Kelp leaf surface area determination method and application thereof

Abstract

The invention provides a method for measuring the surface area of a kelp leaf and application thereof, and relates to the technical field of measurement of the area of the kelp leaf. The measuring method comprises the following steps: (1) shooting original data; (2) preliminary cutting and preprocessing of original data; (3) calibration plate detection; (4) kelp contour detection; and (5) calculating the basic phenotype of the kelp. According to the measuring method, the SAM model based on deep learning is adopted to segment the image, and the error can be close to or even smaller than that of manual measurement.

Description

Technical Field

[0001] The invention belongs to the technical field of seaweed leaf area measurement, and in particular relates to a method for measuring the surface area of ​​kelp leaves and an application thereof. Background Art

[0002] Kelp is a nutritious marine edible vegetable, rich in iron, calcium, protein, vitamins and other nutrients needed by the human body. It is one of the main agricultural varieties in my country and one of the pillar industries of my country's marine aquaculture. In aquaculture practice, understanding the morphological structure and organizational structure of kelp helps to optimize aquaculture conditions to improve aquaculture efficiency and yield. Therefore, in-depth research on the morphological structure and organizational structure of kelp can provide important and precise theoretical basis and practical guidance for the cultivation and artificial breeding of kelp.

[0003] The morphological structure of kelp algae is divided into three parts: the holdfast, the stalk and the blade. The blade is the main part of its nutrient absorption and photosynthesis, and the size of the blade area is extremely important for the cultivation and breeding of kelp. First, the size of the kelp blade area affects its ability to absorb nutrients. A large blade area can more effectively absorb nutrients such as nitrogen and phosphorus in seawater, thereby supporting its rapid growth. Secondly, the blade contains a large number of chloroplasts, which is the main place for photosynthesis. The larger the blade area, the more light energy is absorbed, which promotes the synthesis of more organic matter and increases the growth rate and biomass of kelp. Then, the leaf area of ​​kelp is closely related to the breeding density and breeding method. A reasonable leaf area ensures that there is appropriate space between kelp, avoiding light and nutrient competition caused by too dense, which affects the overall breeding effect. Finally, the size of the leaf area also determines the harvest area of ​​kelp, which directly affects the yield and economic benefits of breeding.

[0004] Chinese patent CN115876127A discloses a plant leaf area measurement method based on AI image recognition, the measurement method includes: S1, collecting leaf images on a white base plate; S2, using image recognition software to identify the edge contour of the leaf and calculate core data, the core data includes the maximum width of the leaf, the maximum length of the leaf and the overall area of ​​the leaf; the specific content of S2 is: S21, converting the leaf image from RGB mode to a grayscale image; S22, Gaussian blurring the grayscale image; S23, performing threshold binarization processing on the grayscale image after Gaussian blurring to obtain a segmented image and image processing data; S24, determining the surrounding relationship of the boundary of the leaf segmentation image and extracting the contour and calculating the maximum width of the leaf, the maximum length of the leaf and the overall area of ​​the leaf. This measurement method involves data export and has complex steps.

[0005] Wang Jing, Zhang Qingquan, Yang Peilin. Research on the measurement method of plant leaf area based on digital image processing [J]. Journal of Shanxi Normal University (Natural Science Edition), 2014, 028(003): 49-52. A plant leaf area measurement scheme based on digital image processing is disclosed. The image enhancement and image segmentation methods are used to process the plant leaf photos containing reference objects, and the edge detection is used to obtain the boundary information of the plant leaf image and the pixel comparison method is used to calculate the leaf area. This measurement method does not consider the influence of factors such as uneven illumination, leaf bending, and shooting angle on the results during image acquisition.

[0006] In view of this, in order to solve the deficiencies of the prior art, the present invention provides a method for measuring the surface area of ​​kelp leaves and its application. Summary of the invention

[0007] The purpose of the invention is to overcome the shortcomings in previous studies and provide a method for measuring the surface area of ​​kelp leaves and an application thereof.

[0008] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0009] In one aspect, the present invention provides a method for determining the surface area of ​​kelp leaves, comprising the following steps:

[0010] (1) Shooting of original data: Lay the calibration plate and kelp leaves flat, then take photos of the calibration plate and kelp leaves to obtain image data;

[0011] (2) Preliminary cropping and preprocessing of raw data: The image data obtained in step (1) is saved to the server, and the position containing the kelp leaves is preliminarily cropped using an image cropping tool, and then the preliminarily cropped image is read, the image is converted into a color space, the brightness channel is enhanced, and the gray world algorithm is used to adjust the white balance of the image;

[0012] (3) Calibration plate detection: Using the pyramid template matching method, the calibration plate is used as the calculation template, and the preprocessed kelp leaf image is used as the target image. After multiple matching, the best matching area is located, and then the width, height and surface area of ​​the calibration plate are calculated based on the external frame information of the calibration plate.

[0013] (4) Kelp contour detection: A real-time semantic segmentation workbench for kelp leaves based on prompt points is constructed. By clicking the position of the kelp leaves, foreground prompt points are provided to the Segment Anything Model (SAM). The SAM model detects the kelp leaf mask based on the foreground prompt points, obtains the kelp leaf mask area, and calculates the outer bounding box of the kelp leaf mask to obtain the width and height.

[0014] (5) Calculation of basic phenotype of kelp: The width, height and surface area of ​​the kelp leaf are calculated by using the width, height and surface area of ​​the calibration plate calculated in step (3), the kelp leaf mask area obtained in step (4), the width and height of the outer frame calculated for the kelp leaf mask, and the actual width, actual height and actual area of ​​the calibration plate to calculate the width, height and surface area of ​​the kelp leaf.

[0015] Preferably, in step (1), the calibration plate is a black and white calibration plate, wherein the number of black and white grids is 7×5, and each grid is 3 cm×3 cm.

[0016] Preferably, in step (1), the device for taking photos is a smart phone with a pixel size of 4608×3456.

[0017] Further preferably, in step (1), the parameters for taking pictures are: sensitivity (ISO) is 80, shutter speed is 1 / 100 second, exposure compensation (EV) is 0, white balance is set to automatic, and the width and height of the image are 4624×2600 pixels respectively.

[0018] Preferably, in step (2), the raw data reading parameters are as follows: the color space is LAB color space, the brightness channel enhancement method is CLAHE, and the white balance adjustment algorithm is gray world.

[0019] Preferably, in step (3), the specific steps of the pyramid template matching method are as follows: using the calibration plate as the calculation template, using the preprocessed kelp leaf image as the target image, sliding the calculation template in the target image, and locating the target object by calculating the similarity between the template and each area of ​​the target image.

[0020] Further preferably, in step (3), locating the target object by calculating the similarity between the template and each area of ​​the target image refers to: using grayscale image processing, generating a similarity score map during matching, scaling the calculated template multiple times and performing multiple matches, and locating the best matching area from the maximum value of the similarity scores of the multiple matches.

[0021] Preferably, step (4) also includes the steps of removing non-kelp leaf background and binarizing the kelp leaf mask area.

[0022] Preferably, in step (5), the formula for calculating the width of the kelp blade is as follows:

[0023]

[0024] Among them, w t is the width of the kelp leaf;

[0025] w k The width of the outer frame calculated for the kelp leaf mask in step (4);

[0026] w c is the width of the calibration plate calculated in step (3);

[0027] w r is the actual width of the calibration plate.

[0028] Preferably, in step (5), the formula for calculating the height of kelp leaves is as follows:

[0029]

[0030] Among them, h t is the height of the kelp blade;

[0031] h k The height obtained by calculating the outer frame of the kelp leaf mask in step (4);

[0032] h c is the width of the calibration plate calculated in step (3);

[0033] h r is the actual width of the calibration plate.

[0034] Preferably, in step (5), the formula for calculating the surface area of ​​kelp leaves is as follows:

[0035]

[0036] Among them, S t is the surface area of ​​kelp leaves;

[0037] S k is the kelp leaf mask area obtained in step (4);

[0038] S c is the surface area of ​​the calibration plate calculated in step (3);

[0039] S r is the actual area of ​​the calibration plate.

[0040] In another aspect, the present invention provides an application of the above determination method in the process of measuring seaweed leaf area. Compared with the prior art, the present invention has the following beneficial effects:

[0041] The present invention discloses a method for measuring the surface area of ​​kelp algae leaves. The method adopts a SAM model based on deep learning to segment images, and can achieve an error close to or even less than that of manual measurement. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is the image data in .jpg format obtained in step (1) in Example 1.

[0043] Figure 2 This is the image data after preliminary cropping in step (2) of Example 1.

[0044] Figure 3 It is the image data after preprocessing in step (2) in Example 1.

[0045] Figure 4 This is the calibration plate detection result of step (3) in Example 1, where the green box in the upper right corner is the detected calibration plate.

[0046] Figure 5 This is the workbench interface for real-time semantic segmentation of kelp based on prompt points in step (4) of Example 1.

[0047] Figure 6 This is the kelp image detected in step (4) of Example 1, wherein the highlighted portion is the kelp.

[0048] Figure 7 This is the image without the non-kelp background removed in step (4) of Example 1.

[0049] Figure 8 This is the kelp mask area image obtained in step (4) in Example 1. DETAILED DESCRIPTION

[0050] The following non-limiting examples can enable those skilled in the art to more fully understand the present invention, but do not limit the present invention in any way. The following content is merely an exemplary description of the scope of protection claimed in this application, and those skilled in the art can make various changes and modifications to the invention of this application based on the disclosed content, which should also fall within the scope of protection claimed in this application.

[0051] The present invention is further described below by way of specific examples. Unless otherwise specified, the various drugs used in the examples of the present invention are obtained through conventional commercial channels.

[0052] Experimental Materials

[0053] The experimental materials include two kelp varieties, “Sanhai” and “Haijia No. 1”. The sampling information is shown in Table 1.

[0054] Table 1 Sampling information table

[0055]

[0056]

[0057] Example 1

[0058] (1) Shooting of raw data: First, make a calibration plate with a known area (7×5 black and white grids, each grid is 3cm×3cm) and fix it on a white background of a 108W flat lamp; then, spread the kelp on the 108W flat lamp, try to keep the leaves horizontal, and keep the kelp stretched and straight as much as possible; finally, use a smartphone with a pixel of 4608×3456 to take a photo of the kelp leaves, ensuring that the calibration plate and the leaves are in the same plane, and the shooting angle is perpendicular to the background plate, and obtain image data in .jpg format. Use a smart camera to shoot under natural light conditions. In order to ensure the same results each time, the shooting parameters are: sensitivity (ISO) is 80, shutter speed is 1 / 100 second, exposure compensation (EV) is 0, white balance is set to automatic, and the width and height of the image are 4624×2600 pixels respectively.

[0059] (2) Preliminary cropping and preprocessing of raw data: Preliminary cropping is to remove redundant background, and preprocessing is to improve images in dark environments or with severe color deviation, providing a better foundation for subsequent image processing. The captured image is saved to the server, and the image cropping tool OpenCV is used to crop out the area containing the kelp and remove the edge background such as the backlight board. The result is as follows: Figure 2 Then read the cropped image, convert the image to LAB color space, use the CLAHE method to enhance the brightness channel, and use the gray world algorithm to adjust the white balance of the image. The result is as follows Figure 3 shown.

[0060] (3) Calibration plate detection: The calibration plate is detected using the pyramid template matching method to adapt to the problem of target size changes. The detection of the surface area, width and height of kelp uses a chessboard as the calibration plate. The width of the chessboard is 37 cm and the height is 16 cm. Template matching locates the target object by sliding a calibration plate template in the preprocessed kelp image and calculating the similarity between the template and each area of ​​the target image. The matching process uses grayscale image processing, and a similarity score map is generated during matching. The template is scaled to 0.5 to 2 times the size of the original template and multiple matches are performed. The best matching area is located from the maximum value of the similarity scores of the multiple matches. After obtaining the best matching area, the width, height and surface area of ​​the calibration plate are calculated based on the external frame information of the calibration plate for subsequent calculation of the basic kelp phenotype. The results of the calibration plate detection are shown as follows. Figure 4 shown.

[0061] (4) Kelp contour detection: Use a cue point-based method for real-time semantic segmentation. The SAM model is used as the baseline model. The model accepts cue points and generates accurate segmentation results in combination with image context. SAM is a Transformer-driven segmentation model that guides segmentation tasks through multimodal cue mechanisms (such as points, boxes, text, etc.). Use the PyQT tool to build a kelp real-time semantic segmentation workbench based on cue points. The workbench interface is as follows: Figure 5 As shown in the figure, by clicking on the kelp position, the SAM model is provided with a foreground cue point, and SAM detects the kelp mask based on the cue point. Usually only 1 to 2 cue points are needed to complete the detection, and the detected kelp will be highlighted, as shown in the figure below. Figure 6 Finally, the non-kelp background is removed according to the detection results, as shown in Figure 7 As shown, the binary kelp mask area is obtained, as Figure 8 shown.

[0062] (5) Calculation of the basic phenotype of kelp. The width w of the calibration plate detected in step (3) c , height h c and surface area s c , and the actual width w of the calibration plate r , actual height h r and the actual area s r The mask area s of the kelp detected by step (4) k , calculate the outer frame of the mask to get the width w k , height h k The actual width of the kelp w t , height h t and surface area s t It can be calculated by the following formula:

[0063]

[0064] Comparative Example 1

[0065] Manual measurement

[0066] The length and width of the kelp leaf were measured with a ruler with an accuracy of 0.1 cm. The leaf width was measured every 10 cm from the base of the algae and recorded as A1, A2, A3, ... A n , the last interval is recorded as H (<10cm); the base is regarded as a triangle. Follow the following calculation formula (Note: the experimental samples were trimmed at the tip and edge of the algae):

[0067] S 叶片表面积 =A1*10 / 2+(A1+A2)*10 / 2+(A2+A3)*10 / 2+……+(A n-1 +A n)*H / 2.

[0068] Comparative Example 2

[0069] Fragility measurement

[0070] Divide the kelp leaf into several regular blocks, such as squares or rectangles, and then use a ruler to measure the area of ​​each block. This method can quickly and accurately obtain the area of ​​each block, and then calculate the total area of ​​the entire leaf, which is recorded as the true area.

[0071] Measurement results

[0072] Example 1

[0073] The surface area of ​​the leaves of the “Sanhai” kelp sample in this experiment is 2031.53±173.69cm 2 ; The surface area of ​​the kelp leaf of “Haijia No. 1” is 2572.94±780.48cm 2 The total sample surface area (the average surface area of ​​all kelp samples) was 2302.24 ± 621.79 cm 2 , the specific results are shown in Table 2.

[0074] Table 2 Surface area of ​​kelp leaves measured in Example 1

[0075]

[0076]

[0077] Comparative Example 1

[0078] The manual measurement results show that the surface area of ​​the leaves of the “Sanhai” kelp sample in this experiment is 2056.38±179.73cm 2 ; The surface area of ​​the kelp leaf of “Haijia No. 1” is 2081.40±348.50cm 2 The total sample surface area (the average surface area of ​​all kelp samples) was 2068.89 ± 273.98 cm 2 , the specific results are shown in Table 3.

[0079] Table 3 Surface area of ​​kelp leaves measured in Comparative Example 1

[0080]

[0081] Comparative Example 2

[0082] Three kelp leaves of each kelp variety were randomly selected for crushing measurement, and the surface area of ​​the leaves of the “Sanhai” kelp was 2107.67±74.23cm 2 ; The surface area of ​​the kelp leaf of “Haijia No. 1” is 2300.00±225.65cm 2The total sample surface area (the average surface area of ​​all kelp samples) was 2203.83 ± 167.50 cm 2 , the specific results are shown in Table 4.

[0083] Table 4 Surface area of ​​kelp leaves measured in comparative example 2

[0084]

[0085] Results and Discussion

[0086] Surface area of ​​“Three Seas” kelp leaves:

[0087] The relative error of Comparative Example 1 is -6.31%, and the relative error of Example 1 is -7.24%. It can be seen that the blade surface areas measured by Comparative Example 1 and Example 1 are relatively close to the actual measured value (Comparative Example 2).

[0088] Surface area of ​​kelp leaves of “Haijia No. 1”:

[0089] The relative error of Comparative Example 1 is -14.49%, while the relative error of Example 1 is only 0.59%. It can be seen that the measurement result of Example 1 is more accurate.

[0090] Therefore, this measurement method uses the SAM model based on deep learning to segment the image, which can achieve an error close to or even less than manual measurement, see Table 5 for details.

[0091] Table 5 Evaluation of blade surface area measurement results

[0092]

[0093]

[0094] Finally, it should be noted that the above content is only used to illustrate the technical solution of the present invention, rather than to limit the scope of protection of the present invention. Simple modifications or equivalent substitutions of the technical solution of the present invention by ordinary technicians in this field do not deviate from the essence and scope of the technical solution of the present invention.

Claims

1. A method for measuring the surface area of ​​kelp leaves, characterized in that: The following steps are involved: (1) Shooting of original data: Lay the calibration plate and kelp leaves flat, then take photos of the calibration plate and kelp leaves to obtain image data; (2) Preliminary cropping and preprocessing of raw data: The image data obtained in step (1) is saved to the server, and the position containing the kelp leaves is preliminarily cropped using an image cropping tool, and then the preliminarily cropped image is read, the image is converted into a color space, the brightness channel is enhanced, and the gray world algorithm is used to adjust the white balance of the image; (3) Calibration plate detection: Using the pyramid template matching method, the calibration plate is used as the calculation template, and the preprocessed kelp leaf image is used as the target image. After multiple matching, the best matching area is located, and then the width, height and surface area of ​​the calibration plate are calculated based on the external frame information of the calibration plate. (4) Kelp contour detection: A real-time semantic segmentation workbench for kelp leaves based on prompt points is constructed. By clicking the position of the kelp leaves, foreground prompt points are provided to the SAM model. The SAM model detects the kelp leaf mask based on the foreground prompt points, obtains the kelp leaf mask area, and calculates the outer bounding box of the kelp leaf mask to obtain the width and height. (5) Calculation of basic phenotype of kelp: The width, height and surface area of ​​the kelp leaf are calculated by using the width, height and surface area of ​​the calibration plate calculated in step (3), the kelp leaf mask area obtained in step (4), the width and height of the outer frame calculated for the kelp leaf mask, and the actual width, actual height and actual area of ​​the calibration plate to calculate the width, height and surface area of ​​the kelp leaf.

2. The measuring method according to claim 1, characterized in that In step (1), the parameters for taking photos are: sensitivity (ISO) is 80, shutter speed is 1 / 100 second, exposure compensation (EV) is 0, white balance is set to automatic, and the width and height of the image are 4624×2600 pixels respectively.

3. The measuring method according to claim 1, characterized in that In step (2), the raw data reading parameters are as follows: the color space is LAB color space, the brightness channel enhancement method is CLAHE, and the white balance adjustment algorithm is gray world.

4. The measuring method according to claim 1, characterized in that In step (3), the specific steps of the pyramid template matching method are as follows: using the calibration plate as the calculation template, using the preprocessed kelp leaf image as the target image, sliding the calculation template in the target image, and locating the target object by calculating the similarity between the template and each area of ​​the target image.

5. The measuring method according to claim 4, characterized in that In step (3), locating the target object by calculating the similarity between the template and each area of ​​the target image means: using grayscale image processing, generating a similarity score map during matching, scaling the calculated template multiple times and performing multiple matches, and locating the best matching area from the maximum value of the similarity scores of the multiple matches.

6. The measuring method according to claim 1, characterized in that Step (4) also includes the steps of removing the non-kelp leaf background and binarizing the kelp leaf mask area.

7. The measuring method according to claim 1, characterized in that In step (5), the formula for calculating the width of the kelp blade is as follows: Among them, w t is the width of the kelp leaf; w k The width of the outer frame calculated for the kelp leaf mask in step (4); w c is the width of the calibration plate calculated in step (3); w r is the actual width of the calibration plate.

8. The measuring method according to claim 1, characterized in that In step (5), the formula for calculating the height of the kelp blade is as follows: Among them, h t is the height of kelp leaves; h k The height obtained by calculating the outer frame of the kelp leaf mask in step (4); h c is the width of the calibration plate calculated in step (3); h r is the actual width of the calibration plate.

9. The measuring method according to claim 1, characterized in that In step (5), the formula for calculating the surface area of ​​kelp leaves is as follows: Among them, S t is the surface area of ​​kelp leaves; S k is the kelp leaf mask area obtained in step (4); S c is the surface area of ​​the calibration plate calculated in step (3); S r is the actual area of ​​the calibration plate.

10. Use of the determination method according to any one of claims 1 to 9 in the process of measuring seaweed leaf area.

Images