Aquaculture area disaster loss assessment method, system, electronic equipment and medium

Abstract

The present invention discloses a method, system, electronic device, and medium for assessing damage in an aquaculture area, relating to the field of aquaculture area damage assessment. The method comprises: acquiring an image of an area covered by an aquaculture flood disaster; extracting, from the image, a first aquaculture area image and a second aquaculture area image of the area covered by the aquaculture flood disaster before the flood disaster occurs, and a third aquaculture area image of the area covered by the aquaculture flood disaster after the flood disaster occurs; calculating, based on the first aquaculture area image, the connectivity of each aquaculture area in the second aquaculture area image and each aquaculture area in the third aquaculture area image using a connectivity algorithm; and comparing the connectivity of each aquaculture area in the second aquaculture area image with the connectivity of each aquaculture area in the third aquaculture area image to determine a damage assessment result for each aquaculture area. The present invention improves the accuracy of aquaculture area damage assessment by improving the extraction accuracy of aquaculture ponds.

Description

Technical Field

[0001] The present invention relates to the field of disaster loss assessment in aquaculture areas, and in particular to a method, system, electronic equipment and medium for disaster loss assessment in aquaculture areas. Background Art

[0002] Aquaculture provides food and a vital source of income for hundreds of millions of people worldwide. Driven by population growth and the need for socioeconomic benefits, aquaculture production has grown rapidly, with an average annual growth rate of 6.9% over the past 30 years, making it one of the fastest-growing sectors in the food industry. Aquaculture ponds have a significant impact on the surrounding ecological environment, leading to environmental problems such as siltation in estuaries and coastal waters and eutrophication. Mapping aquaculture areas is crucial for understanding the spatial distribution of aquaculture ponds, conducting fishery resource surveys, and protecting the ecological environment of aquatic areas. Aquaculture areas are often located near natural water bodies such as lakes and oceans, making them highly vulnerable to natural disasters such as typhoons, floods, and tsunamis. Damage to aquaculture areas often results in the loss of aquaculture products, leading to significant production losses. Assessing damage to aquaculture ponds is crucial for pre-disaster risk prevention, emergency relief efforts, and post-disaster insurance compensation.

[0003] Satellite remote sensing technology, with its large-scale, low-cost, and real-time characteristics, is an effective means for monitoring and studying aquaculture areas. In recent years, remote sensing identification of aquaculture areas has become a key research topic in coastal ecological environments. With the advancement of remote sensing technology, data from an increasing number of different types of remote sensing sensors are being used to observe water bodies. Different types of remote sensing data have their own advantages and characteristics for extracting aquaculture information, corresponding to different application areas and information extraction accuracy. Currently, commonly used methods for identifying aquaculture areas include visual interpretation, information extraction based on ratio index analysis, information extraction based on spatial structure analysis, object-oriented information extraction, and deep learning methods. Based on the commonly used aquaculture area identification methods, technicians in this field have explored practical applications, for example: (1) using high-resolution remote sensing images as the data source, according to the prescribed discrimination rules, various aquaculture water resources are classified and extracted to obtain the spatial distribution data of aquaculture water resources in the country; (2) using the temporal characteristics of SAR and optical satellite data obtained by Sentinel-1 and Sentinel-2, the water body index is calculated respectively to realize the mapping of aquaculture areas in coastal areas of Asian countries; (3) a method for extracting aquaculture land based on texture and spatial characteristics, using texture entropy and water body index to extract aquaculture land, and merging the same type of land objects based on the relationship between adjacent land objects; (4) a method based on Google Earth An aquaculture pond identification method based on the Geographic Information Systems Engine (GEE) platform and the annual Sentinel-1 and Sentinel-2 time series remote sensing data was used to extract aquaculture ponds in the Beibu Gulf coastal zone of Guangxi in 2019, combining multi-threshold segmentation and object-oriented classification; (5) An artificial pond extraction network model that integrates target edge features and semantic information. First, an improved U-Net semantic segmentation network module is used to extract target semantic information from remote sensing images. Then, the above semantic segmentation network is expanded to construct an edge extraction subnetwork to obtain multi-scale edge features of remote sensing images. Finally, the edge features and semantic information are integrated with the encoding-decoding subnetwork to achieve accurate extraction of remote sensing image targets.

[0004] By summarizing previous studies, it was found that the existing technology has the following defects: 1) The current research method is mainly used for extraction in coastal aquaculture areas and is not suitable for extraction in inland aquaculture areas; 2) The aquaculture pond extraction results of the current research method may have the situation where the ponds are interconnected, and the extraction effect is poor; 3) The current research method has difficulty in distinguishing ponds from regular-shaped natural water bodies. Summary of the Invention

[0005] The purpose of the present invention is to provide a method, system, electronic equipment and medium for evaluating damage in an aquaculture area, which can improve the accuracy of damage assessment in aquaculture areas by improving the extraction accuracy of aquaculture ponds.

[0006] To achieve the above object, the present invention provides the following solutions:

[0007] A method for assessing damage in an aquaculture area, comprising:

[0008] Acquire images of the area affected by the aquaculture flooding disaster; the images include first Sentinel-1GRD images of the same length and different time periods, and Sentinel-2MSI remote sensing data and second Sentinel-1GRD images of the same length and different time periods; the first Sentinel-1GRD images include Sentinel-1GRD images before and after the flooding disaster; the Sentinel-2MSI remote sensing data and Sentinel-1GRD images of the same length and different time periods are both images taken before the flooding disaster occurs;

[0009] Extracting a first aquaculture area image of the aquaculture flood disaster coverage area before the flood disaster occurs based on the Sentinel-2MSI remote sensing data and the second Sentinel-1GRD image;

[0010] Extracting, based on the first Sentinel-1GRD image, a second aquaculture area image before the aquaculture flooding disaster occurs and a third aquaculture area image after the aquaculture flooding disaster occurs.

[0011] Calculating, based on the first aquaculture area image, the connectivity of each aquaculture area in the second aquaculture area image and the connectivity of each aquaculture area in the third aquaculture area image using a connectivity algorithm;

[0012] Compare the connectivity of each aquaculture area in the second aquaculture area image with the connectivity of each aquaculture area in the third aquaculture area image to determine the damage assessment result of each aquaculture area within the aquaculture flooding disaster coverage area; the damage assessment result is damaged or not damaged.

[0013] Optionally, extracting a third aquaculture area image of the aquaculture flood disaster coverage area before the flood disaster occurs based on the Sentinel-2MSI remote sensing data and the second Sentinel-1GRD image, specifically comprising:

[0014] Applying a median synthesis method to the second Sentinel-1GRD image to obtain a first median image;

[0015] Performing threshold segmentation on the first median image to determine a water body extraction threshold;

[0016] Calculating the SDWI index of each pixel in the second Sentinel-1GRD image, and taking the pixel whose SDWI index is greater than the water body extraction threshold as the first water body pixel;

[0017] Calculating the water accumulation frequency of the first water body pixel, and taking the pixel whose water accumulation frequency is greater than a first set threshold as the first valid water body;

[0018] Applying a median synthesis method to the Sentinel-2MSI remote sensing data to obtain a second median image;

[0019] Calculating the NDWI index of each pixel in the second median image, and taking the pixel whose NDWI index is greater than a second set threshold as the second effective water body;

[0020] Extracting the intersection of the first effective water body and the second effective water body to obtain an initial water body image;

[0021] Eliminating paddy field images, building shadow images, natural water images, and water images that do not meet an aquaculture pond area threshold range from the initial water image to obtain a water image to be classified;

[0022] Applying a decision tree algorithm to classify the water body image to be classified to obtain an initial water body classification result; the initial classification result includes an initial aquaculture water body image and an initial non-aquaculture water body image;

[0023] Applying a corrosion algorithm to the initial aquaculture water body image and applying a dilation algorithm to the initial non-aquaculture water body image to obtain updated aquaculture water body image and non-aquaculture water body image; the updated aquaculture water body image is a third aquaculture area image of the aquaculture flood disaster coverage area before the flood disaster occurs.

[0024] Optionally, removing paddy field images, building shadow images, natural water images, and water images that do not meet an aquaculture pond area threshold range from the initial water image to obtain a water image to be classified specifically includes:

[0025] Obtain land cover data and natural water body vector data of areas affected by aquaculture flooding disasters;

[0026] The land cover data, the natural water body vector data, and the water body image that does not meet the aquaculture pond area threshold range are eliminated from the initial water body image to obtain a water body image to be classified.

[0027] Optionally, extracting a second aquaculture area image before the flooding disaster occurs and a third aquaculture area image after the flooding disaster occurs from the first Sentinel-1GRD image, specifically including:

[0028] Calculate the first SDWI index of the Sentinel-1GRD image before the flood disaster and the second SDWI index of the Sentinel-1GRD image after the flood disaster;

[0029] The pixels whose first SDWI index is greater than the water body extraction threshold are used as the second aquaculture area image;

[0030] The pixels whose second SDWI index is greater than the water body extraction threshold are used as the third aquaculture area image.

[0031] Optionally, comparing the connectivity of each aquaculture area in the second aquaculture area image with the connectivity of each aquaculture area in the third aquaculture area image to determine a damage assessment result for each aquaculture area within the aquaculture flooding disaster coverage area specifically includes:

[0032] When the connectivity of the aquaculture area in the third aquaculture area image is greater than the connectivity of the corresponding aquaculture area in the second aquaculture area image, it is determined that the aquaculture area is damaged.

[0033] A system for assessing damage in an aquaculture area, applied to the above-mentioned method for assessing damage in an aquaculture area, comprises:

[0034] An acquisition module is configured to acquire images of the area covered by the aquaculture flooding disaster; the images include first Sentinel-1GRD images of the same length and different periods, and Sentinel-2MSI remote sensing data and second Sentinel-1GRD images of the same length and different periods; the first Sentinel-1GRD images include Sentinel-1GRD images before and after the flooding disaster; the Sentinel-2MSI remote sensing data and Sentinel-1GRD images of the same length and different periods are both images taken before the flooding disaster occurs;

[0035] A first extraction module is configured to extract a first aquaculture area image of the aquaculture flood disaster coverage area before the flood disaster occurs based on the Sentinel-2MSI remote sensing data and the second Sentinel-1GRD image;

[0036] A second extraction module is configured to extract, based on the first Sentinel-1GRD image, a second aquaculture area image before the aquaculture flooding disaster occurs and a third aquaculture area image after the flooding disaster occurs.

[0037] a connectivity calculation module, configured to calculate, based on the first aquaculture area image and using a connectivity algorithm, the connectivity of each aquaculture area in the second aquaculture area image and the connectivity of each aquaculture area in the third aquaculture area image;

[0038] An assessment module is used to compare the connectivity of each aquaculture area in the second aquaculture area image with the connectivity of each aquaculture area in the third aquaculture area image, and determine a damage assessment result of each aquaculture area within the aquaculture flooding disaster coverage area; the damage assessment result is damaged or not damaged.

[0039] An electronic device includes a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the above-mentioned aquaculture area disaster loss assessment method.

[0040] A computer-readable storage medium stores a computer program, which implements the above-mentioned aquaculture area disaster loss assessment method when executed by a processor.

[0041] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0042] The present invention provides a method, system, electronic device and medium for assessing damage in an aquaculture area, which obtains a first Sentinel-1GRD image of the same length in different periods and a Sentinel-2MSI remote sensing data and a second Sentinel-1GRD image of the same length in the same period; the first Sentinel-1GRD image includes a Sentinel-1GRD image before the flooding disaster occurs and a Sentinel-1GRD image after the flooding disaster occurs; and the Sentinel-2MSI remote sensing data and the Sentinel-1GRD image of the same length in the same period are obtained. The image before the flood disaster occurred was used, and based on the Sentinel-2MSI remote sensing data and the second Sentinel-1GRD image, the first aquaculture area image of the aquaculture flood disaster coverage area before the flood disaster was extracted, which improved the extraction accuracy of the aquaculture ponds before the flooding. Furthermore, based on the first aquaculture area image, the connectivity algorithm was used to calculate the connectivity of each aquaculture area in the second aquaculture area image and the connectivity of each aquaculture area in the third aquaculture area image. By comparing the connectivity of the aquaculture area before and after the disaster, it was determined whether the aquaculture area was damaged, thereby improving the accuracy of the damage assessment of the aquaculture area. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0044] Figure 1 A flow chart of the aquaculture area disaster loss assessment method provided by the present invention;

[0045] Figure 2 A specific implementation diagram of disaster damage assessment in aquaculture areas provided by an embodiment of the invention;

[0046] Figure 3 is the SDWI median image histogram;

[0047] Figure 4 This is the schematic diagram of water extraction principle of Sentinel-1 time series data;

[0048] Figure 5 This is the principle diagram of corrosion-expansion post-processing;

[0049] Figure 6 The principle of historical-pre-disaster-post-disaster water extraction;

[0050] Figure 7 Schematic diagram for calculating connectivity of aquaculture ponds;

[0051] Figure 8 Schematic diagram of changes in connectivity of aquaculture ponds. DETAILED DESCRIPTION

[0052] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0053] The purpose of the present invention is to provide a method, system, electronic equipment and medium for evaluating damage in an aquaculture area, which can improve the accuracy of damage assessment in aquaculture areas by improving the extraction accuracy of aquaculture ponds.

[0054] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0055] Example 1

[0056] like Figure 1 As shown, the present invention provides a method for assessing damage in an aquaculture area, the method comprising:

[0057] Step 1: Acquire images of the area affected by aquaculture flooding disasters; the images include first Sentinel-1GRD images of the same length in different periods and Sentinel-2MSI remote sensing data and second Sentinel-1GRD images of the same length in the same period; the first Sentinel-1GRD images include Sentinel-1GRD images before and after the flooding disaster; the Sentinel-2MSI remote sensing data and Sentinel-1GRD images of the same length in the same period are both images taken before the flooding disaster occurs.

[0058] This example uses Sri Lanka as the research area. Figure 2As shown in Table 1, we extracted aquaculture areas and assessed the flooded areas before and after the Burevi cyclone disaster. Tropical Cyclone Burevi occurred from November 29 to December 5, 2020. The data used to obtain images of the flooded aquaculture areas included remote sensing data, land cover data, and water body vector data; details are shown in Table 1. Sentinel-1 C-band Synthetic Aperture Radar Ground Range Detected (SAR GRD) remote sensing data was collected from the GEE platform. The imagery has a spatial resolution of 10 meters and has undergone orbit restoration, thermal noise removal, terrain correction, and radiometric calibration preprocessing. This embodiment collects Sentinel-1GRD images covering the study area from November 29, 2019 to November 28, 2020 (one year before the disaster) to extract aquaculture areas; images from November 1 to 28, 2020 (28 days before the disaster) are used as pre-disaster data, and images from December 6, 2020 to January 2, 2021 (28 days after the disaster) are used as post-disaster data; in addition, images from the two years before the disaster, which are in the same time period as the 28 days after the disaster, are collected as historical data to perform historical-pre-disaster-post-disaster water body change detection. Sentinel-2 multispectral (MSI, Multispectral Instrument) time series data are collected from the GEE platform, and the spatial resolution of the visible and near-infrared bands of the data is 10m. All Sentinel-2 MSI (Level-2A) images covering the study area were collected from November 29, 2019, to November 28, 2020, and 28 days after the Burevi cyclone disaster (December 6, 2020, to January 2, 2021). The images were pre-processed to remove clouds using the Sentinel-2 cloud probability product provided by GEE to extract aquaculture areas and verify the results of the flooded area assessment. European Space Agency (ESA) 10m resolution land cover data was collected from the GEE platform, and natural water body vector data of the study area was collected from OpenStreetMap (OSM) to enable detailed extraction of aquaculture areas.

[0059] Table 1. Statistical table of image data of aquaculture flooding disaster coverage area

[0060]

[0061] Specifically, as shown in Table 1, the first Sentinel-1 GRD imagery is from December 6, 2018, to January 2, 2019; from December 6, 2019, to January 2, 2020; from November 1, 2020, to November 28, 2020; and from December 6, 2020, to January 2, 2021. The Sentinel-2 MSI remote sensing data is from November 30, 2019, to November 29, 2020. The second Sentinel-1 GRD imagery is from November 29, 2019, to November 28, 2020. The time series images one year before the disaster include Sentinel-2MSI remote sensing data and the second Sentinel-1GRD imagery; the historical time series images are Sentinel-1GRD images from December 6, 2018 to January 2, 2019 and from December 6, 2019 to January 2, 2020; the pre-disaster time series images are Sentinel-1GRD images from November 1, 2020 to November 28, 2020; and the post-disaster time series images are Sentinel-1GRD images from December 6, 2020 to January 2, 2021.

[0062] Step 2: Extract a first aquaculture area image of the aquaculture flood disaster coverage area before the flood disaster occurs based on the Sentinel-2MSI remote sensing data and the second Sentinel-1GRD image.

[0063] In practical applications, due to the limited resolution of a single water body index, other landforms (such as building shadows) are often misclassified as water bodies. Therefore, this embodiment combines Sentinel-1 and Sentinel-2 time-series imagery to perform water body recognition.

[0064] The aquaculture pond extraction process mainly includes four steps: (1) water body index calculation, calculating the normalized difference water body index of the Sentinel-2 time series data one year before the disaster, and synthesizing the time series data into a median image; calculating the bipolarized water body index of the Sentinel-1 time series data one year before the disaster, synthesizing the time series data into a median image, and using the OTSU threshold segmentation method to calculate the water body segmentation threshold. (2) Water body extraction from Sentinel time series data: Count the frequency of water accumulation in each pixel from the Sentinel-1 time series image, identify pixels with a water accumulation frequency greater than 0.25 as water bodies, and intersect the water body extraction results of the Sentinel-1 time series data with the water body extraction results of the Sentinel-2 time series data; (3) Post-processing of water body extraction results: Eliminate paddy fields, building shadows, and natural water bodies from the water body extraction results, and then eliminate water bodies whose areas do not meet the general rules of aquaculture ponds to minimize the impact of other water bodies on the extraction of aquaculture areas; (4) Aquaculture area extraction: Calculate the shape characteristics of the water body to be classified, and use the decision tree classification algorithm to distinguish aquaculture areas from other water bodies; Based on the spatial agglomeration characteristics of aquaculture areas, eliminate water bodies that are far away from other aquaculture ponds from the aquaculture area classification results, and select water bodies that are close to aquaculture ponds from the other water body classification results and classify them into aquaculture areas.

[0065] As a specific implementation, step 2 specifically includes:

[0066] Step 201: Apply a median synthesis method to the second Sentinel-1 GRD image to obtain a first median image.

[0067] Step 202: Perform threshold segmentation on the first median image to determine a water body extraction threshold.

[0068] Step 203: Calculate the SDWI index of each pixel in the second Sentinel-1GRD image, and use the pixel whose SDWI index is greater than the water body extraction threshold as the first water body pixel.

[0069] Specifically, for the Snetinel-1 SAR image, this embodiment extracts water bodies based on SDWI, and the calculation formula is as follows:

[0070] SDWI=ln(10×VV×VH)-8

[0071] In the formula, VV and VH correspond to the corresponding bands of Sentinel-1 images. The larger the SDWI, the higher the probability that the pixel is a water body. To determine the SDWI segmentation threshold, this example uses the time series data of Sentinel-1 one year before the disaster to synthesize a median image, as shown in the figure below: Figure 3As shown in the figure, the OTSU threshold segmentation method is used to calculate the segmentation threshold of the median image of Sentinel-1 SDWI, and the water body extraction threshold of the SDWI image is obtained as 0.3, that is, the water body extraction threshold is 0.3.

[0072] Step 204: Calculate the water accumulation frequency of the first water body pixel, and use the pixel with the water accumulation frequency greater than the first set threshold as the first effective water body.

[0073] In practical applications, considering that aquaculture areas show spatial agglomeration characteristics and the aquaculture ponds are relatively close to each other. If only the water body is extracted from a single-scene water body index image, affected by factors such as water body changes and noise, it is often difficult to accurately identify the small areas between ponds, resulting in the situation of connection or fusion between different aquaculture ponds. Based on this, in this embodiment, the water accumulation frequency (WF, Water Frequency) of the water body index time series data is calculated to identify the pixels with long-term water storage. WF is defined as the ratio of the number of times a pixel position is identified as a water body within a period of time to the total number of times the pixel is well observed (not covered by clouds and cloud shadows). The calculation formula is as follows:

[0074]

[0075] In the formula, T represents the total number of times a pixel position in the time series image is effectively observed; S represents the binary variable of the pixel position type, with the water body value being 1 and the non-water body being 0.

[0076] Sentinel-2 MSI data belongs to optical data, and its time series images are severely cloud-polluted; while Sentinel-1 GRD data belongs to SAR data, and SAR has the characteristic of penetrating clouds and rain. Therefore, the Sentinel-1 time series data can provide relatively stable observations. Based on this, in this embodiment, the WF of the year before the disaster in the study area is calculated using the SDWI time series data of Sentinel-1 on the GEE platform. Based on the WF values of each pixel position in this year, the water body types are divided into permanent water bodies (WF > 75%), seasonal water bodies (25% < WF ≤ 75%), and temporary water bodies (WF < 25%). Aquaculture ponds belong to seasonal water bodies and permanent water bodies. Therefore, pixels with a water accumulation frequency greater than 25% are defined as the annual effective water bodies. That is, the first set threshold is 25%.

[0077] Step 205: According to the Sentinel-2 MSI remote sensing data, apply the median synthesis method to obtain the second median image.

[0078] Step 206: Calculate the NDWI index of each pixel in the second median image, and use the pixel with the NDWI index greater than the second set threshold as the second effective water body.

[0079] In practical applications, considering that the shortwave infrared (SWIR) band used to calculate mNDWI and AWEI in Sentinel-2 data has a spatial resolution of 20 meters, while the green and near infrared (NIR) bands used to calculate NDWI both have a spatial resolution of 10 meters, a higher-precision extraction effect can be achieved. Therefore, this example calculates the NDWI index of Sentinel-2 time-series images on the GEE platform. The NDWI calculation formula is as follows:

[0080]

[0081] Where ρ Green and ρ NIR They correspond to the B3 and B12 bands of Sentinel-2 images respectively.

[0082] Sentinel-2 time series data is affected by clouds, and observations are unstable. Pixels with a WF greater than 0 are defined as valid water bodies throughout the year. That is, pixels with an NDWI greater than 0 are identified as water bodies. Therefore, the second threshold is set to 0.

[0083] Step 207: extracting the intersection of the first effective water body and the second effective water body to obtain an initial water body image.

[0084] In practical applications, the intersection of the effective water bodies of Sentinel-1 and Sentinel-2 time series data throughout the year is calculated as the water body extraction result, such as Figure 4 shown.

[0085] Step 208: Eliminate paddy field images, building shadow images, natural water images, and water images that do not meet the aquaculture pond area threshold range from the initial water image to obtain a water image to be classified.

[0086] In practical applications, the NDWI water index method based on WF threshold segmentation can extract water bodies with relatively accurate edges to a certain extent. However, paddy fields in agricultural production areas show similar shape characteristics to aquaculture areas, and water bodies with similar shapes to aquaculture areas also exist in natural water bodies. In addition, aquaculture areas are mostly distributed near large-scale natural water bodies, which can easily interfere with subsequent classification operations. Therefore, it is necessary to further process the water body extraction results. In this embodiment, the agricultural production area pixels in the ESA land cover data are overlaid and analyzed with the water body extraction results, and the intersection of the two is directly classified as other water bodies; the line elements and surface elements in the OSM water body vector are successively overlaid and analyzed with the water body extraction results, and the elements that intersect with the OSM water body data are directly classified as other water bodies.

[0087] In order to further reduce the interference of other water bodies on the extraction of aquaculture areas, 200 aquaculture pond samples were selected for water extraction results, and it was found that their area ranged from 300 to 3,000,000 m 2 Therefore, water bodies with an area exceeding the above range are also classified as other water bodies. Elements directly classified as other water bodies will not participate in the extraction of aquaculture areas. In other words, the threshold range of aquaculture pond area is 300~3000000m 2 .

[0088] Step 208 specifically includes:

[0089] Step 2081: Obtain land cover data and natural water body vector data of the aquaculture flood disaster coverage area.

[0090] Step 2082: Remove the land cover data, the natural water body vector data, and the water body image that does not meet the aquaculture pond area threshold range from the initial water body image to obtain a water body image to be classified.

[0091] Step 209: Applying a decision tree algorithm to classify the water body image to be classified, and obtaining an initial water body classification result; the initial classification result includes an initial aquaculture water body image and an initial non-aquaculture water body image.

[0092] In practical applications, the advantage of applying decision trees to aquaculture area identification lies in their strong interpretability compared to other machine learning methods. Based on the rules provided by different nodes in the tree diagram, the shape characteristics that distinguish aquaculture ponds from other water bodies can be inferred. Six shape features commonly used in aquaculture area identification were used as input parameters for the decision tree: area, perimeter, shape landscape index (LSI), compactness, area ratio (convex hull) (Ratio_(convex_hull)), and area ratio (minimum area rectangle) (Ratio_rectangle), as shown in Table 2. 200 samples of aquaculture ponds and 200 samples of non-aquaculture ponds were extracted from the water body extraction results. The number of training samples for each category exceeded 33 times the feature space dimension. The six shape features commonly used in aquaculture area identification were used as input features for the decision tree model, resulting in a feature space dimension of 6. The feature space dimension, which refers to the number of input features, meets the training requirements. The features corresponding to the training samples are input into the decision tree model, and the model parameters are adjusted for training. The trained model is then used to predict the category of the water body to be classified.

[0093] Table 2 Statistics of water body shape characteristics

[0094]

[0095] Step 210: Applying a corrosion algorithm to the initial aquaculture water body image and applying a dilation algorithm to the initial non-aquaculture water body image to obtain an updated aquaculture water body image and a non-aquaculture water body image; the updated aquaculture water body image is an image of a third aquaculture area before the aquaculture flood disaster occurs.

[0096] In practical applications, after decision tree classification, some aquaculture ponds with special shapes or connected together due to insufficient edge segmentation between ponds may be misclassified as non-aquaculture areas, while some natural water bodies with similar shape characteristics to aquaculture ponds may be misclassified as aquaculture areas. Considering that aquaculture areas have spatial agglomeration characteristics, if a certain area shows a high density of aquaculture areas, the probability that the water bodies in this area are aquaculture areas is higher; on the contrary, if a water body is isolated, the probability that the water body is an aquaculture area is lower. Based on the difference in spatial agglomeration characteristics between aquaculture areas and other water bodies, a "corrosion-expansion" post-processing method for water body classification results is proposed, such as Figure 5 As shown, specifically including:

[0097] (1) For the classification results of aquaculture areas, the nearest neighbor distance between each sample is searched, and the classification results with a nearest neighbor distance greater than 100m are eliminated (based on the observation of the classification results and the general distribution pattern of aquaculture ponds, this distance can better distinguish aquaculture ponds from other water bodies), thus realizing the "corrosion" operation of the classification results of aquaculture areas.

[0098] (2) Based on the operation in (1), for the classification results of other water bodies, the nearest neighbor distance between each sample and the classification results of aquaculture areas is searched, and other water bodies with a distance of less than or equal to 100m are corrected to aquaculture ponds. This conforms to the general distribution pattern of aquaculture ponds and has a good extraction effect.

[0099] Step 3: Based on the first Sentinel-1GRD image, extract a second aquaculture area image before the flooding disaster occurs and a third aquaculture area image after the flooding disaster occurs.

[0100] In practical applications, the image of the second aquaculture area before the flood disaster in the aquaculture flood disaster coverage area includes historical and pre-disaster water body images, that is, the image of the second aquaculture area includes historical water body images and pre-disaster water body images, and the image of the third aquaculture area is a post-disaster water body image. The historical-pre-disaster-post-disaster water body extraction is as follows: the SDWI water body index is calculated based on the Sentinel-1GRD image, the median of the water body index of the historical time series image is calculated, and the water body index images before and after the disaster are synthesized according to the time from the Burevi disaster, and the water body is extracted by the threshold segmentation method.

[0101] As a specific implementation, step 3 specifically includes:

[0102] Step 301: Calculate a first SDWI index of the Sentinel-1 GRD image before a flood disaster occurs and a second SDWI index of the Sentinel-1 GRD image after a flood disaster occurs.

[0103] Step 302: Pixels whose first SDWI index is greater than the water body extraction threshold are used as a second aquaculture area image.

[0104] Step 303: Pixels whose second SDWI index is greater than the water body extraction threshold are used as the third aquaculture area image.

[0105] In practical applications, considering the characteristics of cloudy and rainy weather before and after tropical cyclone disasters, the availability of optical remote sensing images is often low. Sentinel-1GRD images are used to synthesize SDWI images using historical, pre-disaster, and post-disaster Sentinel-1 image sets. For pre-disaster and post-disaster images, the pre-disaster SDWI composite images and post-disaster SDWI composite images are generated according to the time length of each image from the time of Cyclone Burevi. The SDWI values ​​of the historical images of the past two years are calculated, and the median calculation is used to synthesize a historical SDWI image of the study area. The principle of water extraction is as follows: Figure 6 shown.

[0106] The flooded area of ​​aquaculture areas is assessed by calculating the connectivity of aquaculture ponds in historical, pre-disaster, and post-disaster water bodies. If the post-disaster water body of an aquaculture pond is connected to other water bodies compared to the historical and pre-disaster water bodies, the aquaculture pond is considered damaged. The details are as follows:

[0107] Step 4: Based on the first aquaculture area image, using a connectivity algorithm, calculate the connectivity of each aquaculture area in the second aquaculture area image and the connectivity of each aquaculture area in the third aquaculture area image.

[0108] Step 5: Compare the connectivity of each aquaculture area in the second aquaculture area image with the connectivity of each aquaculture area in the third aquaculture area image to determine a damage assessment result for each aquaculture area within the aquaculture flooding disaster coverage area; the damage assessment result is either damaged or not damaged. Specifically, when the connectivity of an aquaculture area in the third aquaculture area image is greater than the connectivity of the corresponding aquaculture area in the second aquaculture area image, the aquaculture area is determined to be damaged.

[0109] In practical applications, during disasters such as typhoons, floods, and tsunamis, water levels in aquaculture areas may rise and flood other water bodies, leading to the loss of aquaculture products. Based on this, the damage is assessed by comparing the changes in water connectivity of aquaculture ponds in the past, before, and after the disaster. Water connectivity records the spatial correspondence (one-to-one, many-to-one, or no correspondence) between the center points of aquaculture ponds and other water bodies and the water bodies in the past, before, and after the disaster. Figure 7 If aquaculture ponds are connected to each other or to natural water bodies due to a disaster, the spatial correspondence between the center points of the aquaculture area and non-aquaculture area and the post-disaster water body will change. For an aquaculture area, if the number of aquaculture ponds and non-aquaculture ponds in the corresponding water body increases, then all aquaculture ponds in the water body are considered damaged, such as Figure 8 shown.

[0110] This example also includes an assessment of the area of ​​flooded aquaculture ponds, specifically including:

[0111] 1. Determine the flooded aquaculture ponds based on water body connectivity and obtain the flooded aquaculture ponds.

[0112] 2. Calculate the area of ​​the flooded aquaculture pond vectors and sum them up to get the area of ​​all flooded aquaculture ponds.

[0113] This example uses SAR and visible light time-series imagery from the year before the Burevi disaster (November 29, 2019–November 28, 2020) to extract the distribution of aquaculture areas in the study area. During the observation period, Sri Lanka's aquaculture area totaled 32.83 square kilometers, primarily distributed in coastal areas with scattered distribution in inland areas. The Northwestern Province has an aquaculture area of ​​20.71 square kilometers, accounting for 63.08% of the country's total aquaculture area.

[0114] From the perspective of aquaculture area extraction range, the aquaculture area extraction method proposed in this embodiment can not only identify aquaculture ponds in coastal areas, but also effectively identify aquaculture ponds in inland areas, such as Figure 8As shown in the figure, the literature (Drengstig A, 2020) indicates that the total area of ​​aquaculture areas in Sri Lanka is 45 square kilometers. This value is considered the true value and compared with the extraction results of the present invention and the results calculated using remote sensing data using other methods (Ottinger M et al., 2021). The relative errors were 27.04% and 242.84%, respectively. Furthermore, the aquaculture areas extracted by the present invention do not include abandoned ponds, while the true data do not take this into account.

[0115] To further verify the extraction accuracy of the present invention, 100 aquaculture pond samples and 100 other water body samples were randomly selected from the classification results. All validation points were visually interpreted based on Google Earth high-definition imagery, assigned category attributes, and the accuracy was evaluated by calculating the confusion matrix, as shown in Table 3. The test results demonstrate the high accuracy of the present invention, with an overall accuracy of 0.9250 and a Kappa coefficient of 0.85. The user accuracy for aquaculture ponds was 0.8900, and the mapping accuracy was 0.9570.

[0116] Table 3. Statistics of the accuracy of confusion matrix evaluation for aquaculture ponds and other water bodies

[0117]

[0118] The aquaculture area extraction method proposed in the present invention was applied to Sri Lanka, and the damage to aquaculture areas caused by the Burevi cyclone disaster was assessed using historical, pre-disaster and post-disaster SAR time series images. During the Burevi cyclone disaster, the area of ​​damaged aquaculture areas in Sri Lanka was 5.29 square kilometers, accounting for 16.11% of the country's total aquaculture area. The damaged aquaculture areas were mainly distributed in the western part (North Western Province) and the northern part (Northern Province) of the study area. Among them, the damaged aquaculture ponds in the western part of the study area were more concentrated due to the dense distribution of aquaculture areas and the proximity to the path of the Burevi cyclone; the damaged aquaculture ponds were more concentrated because the Burevi cyclone passed directly through the northern part of the study area. In order to verify the accuracy of the flooded area assessment method, 50 damaged aquaculture pond samples were extracted from the assessment results and tested by visual interpretation on the Sentinel-2 synthetic images before and after the disaster. It was found that the accuracy of the assessment results was 86%.

[0119] This paper proposes a GEE-based method for extracting terrestrial aquaculture areas and assessing flooded areas. This method uses the GEE platform to process remote sensing time-series data and water characteristics, and combines ratio index analysis, spatial structure analysis, and machine learning methods to extract aquaculture ponds. The method also assesses flooded areas based on changes in water connectivity during historical, pre-disaster, and post-disaster periods. This method offers the following advantages:

[0120] (1) Based on the spectral characteristics of water bodies, water body information is extracted from multi-source remote sensing time series data on the GEE platform, effectively distinguishing normal aquaculture ponds from abandoned aquaculture ponds.

[0121] (2) Based on the shape characteristics of aquaculture ponds, a machine learning method was used to classify water bodies. Based on the spatial clustering characteristics of water bodies, a "corrosion-expansion" post-processing method was proposed. This method further improved the accuracy of water body classification and made this research method applicable to the extraction of aquaculture areas in both coastal and inland areas.

[0122] (3) Use historical, pre-disaster, and post-disaster remote sensing time series images to extract water body changes and ensure the reliability of flooded area assessment results. Conduct flooded area assessment based on water body connectivity to effectively identify aquaculture ponds that have experienced losses due to rising water levels and connectivity with other water bodies.

[0123] (4) The aquaculture area extraction and flooded area assessment methods extracted in this study were applied to Sri Lanka and the Burevi cyclone disaster. The results showed that the overall accuracy of the aquaculture area extraction results reached 0.9250, and the Kappa coefficient was 0.85; the accuracy of the aquaculture area flooded area assessment results was 0.86.

[0124] (5) Sri Lanka has a total aquaculture area of ​​32.83 square kilometers, mainly distributed in coastal areas and scattered in inland areas. Cyclone Burevi caused damage to 5.29 square kilometers of aquaculture area in the country, accounting for 16.11% of the total aquaculture area.

[0125] (6) The aquaculture ponds in the study area are small in size, clustered, and irregular in space, resulting in poor segmentation between small aquaculture ponds. Subsequent research will further explore the refined segmentation method of water objects.

[0126] Example 2

[0127] In order to execute the method corresponding to the above embodiment 1 and achieve the corresponding functions and technical effects, a system for assessing damage in an aquaculture area is provided below. The system includes:

[0128] An acquisition module is used to acquire images of the area covered by the aquaculture flooding disaster; the images include first Sentinel-1GRD images of the same length in different periods and Sentinel-2MSI remote sensing data and second Sentinel-1GRD images of the same length in the same period; the first Sentinel-1GRD images include Sentinel-1GRD images before the flooding disaster occurs and Sentinel-1GRD images after the flooding disaster occurs; the Sentinel-2MSI remote sensing data and Sentinel-1GRD images of the same length in the same period are both images taken before the flooding disaster occurs.

[0129] The first extraction module is used to extract a first aquaculture area image of the aquaculture flood disaster coverage area before the flood disaster occurs based on the Sentinel-2MSI remote sensing data and the second Sentinel-1GRD image.

[0130] The second extraction module is used to extract a second aquaculture area image before the aquaculture flooding disaster occurs and a third aquaculture area image after the flooding disaster occurs, based on the first Sentinel-1GRD image.

[0131] The connectivity calculation module is used to calculate the connectivity of each aquaculture area in the second aquaculture area image and the connectivity of each aquaculture area in the third aquaculture area image based on the first aquaculture area image using a connectivity algorithm.

[0132] An assessment module is used to compare the connectivity of each aquaculture area in the second aquaculture area image with the connectivity of each aquaculture area in the third aquaculture area image, and determine a damage assessment result of each aquaculture area within the aquaculture flooding disaster coverage area; the damage assessment result is damaged or not damaged.

[0133] Example 3

[0134] An embodiment of the present invention provides an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the method for privacy protection of security surveillance videos of embodiment 1.

[0135] Optionally, the above-mentioned electronic device may be a server.

[0136] In addition, an embodiment of the present invention further provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the method for privacy protection of security surveillance videos of the first embodiment is implemented.

[0137] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0138] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. A method for assessing damage in an aquaculture area, characterized in that: The method comprises: Acquire images of the area affected by the aquaculture flooding disaster; the images include first Sentinel-1GRD images of the same length and different time periods, and Sentinel-2MSI remote sensing data and second Sentinel-1GRD images of the same length and different time periods; the first Sentinel-1GRD images include Sentinel-1GRD images before and after the flooding disaster; the Sentinel-2MSI remote sensing data and Sentinel-1GRD images of the same length and different time periods are both images taken before the flooding disaster occurs; Extracting a first aquaculture area image of the aquaculture flood disaster coverage area before the flood disaster occurs based on the Sentinel-2MSI remote sensing data and the second Sentinel-1GRD image; Extracting, based on the first Sentinel-1GRD image, a second aquaculture area image before the aquaculture flooding disaster occurs and a third aquaculture area image after the aquaculture flooding disaster occurs. Calculating, based on the first aquaculture area image, the connectivity of each aquaculture area in the second aquaculture area image and the connectivity of each aquaculture area in the third aquaculture area image using a connectivity algorithm; Compare the connectivity of each aquaculture area in the second aquaculture area image with the connectivity of each aquaculture area in the third aquaculture area image to determine the damage assessment result of each aquaculture area within the aquaculture flooding disaster coverage area; the damage assessment result is damaged or not damaged; when the connectivity of the aquaculture area in the third aquaculture area image is greater than the connectivity of the corresponding aquaculture area in the second aquaculture area image, it is determined that the aquaculture area is damaged.

2. The aquaculture area disaster damage assessment method according to claim 1, characterized in that: Extracting a first aquaculture area image of the aquaculture flood disaster coverage area before the flood disaster occurs based on the Sentinel-2MSI remote sensing data and the second Sentinel-1GRD image, specifically including: Applying a median synthesis method to the second Sentinel-1GRD image to obtain a first median image; Performing threshold segmentation on the first median image to determine a water body extraction threshold; Calculating the SDWI index of each pixel in the second Sentinel-1GRD image, and taking the pixel whose SDWI index is greater than the water body extraction threshold as the first water body pixel; Calculating the water accumulation frequency of the first water body pixel, and taking the pixel whose water accumulation frequency is greater than a first set threshold as the first valid water body; Applying a median synthesis method to the Sentinel-2MSI remote sensing data to obtain a second median image; Calculating the NDWI index of each pixel in the second median image, and taking the pixel whose NDWI index is greater than a second set threshold as the second effective water body; Extracting the intersection of the first effective water body and the second effective water body to obtain an initial water body image; Eliminating paddy field images, building shadow images, natural water images, and water images that do not meet an aquaculture pond area threshold range from the initial water image to obtain a water image to be classified; Applying a decision tree algorithm to classify the water body image to be classified to obtain an initial water body classification result; the initial classification result includes an initial aquaculture water body image and an initial non-aquaculture water body image; Applying a corrosion algorithm to the initial aquaculture water body image and applying a dilation algorithm to the initial non-aquaculture water body image to obtain updated aquaculture water body image and non-aquaculture water body image; the updated aquaculture water body image is the first aquaculture area image of the aquaculture flood disaster coverage area before the flood disaster occurs.

3. The aquaculture area disaster damage assessment method according to claim 2, characterized in that: Eliminating paddy field images, building shadow images, natural water images, and water images that do not meet the aquaculture pond area threshold range from the initial water image to obtain a water image to be classified, specifically including: Obtain land cover data and natural water body vector data of areas affected by aquaculture flooding disasters; The land cover data, the natural water body vector data, and the water body image that does not meet the aquaculture pond area threshold range are eliminated from the initial water body image to obtain a water body image to be classified.

4. The method for assessing damage in aquaculture areas according to claim 2, wherein: Extracting a second aquaculture area image before the aquaculture flooding disaster occurs and a third aquaculture area image after the aquaculture flooding disaster occurs from the first Sentinel-1GRD image, specifically including: Calculate the first SDWI index of the Sentinel-1GRD image before the flood disaster and the second SDWI index of the Sentinel-1GRD image after the flood disaster; The pixels whose first SDWI index is greater than the water body extraction threshold are used as the second aquaculture area image; The pixels whose second SDWI index is greater than the water body extraction threshold are used as the third aquaculture area image.

5. A disaster damage assessment system for aquaculture areas, characterized in that: The system comprises: An acquisition module is configured to acquire images of the area covered by the aquaculture flooding disaster; the images include first Sentinel-1GRD images of the same length and different periods, and Sentinel-2MSI remote sensing data and second Sentinel-1GRD images of the same length and different periods; the first Sentinel-1GRD images include Sentinel-1GRD images before and after the flooding disaster; the Sentinel-2MSI remote sensing data and Sentinel-1GRD images of the same length and different periods are both images taken before the flooding disaster occurs; A first extraction module is configured to extract a first aquaculture area image of the aquaculture flood disaster coverage area before the flood disaster occurs based on the Sentinel-2MSI remote sensing data and the second Sentinel-1GRD image; A second extraction module is configured to extract, based on the first Sentinel-1GRD image, a second aquaculture area image before the aquaculture flooding disaster occurs and a third aquaculture area image after the flooding disaster occurs. a connectivity calculation module, configured to calculate, based on the first aquaculture area image and using a connectivity algorithm, the connectivity of each aquaculture area in the second aquaculture area image and the connectivity of each aquaculture area in the third aquaculture area image; An assessment module is used to compare the connectivity of each aquaculture area in the second aquaculture area image with the connectivity of each aquaculture area in the third aquaculture area image, and determine a damage assessment result of each aquaculture area within the aquaculture flooding disaster coverage area; the damage assessment result is damaged or not damaged; when the connectivity of an aquaculture area in the third aquaculture area image is greater than the connectivity of the corresponding aquaculture area in the second aquaculture area image, it is determined that the aquaculture area is damaged.

6. An electronic device, characterized in that: The electronic device comprises a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the aquaculture area damage assessment method according to any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that It stores a computer program, which, when executed by a processor, implements the aquaculture area disaster damage assessment method according to any one of claims 1 to 4.

Images