Optical remote sensing image water body missing data reconstruction method and system

Abstract

The application discloses a kind of water body missing data reconstruction method and system for optical remote sensing image, belong to image processing technical field, wherein, the method includes: vectorization processing and buffer zone processing to GSW data set, obtain the water mask range of the region to be processed;Cloud detection algorithm and topographic shadow detection algorithm are marked to cloud and shadow in optical remote sensing image, obtain marked image;Water classification data in the water mask range of marked image is calculated using Otsu algorithm;Obtain GSW water frequency data, using 98% water frequency threshold to classify it, greater than threshold as reference permanent water, less than threshold as reference seasonal water;Respectively to the intersection of the mask to be reconstructed, reference permanent water, effective data and reference seasonal water are handled, and then reconstruct permanent water and seasonal water.The method makes full use of existing image, while ensuring the stability of algorithm accuracy, improves the reconstruction rate of time series image.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a method and system for reconstructing missing water body data from optical remote sensing images. Background Technology

[0002] Cloud obstruction, partial image coverage, and satellite sensor malfunctions can all lead to significant gaps in water extent data retrieved from optical remote sensing images. Cloud obstruction, in particular, makes it highly susceptible to missing information about surface water extent in optical remote sensing images. Current methods primarily compensate for spatial gaps by sacrificing temporal resolution. This involves mosaicking multiple cloud-removed remote sensing images to create a complete monthly or even yearly image, thereby generating a spatially complete but temporally sparse dataset of surface water extent.

[0003] To fully utilize the information from each optical remote sensing image and achieve high-frequency, spatially complete surface water observation, it is essential to reconstruct surface water data for each image, addressing issues such as cloud cover, partial image coverage, and missing strip data due to satellite sensor malfunctions. Currently, surface water data reconstruction primarily employs the following methods:

[0004] One method is the histogram thresholding method (Zhao and Gao 2018). For example... Figure 1 As shown, this method uses the effective water pixels of the original water classification image to mask the GSW water frequency data, then calculates the statistical histogram of the unmasked water area, and then multiplies the average frequency of the histogram by an adjustment factor of 0.17 to obtain the water frequency threshold corresponding to this value. The GSW water frequency data is then binarized using this threshold, and the water frequency data greater than the threshold is used as supplementary data to reconstruct the masked part.

[0005] Second is the proportional assessment method (Yao et al. 2019). For example... Figure 2 As shown, the method first obtains monthly-scale mosaic images using the median mosaic method, extracts the low and high water level vector ranges of all monthly-scale images, and then performs linear interpolation on the area within the edge of the current data and the area within the high and low water levels to obtain the current water level line position, thereby obtaining the water body range.

[0006] Third is the spatiotemporal proximity reconstruction method (Bai et al. 2022). For example... Figure 3 As shown, it is necessary to find images that are close in time and overlap in space, and there is a possibility that the original data will be modified, resulting in the drawback of overfilling.

[0007] However, the following drawbacks exist: The accuracy of simple and high-reconstruction-rate techniques needs improvement. For example, the histogram thresholding method, while guaranteeing a high reconstruction rate, uses fixed empirical coefficients, leading to unstable accuracy. High-accuracy techniques can only be achieved with sufficient data, easily resulting in a lower reconstruction rate to maintain accuracy. The proportional evaluation method requires filtering images with more than 50% invalid pixels, significantly limiting the number of reconstructable images and only obtaining good data at a monthly scale. Furthermore, the image to be reconstructed must contain valid water body edge data. Reconstructing data using spatiotemporally proximate auxiliary data is superior to using nearby high-quality images, but the time interval can vary significantly, making robustness difficult to guarantee. If the water body extent changes significantly between adjacent images, it can lead to large errors in the method.

[0008] Therefore, there is an urgent need for a method for reconstructing missing water body data that can both improve the accuracy of surface water body range data reconstruction and have the ability to process low-frequency satellite image data. Summary of the Invention

[0009] The present invention aims to at least partially solve one of the technical problems in the related art.

[0010] Therefore, one objective of this invention is to propose a method for reconstructing missing water body data from optical remote sensing images. This method makes full use of existing images and improves the reconstruction rate of time-series images while ensuring the accuracy and stability of the algorithm.

[0011] Another objective of this invention is to propose a system for reconstructing missing water body data from optical remote sensing images.

[0012] To achieve the above objectives, one embodiment of the present invention proposes a method for reconstructing missing water body data from optical remote sensing images, comprising the following steps: Step S1, vectorizing and buffering the GSW dataset (Global Surface Water dataset) to obtain the water body mask range of the area to be processed; Step S2, marking clouds and shadows in the optical remote sensing image using cloud detection and terrain shadow detection algorithms to obtain a marked image; Step S3, using the Otsu algorithm to calculate the water body classification data of the marked image within the water body mask range, wherein the water body classification data includes the mask to be reconstructed and valid data; Step S4, acquiring GSW water body frequency data, classifying it using a 98% water body frequency threshold, and using data greater than the threshold as a reference permanent data. The permanent water body is defined as a reference seasonal water body, with water bodies smaller than a threshold value. Step S5 involves taking the intersection of the mask to be reconstructed and the reference permanent water body to reconstruct the permanent water body. Simultaneously, the intersection of the valid data and the reference seasonal water body is taken to obtain the seasonal water body. The intersection of the valid data and the reference permanent water body is taken to obtain the area of ​​the effective seasonal water body. The areas of the seasonal water body and the effective seasonal water body are iteratively compared to determine the seasonal water body classification threshold. Based on the seasonal water body classification threshold, the reference seasonal water body is binarized to reconstruct the seasonal water body.

[0013] The water body reconstruction method for missing data in optical remote sensing images according to the present invention has the advantages of both histogram thresholding and proportional filling methods. The water body to be reconstructed is divided into two parts, including permanent water body and seasonal water body. The permanent water body is directly replaced and reconstructed using the part of the water body frequency data that is greater than 98%. The reconstruction of seasonal water body adopts a dynamic threshold iteration method based on water body frequency data, thereby achieving stable and more accurate reconstruction of water body range.

[0014] In addition, the water body missing data reconstruction method based on optical remote sensing imagery according to the above embodiments of the present invention may also have the following additional technical features:

[0015] Further, in one embodiment of the present invention, step S1 specifically includes: step S101, obtaining the GSW dataset; step S102, vectorizing the GSW dataset to obtain the maximum water body range; step S103, buffering the maximum water body range to obtain the water body mask range of the area to be processed.

[0016] Furthermore, in one embodiment of the present invention, the marked image includes a valid pixel portion and an invalid pixel portion.

[0017] Furthermore, in one embodiment of the present invention, the effective data includes water masks and land masks.

[0018] Further, in one embodiment of the present invention, step S5 specifically includes: step S501, taking the intersection of the mask to be reconstructed and the reference permanent water body to reconstruct the permanent water body; step S502, taking the intersection of the effective data and the reference seasonal water body to obtain the seasonal water body; step S503, taking the intersection of the water body mask of the effective data and the reference permanent water body to obtain the area of ​​the effective seasonal water body; step S504, assuming the value range of the seasonal water body frequency data is [1-98], using 1 as the starting iteration threshold, calculating the area of ​​seasonal water bodies greater than 1; step S505, comparing the size of the seasonal water body area and the effective seasonal water body area, if the latter is larger, increasing the threshold by 1 step size until the seasonal water body area is less than or equal to the effective seasonal water body area, stopping the iteration, and taking the current threshold as the seasonal water body classification threshold; step S506, performing binarization processing on the reference seasonal water body using the seasonal water body classification threshold to reconstruct the seasonal water body.

[0019] To achieve the above objectives, another embodiment of the present invention proposes a system for reconstructing missing water body data from optical remote sensing images, comprising: a vectorization and buffer processing module for vectorizing and buffering the GSW dataset to obtain the water body mask range of the area to be processed; a labeling module for labeling clouds and shadows in the optical remote sensing image using cloud detection and terrain shadow detection algorithms to obtain a labeled image; a calculation module for calculating water body classification data of the labeled image within the water body mask range using the Otsu algorithm, wherein the water body classification data includes the mask to be reconstructed and valid data; and a threshold classification module for acquiring GSW water body frequency data and using... A 98% water body frequency threshold is used to classify water bodies; those above the threshold are designated as reference permanent water bodies, and those below the threshold are designated as reference seasonal water bodies. A reconstruction module is used to find the intersection of the mask to be reconstructed and the reference permanent water bodies to reconstruct permanent water bodies. Simultaneously, it finds the intersection of the valid data and the reference seasonal water bodies to obtain seasonal water bodies, and finds the intersection of the valid data and the reference permanent water bodies to obtain the area of ​​valid seasonal water bodies. The seasonal water bodies and the area of ​​valid seasonal water bodies are iteratively compared to determine the seasonal water body classification threshold. Based on the seasonal water body classification threshold, the reference seasonal water bodies are binarized to reconstruct seasonal water bodies.

[0020] The water body missing data reconstruction system for optical remote sensing images in this invention has the advantages of both histogram thresholding and proportional filling methods. It divides the water body to be reconstructed into two parts, including permanent water body and seasonal water body. The permanent water body is directly replaced and reconstructed using the portion of the water body frequency data that is greater than 98%. The reconstruction of seasonal water body adopts a dynamic threshold iteration method based on water body frequency data, which achieves stable and more accurate reconstruction of water body range.

[0021] In addition, the water body missing data reconstruction system based on optical remote sensing imagery according to the above embodiments of the present invention may also have the following additional technical features:

[0022] Furthermore, in one embodiment of the present invention, the vectorization and buffer processing module specifically includes: an acquisition unit for acquiring a GSW dataset; a vectorization processing unit for performing vectorization processing on the GSW dataset to obtain the maximum water body range; and a buffer processing unit for performing buffer processing on the maximum water body range to obtain the water body mask range of the area to be processed.

[0023] Furthermore, in one embodiment of the present invention, the marked image includes a valid pixel portion and an invalid pixel portion.

[0024] Furthermore, in one embodiment of the present invention, the effective data includes water masks and land masks.

[0025] Further, in one embodiment of the present invention, the reconstruction module specifically includes: a first intersection and reconstruction unit, used to take the intersection of the mask to be reconstructed and the reference permanent water body to reconstruct the permanent water body; a second intersection unit, used to take the intersection of the effective data and the reference seasonal water body to obtain the seasonal water body; a third intersection unit, used to take the intersection of the water body mask of the effective data and the reference permanent water body to obtain the area of ​​the effective seasonal water body; a preset unit, used to set the value range of the seasonal water body frequency data to [1-98], and use 1 as the starting iteration threshold to calculate the area of ​​seasonal water bodies greater than 1; an iteration unit, used to compare the size of the seasonal water body area and the effective seasonal water body area, and if the latter is larger, increase the threshold by 1 step size until the seasonal water body area is less than or equal to the effective seasonal water body area, stop the iteration, and take the current threshold as the seasonal water body classification threshold; and a binarization and reconstruction unit, used to perform binarization processing on the reference seasonal water body through the seasonal water body classification threshold to reconstruct the seasonal water body.

[0026] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0027] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0028] Figure 1 This is a schematic diagram of reconstruction using the histogram thresholding method;

[0029] Figure 2 This is a schematic diagram of the reconstructed water body area obtained using the proportional assessment method;

[0030] Figure 3 This is a flowchart of spatiotemporal proximity data reconstruction.

[0031] Figure 4 This is a flowchart of a method for reconstructing missing water body data from optical remote sensing images according to an embodiment of the present invention;

[0032] Figure 5 This is a detailed execution block diagram of a method for reconstructing missing water body data from optical remote sensing images according to an embodiment of the present invention;

[0033] Figure 6 This is a schematic diagram of the data processing and verification process of this invention, taking Zhangcuo Lake (latitude and longitude coordinates: 87.44 E, 30.98 N) on the Qinghai-Tibet Plateau as an example;

[0034] Figure 7 This is a schematic diagram of the structure of a water body missing data reconstruction system based on an embodiment of the present invention. Detailed Implementation

[0035] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0036] The following describes, with reference to the accompanying drawings, a method and system for reconstructing missing water body data from optical remote sensing images according to an embodiment of the present invention. First, the method for reconstructing missing water body data from optical remote sensing images according to an embodiment of the present invention will be described with reference to the accompanying drawings.

[0037] Figure 4 This is a flowchart of a method for reconstructing missing water body data from optical remote sensing images according to an embodiment of the present invention.

[0038] like Figure 4 As shown, the method for reconstructing missing water body data from optical remote sensing images includes the following steps:

[0039] In step S1, the GSW dataset is vectorized and buffered to obtain the water mask range of the area to be processed.

[0040] Furthermore, in one embodiment of the present invention, step S1 specifically includes:

[0041] Step S101: Obtain the GSW dataset, which is a pixelated monthly / yearly scale global surface water classification binary classification dataset;

[0042] Step S102: Vectorize the GSW dataset to obtain the maximum water body range. The purpose is that the surface water extraction result of conventional remote sensing image is a binary classification image in pixel format, which is raster data, while the objectified water body is a single surface water body stored with range boundaries, which is vector data.

[0043] Step S103: Perform buffer processing on the maximum water body range to obtain the water body mask range of the area to be processed.

[0044] In step S2, clouds and shadows in the optical remote sensing image are marked using cloud detection and terrain shadow detection algorithms to obtain a marked image. The marked image includes valid pixel portions and invalid pixel portions. Valid pixel portions refer to water body pixels with valid observation values ​​that are not affected by cloud obstruction, mountain shadows, or sensor malfunctions. Invalid pixel portions refer to invalid observation values ​​affected by cloud obstruction, mountain shadows, or sensor malfunctions. The optical remote sensing image can be taken as an example: Landsat TOA image (Landsat Top of Atmosphere (TOA) apparent reflectance data).

[0045] It should be noted that the cloud detection algorithm distinguishes clouds by their high reflectivity in the visible light band and by their brightness; in the thermal infrared band, clouds have lower temperatures and can be distinguished by their temperature. The Normalized Difference Snow Index (NDSI) = (Green – SWIR1) / (Green + SWIR1) can be used to distinguish between clouds and snow, where Green represents the green light band with wavelengths of 500-600nm, SWIR1 represents the short-wave infrared band with wavelengths of 1550-1750nm, and an NDSI greater than 0.4 indicates snow. The terrain shadow detection algorithm generates slope and mountain shadows using 30m resolution radar topographic mapping data SRTM (The Shuttle Radar Topography Mission), image acquisition time, and solar altitude angle, further obtaining the range of shadows generated by the terrain.

[0046] In step S3, the Otsu algorithm is used to calculate the water classification data of the marked image within the water mask area. The water classification data includes the mask to be reconstructed and the effective data, which includes the water mask and the land mask.

[0047] It should be noted that the Otsu algorithm is used to perform binary segmentation of the image histogram, that is, to perform binary segmentation of the labeled image using the Otsu algorithm, thereby obtaining water body classification data within the water body mask area.

[0048] In step S4, GSW water body frequency data is obtained and classified using a 98% water body frequency threshold. Water bodies with frequencies greater than the threshold are used as reference permanent water bodies, and those with frequencies less than the threshold are used as reference seasonal water bodies.

[0049] The water body frequency threshold is a critical value for the binary division of GSW water body frequency data. A value greater than this value is equal to 1, and a value less than this value is equal to 0. In other words, when the GSW water body frequency data is greater than the threshold and equal to 1, it is used as a reference permanent water body, and when it is less than the threshold and equal to 0, it is used as a reference seasonal water body.

[0050] In step S5, the intersection of the mask to be reconstructed and the reference permanent water body is taken to reconstruct the permanent water body. At the same time, the intersection of the effective data and the reference seasonal water body is taken to obtain the seasonal water body. The intersection of the effective data and the reference permanent water body is taken to obtain the area of ​​the effective seasonal water body. The areas of the seasonal water body and the effective seasonal water body are iteratively compared to determine the seasonal water body classification threshold. The reference seasonal water body is binarized according to the seasonal water body classification threshold to reconstruct the seasonal water body.

[0051] It should be noted that regardless of how the remote sensing image is obscured, permanent water bodies always exist on the Earth's surface. Therefore, for the missing water body data, the intersection of the mask to be reconstructed and the reference permanent water body is directly used to reconstruct the permanent water body. Because seasonal water bodies change dynamically over time, if the surface image information for that part is obscured by clouds or is invalid, reconstruction needs to be performed based on the characteristics of the image at that time. Therefore, seasonal water bodies are also a major source of uncertainty in the reconstruction process. Figure 5 As shown, the specific steps are as follows:

[0052] Step S501: Take the intersection of the mask to be reconstructed and the reference permanent water body to reconstruct the permanent water body;

[0053] Step S502: Take the intersection of the valid data and the reference seasonal water body to obtain the seasonal water body;

[0054] Step S503: Take the intersection of the water body mask of the effective data and the reference permanent water body to obtain the effective seasonal water body area;

[0055] Step S504: Set the range of seasonal water body frequency data to [1-98], and use 1 as the starting iteration threshold to calculate the area of ​​seasonal water bodies greater than 1.

[0056] It should be noted that the seasonal water frequency data is based on global surface water frequency data from 1984 to 2019 using Landsat satellite remote sensing imagery and expert classification systems. The value ranges from 0 to 100, where 0 indicates that the probability of detecting water at this location in the multi-year imagery is very low and it is considered non-water. 100 indicates permanent surface water, meaning that all images show that this location is water. The intermediate value indicates seasonal surface water, which is affected by seasonal changes and may be water at certain times.

[0057] Step S505: Compare the size of the seasonal water body area and the effective seasonal water body area. If the seasonal water body area is larger, increase the threshold by 1 step size until the seasonal water body area is less than or equal to the effective seasonal water body area. Stop the iteration and take the current threshold as the seasonal water body classification threshold.

[0058] Step S506: The reference seasonal water body is binarized using the seasonal water body classification threshold to reconstruct the seasonal water body.

[0059] The following specific embodiment further illustrates the proposed method for reconstructing missing water body data from optical remote sensing images.

[0060] like Figure 6 The figure shows the process of reconstructing water body classification data from a Landsat 7 image taken on June 11, 2001, using Zhangcuo Lake in the Qinghai-Tibet Plateau as an example. As can be seen from the figure, the original data contained a large amount of cloud cover, resulting in many missing data points. In the original water body classification data, the area of ​​the water body was 40.95 km². 2 Simultaneously, using GSW water frequency data and a threshold of 98, reference water body data including permanent water bodies (blue) and seasonal water bodies (cyan) was generated. The data is then reconstructed in two parts:

[0061] Step 1: Obtain the permanent water body: Replace the pixels to be reconstructed in the original data with the reference permanent water body to obtain the result of the permanent water body reconstruction, with an area of ​​57.81 km². 2 .

[0062] Step 2, Seasonal Water Bodies: Based on the effective seasonal water bodies in the image data to be reconstructed, reference data is used for the iterative threshold. The water body area is 29.94 km². 2 For seasonal water bodies based on GSW water frequency data, area iteration is used, with an iteration threshold range of [0-98].

[0063] 1. When the threshold is 0, the initial area is 30.26 km². 2 Greater than 29.94 km 2 If the step size is increased by 1, the iteration continues until the threshold of 85 is reached, at which point the area is still greater than 29.94 km². 2 When the threshold is 92, the area is 27.30 km². 2 It is exactly less than 29.94 km 2 Stop iterating.

[0064] 2. Using 92 as the threshold, the reference seasonal water body data is processed to obtain the reconstructed seasonal water body.

[0065] Step three involves merging the reconstructed permanent and seasonal water bodies to obtain a final area of ​​85.11 km². 2 The area obtained by the histogram thresholding method is 88.36 km². 2 The area (81.30 km²) was obtained using the nearest high-resolution cloudless image, namely Landsat 7 on May 10, 2001. 2 The results were verified, showing that the present invention can reconstruct missing water body data well and is superior to the mainstream histogram thresholding method.

[0066] Therefore, the water body missing data reconstruction method proposed in the embodiments of the present invention combines the advantages of the proportional evaluation method for directly preserving permanent water bodies and the histogram threshold method for reconstruction using auxiliary water body frequency data, while avoiding the drawback of the spatiotemporal proximity reconstruction method requiring high-quality proximity data. It has the characteristics of high reconstruction rate, simple method, stable algorithm and better accuracy.

[0067] Next, with reference to the accompanying drawings, a water body missing data reconstruction system based on an embodiment of the present invention is described.

[0068] Figure 7 This is a schematic diagram of the structure of a water body missing data reconstruction system based on an embodiment of the present invention.

[0069] like Figure 7 As shown, the system 10 includes: a vectorization and buffer processing module 100, a labeling module 200, a calculation module 300, a threshold classification module 400, and a reconstruction module 500.

[0070] The system comprises several modules: a vectorization and buffer processing module 100, which performs vectorization and buffer processing on the GSW dataset to obtain the water body mask extent of the area to be processed; a labeling module 200, which labels clouds and shadows in the optical remote sensing image using cloud detection and terrain shadow detection algorithms to obtain labeled images; a calculation module 300, which uses the Otsu algorithm to calculate water body classification data within the water body mask extent of the labeled image, whereby the water body classification data includes the mask to be reconstructed and valid data; and a threshold classification module 400, which acquires GSW water body frequency data and classifies it using a 98% water body frequency threshold, with those above the threshold being used as reference permanent water bodies and those below the threshold being used as reference seasonal water bodies. The reconstruction module 500 is used to take the intersection of the mask to be reconstructed and the reference permanent water body to reconstruct the permanent water body. At the same time, it takes the intersection of the effective data and the reference seasonal water body to obtain the seasonal water body. It takes the intersection of the effective data and the reference permanent water body to obtain the area of ​​the effective seasonal water body. It iteratively compares the areas of the seasonal water body and the effective seasonal water body to determine the seasonal water body classification threshold. Based on the seasonal water body classification threshold, it performs binarization processing on the reference seasonal water body to reconstruct the seasonal water body.

[0071] Furthermore, in one embodiment of the present invention, the vectorization and buffer processing module specifically includes: an acquisition unit for acquiring a GSW dataset; a vectorization processing unit for vectorizing the GSW dataset to obtain the maximum water body range; and a buffer processing unit for buffering the maximum water body range to obtain the water body mask range of the area to be processed.

[0072] Furthermore, in one embodiment of the present invention, the marked image includes a valid pixel portion and an invalid pixel portion.

[0073] Furthermore, in one embodiment of the present invention, the effective data includes water masks and land masks.

[0074] Furthermore, in one embodiment of the present invention, the reconstruction module specifically includes:

[0075] The first intersection and reconstruction unit is used to take the intersection of the mask to be reconstructed and the reference permanent water body to reconstruct the permanent water body;

[0076] The second intersection unit is used to take the intersection of the valid data and the reference seasonal water body to obtain the seasonal water body.

[0077] The third intersection unit is used to take the intersection of the effective data water body mask and the reference permanent water body to obtain the effective seasonal water body area;

[0078] The preset unit is used to set the range of seasonal water body frequency data to [1-98], and calculate the area of ​​seasonal water bodies with a value greater than 1 with 1 as the starting iteration threshold.

[0079] The iterative unit is used to compare the size of the seasonal water body area and the effective seasonal water body area. If the seasonal water body area is larger, the threshold is increased by 1 step size until the seasonal water body area is less than or equal to the effective seasonal water body area. Then the iteration stops and the current threshold is taken as the seasonal water body classification threshold.

[0080] The binarization and reconstruction unit is used to binarize a reference seasonal water body using a seasonal water body classification threshold in order to reconstruct the seasonal water body.

[0081] It should be noted that the foregoing explanation of the method embodiment for reconstructing missing water body data from optical remote sensing images also applies to the system in this embodiment, and will not be repeated here.

[0082] The water body missing data reconstruction system proposed in this embodiment of the invention combines the advantages of the proportional evaluation method for directly preserving permanent water bodies and the histogram threshold method for reconstruction using auxiliary water body frequency data. At the same time, it avoids the drawback of the spatiotemporal proximity reconstruction method requiring high-quality proximity data. It features high reconstruction rate, simple method, stable algorithm, and better accuracy.

[0083] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0084] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0085] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for reconstructing missing water body data from optical remote sensing images, characterized in that, Includes the following steps: Step S1: Vectorize and buffer the GSW dataset to obtain the water mask range of the area to be processed. Step S2: Mark clouds and shadows in the optical remote sensing image using cloud detection algorithm and terrain shadow detection algorithm to obtain marked image; Step S3: The Otsu algorithm is used to calculate the water classification data of the marked image within the water mask area. The water classification data includes the mask to be reconstructed and the effective data, which includes the water mask and the land mask. Step S4: Obtain GSW water body frequency data and classify them using a 98% water body frequency threshold. Water bodies with frequencies greater than the threshold are used as reference permanent water bodies, and those with frequencies less than the threshold are used as reference seasonal water bodies. Step S5: Take the intersection of the mask to be reconstructed and the reference permanent water body to reconstruct the permanent water body. At the same time, take the intersection of the effective data and the reference seasonal water body to obtain the seasonal water body. Take the intersection of the effective data and the reference permanent water body to obtain the area of ​​the effective seasonal water body. Iteratively compare the areas of the seasonal water body and the effective seasonal water body to determine the seasonal water body classification threshold. Binarize the reference seasonal water body according to the seasonal water body classification threshold to reconstruct the seasonal water body.

2. The method for reconstructing missing water body data from optical remote sensing images according to claim 1, characterized in that, Step S1 specifically includes: Step S101: Obtain the GSW dataset; Step S102: Vectorize the GSW dataset to obtain the maximum water body extent; Step S103: Perform buffer processing on the maximum water body range to obtain the water body mask range of the area to be processed.

3. The method for reconstructing missing water body data from optical remote sensing images according to claim 1, characterized in that, The marked image includes valid pixel portions and invalid pixel portions.

4. The method for reconstructing missing water body data from optical remote sensing images according to claim 1, characterized in that, Step S5 specifically includes: Step S501: Take the intersection of the mask to be reconstructed and the reference permanent water body to reconstruct the permanent water body; Step S502: Take the intersection of the valid data and the reference seasonal water body to obtain the seasonal water body; Step S503: Take the intersection of the water mask of the effective data and the reference permanent water body to obtain the effective seasonal water body area; Step S504: Set the range of seasonal water body frequency data to [1-98], and use 1 as the starting iteration threshold to calculate the area of ​​seasonal water bodies greater than 1. Step S505: Compare the size of the seasonal water body area and the effective seasonal water body area. If the seasonal water body area is greater than the effective seasonal water body area, increase the threshold by 1 step size until the seasonal water body area is less than or equal to the effective seasonal water body area, stop the iteration, and take the current threshold as the seasonal water body classification threshold. Step S506: The reference seasonal water body is binarized using the seasonal water body classification threshold to reconstruct the seasonal water body.

5. A system for reconstructing missing water body data from optical remote sensing images, characterized in that, include: The vectorization and buffer processing module is used to perform vectorization and buffer processing on the GSW dataset to obtain the water mask range of the area to be processed. The labeling module is used to label clouds and shadows in optical remote sensing images using cloud detection algorithms and terrain shadow detection algorithms to obtain labeled images; The calculation module is used to calculate the water classification data of the marked image within the water mask area using the Otsu algorithm, wherein the water classification data includes the mask to be reconstructed and the effective data, and the effective data includes the water mask and the land mask; The threshold classification module is used to acquire GSW water body frequency data and classify it using a 98% water body frequency threshold. Water bodies with frequencies greater than the threshold are used as reference permanent water bodies, and those with frequencies less than the threshold are used as reference seasonal water bodies. The reconstruction module is used to take the intersection of the mask to be reconstructed and the reference permanent water body to reconstruct the permanent water body, and simultaneously take the intersection of the effective data and the reference seasonal water body to obtain the seasonal water body. It also takes the intersection of the effective data and the reference permanent water body to obtain the area of ​​the effective seasonal water body. The module iteratively compares the area of ​​the seasonal water body and the area of ​​the effective seasonal water body to determine the seasonal water body classification threshold. Based on the seasonal water body classification threshold, the module binarizes the reference seasonal water body to reconstruct the seasonal water body.

6. The system for reconstructing missing water body data from optical remote sensing images according to claim 5, characterized in that, The vectorization and buffer processing module specifically includes: The acquisition unit is used to acquire the GSW dataset. The vectorization processing unit is used to perform vectorization processing on the GSW dataset to obtain the maximum water body range; A buffer processing unit is used to perform buffer processing on the maximum water body range to obtain the water body mask range of the area to be processed.

7. The system for reconstructing missing water body data from optical remote sensing images according to claim 5, characterized in that, The marked image includes valid pixel portions and invalid pixel portions.

8. The system for reconstructing missing water body data from optical remote sensing images according to claim 5, characterized in that, The reconstruction module specifically includes: The first intersection and reconstruction unit is used to take the intersection of the mask to be reconstructed and the reference permanent water body to reconstruct the permanent water body; The second intersection unit is used to take the intersection of the valid data and the reference seasonal water body to obtain the seasonal water body. The third intersection unit is used to take the intersection of the water body mask of the effective data and the reference permanent water body to obtain the effective seasonal water body area; The preset unit is used to set the range of seasonal water body frequency data to [1-98], and calculate the area of ​​seasonal water bodies with a value greater than 1 with 1 as the starting iteration threshold. An iterative unit is used to compare the size of the seasonal water body area and the effective seasonal water body area. If the seasonal water body area is greater than the effective seasonal water body area, the threshold is increased by 1 step size until the seasonal water body area is less than or equal to the effective seasonal water body area, then the iteration stops and the current threshold is taken as the seasonal water body classification threshold. The binarization and reconstruction unit is used to binarize the reference seasonal water body using the seasonal water body classification threshold in order to reconstruct the seasonal water body.

Images