A method for extracting seagrass bed distribution based on drone images

Through drone image processing technology, the maximum likelihood method and ISO clustering method are used, combined with support vector machine and geographic information system, the precise extraction of seagrass bed distribution is achieved, solving the shortcomings of seagrass bed dynamic monitoring and high-resolution monitoring. , provides efficient seaweed ecological monitoring data support.

CN118552871BInactive Publication Date: 2024-11-05GUANGDONG OCEAN UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411029983.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-30
Publication Date
2024-11-05
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The prior art has little research on dynamic monitoring of seagrass beds, and there are shortcomings in high resolution and flexible deployment, making it difficult to effectively monitor the distribution and growth of seagrass beds.

Method used

The distribution and extraction method of seagrass bed based on drone images is adopted to obtain images through drone aerial photography, and image processing is performed using maximum likelihood method and ISO clustering unsupervised method to extract the distribution status, growth density and coverage of seagrass beds. , combine support vector machine and geographic information system software for classification and analysis.

Benefits of technology

It improves the accuracy and efficiency of accurate extraction and information extraction of seagrass bed distribution, provides reliable data support for seagrass ecological monitoring and protection, and can adapt to seagrass growth under different tide conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118552871B_ABST
    Figure CN118552871B_ABST
Patent Text Reader

Abstract

The present invention belongs to the field of hydrology and ocean technology, and specifically discloses a method for extracting the distribution of seagrass beds based on drone images, comprising the following steps: setting sampling points in the research area, and using drones for aerial photography to obtain drone images of the sampling points; processing the obtained drone images to obtain processed images; using the maximum likelihood method and ISO clustering unsupervised method to extract the areas where seagrass beds are concentrated, the distribution status, growth density and coverage of the seagrass beds from the processed images; based on the areas where seagrass beds are concentrated, the distribution status, growth density and coverage, the distribution characteristics of the seagrass beds are obtained, and the distribution of the seagrass beds is extracted. The present invention improves the accuracy and efficiency of seagrass information extraction, and solves the problems that the existing seagrass bed monitoring technology is less studied in dynamic monitoring, and is insufficient in high resolution and flexible deployment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the field of hydrology and ocean technology, and particularly relates to a design of a seagrass bed distribution extraction method based on unmanned aerial vehicle images. Background Art

[0002] Seagrass is a unique type of plant that usually grows in shallow seas and estuarine waters. It is called the grassland of the ocean and its importance in the marine ecosystem cannot be ignored. Seagrass beds are one of the typical marine ecosystems with extremely high productivity, providing indispensable ecological service functions for coastal areas. In addition, seagrass beds are also the attachment sites of invertebrates, spawning grounds and shelters for herbivorous fish and other organisms, and are often used to measure the health of offshore ecosystems. As the most productive ecosystem, seagrass beds have huge carbon sink capacity, affecting the global C, N, and P cycles, and are responsible for the global nutrient cycle and the stability of the intertidal and subtidal zones. Therefore, although seagrass species are rare, they also play a vital role in the health and stability of marine ecosystems. It is crucial to monitor seagrass ecosystems to maintain their huge emission reduction potential.

[0003] In recent years, domestic scholars have studied the seagrass area in various regions and revealed the reasons for the sharp reduction and degradation of seagrass area. Due to the influence of various human and natural factors, including overfishing, coastal development, water pollution, etc., the seagrass bed area in many regions of my country has been shrinking and the ecological function has gradually weakened. Nearly 1 / 3 of the seagrass in the world has disappeared, and the overall decline rate of seagrass beds has been nearly 2%. In comparison, the regeneration rate of seagrass is only one-tenth of its degradation rate. This rapid decline has not only caused a sharp reduction in the habitat of marine organisms, but also had an unignorable impact on the ecological balance. Therefore, it has become a top priority to protect the seagrass bed ecosystem and promote its regeneration and recovery.

[0004] At present, the monitoring of seagrass beds is mostly based on satellite remote sensing. At the same time, most of the research on seagrass is focused on area distribution, species characteristics and habitat surveys. There are few studies on dynamic monitoring, and there are shortcomings in high-resolution and flexible deployment of marine ecological monitoring. Summary of the invention

[0005] The purpose of the present invention is to solve the problem that there is little research on dynamic monitoring of seagrass bed monitoring technology and that it is insufficient in high resolution and flexible deployment. A seagrass bed distribution extraction method based on drone images is proposed.

[0006] The technical solution of the present invention is: a method for extracting seagrass bed distribution based on drone images, comprising the following steps:

[0007] S1. Set up sampling points in the study area and use drones to take aerial photos and obtain drone images of the sampling points;

[0008] S2. Processing the acquired drone images to obtain multiple block images;

[0009] S3. Use the maximum likelihood method to extract the areas where seagrass beds are concentrated from the block map;

[0010] S4. Using the ISO clustering unsupervised method, the distribution, growth density and coverage of seagrass beds were obtained from the block map;

[0011] S5. Based on the concentrated distribution areas, distribution conditions, growth density and coverage of seagrass beds, the distribution characteristics of seagrass beds are obtained to extract the distribution of seagrass beds.

[0012] The beneficial effects of the present invention are:

[0013] 1. Compare and classify the data monitored by drones to improve the classification accuracy of images and data information in the study area, and use the maximum likelihood method and ISO clustering method for classification, considering the growth and distribution of seagrass meadows under different conditions such as high tide and low tide. On this basis, estimate the coverage information of the extracted seagrass beds to provide a practical and reliable basis for the protection of seagrass ecological environment and the development of seagrass resources.

[0014] 2. Through a series of remote sensing image processing and analysis steps, the present invention realizes the precise extraction of seagrass bed distribution, effectively distinguishes between water and land, defines the suitable growth area of ​​seagrass, and improves the accuracy and efficiency of seagrass information extraction through the comprehensive application of different methods, providing strong data support for seagrass ecological monitoring and protection.

[0015] Preferably, when the drone is used for aerial photography in step S1, the photography is performed during low tide.

[0016] The beneficial effects of the above preferred solution are:

[0017] Choose low tide to take drone photos to ensure that the seaweed is fully exposed above the water surface and improve the quality of observation.

[0018] Preferably, the step S2 specifically comprises the following steps:

[0019] S21. Use drone data processing software to automatically stitch the acquired drone images to obtain orthophotos;

[0020] S22. Use geographic information system software to perform coordinate correction, cropping and block division on the orthophoto to obtain multiple block maps.

[0021] The beneficial effects of the above preferred solution are:

[0022] The accuracy of data analysis is effectively improved by automatically stitching, correcting coordinates, cropping and dividing images.

[0023] Preferably, the step S3 specifically comprises the following steps:

[0024] S31. Use the support vector machine method to perform supervised classification on multiple block diagrams to be classified, and obtain a classified vector data map;

[0025] S32. Use the contour tool in the geographic information system software to divide the classified vector data map to obtain contour blocks;

[0026] S33. Use the feature-to-surface tool in the geographic information system software to convert the generated contour line blocks into surfaces, and sort and number each block to obtain the area where seagrass is concentrated in the block.

[0027] The beneficial effects of the above preferred solution are:

[0028] The maximum likelihood method was used for classification, which enabled accurate extraction of seagrass bed distribution and improved the accuracy and efficiency of seagrass information extraction.

[0029] Preferably, the step S31 specifically includes the following steps:

[0030] S311. Select pixels from the drone image as training samples and mark the categories of the training samples;

[0031] S312. Construct a support vector machine model and set the kernel function type and parameters of the support vector machine model;

[0032] S313. According to the kernel function type and parameters, a classification hyperplane with a maximum margin of the support vector machine model is obtained, and the training samples are input into the support vector machine model for training to obtain a trained support vector machine model;

[0033] S314. Inputting the unclassified multiple block images into the trained support vector machine model for classification to obtain a raster image of the category to which each pixel belongs;

[0034] S315. Obtain a classified vector data map based on the raster image of the category to which each pixel belongs.

[0035] Preferably, the kernel function type of the support vector machine model in step S312 is a Gaussian radial basis kernel function; the expression formula of the Gaussian radial basis kernel function is:

[0036]

[0037] in, represents the Gaussian radial basis kernel function; indicates a seagrass sample; Indicates water sample; represents the bandwidth parameter of the Gaussian function; Represents the Euclidean distance between seagrass samples and water samples.

[0038] The beneficial effects of the above preferred solution are:

[0039] Using the support vector machine model to classify images can clearly describe the differences between different classification results. At the same time, selecting the Gaussian radial basis kernel function can effectively realize the classification of seagrass and water bodies, which helps to accurately extract the distribution information of seagrass beds.

[0040] Preferably, the step S32 specifically includes the following steps:

[0041] S321. Select the vector data map to be classified in the contour line work box, and set the contour line spacing to 1 to divide the contour lines;

[0042] S322. According to the contour lines, select the area where the seaweed distribution is greater than a preset threshold, connect the edge contour lines of the selected area to obtain contour line blocks.

[0043] The beneficial effects of the above preferred solution are:

[0044] The use of contour tools for division improved the classification accuracy of images and data information in the study area and realized the extraction of areas with concentrated seagrass distribution in the block map.

[0045] Preferably, the ISO clustering unsupervised method in step S4 is specifically as follows: according to the spectral reflectance characteristics of the seabed substrate and seaweed in the drone image, the seaweed features in the image are classified and extracted, and the number of classifications is set to 6; and the grayscale co-occurrence matrix method is used to extract texture features from the block map, and the extracted texture features are combined with the original spectral reflectance features to form a feature vector containing spectral reflectance features and texture features.

[0046] The beneficial effects of the above preferred solution are:

[0047] The spectral reflectance characteristics of the seabed and seagrass in the image are used for classification, and the number of classifications is set to 6. Combined with the grayscale co-occurrence matrix method, the interference of other seabed organisms can be avoided, the seagrass growth range can be accurately extracted, the classification effect can be improved, and the seagrass growth range can be accurately extracted.

[0048] Preferably, the step S4 specifically comprises the following steps:

[0049] S41. Convert the color image into a grayscale image, calculate the grayscale co-occurrence matrix and its features using Python, and for the grayscale image, count the frequency of simultaneous occurrence of the grayscale values ​​of each pixel and its neighborhood to form a grayscale co-occurrence matrix;

[0050] S42. Extracting texture features from the gray-level co-occurrence matrix and converting the image data into point features using geographic information system software;

[0051] S43. Adding the texture feature to the image attribute table and integrating it with other attributes of the original image to obtain the image attribute table;

[0052] S44. Using the ISO clustering unsupervised method to classify the block images, 6 raster images with different colors are obtained;

[0053] S45. resample and unify the six grid images with different colors, classify and merge the different colors representing seaweed into the same category, and obtain two grid images with different colors;

[0054] S46. Process the image data using the raster-to-surface tool in the geographic information system software to obtain the pixel area of ​​the seaweed in the image;

[0055] S47. Check the vector data of each image in the image attribute table to obtain the concentrated distribution range of seaweed and the corresponding distribution area, and synthesize the images to obtain the distribution status, growth density and coverage of seaweed in the entire block map.

[0056] The beneficial effects of the above preferred solution are:

[0057] The ISO clustering unsupervised method was used for classification, which achieved accurate extraction of seagrass bed distribution and improved the accuracy and efficiency of seagrass information extraction.

[0058] Preferably, the two grid images of different colors in step S45 include green and blue, the green grid image is seaweed; and the blue grid image is water and other bottom materials.

[0059] The beneficial effects of the above preferred solution are:

[0060] The raster images are divided into two colors: green and blue. Green represents seagrass, and blue represents water and other bottom materials. The seabed bottom materials and seagrass bed types can be clearly distinguished from the image, which helps to accurately extract the distribution information of seagrass beds. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 Shown is a flow chart of a method for extracting seagrass bed distribution based on drone images. DETAILED DESCRIPTION

[0062] Now, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. It should be understood that the embodiments shown and described in the accompanying drawings are only exemplary and are intended to explain the principles and spirit of the present invention, rather than to limit the scope of the present invention.

[0063] Embodiment 1, as Figure 1 As shown, a method for extracting seagrass bed distribution based on drone images includes the following steps:

[0064] S1. Set up sampling points in the study area and use drones for aerial photography. Combined with ground truth data, obtain drone images of the sampling points by setting reasonable flight altitude, speed and shooting parameters;

[0065] S2. Processing the acquired drone images to obtain multiple block images;

[0066] S3. Use the maximum likelihood method to extract the areas where seagrass beds are concentrated from the block map;

[0067] S4. Using the ISO clustering unsupervised method, the distribution, growth density and coverage of seagrass beds were obtained from the block map;

[0068] S5. Based on the concentrated distribution areas, distribution conditions, growth density and coverage of seagrass beds, the distribution characteristics of seagrass beds are obtained to extract the distribution of seagrass beds.

[0069] In this embodiment, when the drone is used for aerial photography in step S1, the photography is performed during low tide.

[0070] In this embodiment, step S2 specifically includes the following steps:

[0071] S21. Use PIX4D software to automatically stitch the acquired drone images to obtain orthophotos;

[0072] S22. Use ArcGIS software to perform coordinate correction, cropping and block division on the orthophoto to obtain multiple block maps.

[0073] In this embodiment, step S3 specifically includes the following steps:

[0074] S31. Using the support vector machine (SVM) method, combined with the collected seaweed and water sample files as training data, supervised classification is performed on the multiple block images to be classified to obtain a classification vector data map;

[0075] S32. Using the contour tool in ArcGIS software to divide the classified vector data map, obtain contour blocks;

[0076] S33. Use the feature-to-polygon tool in ArcGIS software to convert the generated contour line blocks into surfaces, and sort and number each block to obtain the area with concentrated seagrass distribution in the block.

[0077] In this embodiment, step S31 specifically includes the following steps:

[0078] S311. Select pixels from the drone image as training samples and mark the categories of the training samples;

[0079] S312. Construct a support vector machine model and set the kernel function type and parameters of the support vector machine model. The type of the kernel function determines how the support vector machine finds the best classification plane in the feature space;

[0080] S313. According to the kernel function type and parameters, a classification hyperplane with a maximum margin of the support vector machine model is obtained, and the training samples are input into the support vector machine model for training to obtain a trained support vector machine model, wherein the classification hyperplane with a maximum margin can best distinguish the training samples;

[0081] S314. Input the unclassified multiple block images into the trained support vector machine model for classification, and predict the block image category according to the decision function value calculated by the model; map the predicted category of each pixel back to the spatial position of the original image to obtain a raster image of the category to which each pixel belongs, wherein each pixel value represents the category label to which it is classified;

[0082] S315. Obtain a classified vector data map based on the raster image of the category to which each pixel belongs.

[0083] In this embodiment, the kernel function type of the support vector machine model in step S312 is a Gaussian radial basis kernel function; the expression formula of the Gaussian radial basis kernel function is:

[0084]

[0085] in, represents the Gaussian radial basis kernel function; indicates a seagrass sample; Indicates water sample; represents the bandwidth parameter of the Gaussian function; Represents the Euclidean distance between seagrass samples and water samples.

[0086] In this embodiment, step S32 specifically includes the following steps:

[0087] S321. Select the vector data map to be classified in the contour line work box, and set the contour line spacing to 1 to divide the contour lines;

[0088] S322. According to the contour lines, select the area where the seaweed distribution is greater than a preset threshold, connect the edge contour lines of the selected area to obtain contour line blocks.

[0089] In this embodiment, the ISO clustering unsupervised method described in step S4 is specifically as follows: according to the spectral reflectance characteristics of the seabed substrate and seaweed in the drone image, the seaweed features in the image are classified and extracted, and the number of classifications is set to 6; and the gray level co-occurrence matrix (GLCM) method is used to extract texture features from the block map, and the extracted texture features are combined with the original spectral reflectance features to form a feature vector containing spectral reflectance features and texture features, so as to provide more details and discrimination for subsequent classification.

[0090] The growth environment of seagrass is complex. When it is classified, it will be disturbed by other environmental factors such as water reflection, water depth, phytoplankton, etc., which will lead to errors in the spectral reflectance of seagrass. Therefore, field investigations are conducted to investigate the growth status of seagrass in the research area, and the gray-level co-occurrence matrix method is used to obtain its distribution characteristics for classification and extraction of the seagrass growth range. This can avoid the interference of other seabed organisms, accurately extract the seagrass growth range, and achieve better classification results.

[0091] In this embodiment, step S4 specifically includes the following steps:

[0092] S41. Convert the color image into a grayscale image, calculate the grayscale co-occurrence matrix and its features using Python, and for the grayscale image, count the frequency of simultaneous occurrence of the grayscale values ​​of each pixel and its neighborhood to form a grayscale co-occurrence matrix;

[0093] S42. Extracting texture features from the gray-level co-occurrence matrix and converting the image data into point features using ArcGIS software, the texture features include contrast, correlation and entropy; contrast is used to measure the clarity and texture of the image, correlation is used to describe the correlation of pixel value changes in the image, and entropy is used to measure the amount of information contained in the image, that is, the uncertainty or complexity of the image;

[0094] S43. Adding the texture feature to the image attribute table and integrating it with other attributes of the original image to obtain the image attribute table;

[0095] S44. Using the ISO clustering unsupervised method to classify the block images, 6 raster images with different colors are obtained;

[0096] S45. resample and unify the six grid images with different colors, classify and merge the different colors representing seaweed into the same category, and obtain two grid images with different colors;

[0097] S46. Process the image data using the raster-to-surface tool in ArcGIS software to obtain the pixel area of ​​the seaweed in the image;

[0098] S47. Check the vector data of each image in the image attribute table to obtain the concentrated distribution range of seaweed and the corresponding distribution area, and synthesize the images to obtain the distribution status, growth density and coverage of seaweed in the entire block map.

[0099] In this embodiment, the two grid images of different colors in step S45 include green and blue, the green grid image is seaweed; and the blue grid image is water and other bottom materials.

[0100] The specific technical scheme of the present invention is illustrated by taking Xincun Bay and Li'an Lagoon in Hainan Province, China as the research area:

[0101] Example 2, taking Xincun Bay and Li'an Lagoon in Hainan Province, China as the research area, based on the real-time kinematic (RTK) drone data of DJI (PHANTOM 4RTK), combined with field survey data, the seagrass beds in Xincun Bay and Li'an Lagoon were taken as research targets, and the corresponding distribution range was extracted to form a spatial distribution map of seagrass meadows, providing favorable reference data for the monitoring, management and restoration of seagrass beds in Xincun Li'an Lagoon in Hainan Province.

[0102] First, we learned and collected relevant data on Xincun Bay and Li'an Lagoon, conducted field surveys in the study area, and used drone aerial photography to obtain seagrass measurement photos. Combined with field surveys, we studied the growth and ecological environment of lagoon seagrass. We set up observation points in the study area, took fixed-point samples at the sampling points, and used drone aerial photography to obtain drone images of the sampling points. We chose to take drone photos during low tide to ensure that the seagrass was fully exposed to the water surface and improve the quality of observation. By setting reasonable flight altitude, speed, and shooting parameters, we obtained seagrass measurement maps above the sampling points.

[0103] In the second step, the images of each location were classified and processed using Pix4D and ArcGIS software to integrate them into a complete site image map.

[0104] In the third step, the maximum likelihood method was used in combination with visual interpretation to classify the drone site images and extract the area where the seagrass beds were concentrated. The total area of ​​seagrass in this area was 256.8m 2 .

[0105] In the fourth step, the ISO clustering unsupervised method was used to classify and extract the seaweed features in the image to obtain the concentrated distribution range of seaweed and its corresponding distribution area. It can be obtained that the number of patches in the image is 2359, with a total area of ​​294.1m2 ; The images are synthesized to obtain the distribution status, growth density and coverage of seagrass in the entire block map.

[0106] The fifth step is to obtain the distribution characteristics of seagrass beds based on the concentrated distribution areas, distribution conditions, growth density and coverage of seagrass beds, and realize the extraction of seagrass bed distribution.

[0107] By matching the results of UAV remote sensing surveys, the accuracy of seagrass meadow distribution and area extraction is verified, and the extraction precision is verified, so as to explore the correct and applicable method for extracting seagrass distribution area, and then analyze and discuss the various factors affecting seagrass growth in the study area, and put forward reasonable suggestions for the problems.

[0108] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the present invention.

Claims

1. A method for extracting seagrass bed distribution based on drone images, characterized in that: The following steps are involved: S1. Set up sampling points in the study area and use drones to take aerial photos and obtain drone images of the sampling points; S2. Processing the acquired drone images to obtain multiple block images; S3. Use the maximum likelihood method to extract the areas where seagrass beds are concentrated from the block map; S4. Using the ISO clustering unsupervised method, the distribution, growth density and coverage of seagrass beds were obtained from the block map; S5. Based on the concentrated distribution area, distribution status, growth density and coverage of the seagrass beds, the distribution characteristics of the seagrass beds are obtained to extract the distribution of the seagrass beds; The step S3 specifically comprises the following steps: S31. Use the support vector machine method to perform supervised classification on multiple block diagrams to be classified, and obtain a classified vector data map; S32. Use the contour tool in the geographic information system software to divide the classified vector data map to obtain contour blocks; S33. Convert the generated contour blocks into surfaces using the feature-to-surface tool in the geographic information system software, and sort and number each block to obtain the area where seagrass is concentrated in the block; The step S31 specifically includes the following steps: S311. Select pixels from the drone image as training samples and mark the categories of the training samples; S312. Construct a support vector machine model and set the kernel function type and parameters of the support vector machine model; S313. According to the kernel function type and parameters, a classification hyperplane with a maximum margin of the support vector machine model is obtained, and the training samples are input into the support vector machine model for training to obtain a trained support vector machine model; S314. Inputting the unclassified multiple block images into the trained support vector machine model for classification to obtain a raster image of the category to which each pixel belongs; S315. Obtain a classified vector data map according to the raster image of the category to which each pixel belongs; The kernel function type of the support vector machine model in step S312 is a Gaussian radial basis kernel function; the expression formula of the Gaussian radial basis kernel function is: in, represents the Gaussian radial basis kernel function; indicates a seagrass sample; Indicates water sample; represents the bandwidth parameter of the Gaussian function; represents the Euclidean distance between seagrass samples and water samples; The step S32 specifically includes the following steps: S321. Select the vector data map to be classified in the contour line work box, and set the contour line spacing to 1 to divide the contour lines; S322. According to the contour lines, select the area where the seaweed distribution is greater than a preset threshold, connect the edge contour lines of the selected area to obtain the contour line block; The ISO clustering unsupervised method described in step S4 is specifically as follows: according to the spectral reflectance characteristics of the seabed substrate and seaweed in the drone image, the seaweed features in the image are classified and extracted, and the number of classifications is set to 6; and the texture features are extracted from the block map using the gray level co-occurrence matrix method, and the extracted texture features are combined with the original spectral reflectance features to form a feature vector including spectral reflectance features and texture features; The step S4 specifically comprises the following steps: S41. Convert the color image into a grayscale image, calculate the grayscale co-occurrence matrix and its features using Python, and for the grayscale image, count the frequency of simultaneous occurrence of the grayscale values ​​of each pixel and its neighborhood to form a grayscale co-occurrence matrix; S42. Extracting texture features from the gray-level co-occurrence matrix and converting the image data into point features using geographic information system software; S43. Adding the texture feature to the image attribute table and integrating it with other attributes of the original image to obtain the image attribute table; S44. Using the ISO clustering unsupervised method to classify the block images, 6 raster images with different colors are obtained; S45. resampling and uniformly processing the six grid images with different colors, classifying and merging the different colors representing seaweed into the same category, and obtaining two grid images with different colors, wherein the two grid images with different colors include green and blue, the green grid image is seaweed, and the blue grid image is water and other bottom materials; S46. Process the image data using the raster-to-surface tool in the geographic information system software to obtain the pixel area of ​​the seaweed in the image; S47. Check the vector data of each image in the image attribute table to obtain the concentrated distribution range of seaweed and the corresponding distribution area, and synthesize the images to obtain the distribution status, growth density and coverage of seaweed in the entire block map.

2. The method for extracting seagrass bed distribution based on drone images according to claim 1, characterized in that: When the drone is used for aerial photography as described in step S1, the photography is performed during low tide.

3. The method for extracting seagrass bed distribution based on drone images according to claim 1, characterized in that: The step S2 specifically includes the following steps: S21. Use drone data processing software to automatically stitch the acquired drone images to obtain orthophotos; S22. Use geographic information system software to perform coordinate correction, cropping and block division on the orthophoto to obtain multiple block maps.