Surface reflectance data fusion method and system based on multi-factor weight optimization

Abstract

The invention discloses a surface reflectance data fusion method and system based on multi-factor weight optimization, and relates to the technical field of remote sensing image processing. According to the method, radiometric calibration, geometric correction and atmospheric correction are carried out on MODIS and Landsat 8 OLI sensor data, spectral consistency is realized by using a spectral band-pass adjustment factor (SBAF) technology, and spectral response consistency of different sensor data is ensured. Meanwhile, by calculating and comprehensively considering multi-factor weights of cloud, rainfall, aerosol and the like, a comprehensive weight factor screening function is constructed, and a pixel with the minimum comprehensive influence is selected for data fusion, so that surface reflectance data with high consistency and accuracy is generated. According to the method, the problem of remote sensing data fusion deviation caused by spectral response difference and cloud, rainfall and aerosol is effectively solved, and the accuracy and reliability of fused data are remarkably improved.

Description

Technical Field

[0001] The invention belongs to the technical field of remote sensing image processing, and in particular relates to a surface reflectance data fusion method and system based on multi-factor weight optimization. Background Art

[0002] In recent years, remote sensing optical surface reflectance data has been widely used in climate change research, agricultural management, environmental monitoring, and disaster prevention and mitigation. Accurate remote sensing optical surface reflectance data is of great significance for studying surface morphological changes, ecosystem dynamics, water resources management, and weather forecasting. With the continuous development of remote sensing satellite imaging technology, obtaining large-scale reflectance data has gradually become efficient and convenient. The increasing demand for remote sensing optical surface reflectance data has also put forward higher requirements for the parameter settings of newly developed remote sensing satellite imaging systems. High-quality surface reflectance data is also of great significance for the imaging parameter settings of remote sensing satellites.

[0003] Traditional remote sensing optical surface reflectance data acquisition methods mainly rely on data from a single sensor. However, this method has significant limitations. Data from a single sensor is often biased and cannot fully reflect the true situation of the surface. In addition, cloud cover is a major challenge in remote sensing data acquisition. Clouds block satellite observations, resulting in the loss of surface reflectance data. These problems limit the application of traditional remote sensing data in high temporal resolution and continuous monitoring.

[0004] In order to make up for the problem of missing data, researchers usually use a simple splicing method to fill in the missing data. However, this method also has obvious disadvantages. The splicing method is prone to introduce errors, especially when there are differences in observation conditions between different data sources. In addition, the splicing method cannot effectively handle the dynamic changes in surface reflectance, especially in periods and regions where climate conditions change dramatically.

[0005] Based on this, multi-source heterogeneous data fusion methods have gradually received attention. By combining data from multiple sensors and performing bandpass adjustments, more comprehensive and accurate surface reflectance data can be generated. However, this method also has its shortcomings. When generating surface reflectance data, the current multi-source data fusion method usually ignores the differences between spectral response functions (RSR). Different sensors have different spectral response characteristics, which will lead to systematic deviations in the observed surface reflectance data, affecting the accuracy and consistency of the data, resulting in the fused data being unable to truly reflect the actual surface conditions, reducing the reliability of the data in scientific research and applications; second, the combined effects of factors such as aerosols, clouds, and rain are often not fully considered, and the atmosphere and clouds may have an important impact on the reliability of remote sensing data, resulting in missing values ​​and errors in remote sensing data. If this impact cannot be quantified and taken into account, it will affect the data accuracy of the surface reflectance data produced in the end. Summary of the invention

[0006] In order to solve the above technical problems, the present invention proposes a surface reflectance data fusion method and system based on multi-factor weight optimization, adopts spectral band adjustment factor technology (SBAF) to adjust the spectrum, and adopts multi-factor weight fusion technology to comprehensively consider the influence of clouds, rain, and aerosols on imaging, thereby improving the quality and reliability of the final fused data, significantly improving the effect of optical multi-source remote sensing data fusion, and meeting the needs of scientific research and practical applications.

[0007] To achieve the above purpose, the technical solution adopted by the present invention is as follows: In one aspect, the present invention provides a surface reflectance data fusion method based on multi-factor weight optimization, the method comprising the following steps: Step 1: Perform radiometric calibration and geometric correction on MODIS and Landsat8 OLI multispectral data to generate top of atmosphere reflectance data, and generate surface reflectance data of different spectral bands after atmospheric correction; Step 2: Use the spectral bandpass adjustment factor technology to adjust the surface reflectance data of Landsat8 OLI to be consistent with the spectral range of the surface reflectance data of MODIS; Step 3: Calculate the cloud weight factor using the HOT-VBR index method, calculate the rainfall weight factor using the MODRA algorithm, calculate the aerosol weight factor of each pixel using the dark pixel method, and normalize the time of all data to calculate the normalized time weight factor and normalize the cloud weight factor, rainfall weight factor, and aerosol weight factor; Step 4: Calculate the comprehensive weight factor by combining the normalized cloud weight factor, rainfall weight factor, aerosol weight factor, and time weight factor. Filter the comprehensive weight factor through the comprehensive weight factor screening function, and use the pixel with the least impact of the comprehensive weight factor for data fusion.

[0008] On the other hand, the present invention provides a surface reflectance data fusion system based on multi-factor weight optimization, comprising: The data preprocessing unit is used to perform radiometric calibration and geometric correction on MODIS and Landsat8 OLI multispectral data, generate top of atmosphere reflectance data, and generate surface reflectance data of different spectral bands after atmospheric correction; A data adjustment unit is used to adjust the surface reflectance data of Landsat8 OLI to be consistent with the spectral range of the surface reflectance data of MODIS by using the spectral band adjustment factor technology; A weight factor calculation unit is used to calculate the cloud weight factor using the HOT-VBR index method, calculate the rainfall weight factor using the MODRA algorithm, calculate the aerosol weight factor of each pixel using the dark pixel method, calculate the time weight factor and normalize it; The comprehensive weight factor screening unit is used to calculate the comprehensive weight factor by combining the normalized cloud weight factor, rainfall weight factor, aerosol weight factor and time weight factor, and screen the comprehensive weight factor through the comprehensive weight factor screening function, and use the pixel with the least influence of the comprehensive weight factor for data fusion.

[0009] In a third aspect, the present invention provides an electronic device, comprising: one or more processors; a memory for storing one or more programs; wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the aforementioned surface reflectance data fusion method based on multi-factor weight optimization.

[0010] In a fourth aspect, the present invention provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, enables the processor to implement the aforementioned surface reflectance data fusion method based on multi-factor weight optimization.

[0011] The beneficial effects of the present invention are: (1) The present invention adopts the Spectral Band Adjustment Factor (SBAF) technology to ensure the consistency of spectral responses of different sensors, fundamentally solving the data deviation problem caused by spectral response differences, and making the fused data highly consistent and accurate in spectrum.

[0012] (2) The multi-factor weighted fusion technology of the present invention can select the optimal data for fusion by comprehensively considering factors such as clouds, rain, and aerosols, significantly reducing the negative impact of atmospheric and meteorological factors on data quality and ensuring the high reliability of the data.

[0013] (3) The present invention uses an improved data fusion algorithm to screen and fuse the optimal data source in the complexity of multi-source data, thereby improving the accuracy and application effect of the data and meeting the needs of high-precision scientific research and practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 This is a flow chart of the surface reflectance data fusion method based on multi-factor weight optimization of the present invention; Figure 2 It is a schematic diagram of the SBAF technology principle of the present invention; Figure 3 It is a conceptual diagram of the fusion technology based on multi-factor weights of the present invention; Figure 4 This is a schematic diagram of an example of surface reflectivity data of the present invention. DETAILED DESCRIPTION

[0015] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0016] The present invention provides a surface reflectance data fusion method and system based on multi-factor weight optimization, and the main technologies used include: (1) Using Spectral Band Adjustment Factor (SBAF) technology to adjust the spectrum: Different satellite sensors have different spectral responses, which is a hard problem faced by the fusion of different remote sensing data. In view of the differences in spectral response functions of different sensors, the present invention introduces Spectral Band Adjustment Factor (SBAF) technology. SBAF technology can effectively correct the spectral response differences between different sensors, so that the observation data of each sensor are spectrally consistent. This adjustment process ensures that the spectral information of each data source matches each other when multi-source data is fused, thereby improving the accuracy and consistency of the fused data.

[0017] (2) Adopting multi-factor weighted fusion technology to comprehensively consider the impact of clouds, rain, and aerosols on imaging: In order to better deal with the impact of atmospheric and meteorological factors such as clouds, rain, and aerosols on remote sensing imaging, the present invention adopts a multi-factor weighted fusion technology. This technology performs weighted processing on different meteorological and atmospheric factors, comprehensively considers their impact on imaging quality, and selects the data with the least impact for fusion. This method can not only effectively reduce the negative impact of clouds, rain, and aerosols on remote sensing data, but also improve the quality and reliability of the final fused data.

[0018] like Figure 1 FIG. 1 is a flow chart of a surface reflectance data fusion method based on multi-factor weight optimization according to the present invention, comprising the following steps: Step 1: Data preprocessing: Perform radiometric calibration and geometric correction on MODIS and Landsat8 OLI multispectral data to generate top of atmosphere reflectance data, and generate surface reflectance data of different spectral bands after atmospheric correction; The MODIS and Landsat8 OLI data from 2019 to 2023 were calibrated by radiation, geometrically corrected, and atmospherically corrected. The spatial resolution was resampled to 0.01° to obtain the corrected remote sensing surface reflectance data. , where i represents the sensor code, i=1, 2, i=1 represents the MODIS sensor, and i=2 represents the Landsat 8 OLI sensor; j represents the year of the data, j=1, 2, 3, 4, 5 represent 2019, 2020, 2021, 2022, and 2023 respectively. k represents the month of the data, with a value of 1-12, b represents the band of the data source, with a value of 1, 2, ... 8, and m and n are the coordinates of the pixel; Step 2: Use the spectral bandpass adjustment factor technology (SBAF) to adjust the surface reflectance data of Landsat8OLI to be consistent with the spectral range of the surface reflectance data of MODIS; like Figure 2 As shown in the figure, SBAF is used to adjust the surface reflectance data bands of each band of Landsat 8 OLI to be consistent with MODIS. The formula is: , In the formula, Represents the adjusted Landsat 8 OLI surface reflectance data, Indicates the spectral bandpass adjustment factor of each band; in, Simulated reflectivity from MODIS, Landsat 8 OLI sensors , The ratio is calculated as: , Where λ represents the wavelength; the simulated reflectance of the two sensors is obtained using the sample hyperspectral data of the Hyperion sensor and the relative spectral response curve of the sensor itself. The specific method is: Get the average spectral response curve of the entire spectrum from the sample data in the Hyperion sensor Get the relative spectral response curves of Landsat 8 OLI and MODIS sensors ; Calculate simulated reflectivity from Landsat8 OLI sensor , the calculation formula is as follows: , Calculate simulated reflectivity from MODIS sensor : , After the Landsat 8 OLI surface reflectance data is adjusted, the MODIS data does not need to be adjusted, that is: .

[0019] Step 3: Calculate the cloud weight factor using the HOT-VBR index method, calculate the rainfall weight factor using the MODRA algorithm, calculate the aerosol weight factor of each pixel using the dark pixel method, and calculate the time weight factor and normalize it; including: Step 3.1: Calculate the cloud weight factor based on the HOT-VBR index; Cloud Weight Factor The calculation uses the HOT index (Haze optimized transformation) and VBR (visible band ratio) index method. The calculation formula is as follows: , , in, It uses two bands, band 1 and band 3, and the corresponding adjusted surface reflectance data and The calculated HOT index data, It is the VBR index data calculated using the adjusted surface reflectance data corresponding to the three bands: Band 1, Band 2, and Band 3; Using two indexes to calculate cloud weight factor : .

[0020] Step 3.2, calculation of rainfall weight factor based on MODRA index; Rainfall weight factor The MODRA algorithm mainly utilizes thresholds in the red band and the near-infrared band.

[0021] MODRA calculation of rainfall weighting factors The specific algorithm varies according to different thresholds, as follows: , in, Refers to the adjusted surface reflectance data of the 8th band, Refers to the adjusted surface reflectance data of the 3rd band.

[0022] Step 3.3, calculation of aerosol weight factor based on dark pixel method; Used to calculate aerosol factors The dark pixel method of AOD inversion is also called the dense vegetation method. It takes advantage of the low reflectivity of dense vegetation in the red and blue bands. It is currently a widely used AOD inversion algorithm. The so-called dark pixels refer to areas with low reflectivity in remote sensing images, such as trees and shrubs. Studies have shown that the 2.1μm band is not affected by the emitted radiation, its apparent reflectivity is approximately equal to the surface reflectivity, and it is transparent to most aerosol types. Therefore, it can be used to detect dark targets. Analysis of spectral data obtained by satellites and aircraft found that there is an obvious linear relationship between the red (0.66μm), blue (0.47μm) bands and the 2.1μm band: , , in, , are the surface reflectances in the red and blue bands, respectively. Represents the apparent reflectance of the 2.1μm band. Therefore, the surface reflectance of the red and blue bands can be obtained from the apparent reflectance of the mid-infrared band, and then the surface contribution is removed from the apparent reflectance of the red and blue bands to obtain the atmospheric parameters, build a lookup table, and then obtain AOD. The spectral parameters use the red light band, blue light band, and mid-infrared band. The AOD value inverted by the dark pixel method ranges from 0 to 2. The closer the value is to 0, the more transparent the atmosphere is and the less pollution it has. The closer it is to 2, the higher the atmospheric turbidity is and the more serious the pollution is. The specific calculation process is as follows: First, obtain the TOA reflectivity of the data , using the linear relationship between the red and blue bands and the short-wave infrared to obtain the simulated surface reflectivity : , , in, It is the short-wave infrared band data of TOA data. It is the red band data simulating the surface reflectivity. It is the blue band data that simulates the surface reflectivity; Then, the surface contribution is removed from the apparent reflectance of the red and blue bands to obtain the atmospheric parameters, build a lookup table, and then obtain the aerosol weight factor : .

[0023] in, Indicates the TOA reflectivity data of the third band, represents the TOA reflectivity data of the 5th band, and the f(·) function represents the process function of removing the surface contribution from the apparent reflectivity of the red and blue bands, obtaining the atmospheric parameters, constructing the lookup table, and obtaining the aerosol optical thickness.

[0024] Step 3.4, normalize the time weight factor; By combining all the surface reflectance data Normalize according to the time distance to form a time weight factor : Calculate the maximum and minimum range of the time weight factor: , in, and Respectively represent the minimum and maximum values ​​of the time weight factor.

[0025] Normalized time weight factor: , in, Represents the normalized time weight factor, ranging from [0, 1].

[0026] Step 4: Calculate the comprehensive weight factor by combining the normalized cloud weight factor, rainfall weight factor, aerosol weight factor, and time weight factor. Filter the comprehensive weight factor through the comprehensive weight factor screening function, and use the pixel with the least impact of the comprehensive weight factor for data fusion. like Figure 3 As shown, in order to obtain the final fusion data , use the comprehensive weight factor screening function to screen make Optimize in the range [i,j]. That is: , in, Represents the final fusion data, the matrix F is It represents the comprehensive weight factor screening function, which is a standard basis matrix with one element being 1 and other elements being zero. Its calculation method is as follows: exist When (x is within the range of i, y is within the range of j) the maximum value is obtained, then is equal to 1, the others is equal to 0, that is: , Among them, the comprehensive weight factor for: , in, represents the normalized cloud weight factor, represents the normalized rainfall weight factor, represents the normalized aerosol weight factor, represents the normalized time weight factor, w1, w2, w3, and w4 respectively represent the weights corresponding to the normalized cloud weight factor, the normalized rainfall weight factor, the normalized aerosol weight factor, and the normalized time weight factor, and are determined as follows: w1=0.4, w2=0.1, w3=0.2, and w4=0.3.

[0027] like Figure 4 As shown, This is a schematic diagram of the global surface reflectivity data finally obtained under the above conditions.

[0028] On the other hand, the present invention provides a surface reflectance data fusion system based on multi-factor weight optimization, and each unit included in the system can implement each step of the above method. Specifically, the system includes: The data preprocessing unit is used to perform radiometric calibration and geometric correction on MODIS and Landsat8 OLI multispectral data, generate top of atmosphere reflectance data, and generate surface reflectance data of different spectral bands after atmospheric correction; A data adjustment unit is used to adjust the surface reflectance data of Landsat8 OLI to be consistent with the spectral range of the surface reflectance data of MODIS by using the spectral band adjustment factor technology; A weight factor calculation unit is used to calculate the cloud weight factor using the HOT-VBR index method, calculate the rainfall weight factor using the MODRA algorithm, calculate the aerosol weight factor of each pixel using the dark pixel method, calculate the time weight factor and normalize it; The comprehensive weight factor screening unit is used to calculate the comprehensive weight factor by combining the normalized cloud weight factor, rainfall weight factor, aerosol weight factor and time weight factor, and screen the comprehensive weight factor through the comprehensive weight factor screening function, and use the pixel with the least influence of the comprehensive weight factor for data fusion.

[0029] In a third aspect, the present invention provides an electronic device, comprising: one or more processors; a memory for storing one or more programs; wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the aforementioned surface reflectance data fusion method based on multi-factor weight optimization.

[0030] In a fourth aspect, the present invention provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, enables the processor to implement the aforementioned surface reflectance data fusion method based on multi-factor weight optimization.

[0031] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A surface reflectance data fusion method based on multi-factor weight optimization, characterized in that: The method comprises the following steps: Step 1: Perform radiometric calibration and geometric correction on MODIS and Landsat8 OLI multispectral data to generate top of atmosphere reflectance data, and generate surface reflectance data of different spectral bands after atmospheric correction; Step 2: Use the spectral bandpass adjustment factor technology to adjust the surface reflectance data of Landsat8 OLI to be consistent with the spectral range of the surface reflectance data of MODIS; Step 3: Calculate the cloud weight factor using the HOT-VBR index method, calculate the rainfall weight factor using the MODRA algorithm, calculate the aerosol weight factor of each pixel using the dark pixel method, calculate the time weight factor and normalize it; Step 4: Calculate the comprehensive weight factor by combining the normalized cloud weight factor, rainfall weight factor, aerosol weight factor, and time weight factor. Filter the comprehensive weight factor through the comprehensive weight factor screening function, and use the pixel with the least impact of the comprehensive weight factor for data fusion.

2. The surface reflectance data fusion method based on multi-factor weight optimization according to claim 1 is characterized in that: The second step comprises: The spectral bandpass adjustment factor technology is used to adjust the surface reflectance data of Landsat8 OLI to be consistent with the spectral range of the surface reflectance data of MODIS: ， In the formula, the superscript i represents the sensor code, which takes the value of 1 or 2, i=1 represents the MODIS sensor, and i=2 represents the Landsat 8 OLI sensor; j represents the year of the data, k represents the month of the data, b represents the band of the data source, which takes the value of 1, 2, ... 8, m and n are the coordinates of the pixel, Represents the adjusted Landsat 8 OLI surface reflectance data, Represents the Landsat 8 OLI surface reflectance data before adjustment, Represents the spectral bandpass adjustment factor for each band.

3. The surface reflectance data fusion method based on multi-factor weight optimization according to claim 1 is characterized in that: The step three comprises: The cloud weight factor calculation formula is as follows: ， ， ， in, represents the cloud weight factor, Indicates the adjusted Landsat8 OLI surface reflectance data corresponding to the first and third bands and The calculated HOT index data, It is the VBR index data calculated using the adjusted Landsat8 OLI surface reflectance data corresponding to the three bands: Band 1, Band 2, and Band 3.

4. The surface reflectance data fusion method based on multi-factor weight optimization according to claim 1 is characterized in that: The step three comprises: The MODRA algorithm calculates the rainfall weight factor using thresholds in the red and near-infrared bands. : ， in, Refers to the adjusted Landsat8 OLI surface reflectance data of Band 8, Refers to the adjusted Band 3 Landsat8 OLI surface reflectance data.

5. The surface reflectance data fusion method based on multi-factor weight optimization according to claim 1 is characterized in that: The step three comprises: Aerosol weight factor Calculate according to the following formula: ， in, Indicates the TOA reflectivity data of the third band, represents the TOA reflectance data of band 5, and the f(·) function represents the removal of the surface contribution from the apparent reflectance of the red and blue bands; The time of all surface reflectance data Normalize according to the time distance to form a time weight factor , and normalize it: ， in, is the normalized time weight factor, and Respectively represent the minimum and maximum values ​​of the time weight factor.

6. The surface reflectance data fusion method based on multi-factor weight optimization according to claim 1 is characterized in that: The step 4 comprises: Using the comprehensive weight factor screening function to screen the adjusted Landsat8 OLI surface reflectance data To optimize it in the range [i,j]: ， in, Represents the final fused data, matrix Represents the comprehensive weight factor screening function, which is a standard basis matrix with one element being 1 and other elements being zero. exist When the maximum value is obtained, the element is equal to 1, the other elements is equal to 0: ， Among them, i∈x, j∈y.

7. The surface reflectance data fusion method based on multi-factor weight optimization according to claim 6 is characterized in that: The comprehensive weight factor for: ， In the formula, represents the normalized cloud weight factor, represents the normalized rainfall weight factor, represents the normalized aerosol weight factor, represents the normalized time weight factor, w1, w2, w3, and w4 represent the weights corresponding to the normalized cloud weight factor, the normalized rainfall weight factor, the normalized aerosol weight factor, and the normalized time weight factor, respectively.

8. A surface reflectance data fusion system based on multi-factor weight optimization, characterized in that: include: The data preprocessing unit is used to perform radiometric calibration and geometric correction on MODIS and Landsat8 OLI multispectral data, generate top of atmosphere reflectance data, and generate surface reflectance data of different spectral bands after atmospheric correction; A data adjustment unit is used to adjust the surface reflectance data of Landsat8 OLI to be consistent with the spectral range of the surface reflectance data of MODIS by using the spectral band adjustment factor technology; A weight factor calculation unit is used to calculate the cloud weight factor using the HOT-VBR index method, calculate the rainfall weight factor using the MODRA algorithm, calculate the aerosol weight factor of each pixel using the dark pixel method, calculate the time weight factor and normalize it; The comprehensive weight factor screening unit is used to calculate the comprehensive weight factor by combining the normalized cloud weight factor, rainfall weight factor, aerosol weight factor and time weight factor, and screen the comprehensive weight factor through the comprehensive weight factor screening function, and use the pixel with the least influence of the comprehensive weight factor for data fusion.

9. An electronic device, characterized in that: include: one or more processors; A memory for storing one or more programs; Wherein, when one or more programs are executed by the one or more processors, the one or more processors implement the surface reflectance data fusion method based on multi-factor weight optimization as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: Executable instructions are stored thereon, and when the instructions are executed by the processor, the processor can implement the surface reflectance data fusion method based on multi-factor weight optimization as described in any one of claims 1 to 7.

Images