A device and method for rapid acquisition of intertidal zone seaweed biomass

Abstract

The present application relates to the technical field of intertidal zone ecological monitoring and computer vision application, and discloses a device and method for rapidly obtaining intertidal zone seaweed biomass. The device comprises: a positioning calibration module, which aligns the projection laser mark auxiliary image acquisition area with the quadrat area; an image acquisition module, which acquires seaweed images in the quadrat; and a data processing module, which calls a deep learning segmentation model to perform pixel-level segmentation and species identification on the images, calculates the coverage of each type of seaweed according to a preset scale, and combines the species correction coefficient K in the density correction coefficient database to calculate the biomass of each type of seaweed and obtain the total biomass of the quadrat. The device replaces the traditional in-situ acquisition with non-contact image acquisition, realizes pixel-level identification of seaweed through a deep learning segmentation model, rapidly calculates the biomass in combination with the density correction coefficient, significantly improves the investigation efficiency, reduces the labor cost and the safety hidden danger of operation, and realizes non-destructive observation of seaweed communities.

Description

Technical Field

[0001] This invention relates to the field of intertidal ecological monitoring and computer vision application technology, and in particular to a device and method for rapidly acquiring intertidal algal biomass. Background Technology

[0002] Seaweed biomass is one of the core indicators in intertidal marine ecosystem surveys, and its data accuracy directly affects the assessment of marine ecological environment health, biological resource reserves, and community structure stability. Current methods for obtaining intertidal seaweed biomass primarily involve the following process: a sampling quadrat is established within a selected survey area; all seaweed within the quadrat is collected; the samples are brought back to the laboratory for sorting, washing, and drying; the wet weight of each type of seaweed is then measured using a balance, and the seaweed biomass is calculated. This method has several drawbacks: First, sample collection, transportation, sorting, and weighing require significant manpower and time, and the sorting process relies on experienced professionals, leading to overall inefficiency. Second, the intertidal zone is affected by tides and must be surveyed during the lowest tide period, especially in low tide areas where rushed work can easily result in incomplete sample collection, and sampling personnel face safety hazards in inclement weather. Finally, this sampling method is typically destructive, directly damaging the seaweed community structure and making long-term continuous observation of the same area difficult. Therefore, the need for efficient, safe, and non-destructive surveys of intertidal seaweed biomass is urgent. Summary of the Invention

[0003] To address the technical problems of low survey efficiency, significant operational safety hazards, destructive sampling, and difficulty in continuous observation in existing technologies, this invention provides a device and method for rapidly acquiring intertidal algal biomass.

[0004] The first aspect of this invention discloses a device for rapidly obtaining intertidal algal biomass, comprising: The positioning and calibration module is used to project laser marks onto a preset sample plot area to assist in aligning the image acquisition area with the preset sample plot area; The image acquisition module is used to acquire images of seaweed within the sample plots; The data processing module is used to call a deep learning segmentation model to perform pixel-level segmentation and species identification of the seaweed image, calculate the coverage of various seaweeds based on a preset image scale, and calculate the biomass of various seaweeds based on the coverage and the density correction coefficient K of the corresponding seaweed species in a pre-established density correction coefficient database, and sum them up to obtain the total biomass of the quadrat.

[0005] As an optional implementation, in the first aspect of the present invention, the image acquisition module is a multispectral camera, used to simultaneously acquire visible light band images and near-infrared band images within the sample plot; the data processing module is further used to register the near-infrared band images and the visible light band images, calculate vegetation index features, and input the vegetation index features and the visible light band images together into the deep learning segmentation model.

[0006] As an optional implementation, in the first aspect of the present invention, the device further includes a flushing module and a turbidity detection module; The flushing module is used to flush away debris attached to the surface of seaweed in the sample plot using low-pressure water flow before image acquisition. The turbidity detection module is used to detect the turbidity of the residual water in the sample plot in real time after flushing. The data processing module is also used to trigger the image acquisition module to perform an image acquisition operation when the turbidity of the water body is lower than a preset turbidity threshold.

[0007] As an optional implementation, in the first aspect of the present invention, the data processing module is further configured to calculate a comprehensive quality score for each frame of the acquired image, including sharpness, illumination uniformity, and algae occlusion rate; when the comprehensive quality score of all acquired images is lower than a preset quality threshold, a re-acquisition instruction is sent to the image acquisition module.

[0008] As an optional implementation, in the first aspect of the present invention, the deep learning segmentation model is a network model based on an improved YOLOv8.

[0009] As an optional implementation, in the first aspect of the present invention, the image acquisition module is further configured to acquire multiple frames of images of the same plot at a preset angle sequence; the data processing module is further configured to run a motion recovery structure algorithm based on the multiple frames of images to generate a three-dimensional point cloud of the seaweed community in the plot, extract the canopy height of various seaweeds, and introduce a thickness correction function of the canopy height when calculating the biomass of various seaweeds, wherein the value of the thickness correction function is 1 when the canopy height is zero.

[0010] As an optional implementation, in the first aspect of the present invention, the density correction coefficient K is obtained by simultaneously acquiring the measured wet weight and image recognition area of ​​various algae in the sample plot under different intertidal zone partitions and seasonal conditions, and then performing linear fitting.

[0011] A second aspect of this invention discloses a method for rapidly obtaining intertidal algal biomass, comprising: Align the image acquisition area with the preset sample plot area and acquire images of seaweed within the sample plot; A deep learning segmentation model is invoked to perform pixel-level segmentation and species identification on the seaweed image; Based on a preset image scale, the coverage of various types of seaweed in the quadrat is calculated. Based on the coverage and the density correction coefficient K of the corresponding seaweed species in the pre-established density correction coefficient database, the biomass of each type of seaweed is calculated and summed to obtain the total biomass of the quadrat.

[0012] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: The device provided by this invention uses a positioning calibration module to project laser markers to assist in the positioning of sample plots, an image acquisition module to perform non-contact image acquisition of seaweed within the sample plots, and a data processing module to call a deep learning segmentation model to complete pixel-level seaweed species identification. Based on a preset image scale, the coverage of each type of seaweed is calculated, and the biomass of each type of seaweed is calculated based on the density correction coefficient K corresponding to the seaweed species in the coverage and density correction coefficient database. These results in the total biomass of the sample plot. This device can replace traditional in-situ sampling with non-contact image acquisition, achieve pixel-level seaweed species identification through a deep learning segmentation model, and quickly calculate the biomass of each type of seaweed using a density correction coefficient database. This significantly improves survey efficiency, reduces labor costs and operational safety hazards, and enables non-destructive observation of seaweed communities.

[0013] It should be understood that the above general description and the following detailed description are merely exemplary and do not limit the invention. Attached Figure Description

[0014] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0015] Figure 1 This is a block diagram illustrating a device for rapidly acquiring intertidal algal biomass according to an exemplary embodiment; Figure 2 This is a training curve diagram of a deep learning segmentation model illustrated according to an exemplary embodiment; Figure 3 This is a comparison chart of the recognition performance of a deep learning segmentation model according to an exemplary embodiment; Figure 4 This is a linear fitting diagram of the relationship between the coverage and biomass of Ulva perforatum according to an exemplary embodiment; Figure 5 A block diagram of a device for rapidly acquiring intertidal algal biomass, according to another exemplary embodiment; Figure 6 This is a flowchart illustrating a method for rapidly obtaining intertidal algal biomass according to an exemplary embodiment. Detailed Implementation

[0016] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the invention as detailed in the appended claims.

[0017] Figure 1 This is a block diagram illustrating a device for rapidly acquiring intertidal algal biomass according to an exemplary embodiment. Figure 1 As shown, this device includes: The positioning calibration module 101 is used to project laser marks onto the preset sample plot area to assist the image acquisition module 102 in aligning the image acquisition area with the preset sample plot area during acquisition.

[0018] Specifically, the positioning calibration module 101 may include a laser emitting unit that projects a crosshair laser line onto the ground. Based on the laser markings, the device quickly aligns with the sample plot area, while automatically calibrating the shooting angle to ensure the image is free of tilt distortion. This achieves precise positioning of the 0.25×0.25m² preset sample plot area without the need for physical placement of the sample plot frame. Furthermore, the positioning calibration module 101 may also include a GPS positioning unit for real-time recording of the geographic coordinates of each sampling point and sending the coordinate information to the data processing module 103 to achieve spatial matching and long-term continuous observation of the sampling data.

[0019] Image acquisition module 102 is used to acquire images of seaweed in the sample plot and send them to data processing module 103.

[0020] In this embodiment of the invention, the image acquisition module 102 can be fixed to a telescopic bracket. The bracket has a horizontal adjustment base at its bottom to ensure that the lens is perpendicular to the sampling plane during shooting, guaranteeing a known and stable image scale. Specifically, the image acquisition module 102 uses a high-definition camera with waterproof and anti-backlight capabilities. The camera lens is equipped with a quickly detachable transparent protective cover to prevent seawater and sediment from contaminating the lens. For example, 3 to 5 frames can be captured for each sample plot, and the image with the highest clarity and no obstructions is selected as the valid image.

[0021] The data processing module 103 is used to call a deep learning segmentation model to perform pixel-level segmentation and species identification of seaweed images, calculate the coverage of various seaweeds based on a preset image scale, and calculate the biomass of various seaweeds based on the coverage and the density correction coefficient K of the corresponding seaweed species in a pre-established density correction coefficient database, and sum them up to obtain the total biomass of the quadrat.

[0022] In this embodiment of the invention, specifically, the data processing module 103 can have a built-in embedded processor, pre-store the trained deep learning segmentation model, receive the seaweed image data transmitted by the image acquisition module 102 in real time, complete the seaweed species identification and area calculation, and synchronously store GPS positioning information and calculation results; in addition, the data processing module 103 can also support local data caching and wireless transmission (such as 4G / 5G, Bluetooth), and can synchronize the calculated biomass results and image data to the mobile terminal or the backend server.

[0023] In this embodiment of the invention, the derivation process of the density correction coefficient K is as follows. Assume a specific type of seaweed within the sample plot... The coverage is seaweed in unit plot The quality is The density is The volume is ,but Because in a specific sample frame (e.g., the area can be denoted as...), In the given scenario, at low tide, the seaweed is in a collapsed state, and its thickness (<0.1m) is negligible. Therefore, the volume... Approximately equal to the area of ​​the perpendicular imaging plane ,Right now Assume that the seaweed in the unit sample plot... The biomass is ,but: ; From the above equation (1), it can be seen that the seaweed in a unit sample plot biomass and its cover Proportional, density correction factor K (i.e. (Unit: g / m²) is a constant. Therefore, the coverage percentage of various algae within the plot is further calculated, and then multiplied by the corresponding density correction factor K to obtain its biomass data. The total biomass B of the plot is: ; In equation (2), n represents the total number of seaweed species in a unit quadrat.

[0024] In this embodiment of the invention, optionally, the aforementioned deep learning segmentation model is a network model based on an improved YOLOv8, employing a YOLOv8 backbone network and a feature pyramid network as feature extraction modules, and a segmentation decoding head as the output module. It is trained on a labeled dataset containing thousands of images of intertidal algae, covering 10 common species such as Sargassum, Sargassum fusiforme, and Ulva procumbens. Images are manually annotated at the pixel level, with each image generating a corresponding pixel-level label image. The labeled categories include various types of algae and background (mud, rocks, water). The dataset is divided into training, validation, and test sets in an 8:1:1 ratio, and the classification accuracy after training is no less than 92%. Figure 2 As shown, the curves displaying the loss function and accuracy changes during the training process of the YOLOv8 instance segmentation model demonstrate good model convergence. In practical application tests, as shown... Figure 3 The image shows a comparison between the original seaweed images and the test set, demonstrating the model's recognition and segmentation performance in the complex intertidal environment (containing densely distributed seaweed, rocks, etc., with complex textures and diverse target morphologies). Comparisons of the original images and predicted images from multiple test sets show that for targets such as Sargassum, Ulva procumbens, and Sargassum fusiforme, the confidence scores of the model's output bounding boxes are mostly above 0.85, with many approaching 0.95. Simultaneously, the segmentation masks output by the model accurately correspond to the pixel-level regions of the targets, exhibiting a high degree of fit with the actual contours of the targets in the original images. This provides an extremely reliable data foundation for subsequent high-precision seaweed coverage calculations.

[0025] Furthermore, the aforementioned density correction coefficient K can be obtained by simultaneously acquiring the measured wet weight and image recognition area of ​​various algae within the sample plot under different intertidal zones and seasonal conditions, and then performing linear fitting. This embodiment of the invention does not limit this. Specifically, under various combinations of conditions in the high tide zone, mid-tide zone, low tide zone, and spring, summer, autumn, and winter seasons, the wet weight of various algae is measured, and linear fitting is performed using the corresponding image recognition area of ​​the sample plot to obtain the density correction coefficient K for each type of algae under each condition. A density correction coefficient database is then constructed based on this. The data processing module 103 automatically calls the corresponding sub-database according to the zone to which the sampling point belongs and the current season during biomass calculation. Taking *Ulva pertusa* as an example, its linear fitting equation for coverage and biomass is y=2.6328x-15.843 (R²=0.9206), and the fitting effect is as follows: Figure 4 As shown, the rationality of the linear relationship in equation (1) is verified.

[0026] As an optional embodiment, the data processing module 103 can also be used to calculate a comprehensive quality score for each acquired image frame, including sharpness, illumination uniformity, and algae occlusion rate. When the comprehensive quality score of all acquired images is lower than a preset quality threshold, a re-acquisition command is sent to the image acquisition module 102, and the image acquisition is retried after adjusting the supplementary light intensity of the supplementary light module.

[0027] In this embodiment of the invention, the comprehensive quality score can be obtained by weighted summation of the sharpness score, illumination uniformity score, and algae occlusion rate score, and all scores are normalized to the 0-1 range. Specifically, the sharpness score is obtained by calculating and normalizing the image gradient variance; the illumination uniformity score is obtained by calculating and inverting the standard deviation of the image brightness channel and normalizing it, with a smaller standard deviation resulting in a higher score; the algae occlusion rate score is the complement of the proportion of the area obscured by foreign objects, i.e., 1 minus the proportion of the obscured area to the total image area. When the comprehensive quality score of all acquired images is lower than a preset quality threshold, a re-acquisition command is sent to the image acquisition module 102, triggering an adjustment of the supplementary lighting module's supplementary lighting intensity before re-triggering image acquisition to ensure that the image quality entering subsequent processing meets the requirements.

[0028] As another optional embodiment, the image acquisition module 102 described above can be a multispectral camera, used to simultaneously acquire visible light and near-infrared images within the sample plot. Further optionally, the data processing module 103 can also be used to register the near-infrared and visible light images, calculate vegetation index features, and input the vegetation index features and visible light images together into a deep learning segmentation model.

[0029] In this embodiment of the invention, the high reflectivity of vegetation in the near-infrared channel is used to help distinguish algae-covered areas from abiotic backgrounds such as mud, sand, and bare beaches, which can effectively improve the pixel classification accuracy in complex intertidal environments. Specifically, the above-mentioned vegetation index features can be calculated using the Normalized Difference Vegetation Index (NDVI = (NIR - Red) / (NIR + Red)). After multispectral image registration, the NDVI feature map is used as an auxiliary channel and combined with the visible light band image to form a multi-channel input deep learning segmentation model. Here, NIR represents the near-infrared band reflectivity, reflecting the vegetation's ability to reflect near-infrared light; Red represents the red light band reflectivity, reflecting the vegetation's absorption of red light.

[0030] As an optional implementation, the image acquisition module 102 is further configured to acquire multiple frames of images from the same plot at a preset angle sequence. The data processing module is further configured to run a Structure from Motion (SfM) algorithm based on the multiple frames of images to generate a three-dimensional point cloud of the algal community within the plot, extract the canopy height of various algae, and introduce a canopy height thickness correction function when calculating the biomass of various algae. Specifically, when the canopy height... When the thickness is less than or equal to the preset thickness threshold, ;when When the thickness exceeds the preset threshold, Follow It increases with the increase of canopy height; in particular, when the canopy height increases. When zero, the thickness correction function The value is 1. Specifically, equation (1) is modified as follows: ; in The thickness correction function is a linear or piecewise function model based on canopy height. Due to the different three-dimensional morphologies of various large seaweed species, such as Sargassum and Sargassum fusiforme, their spatial volume increases at different rates with canopy height. When the canopy height... When it is zero, the thickness correction function When the value of the trend is 1, equation (3) degenerates into equation (1). This formula (3) can be used in scenarios with thick canopies of large algae and large approximate errors in planar coverage, and can significantly improve the accuracy of biomass estimation.

[0031] In this embodiment of the invention, the aforementioned preset angle sequence can be based on the center of the quadrat as the axis, with one image frame acquired every 30° rotation, for a total of 12 frames, covering a 360-degree full field of view. Preferably, the image acquisition module 102 can be fixed on a mechanical rotary table and driven to rotate by a stepper motor, thereby improving image acquisition efficiency and accuracy.

[0032] It is evident that implementation Figure 1 The described rapid intertidal algae biomass acquisition device can achieve rapid and accurate alignment between the image acquisition area and the preset sample plot area through the laser marking assistance function of the positioning calibration module. After the image acquisition module acquires algae images within the sample plot, the data processing module calls a deep learning segmentation model to perform pixel-level segmentation and species identification of the images, and calculates the coverage of various algae based on the preset image scale. Then, based on the density correction coefficient K of the corresponding algae species in the coverage and density correction coefficient database, the biomass of various algae is calculated and accumulated to obtain the total biomass of the sample plot. This device combines on-site image acquisition with intelligent data processing, eliminating the need for on-site collection, transportation, and laboratory sorting and weighing, significantly improving survey efficiency, reducing labor costs and operational safety hazards, and achieving non-destructive observation.

[0033] Figure 5 This is a block diagram illustrating another device for rapidly acquiring intertidal algal biomass according to an exemplary embodiment. Figure 5 The rapid intertidal algae biomass acquisition device shown is composed of Figure 1 The device shown was further optimized. Figure 1 Compared to the device shown, in Figure 5 The rapid intertidal algal biomass acquisition device shown may also include: The flushing module 104 is used to flush away debris attached to the surface of seaweed in the sample plot using low-pressure water flow before image acquisition.

[0034] In this embodiment of the invention, the flushing module 104 specifically consists of a small waterproof submersible pump, a flexible water pipe, and a flow regulating valve. During operation, the water pipe is aimed at the seaweed area in the sample plot. After the pump is turned on, low-pressure water flow is used to gently flush the mud, sand, dust, and other debris attached to the seaweed surface. The flow regulating valve can control the water flow intensity to avoid high-pressure water flow damaging the seaweed. After flushing, the surface is left to stand for a moment to allow the surface water stains to dissipate, triggering the turbidity detection module 105 to perform detection. Based on the detection results, image acquisition is then performed to improve the distinction between the seaweed area and the background.

[0035] The turbidity detection module 105 is used to detect the turbidity of the residual water in the sample plot after flushing in real time and transmit the detection results to the data processing module 103 in real time.

[0036] The aforementioned data processing module 103 is also used to trigger the image acquisition module 102 to perform image acquisition operations when the turbidity of the water body is lower than a preset turbidity threshold.

[0037] In this embodiment of the invention, the linkage control between the turbidity detection module 105 and the image acquisition module 102 realizes automated timing control of the rinsing process and image acquisition, eliminating the need for manual judgment and waiting, and further improving the standardization and operational efficiency of field operations.

[0038] It is evident that implementation Figure 5The described rapid intertidal algae biomass acquisition device can achieve rapid and accurate alignment between the image acquisition area and the preset sample plot area through the laser marking assistance function of the positioning calibration module. After the image acquisition module acquires algae images within the sample plot, the data processing module calls a deep learning segmentation model to perform pixel-level segmentation and species identification of the images, and calculates the coverage of various algae based on the preset image scale. Then, based on the density correction coefficient K of the corresponding algae species in the coverage and density correction coefficient database, the biomass of various algae is calculated and accumulated to obtain the total biomass of the sample plot. This device combines on-site image acquisition with intelligent data processing, eliminating the need for on-site collection, transportation, and laboratory sorting and weighing, significantly improving survey efficiency, reducing labor costs and operational safety hazards, and achieving non-destructive observation.

[0039] Figure 6 This is a flowchart illustrating a method for rapidly obtaining intertidal algal biomass according to an exemplary embodiment. Figure 6 As shown, the method includes the following steps: Step 201: Align the image acquisition area with the preset sample plot area and acquire images of seaweed within the sample plot.

[0040] In this embodiment of the invention, the image acquisition area can be aligned with the preset sample plot area by the positioning calibration module, and the image acquisition module can be used to acquire images of seaweed in the sample plot.

[0041] Step 202: Call the deep learning segmentation model to perform pixel-level segmentation and species identification of the seaweed image.

[0042] In this embodiment of the invention, specifically, effective seaweed images can be preprocessed (including at least one of grayscale conversion, denoising, and image enhancement), and then a pre-trained deep learning segmentation model can be called to perform pixel-level segmentation of the seaweed in the image, distinguishing between different types of seaweed-covered areas and non-seaweed backgrounds (mud, rocks, etc.).

[0043] Step 203: Calculate the coverage of various types of seaweed in the quadrat based on the preset image scale.

[0044] Step 204: Based on the coverage and the density correction coefficient K of the corresponding seaweed species in the pre-established density correction coefficient database, calculate the biomass of each type of seaweed and sum them up to obtain the total biomass of the quadrat.

[0045] It is evident that implementation Figure 6The described method for rapidly acquiring intertidal algae biomass enables rapid and accurate alignment between the image acquisition area and the preset sample plot area through the laser marking assistance function of the positioning and calibration module. After acquiring algae images within the sample plot through the image acquisition module, the data processing module calls a deep learning segmentation model to perform pixel-level segmentation and species identification on the images, and calculates the coverage of various algae based on the preset image scale. Then, based on the density correction coefficient K of the corresponding algae species in the coverage and density correction coefficient database, the biomass of various algae is calculated and accumulated to obtain the total biomass of the sample plot. This method combines on-site image acquisition with intelligent data processing, eliminating the need for on-site collection, transportation, and laboratory sorting and weighing, significantly improving survey efficiency, reducing labor costs and operational safety hazards, and achieving non-destructive observation.

[0046] It should be understood that the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A device for rapidly acquiring intertidal seaweed biomass, characterized in that, include: The positioning and calibration module is used to project laser marks onto a preset sample plot area to assist in aligning the image acquisition area with the preset sample plot area; The image acquisition module is used to acquire images of seaweed within the sample plots; The data processing module is used to call a deep learning segmentation model to perform pixel-level segmentation and species identification of the seaweed image, calculate the coverage of various seaweeds based on a preset image scale, and calculate the biomass of various seaweeds based on the coverage and the density correction coefficient K of the corresponding seaweed species in a pre-established density correction coefficient database, and sum them up to obtain the total biomass of the quadrat.

2. The apparatus according to claim 1, characterized in that, The image acquisition module is a multispectral camera used to simultaneously acquire visible light and near-infrared images within the sample plot; the data processing module is also used to register the near-infrared images and visible light images, calculate vegetation index features, and input the vegetation index features and visible light images into the deep learning segmentation model.

3. The apparatus according to claim 1, characterized in that, It also includes a flushing module and a turbidity detection module; The flushing module is used to flush away debris attached to the surface of seaweed in the sample plot using low-pressure water flow before image acquisition. The turbidity detection module is used to detect the turbidity of the residual water in the sample plot in real time after flushing. The data processing module is also used to trigger the image acquisition module to perform an image acquisition operation when the turbidity of the water body is lower than a preset turbidity threshold.

4. The apparatus according to claim 1, characterized in that, The data processing module is also used to calculate a comprehensive quality score for each frame of the acquired image, including sharpness, illumination uniformity, and seaweed occlusion rate; when the comprehensive quality score of all acquired images is lower than a preset quality threshold, a re-acquisition command is sent to the image acquisition module.

5. The apparatus according to claim 1, characterized in that, The deep learning segmentation model is a network model based on an improved version of YOLOv8.

6. The apparatus according to claim 1, characterized in that, The image acquisition module is also used to acquire multiple frames of images of the same plot at a preset angle sequence; the data processing module is also used to run the motion recovery structure algorithm based on the multiple frames of images to generate a three-dimensional point cloud of the seaweed community in the plot, extract the canopy height of various seaweeds, and introduce the thickness correction function of the canopy height when calculating the biomass of various seaweeds. When the canopy height is zero, the value of the thickness correction function is 1.

7. The apparatus according to claim 1, characterized in that, The density correction coefficient K is obtained by simultaneously acquiring the measured wet weight and image recognition area of ​​various algae in the sample plot under different intertidal zone zoning and seasonal conditions, and then performing linear fitting.

8. A method for rapidly obtaining intertidal algal biomass, characterized in that, include: Align the image acquisition area with the preset sample plot area and acquire images of seaweed within the sample plot; A deep learning segmentation model is invoked to perform pixel-level segmentation and species identification on the seaweed image; Based on a preset image scale, the coverage of various types of seaweed in the quadrat is calculated. Based on the coverage and the density correction coefficient K of the corresponding seaweed species in the pre-established density correction coefficient database, the biomass of each type of seaweed is calculated and summed to obtain the total biomass of the quadrat.

Images