polarization backscatter ratio and the phycocyanin characteristic index value of the training sample pixels are counted, and a scatter plot of the polarization backscatter ratio and the phycocyanin characteristic index value is drawn to find that there is a logarithmic relationship between them. The least squares method is used to fit and process to construct a nonlinear regression model containing a logarithmic function and a linear coefficient. The least squares method solves the model parameters by minimizing the sum of squares of residuals between the predicted values and the actual observed values. The model form is that the radar inversion phycocyanin index equals the first coefficient multiplied by the polarization backscatter ratio plus the natural logarithm of the second coefficient plus the third coefficient. The first coefficient, the second coefficient, and the third coefficient values that minimize the sum of squares of residuals are solved by iterative calculation to obtain the regression coefficients of the nonlinear regression model. When the polarization backscatter ratio of the target lake at the current time is substituted into the nonlinear regression model for logarithmic operation and linear transformation processing, first, the polarization backscatter ratio is added to the second coefficient, the sum is calculated, the logarithm of the result is calculated, the logarithm result is multiplied by the first coefficient, and the third coefficient is added. The radar inversion phycocyanin index is calculated pixel by pixel, and the radar inversion phycocyanin index image is obtained by assigning the radar inversion phycocyanin index to the corresponding pixel spatial position. When screening the pixels in the radar inversion phycocyanin index image that are greater than the phycocyanin discrimination threshold, the phycocyanin discrimination threshold remains consistent with the phycocyanin characteristic index threshold determined in step three. All pixels in the radar inversion phycocyanin index image are traversed, and it is determined whether the radar inversion phycocyanin index value of each pixel is greater than the phycocyanin discrimination threshold. If it is greater than the threshold, the pixel is screened. When performing connected component labeling and area screening processing on the screened pixels, the same eight-neighbor connected component labeling method as in step three is used to aggregate spatially connected screened pixels into connected components. The total area of each connected component is calculated, and connected components with a total area greater than zero point five square kilometers are retained to obtain the radar bloom distribution mask.When the radar water bloom distribution mask and the optical water bloom distribution mask are superimposed and fused, the pixels at corresponding positions of the two masks in the image space coordinate system are subjected to logical OR operation. If a pixel is marked as a water bloom pixel in the optical water bloom distribution mask or in the radar water bloom distribution mask, the pixel is marked as a water bloom pixel in the fusion result. After the fusion of all pixels is completed, the cyanobacterial water bloom monitoring result is obtained. The monitoring result integrates the spectral feature information of the optical remote sensing image and the all-weather observation capability of the synthetic aperture radar image. Under sunny conditions, the monitoring result mainly depends on the optical water bloom distribution mask, and under cloud cover conditions, the radar water bloom distribution mask is used to supplement the spatial distribution information of the water bloom.

[0033] In a specific embodiment, step S1 comprises:

[0034] The multispectral remote sensing image original data of the target lake is acquired, radiation calibration is performed on the multispectral remote sensing image original data to obtain atmospheric top reflectance data, and atmospheric correction is performed on the atmospheric top reflectance data to obtain water surface off-water reflectance data.

[0035] The synthetic aperture radar image original data of the target lake is acquired, radiation calibration is performed on the synthetic aperture radar image original data to obtain backscattering intensity data, and terrain correction is performed on the backscattering intensity data to obtain backscattering coefficient data.

[0036] The phycocyanobilin absorption characteristic band, the chlorophyll absorption characteristic band, the chlorophyll fluorescence characteristic band and the near-infrared characteristic band in the water surface off-water reflectance data are extracted as a plurality of characteristic band reflectances, and the vertical polarization component and the cross-polarization component in the backscattering coefficient data are extracted as dual-polarization backscattering coefficients.

[0037] Specifically, when acquiring the multispectral remote sensing image original data of the target lake, the satellite sensor records the digital quantization value DN value after analog-digital conversion. The size of the DN value depends on the intensity of the received radiation energy, but the DN value itself has no physical meaning and needs to be converted into the physically meaningful radiation brightness or reflectivity through radiation calibration processing. The radiation calibration processing is based on the calibration coefficient obtained by the on-orbit calibration of the sensor. The calibration coefficient includes the gain coefficient and the offset coefficient. A linear transformation formula is applied to the DN value of each band. The DN value is multiplied by the gain coefficient of the band and then added to the offset coefficient to obtain the radiation brightness of the band at the entrance pupil of the sensor. Then, according to the solar irradiance and the solar zenith angle, the radiation brightness is converted into the top-of-atmosphere reflectance data. During the conversion process, the variation of the distance from the sun to the earth with time needs to be considered, and the geodetic distance correction factor is introduced. The top-of-atmosphere reflectance data represents the total reflection ratio of the solar radiation by the earth's surface and the atmospheric system, and contains the comprehensive information of atmospheric molecular scattering, aerosol scattering and real surface reflection. When performing atmospheric correction processing on the top-of-atmosphere reflectance data, a radiative transfer model is used to simulate the transmission process of sunlight in the atmosphere. The radiative transfer model divides the atmosphere into multiple uniform layers, calculates the direct radiation and scattered radiation of sunlight passing through the atmosphere layer to reach the ground, and the proportion of absorption and scattering of the reflected light from the ground passing through the atmosphere layer to reach the sensor. The input parameters of the model include the solar zenith angle at the imaging time, which reflects the incident direction of sunlight; the observation zenith angle, which reflects the observation direction of the sensor; the relative azimuth angle, which reflects the relative position relationship between the sun and the sensor; and the aerosol optical depth, which reflects the concentration of aerosol particles in the atmosphere. The aerosol optical depth is usually estimated from meteorological observation data or remote sensing image itself. The radiative transfer model calculates the conversion relationship between the top-of-atmosphere reflectance and the water surface reflectance, and converts the top-of-atmosphere reflectance data into the water surface reflectance data through a lookup table or an analytical expression. The water surface reflectance data is the reflection ratio of the light propagating from the water surface to the atmosphere relative to the incident solar radiation, which eliminates the influence of atmospheric scattering and absorption and truly reflects the spectral reflectance characteristics of the water body itself. This data is the basis for subsequent calculation of the phycocyanin characteristic index and the chlorophyll fluorescence peak index.

[0038] When acquiring the synthetic aperture radar image original data of a target lake, the radar sensor records the amplitude and phase information of the backscattering echo signal of the ground target to the radar wave. The original data is stored in the form of complex number, containing real part and imaginary part. After the pulse compression processing in the range direction and the azimuth direction, the single-view complex image is formed. The modulus value of each pixel value of the single-view complex image reflects the backscattering intensity of the ground target corresponding to the pixel. When performing the radiation scaling processing on the synthetic aperture radar image original data, the backscattering intensity is converted into the normalized backscattering coefficient according to the calibration parameters of the radar system. The radiation scaling needs to consider the influence of the radar transmitting power, the antenna gain, the propagation distance, the incident angle and other factors on the echo signal. The radiation scaling formula divides the backscattering intensity by the system constant and the distance attenuation factor in the radar equation to obtain the backscattering coefficient data. The backscattering coefficient is usually expressed in decibels, that is, the linear value is taken as the logarithm with base ten and then multiplied by ten. The decibel value is convenient for compressing the data dynamic range and intuitively comparing the scattering characteristics of different targets. When performing the terrain correction processing on the backscattering intensity data, the terrain fluctuation causes the difference in the incident angle of the radar wave at different terrain positions. The change of the incident angle causes the change of the backscattering coefficient. At the same time, the terrain fluctuation causes the geometric distortion of the radar image, including the slope shortening and stretching and the shadow area. The terrain correction processing adopts the range-Doppler terrain correction algorithm. The algorithm uses the digital elevation model to provide the terrain elevation information. The digital elevation model records the elevation value of each position on the ground. The algorithm first establishes the mapping relationship between the radar image coordinates and the geographic coordinates according to the radar imaging geometry and the digital elevation model. For each pixel in the radar image, the corresponding ground position and elevation are calculated according to its position in the radar coordinate system. Then, the radar incident angle corresponding to the pixel is calculated according to the ground position and the elevation. The radiation normalization processing is performed on the backscattering coefficient according to the incident angle, so as to eliminate the influence of the terrain fluctuation on the incident angle on the backscattering coefficient. At the same time, the radar image is resampled from the radar coordinate system to the geographic coordinate system, so as to eliminate the geometric distortion and obtain the backscattering coefficient data. The data has a unified spatial reference in the geographic coordinate system, and the backscattering coefficient value is normalized to the scattering characteristics under the standard incident angle condition.

[0039] When the algal blue pigment absorption characteristic band, the chlorophyll absorption characteristic band, the chlorophyll fluorescence characteristic band and the near-infrared characteristic band in the water surface off-water reflectivity data are extracted as multiple characteristic band reflectivity, the water surface off-water reflectivity data contains reflectivity information of all bands of the sensor, and specific bands are selected according to the spectral absorption characteristics of blue-green algae pigments for extraction. Algal blue pigment is a specific auxiliary pigment of blue-green algae cells, and there is an absorption peak in the visible red light band. The band with a center wavelength of 620 nm is selected as the algal blue pigment absorption characteristic band. The reflectivity value of this band is significantly affected by the concentration of algal blue pigment in blue-green algae cells. When the concentration of algal blue pigment increases, the reflectivity of this band decreases. Chlorophyll a is a photosynthetic pigment common to all algae, and there are absorption peaks in the blue light band and the red light band. The band with a center wavelength of 665 nm is selected as the chlorophyll absorption characteristic band. This band is located at the red light absorption peak position of chlorophyll a. Chlorophyll a emits fluorescence after being excited by sunlight, and the fluorescence emission peak is located at the edge of the near-infrared band. The band with a center wavelength of 681 nm is selected as the chlorophyll fluorescence characteristic band. The reflectivity value of this band contains the contribution of chlorophyll a fluorescence emission. The band with a center wavelength of 709 nm is selected as the near-infrared characteristic band. This band is located outside the absorption peaks of chlorophyll and algal blue pigment, and the reflectivity value is mainly affected by the scattering of water body suspended matter and is not sensitive to pigment absorption. The extraction operation is to read the reflectivity values of the four bands from the water surface off-water reflectivity data to form a data set of multiple characteristic band reflectivity. This data set contains reflectivity information of each pixel in the four characteristic bands. When the vertical polarization component and the cross-polarization component in the backscattering coefficient data are extracted as dual-polarization backscattering coefficients, the backscattering coefficient data contains backscattering coefficient information of multiple polarization channels obtained by the radar sensor. The polarization of radar waves is divided into vertical polarization and horizontal polarization. The channel that transmits vertical polarization radar waves and receives vertical polarization backscattering is called the vertical polarization channel, and the recorded backscattering coefficient is the vertical polarization component. The channel that transmits vertical polarization radar waves and receives horizontal polarization backscattering is called the cross-polarization channel, and the recorded backscattering coefficient is the cross-polarization component. The vertical polarization component reflects the scattering characteristics of the same polarization radar wave by the ground target, which is mainly affected by the target surface roughness and dielectric constant. The cross-polarization component reflects the scattering characteristics of the cross-polarization radar wave by the ground target, which is mainly affected by the target internal structure and volume scattering. When there is a blue-green algae biofilm on the surface of the water body, the rough surface and internal bubble structure formed by the biofilm enhance the cross-polarization scattering, resulting in an increase in the cross-polarization component. The extraction operation is to read the backscattering coefficient values of the vertical polarization channel and the cross-polarization channel from the backscattering coefficient data to form a data set of dual-polarization backscattering coefficients. This data set contains vertical polarization backscattering coefficient and cross-polarization backscattering coefficient information of each pixel.

[0040] In a specific embodiment, step S2 comprises:

[0041] extracting phycocyanin absorption characteristic band reflectivity and near-infrared characteristic band reflectivity from the multiple characteristic band reflectivities, and performing difference normalization processing on the phycocyanin absorption characteristic band reflectivity and the near-infrared characteristic band reflectivity to obtain a phycocyanin characteristic index;

[0042] extracting chlorophyll absorption characteristic band reflectivity, chlorophyll fluorescence characteristic band reflectivity, and near-infrared characteristic band reflectivity from the multiple characteristic band reflectivities, performing linear baseline construction processing on the chlorophyll absorption characteristic band reflectivity and the near-infrared characteristic band reflectivity to obtain a baseline reflectivity value, and performing difference calculation processing on the chlorophyll fluorescence characteristic band reflectivity and the baseline reflectivity value to obtain a chlorophyll fluorescence peak index;

[0043] performing per-pixel mapping processing on the phycocyanin characteristic index and the chlorophyll fluorescence peak index to obtain a phycocyanin characteristic index image and a chlorophyll fluorescence peak index image.

[0044] Specifically, when the phycocyanin absorption characteristic band reflectance and the near-infrared characteristic band reflectance are extracted from the plurality of characteristic band reflectances, the plurality of characteristic band reflectance data sets contain reflectance values of a 620-nanometer phycocyanin absorption characteristic band, a 665-nanometer chlorophyll absorption characteristic band, a 681-nanometer chlorophyll fluorescence characteristic band, and a 709-nanometer near-infrared characteristic band, and the extraction operation selects the 620-nanometer band reflectance as the phycocyanin absorption characteristic band reflectance and selects the 709-nanometer band reflectance as the near-infrared characteristic band reflectance. When the phycocyanin absorption characteristic band reflectance and the near-infrared characteristic band reflectance are subjected to difference normalization processing, the difference normalization processing is a commonly used spectral index calculation method, which enhances the target feature and suppresses the background interference by calculating the ratio of the difference value and the sum value of the reflectance of two bands. The specific calculation process first calculates the difference value obtained by subtracting the near-infrared characteristic band reflectance from the phycocyanin absorption characteristic band reflectance, which reflects the absolute difference between the reflectance of two bands. Then, the sum value obtained by adding the phycocyanin absorption characteristic band reflectance and the near-infrared characteristic band reflectance is calculated, which reflects the overall level of the reflectance of two bands. The phycocyanin feature index is obtained by dividing the difference value by the sum value, and the index value ranges from negative one to positive one. The normalization processing eliminates the influence of the absolute value of the reflectance, so that the index under different water body conditions is comparable. The difference normalization processing utilizes the strong absorption characteristic of phycocyanin of blue-green algae at the 620-nanometer band. When the concentration of blue-green algae in the water body is low, the reflectance at the 620-nanometer band is close to the reflectance at the 709-nanometer band, the difference value is close to zero, and the phycocyanin feature index is close to zero. When the concentration of blue-green algae in the water body increases, the absorption of phycocyanin to light at the 620-nanometer band is enhanced, resulting in a significant decrease in the reflectance at the 620-nanometer band. The 709-nanometer band is outside the phycocyanin absorption peak, and the reflectance is less affected by blue-green algae. Therefore, the reflectance at the 620-nanometer band is less than the reflectance at the 709-nanometer band, the difference value is negative and has a large absolute value, and the phycocyanin feature index is negative and has a large absolute value.

[0045] When extracting the chlorophyll absorption characteristic band reflectance, the chlorophyll fluorescence characteristic band reflectance and the near-infrared characteristic band reflectance from the multiple characteristic band reflectances, the extraction operation selects the 665 nm band reflectance as the chlorophyll absorption characteristic band reflectance, selects the 681 nm band reflectance as the chlorophyll fluorescence characteristic band reflectance, and selects the 709 nm band reflectance as the near-infrared characteristic band reflectance. When performing the linear baseline construction processing on the chlorophyll absorption characteristic band reflectance and the near-infrared characteristic band reflectance, the linear baseline construction processing is a basic method for extracting the fluorescence peak height. The linear baseline is constructed based on the 665 nm and 709 nm bands to estimate the reflectance baseline value of the 681 nm band in the case of no fluorescence emission. The linear baseline assumes that the reflectance changes linearly with the wavelength between the 665 nm and 709 nm bands. The 665 nm band reflectance and the 709 nm band reflectance are taken as two endpoints of a linear function. The 665 nm band corresponds to the abscissa 665 and the ordinate is the reflectance of the band. The 709 nm band corresponds to the abscissa 709 and the ordinate is the reflectance of the band. A straight line is formed by connecting the two points. The ordinate value of the straight line at the abscissa 681 nm is calculated as the baseline reflectance value. Since the 681 nm wavelength is between the 665 nm and 709 nm wavelengths, the baseline reflectance value is calculated by using a linear interpolation method. The wavelength difference from 681 nm to 665 nm is divided by the wavelength difference from 709 nm to 665 nm to obtain an interpolation weight. The reflectance change amount is obtained by subtracting the 665 nm band reflectance from the 709 nm band reflectance. The reflectance increment at 681 nm is obtained by multiplying the reflectance change amount by the interpolation weight. The baseline reflectance value is obtained by adding the reflectance increment to the 665 nm band reflectance. The baseline reflectance value represents the reflectance level of the 681 nm band without considering the chlorophyll fluorescence emission.

[0046] When the phycocyanin characteristic index and the chlorophyll fluorescence peak index are processed by the pixel-by-pixel mapping, the pixel-by-pixel mapping is to assign the index value calculated by each pixel to the corresponding position in the image space, to generate an index image with the same spatial resolution and geographical range as the original remote sensing image, and the mapping process traverses all the pixels in the multiple characteristic band reflectivity data sets, and performs the phycocyanin characteristic index calculation and the chlorophyll fluorescence peak index calculation for each pixel in turn, reads the 620 nm band reflectivity and the 709 nm band reflectivity of the pixel, calculates the phycocyanin characteristic index value of the pixel according to the difference normalization processing method, stores the calculation result in the spatial position corresponding to the pixel in the phycocyanin characteristic index image, reads the 665 nm band reflectivity, the 681 nm band reflectivity and the 709 nm band reflectivity of the pixel, calculates the chlorophyll fluorescence peak index value of the pixel according to the linear baseline construction processing and the difference calculation processing method, and stores the calculation result in the spatial position corresponding to the pixel in the chlorophyll fluorescence peak index image. After the traversal and calculation of all the pixels are completed, the phycocyanin characteristic index image and the chlorophyll fluorescence peak index image are obtained. The value of each pixel in the two index images respectively records the spectral characteristic intensity of the phycocyanin of the cyanobacteria and the fluorescence emission intensity of the chlorophyll a at the spatial position. The index image retains the spatial structure information of the original remote sensing image, the spatial proximity relationship between the pixels is consistent with the original image, the geographical coordinates of the index image are the same as those of the original image, and each pixel coordinate in the index image corresponds to the pixel position with the same coordinate in the original image.

[0047] In a specific embodiment, step S3 comprises:

[0048] The phycocyanin characteristic index image and the chlorophyll fluorescence peak index image are subjected to double-threshold discrimination processing, and the pixels that satisfy both the phycocyanin characteristic index threshold and the chlorophyll fluorescence peak index threshold are screened to obtain a suspected water bloom pixel set;

[0049] The suspected water bloom pixel set is subjected to eight-neighbor connected component labeling processing to obtain a plurality of suspected water bloom pixel clusters, and the number of pixels in each suspected water bloom pixel cluster is calculated.

[0050] The plurality of suspected water bloom pixel clusters are subjected to area screening processing according to the number of pixels, and the suspected water bloom pixel clusters whose number of pixels satisfy the area threshold are retained to obtain an optical water bloom distribution mask.

[0051] Specifically, the cyanobacterial bloom pixel is identified by simultaneously constraining the phycocyanin characteristic index and the chlorophyll fluorescence peak index, which requires determining the phycocyanin characteristic index threshold value and the chlorophyll fluorescence peak index threshold value in advance. The threshold value is determined based on statistical analysis of sample pixels in known cyanobacterial bloom areas. Sample pixels in cyanobacterial bloom areas in lakes are collected through historical monitoring or artificial investigation confirmation. The values of these sample pixels in the phycocyanin characteristic index image are extracted. The arithmetic mean of the phycocyanin characteristic index values of all sample pixels is calculated as the mean value. The average of the square deviations of each sample pixel value from the mean value is calculated and then the square root is taken as the standard deviation. The mean value and the standard deviation of the chlorophyll fluorescence peak index values are also calculated in the same way. The threshold value is determined using the standard deviation multiple method in statistics. The mean value of the phycocyanin characteristic index values is subtracted by twice the standard deviation to obtain the phycocyanin characteristic index threshold value, which covers about 95% of the sample data based on the assumption of normal distribution. The mean value of the chlorophyll fluorescence peak index values is subtracted by 1.5 times the standard deviation to obtain the chlorophyll fluorescence peak index threshold value. Different multiple coefficients reflect the difference in strictness of the discrimination of the two indexes. When screening the pixels that meet the phycocyanin characteristic index threshold value and the chlorophyll fluorescence peak index threshold value at the same time, each pixel in the phycocyanin characteristic index image and the chlorophyll fluorescence peak index image is traversed. The value of the pixel in the phycocyanin characteristic index image is read to determine whether it is greater than the phycocyanin characteristic index threshold value. The value of the pixel in the chlorophyll fluorescence peak index image is read to determine whether it is greater than the chlorophyll fluorescence peak index threshold value. Only when both conditions are met, the pixel is marked as a suspected bloom pixel and added to the suspected bloom pixel set. After the traversal is completed, the suspected bloom pixel set contains all pixels that meet the double threshold conditions.

[0052] When the eight-neighbor connected component labeling processing is performed on the suspected water bloom pixel set, the eight-neighbor connected component labeling processing is an image connectivity analysis algorithm, which is used to aggregate the spatially adjacent suspected water bloom pixels into independent connected domains, i.e. suspected water bloom pixel clusters. The eight-neighbor refers to eight adjacent pixels in eight directions around the pixel, including four orthogonal directions of up, down, left and right, and four diagonal directions of upper left, upper right, lower left and lower right. The connected component labeling processing adopts a two-pass scanning algorithm. In the first pass, the scanning starts from the top left corner of the image and traverses all suspected water bloom pixels in row and column order. For the current pixel, it is checked whether the left and upper pixels also belong to the suspected water bloom pixel set. If neither the left nor the upper pixel is a suspected water bloom pixel, a new connected domain label is assigned to the current pixel. If one of the left or upper pixels is a suspected water bloom pixel, the connected domain label of the adjacent pixel is assigned to the current pixel. If both the left and upper pixels are suspected water bloom pixels but have different labels, the two labels are recorded as an equivalence relation. In the second pass, all suspected water bloom pixels are traversed again, and the labels with equivalence relations are merged into the same connected domain label according to the equivalence relation table. After two passes of scanning, the suspected water bloom pixels with the same connected domain label form a suspected water bloom pixel cluster, and different connected domain labels correspond to different suspected water bloom pixel clusters, obtaining multiple suspected water bloom pixel clusters.

[0053] When the area screening processing is performed on the multiple suspected water bloom pixel clusters according to the number of pixels, the area screening processing eliminates noise interference and misjudgment pixels by removing suspected water bloom pixel clusters with too small areas, and retains suspected water bloom pixel clusters with larger areas as real cyanobacterial water bloom patches. The area calculation multiplies the number of pixels of each suspected water bloom pixel cluster by the area of a single pixel. The area of a single pixel is determined by the spatial resolution of the remote sensing image. The spatial resolution of Sentinel-3 / OLCI image is 300 meters, i.e. the ground area corresponding to a single pixel is 300 meters by 300 meters, equal to 90,000 square meters, i.e. 0.09 square kilometers. The area threshold is set to 0.5 square kilometers, corresponding to a pixel number threshold of 0.5 divided by 0.09, approximately equal to 5.56 pixels, rounded to 6 pixels. The area screening determines whether the number of pixels of each suspected water bloom pixel cluster is greater than or equal to 6 pixels. Suspected water bloom pixel clusters with a number of pixels greater than or equal to 6 pixels are retained, and suspected water bloom pixel clusters with a number of pixels less than 6 pixels are removed. The retained suspected water bloom pixel clusters constitute an optical water bloom distribution mask. The optical water bloom distribution mask is a binary image. The pixels in the retained suspected water bloom pixel clusters are marked as 1 in the mask, indicating water bloom pixels, and the remaining pixels are marked as 0 in the mask, indicating non-water bloom pixels. The optical water bloom distribution mask records the spatial distribution of cyanobacterial water bloom identified based on the optical remote sensing image double-parameter discrimination model.

[0054] In a specific embodiment, the phycocyanin characteristic index image and the chlorophyll fluorescence peak index image are subjected to double-threshold discrimination processing, and the suspected water bloom pixel set is obtained by screening the pixels that simultaneously satisfy the phycocyanin characteristic index threshold and the chlorophyll fluorescence peak index threshold, including:

[0055] The phycocyanin characteristic index values and the chlorophyll fluorescence peak index values of the sample pixels in the known cyanobacterial water bloom area of the target lake are counted, and the mean and standard deviation of the phycocyanin characteristic index values and the mean and standard deviation of the chlorophyll fluorescence peak index values are calculated.

[0056] The phycocyanin characteristic index threshold is determined by the mean and standard deviation of the phycocyanin characteristic index values, and the chlorophyll fluorescence peak index threshold is determined by the mean and standard deviation of the chlorophyll fluorescence peak index values.

[0057] All pixels of the phycocyanin characteristic index image and the chlorophyll fluorescence peak index image are traversed, and the suspected water bloom pixel set is obtained by screening the pixels that are simultaneously greater than the phycocyanin characteristic index threshold and the chlorophyll fluorescence peak index threshold.

[0058] Specifically, when counting the phycocyanin characteristic index values and the chlorophyll fluorescence peak index values of the sample pixels in the known cyanobacterial water bloom area of the target lake, the sample pixels in the known cyanobacterial water bloom area are the pixels in the cyanobacterial water bloom outbreak area confirmed by historical monitoring data or field investigation, which are explicitly marked as cyanobacterial water bloom outbreak areas in the monitoring records and have geographic coordinate information. The sample pixels are located in the phycocyanin characteristic index image and the chlorophyll fluorescence peak index image according to the geographic coordinates, the values of each sample pixel in the phycocyanin characteristic index image are read to form a phycocyanin characteristic index value sample set, and the values of each sample pixel in the chlorophyll fluorescence peak index image are read to form a chlorophyll fluorescence peak index value sample set. When calculating the mean of the phycocyanin characteristic index values, all sample values in the phycocyanin characteristic index value sample set are added up, and the sum is divided by the number of samples to obtain the mean, which reflects the central tendency of the sample values, i.e., the phycocyanin characteristic index level of the typical water bloom area. When calculating the standard deviation of the phycocyanin characteristic index values, each sample value is subtracted from the mean to obtain a deviation, all deviations are squared and added up, the sum is divided by the number of samples to obtain a variance, and the square root of the variance is obtained to obtain the standard deviation, which reflects the dispersion of the sample values, i.e., the variation range of the phycocyanin characteristic index of the water bloom area. The mean and standard deviation of the chlorophyll fluorescence peak index values are calculated in the same way, i.e., all sample values in the chlorophyll fluorescence peak index value sample set are added up and divided by the number of samples to obtain the mean, the deviation of each sample value from the mean is squared, added up, divided by the number of samples, and the square root is taken to obtain the standard deviation.

[0059] When the threshold value determination process is performed according to the mean and standard deviation of the phycocyanin characteristic index value, the threshold value determination process sets the discrimination boundary by using the method of subtracting the multiple of the standard deviation from the mean, and the selection of the multiple coefficient is based on the normal distribution theory in statistics. It is assumed that the phycocyanin characteristic index value in the water bloom region is normally distributed, and the lower tail boundary of the distribution corresponding to the mean minus twice the standard deviation covers about 95 percent of the sample data. The phycocyanin characteristic index threshold value is obtained by subtracting twice the standard deviation from the mean of the phycocyanin characteristic index value. This threshold value is used as the lower limit condition for discriminating whether the pixel belongs to the water bloom. The pixel with a phycocyanin characteristic index value greater than the threshold value has a similar phycocyanin spectral feature to the known water bloom region. When the threshold value determination process is performed according to the mean and standard deviation of the chlorophyll fluorescence peak index value, the chlorophyll fluorescence peak index threshold value is obtained by subtracting one and a half times the standard deviation from the mean of the chlorophyll fluorescence peak index value. The one and a half times coefficient is more relaxed than the two times coefficient, which reflects that the variation of the chlorophyll fluorescence peak index in the water bloom region is larger. The use of different multiple coefficients is based on the difference in the statistical distribution characteristics of the two indexes in the water bloom region. The phycocyanin characteristic index, as a specific pigment index of blue-green algae, has a higher discrimination degree, and a more stringent two times standard deviation is used. The chlorophyll fluorescence peak index, as a common chlorophyll index of all algae, uses a relatively relaxed one and a half times standard deviation. The cooperative constraint of the two threshold values ensures the specificity of blue-green algae and the universality of chlorophyll.

[0060] When all the pixels of the phycocyanin characteristic index image and the chlorophyll fluorescence peak index image are traversed, the traversal operation starts from the first row and the first column of the image and scans row by row and column by column. For the currently scanned pixel, the value of the pixel at the corresponding row and column position in the phycocyanin characteristic index image is read as the phycocyanin characteristic index value of the pixel, and the value of the pixel at the corresponding row and column position in the chlorophyll fluorescence peak index image is read as the chlorophyll fluorescence peak index value of the pixel. When screening the pixels greater than both the phycocyanin characteristic index threshold value and the chlorophyll fluorescence peak index threshold value, it is determined whether the phycocyanin characteristic index value of the pixel is greater than the phycocyanin characteristic index threshold value, and whether the chlorophyll fluorescence peak index value of the pixel is greater than the chlorophyll fluorescence peak index threshold value. The screening operation is only executed when both conditions are true. The screening operation records the row and column position coordinates of the pixel to the suspected water bloom pixel set. If either condition is false, the pixel is skipped and the next pixel is scanned. After the traversal and judgment of all pixels are completed, the suspected water bloom pixel set contains the position information of all pixels that satisfy both threshold conditions. This set is the input data for the subsequent connected component labeling and area screening.

[0061] In a specific embodiment, step S4 comprises:

[0062] extract the vertical polarization backscattering coefficient and the cross polarization backscattering coefficient from the dual polarization backscattering coefficient, and calculate a polarization backscattering ratio of the cross polarization backscattering coefficient to the vertical polarization backscattering coefficient;

[0063] The polarization backscattering ratio is substituted into a nonlinear regression model for logarithmic regression calculation to obtain a radar inversion phycocyanin index, and the radar inversion phycocyanin index is subjected to a per-pixel mapping process to obtain a radar inversion phycocyanin index image.

[0064] The radar inversion phycocyanin index image is screened for pixels greater than a phycocyanin discrimination threshold, the screened pixels are subjected to connected domain labeling and area screening to obtain a radar water bloom distribution mask, and the radar water bloom distribution mask is superimposed and fused with an optical water bloom distribution mask to obtain a cyanobacterial water bloom monitoring result.

[0065] Specifically, when the vertical polarization backscattering coefficient and the cross polarization backscattering coefficient are extracted from the dual polarization backscattering coefficient, the dual polarization backscattering coefficient data contains backscattering coefficient information of two polarization channels, the vertical polarization backscattering coefficient corresponds to a VV channel, i.e., a backscattering coefficient of a radar wave with vertical transmission and vertical reception, and the cross polarization backscattering coefficient corresponds to a VH channel, i.e., a backscattering coefficient of a radar wave with vertical transmission and horizontal reception. The extraction operation reads the VV channel value of each pixel from the dual polarization backscattering coefficient data as the vertical polarization backscattering coefficient, and reads the VH channel value of each pixel as the cross polarization backscattering coefficient. When the polarization backscattering ratio of the cross polarization backscattering coefficient to the vertical polarization backscattering coefficient is calculated, the polarization backscattering ratio calculation divides the cross polarization backscattering coefficient by the vertical polarization backscattering coefficient. This ratio reflects the intensity relationship of cross polarization scattering relative to co-polarization scattering. When the water surface is smooth, co-polarization scattering dominates and cross polarization scattering is weak, resulting in a small ratio. When the water surface has a blue-green biological membrane, the rough surface and internal bubble structure of the biological membrane enhance volume scattering and multiple scattering, which enhances cross polarization scattering, resulting in an increased ratio. The polarization backscattering ratio serves as a substitute indicator for the roughness of the blue-green biological membrane on the water surface, and the polarization backscattering ratio data is obtained by per-pixel calculation.

[0066] When the polarized backscatter ratio is substituted into the nonlinear regression model for logarithmic regression calculation and processing, the nonlinear regression model is a mathematical relationship model between the polarized backscatter ratio and the phycocyanin characteristic index, which is established in advance through a training data set. The training data set comes from historical multi-spectral remote sensing images and historical synthetic aperture radar images obtained at the same time under clear sky conditions. The phycocyanin characteristic index of the pixels in the confirmed cyanobacterial bloom area is calculated from the historical multi-spectral remote sensing images as the true value, and the polarized backscatter ratio of the pixels at the same position is extracted from the historical synthetic aperture radar images as the input. The least squares method is used to fit the relationship between the polarized backscatter ratio and the phycocyanin characteristic index. It is found that the two are in logarithmic relationship during the fitting process. The nonlinear regression model is established in the form of radar inversion phycocyanin index equals to the first coefficient multiplied by the polarized backscatter ratio plus the natural logarithm of the second coefficient plus the third coefficient. The least squares method solves the optimal value of the three coefficients by minimizing the sum of squared residuals between the model predicted value and the true value. When the logarithmic regression calculation and processing are executed, the polarized backscatter ratio of the current pixel is read, the ratio is added to the second coefficient, the sum is calculated as the natural logarithm, the logarithmic result is multiplied by the first coefficient, and the product is added to the third coefficient to obtain the radar inversion phycocyanin index of the pixel. When the radar inversion phycocyanin index is processed pixel by pixel, the radar inversion phycocyanin index value calculated for each pixel is assigned to the spatial position corresponding to the pixel in the radar inversion phycocyanin index image. After the calculation and mapping of all pixels are completed, the radar inversion phycocyanin index image is obtained.

[0067] In the screening of the radar-retrieved phycocyanin index image, the phycocyanin threshold value is consistent with the phycocyanin characteristic index threshold value determined in the generation of the optical water bloom distribution mask, all pixels of the radar-retrieved phycocyanin index image are traversed, the radar-retrieved phycocyanin index value of each pixel is read, and it is judged whether the value is greater than the phycocyanin threshold value. If it is greater than the threshold value, the pixel is screened. In the connected domain labeling and area screening processing of the screened pixels, the eight-neighbor connected domain labeling algorithm is used to aggregate the spatially adjacent screened pixels into a connected domain, the number of pixels contained in each connected domain is calculated, and it is judged whether the area threshold is satisfied. The connected domain satisfying the area threshold is retained, and the radar water bloom distribution mask is obtained. The radar water bloom distribution mask records the spatial distribution of the blue-green algae water bloom retrieved based on the synthetic aperture radar image. In the superimposition and fusion processing of the radar water bloom distribution mask and the optical water bloom distribution mask, the information of the two masks is integrated by using the logic OR operation. In the image spatial coordinate system, if the pixel corresponding to the same spatial position in the two masks is marked as a water bloom pixel or a water bloom pixel in the radar water bloom distribution mask, the pixel is marked as a water bloom pixel in the fusion result. If the pixel is marked as a non-water bloom pixel in the two masks, the pixel is marked as a non-water bloom pixel in the fusion result. After the logic OR operation of all pixels is completed, the blue-green algae water bloom monitoring result is obtained. The monitoring result integrates the spectral feature recognition capability of the optical remote sensing image and the all-weather observation capability of the synthetic aperture radar image, and solves the problem of missing optical remote sensing data caused by cloud cover.

[0068] In a specific embodiment, the polarized backscatter ratio is substituted into a nonlinear regression model for logarithmic regression calculation to obtain a radar-retrieved phycocyanin index. The radar-retrieved phycocyanin index is processed by pixel-by-pixel mapping to obtain a radar-retrieved phycocyanin index image, including:

[0069] The historical multispectral remote sensing image and the historical synthetic aperture radar image obtained under clear sky conditions at the same time are selected as the training data set. Training sample pixels of the confirmed blue-green algae water bloom region in the training data set are extracted, and the polarized backscatter ratio and the phycocyanin characteristic index value of the training sample pixels are counted.

[0070] The polarized backscatter ratio and the phycocyanin characteristic index value of the training sample pixels are subjected to least squares fitting processing to construct a nonlinear regression model containing a logarithmic function and a linear coefficient, and the regression coefficient of the nonlinear regression model is solved.

[0071] The polarized backscatter ratio of the target lake at the current time is substituted into the nonlinear regression model for logarithmic operation and linear transformation processing, and the radar-retrieved phycocyanin index is calculated pixel by pixel to obtain a radar-retrieved phycocyanin index image.

[0072] Specifically, when the historical multi-spectral remote sensing image and the historical synthetic aperture radar image obtained simultaneously under clear sky conditions are selected as the training data set, the clear sky condition ensures that the historical multi-spectral remote sensing image is not affected by cloud cover, and the simultaneous acquisition means that the imaging time interval of the two types of images is within a few hours, ensuring that the two types of images reflect consistent lake water body states. The historical image refers to the accumulated remote sensing image data in the past monitoring period, and the images that meet the clear sky and simultaneous acquisition conditions are selected from the historical image archives to form the training data set. When the training sample pixels in the confirmed cyanobacterial bloom area in the training data set are extracted, the confirmed cyanobacterial bloom area is determined from the historical monitoring records or artificial investigation to determine the bloom outbreak area. These areas have been identified as cyanobacterial bloom areas in the historical multi-spectral remote sensing image through a double-parameter discrimination model of phycocyanin characteristic index and chlorophyll fluorescence peak index. According to the geographical coordinates of these areas, the corresponding pixels in the historical multi-spectral remote sensing image and the historical synthetic aperture radar image are located, and these pixels are used as training sample pixels. When the polarized backscatter ratio and the phycocyanin characteristic index value of the training sample pixels are counted, for each training sample pixel, the vertical polarization backscatter coefficient and the cross-polarization backscatter coefficient of the pixel are read from the historical synthetic aperture radar image, the cross-polarization backscatter coefficient is divided by the vertical polarization backscatter coefficient to obtain the polarized backscatter ratio of the pixel, and the phycocyanin characteristic index value of the pixel is read from the historical multi-spectral remote sensing image. The polarized backscatter ratio and the phycocyanin characteristic index value of each training sample pixel are paired and recorded to form a set of training sample data pairs.

[0073] When the polarized backscattering ratio of the training sample pixel is least square fitted with the phycocyanobilin characteristic index value, the least square fitting is a parameter estimation method, and the optimal value of the model parameter is determined by minimizing the residual sum of squares between the model predicted value and the observed true value. The fitting is based on the analysis of the scatter plot of the training sample data pairs, and it is found that there is a logarithmic relationship between the polarized backscattering ratio and the phycocyanobilin characteristic index. When a nonlinear regression model containing a logarithmic function and a linear coefficient is constructed, the mathematical form of the nonlinear regression model is set as the radar inversion phycocyanobilin index equal to the first regression coefficient multiplied by the polarized backscattering ratio plus the natural logarithm of the second regression coefficient plus the third regression coefficient. The model contains three regression coefficients to be solved, the first regression coefficient controls the amplitude scaling of the logarithmic function, the second regression coefficient controls the translation of the independent variable of the logarithmic function, and the third regression coefficient controls the vertical translation of the entire function. When the regression coefficients of the nonlinear regression model are solved, for each data pair in the training sample data pair set, the polarized backscattering ratio is added to the second regression coefficient, the natural logarithm of the sum is calculated, the logarithm result is multiplied by the first regression coefficient, and the product result is added to the third regression coefficient to obtain the model predicted value of the training sample. The difference between the model predicted value and the true value of the phycocyanobilin characteristic index of the training sample is calculated as the residual. The residual squares of all training samples are accumulated to obtain the residual sum of squares. The values of the three regression coefficients are adjusted by using an iterative optimization algorithm to minimize the residual sum of squares. The iterative optimization algorithm includes gradient descent method or Newton method. The optimal value of the three regression coefficients that minimizes the residual sum of squares is obtained by multiple iterations.

[0074] When the current polarization backscattering ratio of the target lake is substituted into the nonlinear regression model for logarithmic operation and linear transformation processing, the current polarization backscattering ratio is from the current synthetic aperture radar image obtained, for each pixel in the current synthetic aperture radar image, the polarization backscattering ratio of the pixel is read, the ratio is added to the second regression coefficient of the nonlinear regression model, the natural logarithm of the addition result is calculated by logarithmic operation, the base of the natural logarithm is Euler number, which is about 2.718, the logarithmic operation converts the multiplication relationship into an addition relationship, the logarithmic result is multiplied by the first regression coefficient and added to the third regression coefficient by linear transformation processing, multiplication by the first regression coefficient realizes amplitude scaling, and addition of the third regression coefficient realizes vertical translation. When the radar retrieval phycocyanin index is calculated pixel by pixel, the above logarithmic operation and linear transformation processing are repeated for each pixel of the current synthetic aperture radar image, and the radar retrieval phycocyanin index value of the pixel is calculated. The radar retrieval phycocyanin index value reflects the concentration level of blue-green algae phycocyanin inferred based on the radar backscattering characteristics at the pixel position. When the radar retrieval phycocyanin index is assigned to the corresponding pixel spatial position, the row number and column number of the current calculation pixel in the synthetic aperture radar image are read, and the radar retrieval phycocyanin index value calculated by the pixel is stored in the same row number and column number position in the radar retrieval phycocyanin index image. After the calculation and assignment of all pixels are completed, the radar retrieval phycocyanin index image is obtained, and the spatial resolution and geographical range of the image are consistent with the current synthetic aperture radar image.

[0075] The lake blue-green algae bloom detection method of fusing remote sensing images in the embodiments of the application is described above, and the lake blue-green algae bloom detection system of fusing remote sensing images in the embodiments of the application is described below. Please refer to Figure 2 An embodiment of the lake blue-green algae bloom detection system of fusing remote sensing images in the embodiments of the application includes:

[0076] The extraction module is configured to obtain a multispectral remote sensing image and a synthetic aperture radar image of a target lake, the multispectral remote sensing image includes a plurality of characteristic band reflectances, and the synthetic aperture radar image includes a dual-polarization backscattering coefficient.

[0077] The calculation module is configured to calculate an algal blue pigment characteristic index and a chlorophyll fluorescence peak index according to the plurality of characteristic band reflectances.

[0078] The screening module is configured to screen pixels that satisfy the algal blue pigment characteristic index and the chlorophyll fluorescence peak index to obtain an optical water bloom distribution mask.

[0079] The fusion module is configured to calculate a radar inversion phycocyanin index according to the dual-polarized backscatter coefficient through a nonlinear regression model, screen a radar water bloom distribution mask by screening a pixel whose radar inversion phycocyanin index meets a discrimination condition, and obtain a blue-green algae water bloom monitoring result by fusing the optical water bloom distribution mask and the radar water bloom distribution mask.

[0080] The above Figure 2 The lake blue-green algae water bloom detection system for fusing remote sensing images in the embodiment of the present application is described in detail from the perspective of modular functional entities, and the lake blue-green algae water bloom detection device for fusing remote sensing images in the embodiment of the present application is described in detail from the perspective of hardware processing.

[0081] With reference to Figure 3 In the embodiment of the present application, a lake blue-green algae water bloom detection device for fusing remote sensing images is also provided, which can be a server, and the internal structure thereof can be as shown in Figure 3 The lake blue-green algae water bloom detection device for fusing remote sensing images comprises a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. The processor of the computer is configured to provide computing and control capabilities. The memory of the lake blue-green algae water bloom detection device for fusing remote sensing images comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The database of the lake blue-green algae water bloom detection device for fusing remote sensing images is configured to store corresponding data in the embodiment. The network interface of the lake blue-green algae water bloom detection device for fusing remote sensing images is configured to communicate with an external terminal through a network connection. The computer program is executed by the processor to implement the above method.

[0082] Those skilled in the art can understand, Figure 3 The structure shown in the above

[0083] The present application also provides a computer readable storage medium, which can be a non-volatile computer readable storage medium or a volatile computer readable storage medium. The computer readable storage medium stores instructions, and when the instructions are run on a computer, the computer executes the steps of the lake blue-green algae water bloom detection method for fusing remote sensing images.

[0084] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, system and unit can refer to the corresponding processes in the foregoing method embodiments, and will not be described here.

[0085] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application or the entire or part of the technical solutions that essentially contribute to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a lake cyanobacterial bloom detection device integrated with remote sensing images (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various program code storage media.

[0086] The above embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for detecting cyanobacterial blooms in lakes by fusing remote sensing imagery, characterized in that, The method includes: Step S1: Acquire multispectral remote sensing imagery and synthetic aperture radar (SAR) imagery of the target lake. The multispectral remote sensing imagery includes multiple characteristic band reflectance, and the SAR imagery includes dual-polarization backscattering coefficients. This includes: acquiring raw multispectral remote sensing imagery data of the target lake; performing radiometric calibration on the raw multispectral remote sensing imagery data to obtain atmospheric top reflectance data; performing atmospheric correction on the atmospheric top reflectance data to obtain water surface reflectance data; acquiring raw SAR imagery data of the target lake; performing radiometric calibration on the raw SAR imagery data to obtain backscattering intensity data; performing terrain correction on the backscattering intensity data to obtain backscattering coefficient data; extracting the phycocyanin absorption characteristic band, chlorophyll absorption characteristic band, chlorophyll fluorescence characteristic band, and near-infrared characteristic band from the water surface reflectance data as the multiple characteristic band reflectance; and extracting the vertical polarization component and cross-polarization component from the backscattering coefficient data as the dual-polarization backscattering coefficients. Step S2: Calculate the phycocyanin characteristic index and chlorophyll fluorescence peak index based on the reflectance of the multiple characteristic bands; Step S3: Select pixels that simultaneously meet the discrimination conditions of the phycocyanin characteristic index and the chlorophyll fluorescence peak index to obtain an optical bloom distribution mask; Step S4: Calculate the radar inversion phycocyanin index based on the dual-polarization backscattering coefficient using a nonlinear regression model, select pixels whose radar inversion phycocyanin index meets the discrimination criteria to obtain a radar bloom distribution mask, and fuse the optical bloom distribution mask and the radar bloom distribution mask to obtain the cyanobacterial bloom monitoring results.

2. The method for detecting cyanobacterial blooms in lakes by fusing remote sensing images according to claim 1, characterized in that, Step S2 includes: The phycocyanin absorption characteristic band reflectance and near-infrared characteristic band reflectance are extracted from the multiple characteristic band reflectances. The difference between the phycocyanin absorption characteristic band reflectance and the near-infrared characteristic band reflectance is normalized to obtain the phycocyanin characteristic index. Chlorophyll absorption characteristic band reflectance, chlorophyll fluorescence characteristic band reflectance, and near-infrared characteristic band reflectance are extracted from the multiple characteristic band reflectance. The chlorophyll absorption characteristic band reflectance and the near-infrared characteristic band reflectance are linearly baseline constructed to obtain the baseline reflectance value. The difference between the chlorophyll fluorescence characteristic band reflectance and the baseline reflectance value is calculated to obtain the chlorophyll fluorescence peak index. The phycocyanin characteristic index and the chlorophyll fluorescence peak index are processed pixel-by-pixel mapping to obtain phycocyanin characteristic index images and chlorophyll fluorescence peak index images.

3. The method for detecting cyanobacterial blooms in lakes by fusing remote sensing images according to claim 2, characterized in that, Step S3 includes: The phycocyanin characteristic index image and the chlorophyll fluorescence peak index image are subjected to dual threshold discrimination processing, and pixels that simultaneously meet the phycocyanin characteristic index threshold and the chlorophyll fluorescence peak index threshold are selected to obtain a set of suspected algal bloom pixels. The suspected algal bloom pixel set is subjected to eight-neighbor connected component labeling to obtain multiple suspected algal bloom pixel clusters, and the number of pixels in each suspected algal bloom pixel cluster is calculated. The multiple suspected bloom pixel clusters are screened by area based on the number of pixels, and the suspected bloom pixel clusters whose number of pixels meets the area threshold are retained to obtain the optical bloom distribution mask.

4. The method for detecting cyanobacterial blooms in lakes by fusing remote sensing images according to claim 3, characterized in that, The process of performing dual-threshold discrimination processing on the phycocyanin characteristic index image and the chlorophyll fluorescence peak index image, and selecting pixels that simultaneously meet the phycocyanin characteristic index threshold and the chlorophyll fluorescence peak index threshold to obtain a set of suspected algal bloom pixels, including: The phycocyanin characteristic index and chlorophyll fluorescence peak index values ​​of sample pixels in the known cyanobacterial bloom area of ​​the target lake are statistically analyzed, and the mean and standard deviation of the phycocyanin characteristic index values ​​and the chlorophyll fluorescence peak index values ​​are calculated. The phycocyanin characteristic index threshold is obtained by performing threshold determination processing based on the mean and standard deviation of the phycocyanin characteristic index values, and the chlorophyll fluorescence peak index threshold is obtained by performing threshold determination processing based on the mean and standard deviation of the chlorophyll fluorescence peak index values. By traversing all pixels of the phycocyanin characteristic index image and the chlorophyll fluorescence peak index image, pixels that are simultaneously greater than the phycocyanin characteristic index threshold and the chlorophyll fluorescence peak index threshold are selected to obtain the suspected algal bloom pixel set.

5. The method for detecting cyanobacterial blooms in lakes by fusing remote sensing images according to claim 1, characterized in that, Step S4 includes: Extract the vertical polarization backscattering coefficient and the cross-polarization backscattering coefficient from the dual-polarization backscattering coefficient, and calculate the polarization backscattering ratio of the cross-polarization backscattering coefficient to the vertical polarization backscattering coefficient; The polarization backscattering ratio is substituted into the nonlinear regression model for logarithmic regression calculation to obtain the radar inverted phycocyanin index. The radar inverted phycocyanin index is then processed pixel-by-pixel mapping to obtain the radar inverted phycocyanin index image. Pixels in the radar-retrieved phycocyanin index image that are greater than the phycocyanin discrimination threshold are selected. Connectivity labeling and area filtering are performed on the selected pixels to obtain the radar bloom distribution mask. The radar bloom distribution mask and the optical bloom distribution mask are superimposed and fused to obtain the cyanobacterial bloom monitoring results.

6. The method for detecting cyanobacterial blooms in lakes by fusing remote sensing images according to claim 5, characterized in that, The process of substituting the polarization backscattering ratio into the nonlinear regression model for logarithmic regression calculation to obtain the radar-retrieved phycocyanin index, and performing pixel-by-pixel mapping processing on the radar-retrieved phycocyanin index to obtain the radar-retrieved phycocyanin index image includes: Historical multispectral remote sensing images and historical synthetic aperture radar images acquired simultaneously under clear sky conditions were selected as training datasets. Training sample pixels of confirmed cyanobacterial bloom areas were extracted from the training datasets, and the polarization backscattering ratio and phycocyanin characteristic index values ​​of the training sample pixels were statistically analyzed. The polarization backscattering ratio of the training sample pixels is fitted with the phycocyanin feature index using the least squares method to construct the nonlinear regression model containing a logarithmic function and linear coefficients, and the regression coefficients of the nonlinear regression model are solved. The polarization backscattering ratio of the target lake at the current moment is substituted into the nonlinear regression model for logarithmic operation and linear transformation processing. The radar inversion phycocyanin index is calculated pixel by pixel. The radar inversion phycocyanin index is then assigned to the corresponding pixel spatial position to obtain the radar inversion phycocyanin index image.

7. A lake cyanobacterial bloom detection system integrating remote sensing imagery, characterized in that, The method for detecting cyanobacterial blooms in lakes using fused remote sensing images as described in any one of claims 1 to 6, wherein the fused remote sensing image system for detecting cyanobacterial blooms in lakes comprises: An extraction module is used to acquire multispectral remote sensing images and synthetic aperture radar (SAR) images of a target lake. The multispectral remote sensing images include reflectance in multiple characteristic bands, and the SAR images include dual-polarization backscattering coefficients. The module includes: acquiring raw multispectral remote sensing image data of the target lake; performing radiometric calibration on the raw multispectral remote sensing image data to obtain atmospheric top reflectance data; performing atmospheric correction on the atmospheric top reflectance data to obtain water surface reflectance data; acquiring raw SAR image data of the target lake; performing radiometric calibration on the raw SAR image data to obtain backscattering intensity data; performing terrain correction on the backscattering intensity data to obtain backscattering coefficient data; extracting the phycocyanin absorption characteristic band, chlorophyll absorption characteristic band, chlorophyll fluorescence characteristic band, and near-infrared characteristic band from the water surface reflectance data as the multiple characteristic band reflectances; and extracting the vertical polarization component and cross-polarization component from the backscattering coefficient data as the dual-polarization backscattering coefficients. The calculation module is used to calculate the phycocyanin characteristic index and the chlorophyll fluorescence peak index based on the reflectance of the multiple characteristic bands; The filtering module is used to filter pixels that simultaneously meet the discrimination conditions of the phycocyanin characteristic index and the chlorophyll fluorescence peak index to obtain an optical bloom distribution mask. The fusion module is used to calculate the radar-retrieved phycocyanin index based on the dual-polarization backscattering coefficient through a nonlinear regression model, screen pixels whose radar-retrieved phycocyanin index meets the discrimination conditions to obtain a radar bloom distribution mask, and fuse the optical bloom distribution mask and the radar bloom distribution mask to obtain the cyanobacterial bloom monitoring results.

8. A lake cyanobacterial bloom detection device integrating remote sensing imagery, characterized in that, The method includes a memory and a processor, wherein the memory stores a computer program that can run on the processor, and the processor executes the computer program to implement the lake cyanobacterial bloom detection method based on fused remote sensing images as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by the processor, it causes the processor to execute the lake cyanobacterial bloom detection method based on fused remote sensing images as described in any one of claims 1 to 6.

Images