Satellite flood detection method, device and equipment and storage medium

Abstract

The invention discloses a satellite flood detection method, device and equipment and a storage medium, and relates to the technical field of computers, and the method comprises the steps: obtaining the channel data of a satellite MERSI load; preprocessing the channel data, and constructing spectral features and remote sensing index features suitable for water body detection based on the preprocessed channel data; using a flood product loaded by a satellite MODIS as a truth value training sample, and training the spectral features and the remote sensing index features to obtain a naive Bayesian water body recognition model; after the trained naive Bayes water body recognition model is used for recognizing a water body, a spherical geometric model is used for removing cloud shadows in the water body, and a flood detection result is obtained.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a satellite flood detection method, device, equipment and storage medium. Background Art

[0002] Flood disasters have a great impact on human life safety and ecological environment. With the continuous development of remote sensing technology, the use of remote sensing technology to detect flood disasters has become one of the current research hotspots. Optical and radar sensors are used for flood disaster detection. Among them, synthetic aperture radar technology has many limitations in water body recognition, including image noise and speckle effect, confusion between ground objects and water bodies, changes in waves and water surface roughness, imaging geometry and perspective effects, data processing complexity, resolution limitations, and polarization mode selection. Therefore, optical imaging is increasingly used in flood detection.

[0003] At present, the main idea of ​​flood detection methods based on optical sensors is to conduct in-depth analysis of the extracted optical remote sensing water body characteristic index, which can be summarized into the following two types: 1) Time series method anomaly or change detection, mainly including unsupervised change detection proposed by REZKI MOSTEFA and spatiotemporal model for anomaly detection proposed by Liu Xiaoyi. These methods detect changes caused by abnormal events by performing complex statistical analysis or modeling on the changes in water body characteristics of sequence remote sensing images, thereby achieving the purpose of flood detection; 2) Traditional physical models, machine learning or deep learning models, such as Zhu Hui et al. proposed to use the regional growing method to quickly and accurately extract the scope of flood disasters; Zhang Hui et al. used the random forest model for water body extraction and proved that the extraction accuracy of this method is higher than that of the traditional threshold method and water body index method; Liu Yong et al. and Dong Zhenjun introduced deep learning semantic segmentation and change detection neural network-HRNet to detect floods and proved the effectiveness of the method; Wang Yanting et al. used the random forest model of multi-feature fusion to identify water bodies by fusing water body features extracted from multi-source remote sensing images, thereby achieving the purpose of dynamic monitoring of floods.

[0004] Since machine learning or deep learning models can better explore the deep laws of water spectral feature distribution, they are widely used in water body detection. However, the main idea of ​​most current methods is to analyze the characteristics of water body remote sensing indexes, ignoring the cloud shadows that are similar to the spectral characteristics of water bodies, and the influence of the snow underlying surface that affects the spectral characteristics. Summary of the invention

[0005] The purpose of this application is to provide a satellite flood detection method, device, equipment and storage medium.

[0006] To achieve the above objectives, this application provides the following solutions:

[0007] In a first aspect, the present application provides a satellite flood detection method, comprising:

[0008] Obtain channel data of the satellite MERSI payload;

[0009] Preprocessing the channel data, and constructing spectral features and remote sensing index features suitable for water body detection based on the preprocessed channel data;

[0010] Using the flood products of satellite MODIS payload as true value training samples, the spectral features and the remote sensing index features are trained to obtain a naive Bayesian water body recognition model;

[0011] After the water body is identified by using the trained naive Bayesian water body identification model, the cloud shadow in the water body is removed by using a spherical geometry model to obtain a flood detection result.

[0012] Optionally, the step of acquiring channel data of a satellite MERSI payload includes:

[0013] A spatial matching threshold with MODIS data is set according to the data spatial resolution to obtain successfully matched channel data, wherein the matched data at least includes: visible light channel data, near infrared channel data, shortwave infrared channel data, longitude and latitude, solar zenith angle and MODIS MCDWD_L3_NRT data, wherein the wavelength of the visible light channel data is 0.67um, the wavelength of the infrared channel data is 0.86um, and the wavelength of the shortwave infrared channel data is 1.64um.

[0014] Optionally, the step of preprocessing the channel data includes:

[0015] Perform calibration calculation on the successfully matched channel data to obtain the channel data within the effective range;

[0016] The daytime data with a solar zenith angle less than 85° is screened from the channel data within the effective range.

[0017] Optionally, the step of constructing spectral features and remote sensing index features suitable for water body detection based on the preprocessed channel data includes:

[0018] Spectral features and remote sensing index features are constructed based on the screened channel data, the spectral features are 0.67um, 0.86um and 1.64um channel features, and the remote sensing index features include at least the normalized difference water index, the normalized difference vegetation index and the normalized difference ice and snow index.

[0019] Optionally, the step of using the flood products of satellite MODIS payload as true value training samples to train the naive Bayesian water body recognition model includes:

[0020] Select the MODIS flood data corresponding to the MERSI data time;

[0021] Using MODIS flood data as true training samples, the extracted spectral features and remote sensing index features are trained to obtain a naive Bayesian water body recognition model.

[0022] Optionally, the step of removing cloud shadows in the water body using a spherical geometric model to obtain a flood detection result includes:

[0023] Establish a spherical geometric model between cloud shadows and clouds;

[0024] Based on the assumption that a cloud pixel casts at most one cloud shadow pixel, a one-to-one relationship between cloud shadows and clouds is iteratively established to calculate the cloud shadow position;

[0025] The cloud shadows in the water body are removed based on the cloud shadow positions to obtain a flood detection result.

[0026] Optionally, after the step of removing the cloud shadows in the identified water body by using the spherical geometric model to obtain the flood detection result, the method further includes:

[0027] The flood detection results of the naive Bayesian water body identification model at multiple times are selected and compared with the flood detection results of Real Earth to verify the accuracy of the naive Bayesian water body identification model.

[0028] In a second aspect, the present application provides a satellite flood detection device, comprising:

[0029] Data preprocessing module, used to obtain channel data of satellite MERSI payload;

[0030] Preprocessing the channel data, and constructing spectral features and remote sensing index features suitable for water body detection based on the preprocessed channel data;

[0031] A training module, for using the flood products of the satellite MODIS payload as true value training samples, training the spectral features and the remote sensing index features to obtain a naive Bayesian water body recognition model;

[0032] The prediction module is used to obtain flood detection results by removing cloud shadows in the water body using a spherical geometry model after identifying the water body using the trained naive Bayesian water body recognition model.

[0033] Optionally, the data preprocessing module is further used to:

[0034] Get the channel data of the satellite MERSI payload within the specified time period;

[0035] A spatial matching threshold with MODIS data is set according to the data spatial resolution to obtain successfully matched channel data, wherein the matched data at least includes: visible light channel data, near infrared channel data, shortwave infrared channel data, longitude and latitude, solar zenith angle and MODIS MCDWD_L3_NRT data, wherein the wavelength of the visible light channel data is 0.67um, the wavelength of the infrared channel data is 0.86um, and the wavelength of the shortwave infrared channel data is 1.64um.

[0036] Optionally, the data preprocessing module is further used to:

[0037] Perform calibration calculation on the successfully matched channel data to obtain the channel data within the effective range;

[0038] The daytime data with a solar zenith angle less than 85° is screened from the channel data within the effective range.

[0039] Optionally, the training module is further used to:

[0040] Spectral features and remote sensing index features are constructed based on the screened channel data, the spectral features are 0.67um, 0.86um and 1.64um channel features, and the remote sensing index features include at least the normalized difference water index, the normalized difference vegetation index and the normalized difference ice and snow index.

[0041] Optionally, the training module is further used to:

[0042] Select the MODIS flood data corresponding to the MERSI data time;

[0043] Using MODIS flood data as true training samples, the extracted spectral features and remote sensing index features are trained to obtain a naive Bayesian water body recognition model.

[0044] Optionally, the prediction module is further used to:

[0045] Establish a spherical geometric model between cloud shadows and clouds;

[0046] Based on the assumption that a cloud pixel casts at most one cloud shadow pixel, a one-to-one relationship between cloud shadows and clouds is iteratively established to calculate the cloud shadow position;

[0047] The cloud shadows in the water body are removed based on the cloud shadow positions to obtain a flood detection result.

[0048] Optionally, the prediction module is further used to:

[0049] The flood detection results of the naive Bayesian water body identification model at multiple times are selected and compared with the flood detection results of Real Earth to verify the accuracy of the naive Bayesian water body identification model.

[0050] In a third aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of any one of the satellite flood detection methods described above.

[0051] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any one of the satellite flood detection methods described above.

[0052] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of any one of the satellite flood detection methods described above.

[0053] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0054] The present application provides a satellite flood detection method, device, equipment and storage medium. By adopting a naive Bayesian water body recognition model based on satellite MERSI payload data and MODIS flood data as true value training samples, and combining a spherical geometry model to remove cloud shadows, the flood disaster monitoring technology has been significantly improved, and the accuracy of flood detection has been effectively improved. By accurately identifying water bodies and eliminating cloud shadow interference, the reliability of monitoring results is ensured. At the same time, the real-time data of the satellite MERSI payload is used to enhance the real-time nature of flood monitoring, so that the disaster situation can be grasped in a timely manner. In addition, the automated processing process reduces manual intervention, improves monitoring efficiency, and provides strong support for emergency response. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0056] Figure 1 A schematic diagram of a flow chart of a satellite flood detection method provided in one embodiment of the present application;

[0057] Figure 2 This is one of the schematic diagrams showing the effect of a satellite flood detection method provided in one embodiment of the present application;

[0058] Figure 3 One of the logical schematic diagrams of a satellite flood detection method provided in one embodiment of the present application;

[0059] Figure 4 A second logic diagram of a satellite flood detection method provided in an embodiment of the present application;

[0060] Figure 5 A second schematic diagram of the effect of a satellite flood detection method provided in an embodiment of the present application;

[0061] Figure 6 A third schematic diagram of the effect of a satellite flood detection method provided in an embodiment of the present application;

[0062] Figure 7 A schematic diagram of functional modules of a satellite flood detection device provided in one embodiment of the present application;

[0063] Figure 8 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0064] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0065] like Figure 1 As shown, some embodiments of the present application provide a satellite flood detection method, which includes the following steps 101 to 104. Among them:

[0066] Step 101, obtaining channel data of the satellite MERSI payload;

[0067] In an embodiment of the present application, the system can obtain data from the medium resolution spectral imager (MERSI) carried by the Fengyun-3D satellite (FY-3D) in the visible, near-infrared and far-wave infrared channels. These data contain rich surface spectral information, which is essential for water body detection. Continuous data over a period of time is usually selected for monitoring, such as data from August 1, 2023 to August 31, 2023.

[0068] Step 102, preprocessing the channel data, and constructing spectral features and remote sensing index features suitable for water body detection based on the preprocessed channel data;

[0069] In an embodiment of the present application, the acquired MERSI load channel data is calibrated to ensure the accuracy and consistency of the data, thereby obtaining load channel data within the effective range. Screening is performed according to the solar zenith angle, and only daytime data (such as solar zenith angle less than 85°) is retained to reduce the impact of lighting conditions on water body detection. The MERSI data is spatially matched with the corresponding MODIS (Moderate Resolution Imaging Spectroradiometer) data, and the matched data is obtained according to the spatial resolution of the two (such as setting the matching threshold to 250m), including visible light channel data, near infrared channel data, shortwave infrared channel data, longitude and latitude, solar zenith angle, etc.

[0070] According to the spectral characteristics reflected by the water body, the spectral characteristics and remote sensing index characteristics suitable for water body detection are selected, such as 0.67um, 0.86um and 1.64um channel characteristics, as well as remote sensing index characteristics such as normalized vegetation index (NDVI), normalized water index (NDWI) and normalized difference ice and snow index (NDSI). Based on the screened MERSI channel data, these spectral characteristics and remote sensing index characteristics are calculated for subsequent water body identification.

[0071] Step 103, using the flood products of satellite MODIS payload as true value training samples, training the spectral features and the remote sensing index features to obtain a naive Bayesian water body recognition model;

[0072] In the embodiment of the present application, MODIS flood products (such as static water and flood data in MODIS MCDWD_L3_NRT data) of the corresponding time period are selected as true value training samples to ensure the consistency of MODIS data and MERSI data in time and space for model training.

[0073] The Naive Bayesian model is used as the classifier because it is simple, effective and suitable for large-scale data processing. The extracted spectral features and remote sensing index features are used as input, and the MODIS flood data is used as the true value label to train the Naive Bayesian model. After multiple iterations and optimizations, a trained Naive Bayesian water body recognition model is obtained.

[0074] Step 104, after the water body is identified using the trained naive Bayesian water body identification model, a spherical geometry model is used to remove cloud shadows in the water body to obtain a flood detection result.

[0075] In an embodiment of the present application, the preprocessed MERSI channel data is input into the trained naive Bayes water body recognition model to obtain a preliminary water body recognition result. Since cloud shadows and water bodies have similar spectral characteristics in optical images, they are difficult to distinguish in preliminary recognition. Therefore, a spherical geometric model is used to construct a spherical geometric relationship between cloud shadows and clouds, and the cloud shadows are removed from the water body recognition results by calculating their positions and ranges. After the cloud shadows are removed, the final flood detection results are obtained. By comparing the detection results at different times, the dynamic changes of the flood disaster can be monitored.

[0076] The embodiment of the present application uses a naive Bayesian water body recognition model based on satellite MERSI payload data and MODIS flood data as true value training samples, and combines a spherical geometry model to remove cloud shadows, thereby significantly improving flood disaster monitoring technology and effectively improving the accuracy of flood detection. By accurately identifying water bodies and removing cloud shadow interference, the reliability of monitoring results is ensured. At the same time, the real-time data of the satellite MERSI payload is used to enhance the real-time nature of flood monitoring, so that disaster situations can be grasped in a timely manner. In addition, the automated processing process reduces manual intervention, improves monitoring efficiency, and provides strong support for emergency response.

[0077] Optionally, the step 101 includes:

[0078] Step 1011, obtaining channel data of the satellite MERSI payload within a specified time period;

[0079] In the embodiment of the present application, in order to carry out subsequent flood detection, it is necessary to obtain the channel data of the medium-resolution spectral imager (MERSI) carried by the Fengyun-3D satellite (FY-3D) within a specific time period (such as 20230801-20230831). These data contain the spectral information of the surface and are the basis for subsequent analysis and modeling. The channel data mainly includes visible light channel data, near-infrared channel data, and short-wave infrared channel data. These data can reflect the spectral characteristics of surface objects and are particularly important for water body detection.

[0080] Step 1012, according to the data spatial resolution, a spatial matching threshold with the MODIS data is set to obtain successfully matched channel data, wherein the matched data at least includes: visible light channel data, near infrared channel data, short-wave infrared channel data, longitude and latitude, solar zenith angle, and MODIS MCDWD_L3_NRT data, wherein the wavelength of the visible light channel data is 0.67um, the wavelength of the infrared channel data is 0.86um, and the wavelength of the short-wave infrared channel data is 1.64um.

[0081] In the embodiment of the present application, since the FY-3D MERSI load data and the MODIS MCDWD_L3_NRT data have different spatial resolutions, in order to effectively combine the two, a spatial matching threshold needs to be set to ensure the consistency of the data in space. According to the spatial resolution of the FY-3D MERSI load data and the MODIS MCDWD_L3_NRT data, a reasonable matching threshold (such as 250m) is set. This threshold determines the maximum distance or range in which the two can match in space.

[0082] The FY-3D MERSI payload data and MODIS MCDWD_L3_NRT data are spatially matched, and the matching data pairs in spatial position are screened out according to the set matching threshold. The successfully matched channel data will be used for subsequent modeling and analysis. These data not only include the visible light, near infrared and short-wave infrared channel data of the FY-3D MERSI payload, but also include MODIS flood data (MCDWD_L3_NRT), as well as additional information such as longitude and latitude information and solar zenith angle.

[0083] Among them, the visible light channel data reflects the reflection characteristics of surface objects in the visible light band, which is of great significance for water body detection. The near-infrared channel data provides reflection information of surface objects in the near-infrared band, which helps to distinguish water bodies from other ground objects. The short-wave infrared channel data contains the reflection characteristics of surface objects in the short-wave infrared band, which is helpful for identifying specific types of ground objects (such as water bodies, vegetation, etc.). Longitude and latitude provide geographical location information for data to ensure the spatial accuracy of the data. The solar zenith angle reflects the angle between the sun's rays and the surface normal, which is crucial for subsequent steps such as geometric modeling and cloud shadow removal. MODIS MCDWD_L3_NRT data is used as a true value training sample to train the naive Bayes water body recognition model and improve the accuracy of water body recognition.

[0084] Optionally, the steps of step 102 include:

[0085] Step 1021, performing calibration calculation on the successfully matched channel data to obtain channel data within a valid range;

[0086] In the embodiment of the present application, on the basis of calibration calculation, channel data within the effective range are further screened out to ensure the accuracy and reliability of subsequent analysis. According to the result of calibration calculation, a reasonable threshold range is set to screen out channel data that meet the requirements. These thresholds are usually determined based on factors such as sensor characteristics, surface coverage type, and weather conditions. Channel data within the effective range are obtained, and these data have higher accuracy and reliability in terms of radiation brightness or reflectivity, providing better data guarantee for subsequent water body detection.

[0087] Step 1022, filtering daytime data with a solar zenith angle less than 85° from the channel data within the effective range.

[0088] In an embodiment of the present application, the solar zenith angle is the angle between the sun's rays and the surface normal, which reflects the position of the sun in the sky. For water body detection, data with a small solar zenith angle during the day usually have better lighting conditions and lower cloud shadow effects, and are therefore more suitable for subsequent analysis and modeling. In the channel data within the effective range, data less than 85° are screened out based on the information of the solar zenith angle. The selection of this threshold is based on a comprehensive consideration of the effects of sunlight conditions and cloud shadows. A solar zenith angle of less than 85° usually means that the data is acquired during the day when the sun is relatively high, and therefore has better lighting conditions. Daytime data with a solar zenith angle of less than 85° are screened out, and these data are more superior in terms of lighting conditions and cloud shadow effects, providing better data support for subsequent water body detection and modeling.

[0089] Optionally, the step 102 includes: constructing spectral features and remote sensing index features based on the filtered channel data, the spectral features are 0.67um, 0.86um and 1.64um channel features, and the remote sensing index features include at least normalized difference water index, normalized difference vegetation index and normalized difference ice and snow index.

[0090] In the embodiment of the present application, the Normalized Difference Water Index (NDWI) is a remote sensing index feature used to enhance water body information. It can highlight the water area by performing difference calculation and normalization on the reflectivity of the near infrared band and the short wave infrared band. In flood detection, NDWI is an important tool for identifying water bodies.

[0091] The Normalized Difference Vegetation Index (NDVI) is a remote sensing index feature used to monitor vegetation coverage and growth. It reflects the lushness and health of vegetation by performing differential calculations and normalizing the reflectance of the red band and the near-infrared band. Although NDVI is not directly used to identify water bodies in flood detection, it can help identify vegetation coverage areas, thereby indirectly assisting water body identification.

[0092] The Normalized Difference Snow and Ice Index (NDSI) is a remote sensing index feature used to identify snow-covered areas. It can highlight snow-covered areas by performing differential calculations on the reflectance of the shortwave infrared band and the visible light band and normalizing them. In flood detection, considering the impact of snow and ice on water body identification (snow and ice may be similar to water bodies in spectral characteristics), the introduction of NDSI can help eliminate snow and ice interference and improve the accuracy of water body identification.

[0093] By constructing these spectral features and remote sensing index features, researchers can extract richer and more accurate surface cover information from satellite remote sensing data. In flood detection, these features can not only be used to identify water areas, but also help eliminate interference factors such as ice, snow, and cloud shadows, thereby improving the accuracy and reliability of detection. At the same time, these features can also provide important data support for subsequent model construction and optimization.

[0094] Optionally, the step 103 includes:

[0095] Step 1031, selecting MODIS flood data corresponding to the MERSI data time;

[0096] In the embodiment of the present application, in order to obtain MODIS flood data that is synchronized with the MERSI data time. Since MODIS and MERSI belong to different satellite systems, their data collection time and frequency may be different. Therefore, it is necessary to select MODIS flood data that corresponds to or is close to the MERSI data time window to ensure the temporal and spatial consistency of the data. This allows the MODIS flood data to be used as a reference or true value for the MERSI data for subsequent model training and verification.

[0097] Step 1032, using MODIS flood data as true value training samples, the extracted spectral features and remote sensing index features are trained to obtain a naive Bayes water body recognition model.

[0098] In the embodiment of the present application, after obtaining the MODIS flood data corresponding to the MERSI data time, the MERSI data needs to be preprocessed, including calibration calculation, data screening, quality control and other steps to obtain channel data within the effective range.

[0099] Based on the processed MERSI channel data, spectral features and remote sensing index features are constructed. The spectral features are 0.67um, 0.86um and 1.64um channel features, and the remote sensing index features include the normalized difference water index (NDWI), the normalized difference vegetation index (NDVI) and the normalized difference ice and snow index (NDSI), etc. These features can highlight the information of land features such as water bodies, vegetation, ice and snow.

[0100] MODIS flood data is used as true value or reference data to train the naive Bayesian water body identification model. Since MODIS data has high spatial resolution and temporal resolution, and has been used and verified for a long time, its flood detection results have high accuracy and reliability. Therefore, using MODIS flood data as true value training samples can make the trained naive Bayesian water body identification model more accurate and reliable.

[0101] The extracted spectral features and remote sensing index features are used as input data and input into the naive Bayes model for training. During the training process, the model will learn the spectral features and distribution patterns of water bodies based on the input features and data labels (i.e. MODIS flood data), thereby constructing a naive Bayes water body recognition model that can identify water bodies.

[0102] The naive Bayes water body recognition model obtained after training can be used to detect and identify flood disasters in new MERSI data. In practical applications, the model can be applied to real-time satellite remote sensing data to achieve rapid response and accurate monitoring of flood disasters.

[0103] At the same time, in order to verify the accuracy and reliability of the model, the prediction results of the model can also be compared and analyzed with MODIS flood data. By continuously adjusting and optimizing the model parameters and algorithms, the recognition accuracy and generalization ability of the model can be further improved.

[0104] Optionally, the step 104 includes:

[0105] Step 1041, establishing a spherical geometric model between the cloud shadow and the cloud;

[0106] Step 1042, based on the assumption that one cloud pixel casts at most one cloud shadow pixel, iteratively establish a one-to-one relationship between the cloud shadow and the cloud, and calculate the cloud shadow position;

[0107] Step 1043: removing the cloud shadows in the water body based on the cloud shadow positions to obtain flood detection results.

[0108] In the application embodiment, cloud shadow is a shadow area formed on the ground after the cloud blocks the sunlight, and its position is related to the position of the cloud, the position of the sun, and the shape of the earth. By establishing a spherical geometry model between the cloud shadow and the cloud, the position of the cloud shadow on the ground can be accurately calculated. The spherical geometry model takes into account the fact that the earth is an approximate sphere. Through parameters such as the position of the cloud, the position of the sun (usually expressed by the solar zenith angle and azimuth), and the radius of the earth, the specific position of the cloud shadow on the ground can be derived.

[0109] One cloud pixel casts at most one cloud shadow pixel. This means that in the spherical geometry model, the shadow area formed by a cloud block on the surface is unique, and there will not be a situation where a cloud block casts multiple shadow areas. For each cloud pixel in the image, the corresponding cloud shadow position is calculated using the spherical geometry model based on its position and the position of the sun. These cloud shadow positions are then matched with the actual shadow areas in the image to establish a one-to-one relationship between cloud shadows and clouds. This process needs to be iterated until the cloud shadow positions of all cloud pixels are accurately calculated and matched. After determining the one-to-one relationship between cloud shadows and clouds, the spherical geometry model can be used to accurately calculate the shadow position of each cloud pixel on the surface.

[0110] After obtaining the cloud shadow positions, the pixels corresponding to these positions can be removed from the water body identification results, thereby eliminating the interference of cloud shadows on flood detection results. This process usually involves image processing techniques, such as pixel replacement and image filtering. By removing cloud shadows, water body areas can be identified more accurately, thereby improving the accuracy of flood detection. After removing the interference of cloud shadows, the final flood detection results will be closer to the actual situation. This result can be used in subsequent disaster assessment, emergency response and other work.

[0111] Optionally, after step 104, the method further includes: selecting flood detection results of the naive Bayesian water body recognition model at multiple times, and comparing them with the Real Earth flood detection results to verify the accuracy of the naive Bayesian water body recognition model.

[0112] In the embodiment of the present application, the trained naive Bayesian water body recognition model is applied to the actual FY-3D satellite remote sensing data to detect flood disasters at multiple times. The selection of these times is usually to cover different weather conditions, geographical regions and flood severity to ensure the comprehensiveness and representativeness of the verification. From the results of the model application, the flood detection results of multiple times are selected. These results will be used for subsequent comparative analysis to evaluate the accuracy and reliability of the model.

[0113] Real Earth flood detection results were used as comparative data. Real Earth may be a platform that provides real-time or historical environmental data worldwide, and its flood detection results are considered to be a more accurate and reliable reference standard. Selecting Real Earth flood detection results as comparative data can ensure the objectivity and fairness of the verification process. At the same time, Real Earth data may cover a wider geographical area and a longer time span, which is helpful for comprehensively evaluating the performance of the model.

[0114] Compare the flood detection results of the Naive Bayesian water body identification model at multiple times with the flood detection results of the corresponding times provided by Real Earth. The comparison may include key indicators such as the region, area, and intensity of the flood. Through comparative analysis, the accuracy of the Naive Bayesian water body identification model in flood detection can be evaluated. If the model results are highly consistent with the Real Earth data, it means that the model has a high accuracy; if there are significant differences, it is necessary to further analyze the reasons and improve the model.

[0115] For example, refer to Figure 2 This is a comparison chart between the flood identification results of Real Earth and the flood prediction results obtained through the steps of the satellite flood detection method provided in some embodiments of the present application. It can be seen that the flood prediction results are highly consistent with the actual flood identification results of Real Earth. The recognition accuracy of cloud shadows is high, which can provide more accurate support for subsequent flood identification.

[0116] Exemplarily, in an achievable embodiment of the present application:

[0117] Step 1: Reference Figure 3 , obtain the MERSI payload channel data of FY-3D satellite from 20230801 to 20230831, and search for the MODIS data of the corresponding date based on the FY-3D data, and set the spatial matching threshold of the FY-3D MERSI payload channel data and MODIS MCDWD_L3_NRT data to 250m according to the spatial resolution of the two, so as to obtain the matched data of FY-3D satellite and MODIS, including the visible light channel data, near infrared channel data, shortwave infrared channel data, longitude and latitude, solar zenith angle and MODIS MCDWD_L3_NRT data of FY-3D satellite MERSI payload;

[0118] Step 2: Preprocess the matched data of the MERSI payload of the Fengyun-3D satellite, perform calibration calculations on it, obtain matching data within the effective range, and perform quality control on the processed data. Only the matching data with a solar zenith angle less than 85° (daytime) are screened, and remote sensing index features suitable for water body detection are constructed based on the screened data;

[0119] Step 3: Reference Figure 4 , a naive Bayes model is constructed, in which MODIS Flood 3-Day 250m data is used as the true value sample, the selected spectral features and remote sensing index features are trained, and the naive Bayes water body recognition model is obtained;

[0120] Step 4: Although the spectral characteristics of water surface are different from other land types in optical images, cloud shadows and water have similar spectral characteristics in visible light, near infrared, far-wave infrared and thermal infrared channels, which makes them difficult to distinguish in water body identification based on the spectral characteristics of optical images. Therefore, geometry can be used to construct a spherical geometric model to remove cloud shadows. This method first establishes a spherical geometric model between cloud shadows and clouds, and then iteratively establishes a one-to-one relationship between cloud shadows and clouds based on the assumption that one cloud pixel projects at most one cloud shadow pixel.

[0121] Assume that cloud pixel B (Lon b ,Lat b ) in the optical image is known, the sensor azimuth can be used Calculate the arc to locate the actual position P (Lon p ,Lat p ), the calculation formula is as follows:

[0122]

[0123] Where R is the radius of the earth, arc is the parallax distance between the real cloud pixel position P and the cloud position in the satellite image B.

[0124] At a given cloud pixel position P(Lon p ,Lat p ) in the case of Consider cloud shadow length The cloud shadow position A(Lon a ,Lat a ), the calculation formula is as follows:

[0125]

[0126]

[0127] Among them, the shadow angle α and the parallax angle β can be used to and It is calculated as an arc along a circle with a radius of R. The shadow angle α and the parallax angle β can be calculated using the following equations:

[0128]

[0129] Where δ represents the shadow angle α or the parallax angle β, R is the radius of the earth, h is the cloud height, and θ is the zenith angle.

[0130] Some embodiments of the present application provide a satellite flood detection method that uses a naive Bayesian model as a classifier, selects spectral features and remote sensing index features suitable for water body detection from FY3D / MERSI satellite observation data, uses MODIS flood products (static water, flood) as true value training samples, and constructs a water body recognition model. The following figures are probability distribution diagrams of selected spectral features and remote sensing index features, respectively. Figure 5 It can be seen that the selected spectral features and remote sensing index features have a relatively obvious distinction between the probability distribution intervals of floods and non-floods, especially the three remote sensing index features of NDVI, NDWI and NDSI. In addition, different water body spectral features also have different sensitive areas for floods and non-floods. Therefore, selecting VIS, NIR, SWIR, NDVI, NDWI and NDSI multi-water body spectral features to construct a naive Bayesian water body recognition model is more conducive to identifying and distinguishing flood and non-flood information.

[0131] In addition, considering that the extraction of water body information using only spectral features is greatly affected by ice, snow, and cloud shadows, ice, snow, and cloud shadows are removed after identification to improve the accuracy of flood detection. Figure 6 , showing the secondary cloud shadow recognition results. The results show that the recognition accuracy of cloud shadows is high, which can provide more accurate support for subsequent flood identification.

[0132] Based on the same inventive concept, the embodiment of the present application also provides a satellite flood detection device for implementing the satellite flood detection method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more satellite flood detection device embodiments provided below can refer to the limitations of the satellite flood detection method above, and will not be repeated here.

[0133] In an exemplary embodiment, Figure 7 As shown, a satellite flood detection device 20 is provided, comprising:

[0134] The data preprocessing module 201 is used to obtain the channel data of the satellite MERSI payload;

[0135] Preprocessing the channel data, and constructing spectral features and remote sensing index features suitable for water body detection based on the preprocessed channel data;

[0136] The training module 202 is used to use the flood products of the satellite MODIS payload as true value training samples to train the spectral features and remote sensing index features to obtain a naive Bayesian water body recognition model;

[0137] The prediction module 203 is used to obtain flood detection results by removing cloud shadows in the water body using a spherical geometry model after identifying the water body using the trained naive Bayesian water body recognition model.

[0138] Optionally, the data preprocessing module 201 is further used for:

[0139] Get the channel data of the satellite MERSI payload within the specified time period;

[0140] A spatial matching threshold with MODIS data is set according to the data spatial resolution to obtain successfully matched channel data, wherein the matched data at least includes: visible light channel data, near infrared channel data, shortwave infrared channel data, longitude and latitude, solar zenith angle and MODIS MCDWD_L3_NRT data, wherein the wavelength of the visible light channel data is 0.67um, the wavelength of the infrared channel data is 0.86um, and the wavelength of the shortwave infrared channel data is 1.64um.

[0141] Optionally, the data preprocessing module 201 is further used for:

[0142] Perform calibration calculation on the successfully matched channel data to obtain the channel data within the effective range;

[0143] The daytime data with a solar zenith angle less than 85° is screened from the channel data within the effective range.

[0144] Optionally, the training module 202 is further used to:

[0145] Spectral features and remote sensing index features are constructed based on the screened channel data, the spectral features are 0.67um, 0.86um and 1.64um channel features, and the remote sensing index features include at least the normalized difference water index, the normalized difference vegetation index and the normalized difference ice and snow index.

[0146] Optionally, the training module 202 is further used to:

[0147] Select the MODIS flood data corresponding to the MERSI data time;

[0148] Using MODIS flood data as true training samples, the extracted spectral features and remote sensing index features are trained to obtain a naive Bayesian water body recognition model.

[0149] Optionally, the prediction module 203 is further used for:

[0150] Establish a spherical geometric model between cloud shadows and clouds;

[0151] Based on the assumption that a cloud pixel casts at most one cloud shadow pixel, a one-to-one relationship between cloud shadows and clouds is iteratively established to calculate the cloud shadow position;

[0152] The cloud shadows in the water body are removed based on the cloud shadow positions to obtain a flood detection result.

[0153] Optionally, the prediction module 203 is further used for:

[0154] The flood detection results of the naive Bayesian water body identification model at multiple times are selected and compared with the flood detection results of Real Earth to verify the accuracy of the naive Bayesian water body identification model.

[0155] The embodiment of the present application uses a naive Bayesian water body recognition model based on satellite MERSI payload data and MODIS flood data as true value training samples, and combines a spherical geometry model to remove cloud shadows, thereby significantly improving flood disaster monitoring technology and effectively improving the accuracy of flood detection. By accurately identifying water bodies and removing cloud shadow interference, the reliability of monitoring results is ensured. At the same time, the real-time data of the satellite MERSI payload is used to enhance the real-time nature of flood monitoring, so that disaster situations can be grasped in a timely manner. In addition, the automated processing process reduces manual intervention, improves monitoring efficiency, and provides strong support for emergency response.

[0156] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 8 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes 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 operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store satellite flood detection data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a satellite flood detection method is implemented.

[0157] Those skilled in the art will understand that Figure 8The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0158] In an exemplary embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0159] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0160] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0161] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0162] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0163] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0164] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0165] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A satellite flood detection method, characterized in that: The satellite flood detection method comprises: Obtain channel data of satellite MERSI payload; Preprocessing the channel data, and constructing spectral features and remote sensing index features suitable for water body detection based on the preprocessed channel data; Using the flood products of satellite MODIS payload as true value training samples, the spectral features and the remote sensing index features are trained to obtain a naive Bayesian water body recognition model; After the water body is identified by using the trained naive Bayesian water body identification model, the cloud shadow in the water body is removed by using a spherical geometry model to obtain a flood detection result.

2. The satellite flood detection method according to claim 1, characterized in that: The step of obtaining channel data of the satellite MERSI payload includes: Get the channel data of the satellite MERSI payload within the specified time period; A spatial matching threshold with MODIS data is set according to the data spatial resolution to obtain successfully matched channel data, wherein the matched data at least includes: visible light channel data, near infrared channel data, shortwave infrared channel data, longitude and latitude, solar zenith angle and MODIS MCDWD_L3_NRT data, wherein the wavelength of the visible light channel data is 0.67um, the wavelength of the infrared channel data is 0.86um, and the wavelength of the shortwave infrared channel data is 1.64um.

3. The satellite flood detection method according to claim 2, characterized in that: The step of preprocessing the channel data comprises: Perform calibration calculation on the successfully matched channel data to obtain the channel data within the effective range; The daytime data with a solar zenith angle less than 85° is screened from the channel data within the effective range.

4. The satellite flood detection method according to claim 3, characterized in that: The step of constructing spectral features and remote sensing index features suitable for water body detection based on the preprocessed channel data includes: Spectral features and remote sensing index features are constructed based on the screened channel data, the spectral features are 0.67um, 0.86um and 1.64um channel features, and the remote sensing index features include at least the normalized difference water index, the normalized difference vegetation index and the normalized difference ice and snow index.

5. The satellite flood detection method according to claim 1, characterized in that: The step of using the flood products of satellite MODIS payload as true value training samples to train the naive Bayesian water body recognition model includes: Select the MODIS flood data corresponding to the MERSI data time; Using MODIS flood data as true training samples, the extracted spectral features and remote sensing index features are trained to obtain a naive Bayesian water body recognition model.

6. The satellite flood detection method according to claim 1, characterized in that: The step of removing cloud shadows in the water body by using a spherical geometric model to obtain a flood detection result comprises: Establish a spherical geometric model between cloud shadows and clouds; Based on the assumption that a cloud pixel casts at most one cloud shadow pixel, a one-to-one relationship between cloud shadows and clouds is iteratively established to calculate the cloud shadow position; The cloud shadows in the water body are removed based on the cloud shadow positions to obtain a flood detection result.

7. The satellite flood detection method according to claim 6, characterized in that: After the step of removing the cloud shadows in the identified water body by using the spherical geometric model to obtain the flood detection result, the method further includes: The flood detection results of the naive Bayesian water body identification model at multiple times are selected and compared with the flood detection results of Real Earth to verify the accuracy of the naive Bayesian water body identification model.

8. A satellite flood detection device, characterized in that: The satellite flood detection device comprises: Data preprocessing module, used to obtain channel data of satellite MERSI payload; Preprocessing the channel data, and constructing spectral features and remote sensing index features suitable for water body detection based on the preprocessed channel data; A training module, for using the flood products of the satellite MODIS payload as true value training samples, training the spectral features and the remote sensing index features to obtain a naive Bayesian water body recognition model; The prediction module is used to obtain flood detection results by removing cloud shadows in the water body using a spherical geometry model after identifying the water body using the trained naive Bayesian water body recognition model.

9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the satellite flood detection method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the satellite flood detection method according to any one of claims 1 to 7 are implemented.

Images