A method and device for measuring biomass in a sample frame, a medium and an electronic device

Abstract

The application provides a method and device for measuring biomass in a sample frame, a medium and an electronic device. The application can highlight the plant features in the original image including the sample frame by using a near-infrared camera, so as to facilitate the identification of plants in the image during image recognition. The height of plants in the sample frame and the coverage in the sample frame are obtained from the original image, and then the height and the coverage are applied to a biomass model to obtain the biomass in the sample frame. The embodiments of the application automatically measure the biomass in the sample frame above the grassland monitoring point by means of artificial intelligence, improve the monitoring efficiency, and meet the demand for measuring biomass on a large scale.

Description

Technical Field

[0001] This application relates to the field of agricultural technology, and more specifically, to a method, apparatus, medium, and electronic equipment for measuring biomass within a quadrat. Background Technology

[0002] Timely and effective access to information on grassland resource status, ecological condition, and utilization is crucial for improving the level of refined grassland management and is an important part of the country's development of high-level agriculture. It can also provide data support for comprehensively deepening the reform of the grassland ecological civilization system.

[0003] Biomass refers to the total amount of living organic matter (dry or wet weight) per unit area at a given moment (including the weight of food stored within organisms), measured in kg / m². 2 or t / hm 2 This indicates that, for the needs of agricultural and pastoral production and daily life, surveys and statistics on aboveground plant biomass in grasslands are frequently conducted to determine the proportion of biomass of various plant communities in the total biomass within a sample plot.

[0004] Currently, the method for measuring biomass involves setting up quadrats at selected locations. These quadrats are square areas enclosed by a single-color material (e.g., white), creating a striking contrast with the green of the grassland for easy identification. Biomass within the quadrats is then estimated manually. However, this method becomes very inefficient and unsuitable for large-scale measurements when multiple quadrats need to be set up at different locations (e.g., every 50 meters) across a grassland.

[0005] Therefore, this application provides a method for measuring biomass within a quadrat to solve one of the aforementioned technical problems. Summary of the Invention

[0006] The purpose of this application is to provide a method, apparatus, medium, and electronic device for measuring biomass within a quadrat frame, which can solve at least one of the aforementioned technical problems. The specific solution is as follows:

[0007] According to a specific embodiment of this application, in a first aspect, this application provides a method for measuring biomass within a quadrat frame, comprising:

[0008] Raw images are acquired using near-infrared cameras at grassland monitoring points, wherein the raw images include images of pre-defined sample boxes;

[0009] The height of the plants within the quadrat frame and the canopy coverage within the quadrat frame are obtained based on the original image.

[0010] The height and the cover are applied to the biomass model to obtain the biomass within the quadrat.

[0011] According to a specific embodiment of this application, in a second aspect, this application provides a device for measuring biomass within a quadrat frame, comprising:

[0012] The acquisition unit is used to acquire raw images through a near-infrared camera at a grassland monitoring point, wherein the raw images include images of preset sample boxes;

[0013] Analysis unit, used to obtain the height of the plants within the quadrat frame and the canopy coverage within the quadrat frame based on the original image;

[0014] A calculation unit is used to apply the height and the cover to a biomass model to obtain the biomass within the quadrat frame.

[0015] According to a specific embodiment of this application, in a third aspect, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for measuring biomass within a quadrat frame as described in any of the preceding claims.

[0016] According to a specific embodiment of this application, in a fourth aspect, this application provides an electronic device, including: one or more processors; and a storage device for storing one or more programs, which, when executed by the one or more processors, cause the one or more processors to implement the method for measuring biomass within a quadrat frame as described in any of the preceding claims.

[0017] Compared with the prior art, the above-described solutions of this application have at least the following beneficial effects:

[0018] This application provides a method, apparatus, medium, and electronic device for measuring biomass within a quadrat frame. This application uses a near-infrared camera to acquire original images including the quadrat frame, which better highlights plant features and facilitates plant identification during image recognition. The height of the plants within the quadrat frame and the canopy cover are obtained from the original image, and then the height and canopy cover are applied to a biomass model to obtain the biomass within the quadrat frame. Embodiments of this application automatically measure the biomass within the quadrat frame above grassland monitoring points using artificial intelligence, improving monitoring efficiency and suitable for large-scale biomass measurement needs. Attached Figure Description

[0019] Figure 1 A flowchart illustrating a method for measuring biomass within a quadrat frame according to an embodiment of this application is shown.

[0020] Figure 2 A schematic diagram of a grassland monitoring point according to an embodiment of this application is shown;

[0021] Figure 3 A schematic diagram of a sample block according to an embodiment of this application is shown;

[0022] Figure 4 A schematic diagram of the structure of the coverage classification model according to an embodiment of this application is shown;

[0023] Figure 5 A schematic diagram showing sample images according to an embodiment of this application is provided;

[0024] Figure 6 A schematic diagram of fitting a wet weight model according to an embodiment of this application is shown;

[0025] Figure 7 A unit block diagram of a biomass measurement device within a quadrat frame according to an embodiment of this application is shown;

[0026] Explanation of reference numerals in the attached figures

[0027] 1-Near-infrared camera, 2-Height measuring rod, 3-Square frame, 4-Plant, 5-Sample image. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0029] The terminology used in the embodiments of this application is for the purpose of describing particular embodiments only and is not intended to limit the application. The singular forms “a,” “said,” and “the” used in the embodiments of this application and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0030] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0031] It should be understood that although the terms first, second, third, etc., may be used in the embodiments of this application, these descriptions should not be limited to these terms. These terms are only used to distinguish the descriptions. For example, first may also be referred to as second without departing from the scope of the embodiments of this application, and similarly, second may also be referred to as first.

[0032] Depending on the context, the words “if” or “suppose” as used here can be interpreted as “when” or “in response to determination” or “in response to detection.” Similarly, depending on the context, the phrases “if determination” or “if detection (of the stated condition or event)” can be interpreted as “when determination” or “in response to determination” or “when detection (of the stated condition or event)” or “in response to detection (of the stated condition or event).”

[0033] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that an article or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such an article or device. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the article or device that includes said element.

[0034] It should be noted that any symbols and / or numbers present in the specification that are not marked in the accompanying drawings are not reference numerals.

[0035] The optional embodiments of this application are described in detail below with reference to the accompanying drawings.

[0036] Example 1

[0037] The embodiments provided in this application are embodiments of a method for measuring biomass within a quadrat frame.

[0038] The following is combined with Figure 1 The embodiments of this application will be described in detail.

[0039] Step S101: Acquire the original image using the near-infrared camera 1 at the grassland monitoring point.

[0040] In this application, the selection of grassland monitoring sites aims to maximize the variety of grassland samples, ensuring they represent the local grassland types and the growth stages of as many plants as possible. Figure 2 As shown, a near-infrared camera 1, an elevation measuring rod 2, and a quadrat frame 3 were set up at the grassland monitoring point.

[0041] like Figure 3 As shown, quadrat 3 is composed of squares of a single color. For example, since grassland is predominantly green, quadrat 3 is white, creating a strong contrast with the grassland color and facilitating image recognition. The grassland samples within quadrat 3 should be as diverse as possible to represent the local grassland type and the growth stages of as many plants 4 as possible. For example, quadrat 3 may include: stones, animal excrement, and dry or rain-soaked ground.

[0042] Near-infrared camera 1 is mounted on a tall pole to take images of sample frame 3 from above, and to make sample frame 3 as square as possible in the image.

[0043] The height measuring scale 2 is set inside the quadrat frame 3. The image captured by the near-infrared camera 1 includes the image of the height measuring scale 2. By observing the positional relationship between the plant 4 and the scale on the height measuring scale 2 in the image, the height of the plant 4 can be determined.

[0044] Near-infrared (NIR) light is electromagnetic radiation between visible light (VIS) and mid-infrared (MIR), referring to electromagnetic waves with wavelengths ranging from 780 to 2526 nm. It includes both short-wave (780–1100 nm) and long-wave (1100–2526 nm) near-infrared regions. The near-infrared region was the first non-visible light region discovered. Near-infrared light includes cosmic rays and gamma rays.

[0045] Images captured by a camera include multiple color channels, which determine the colors in the image. The number of color channels in an image depends on its color mode; that is, the color mode of an image determines the number of its color channels. For example, RGB and Lab color modes have 3 color channels, while CMYK color modes have 4 color channels. In this embodiment, near-infrared light occupies any color channel in the color mode to highlight near-infrared light information in the image.

[0046] Because healthy plants 4 contain abundant chlorophyll, they reflect more near-infrared and green light while absorbing more red and blue light. In some specific implementations, the near-infrared camera 1 uses the RGB image color mode, with three color channels: a near-infrared light channel, a green channel, and a red channel. The near-infrared light channel is used to represent near-infrared light information in the image, the green channel is used to represent green light information, and the red channel is used to represent red light information. In other words, the near-infrared light channel occupies the blue channel in the RGB image color mode. Therefore, the original image acquired by the near-infrared camera 1 includes a fusion of near-infrared, green, and red light information. Of course, this application is not limited to this; the near-infrared light channel can also occupy the red channel in the RGB image color mode, and the original image includes near-infrared, green, and blue light information. The original image obtained by the near-infrared camera 1 can better highlight the characteristics of the plant 4, making it easier to identify the plant 4 in the image during image recognition.

[0047] The original image includes an image of a preset voxel frame 3 and a target image within the voxel frame 3. The target image includes images of various plants 4, and the various plants 4 include at least the target plant 4. By using a target detection model (such as the YOLOv algorithm model) to perform image recognition on the original image, the target image within the voxel frame 3 can be obtained.

[0048] The target plant 4 refers to plants beneficial to agricultural and pastoral production and daily life, such as crested wheatgrass, ice grass, and alfalfa. The target plant 4 is related to the coverage of the plants 4 within the measurement quadrat 3. However, some non-target plants 4 that are not beneficial to agricultural and pastoral production and daily life are not related to the coverage of the plants 4 within the measurement quadrat 3, such as wolfsbane.

[0049] The target image refers to the image within the quadratic frame 3. The target image includes at least an image of the target plant 4, and may also include images of non-target plants 4.

[0050] Step S102: Based on the original image, obtain the height of the plant 4 within the quadrat 3 and the coverage within the quadrat 3.

[0051] In some specific embodiments, the original image also includes the target image within the quadratic frame 3 and the scale image of the height measuring scale 2 set within the quadratic frame 3. The target image includes images of various plants 4, and the various plants 4 include the target plant.

[0052] Accordingly, obtaining the height of the plant 4 within the sample box 3 based on the original image includes the following steps:

[0053] Step S102a: In the original image, based on the positional relationship between the images of the various plants 4 and the scale in the ruler image, determine the height of the plant 4 within the quadratic frame 3.

[0054] In some specific embodiments, obtaining the plant cover 4 within the quadratic frame 3 based on the original image includes the following steps:

[0055] Step S102b-1: Based on the target image in the original image and combined with the preset latitude and longitude information and preset altitude information of the grassland monitoring point, obtain the grassland type within the sample box 3.

[0056] This application utilizes image classification models (such as the ResNet50 model) to perform grassland type analysis on target images.

[0057] In some specific implementations, the grassland types include: temperate grassland type, alpine grassland type, temperate desert type, alpine desert type, warm shrub-grassland type, tropical shrub-grassland type, lowland meadow type, mountain meadow type and / or alpine meadow type.

[0058] Since information such as temperate, alpine, warm, hot, lowland, and mountainous grassland types is closely related to the geographical location and altitude of the grassland monitoring point, this application pre-sets the latitude, longitude, and altitude information of the grassland monitoring point to assist in determining the grassland type. For example, the pre-set latitude and longitude information of the grassland monitoring point is 99°58'E, 25°36'N, and the pre-set altitude information is 2800m; that is, the grassland monitoring point is located in Dali on the Yunnan-Guizhou Plateau; the obtained grassland type is temperate grassland.

[0059] Step S102b-2: Determine the parameter values ​​of multiple model parameters in the cover classification model based on the grassland type.

[0060] The coverage classification model includes a trained convolutional neural network model.

[0061] A Convolutional Neural Network (CNN) is a feedforward neural network model composed of multidimensional neurons with learnable weights and constant biases. Each neuron receives some input and performs some dot product calculations, outputting a score for each category. The artificial neurons in a CNN can respond to a portion of the surrounding units within their coverage area. Each neuron includes multiple dimensions of model parameters, such as width, height, and depth, as well as kernel parameters (e.g., convolution kernel parameters, pooling kernel parameters), stride parameters, node parameters, and probability parameters.

[0062] A CNN is a deep learning model, and only a trained CNN can be used. During training, a large number of historical target images are used to train the CNN until the trained CNN can recognize historical target images with a preset accuracy. The training process of the CNN is not detailed in this embodiment; various implementation methods in the prior art can be referred to.

[0063] CNNs include: googleNet model, ResNet model, VGG-16 model, AlexNet model, and LeNet-5 model.

[0064] In some specific implementations, the coverage classification model includes at least two convolutional layers for input sample image 5 and one fully connected layer after the two convolutional layers; each convolutional layer is followed by a pooling layer, and each pooling layer is followed by a normalization layer; the fully connected layer is followed by a dropout layer, and the dropout layer is followed by a Softmax layer that outputs the classification result.

[0065] A convolutional layer consists of several convolutional units, and the operational parameters of each convolutional unit are optimized using the backpropagation algorithm. The purpose of convolution is to extract different features from the input. The two convolutional layers include a first convolutional layer and a second convolutional layer. The first convolutional layer is used to extract low-level features of the target plant, such as its edges, contour lines, and angles. The second convolutional layer is used to extract the biological features of the target plant.

[0066] A fully connected layer is a layer in which every node is connected to all nodes in the previous layer, used to combine the extracted features for classification. Due to its fully connected nature, a fully connected layer generally has the most computational parameters.

[0067] Pooling layers are a form of downsampling. They use a non-linear pooling function to divide the input image into several rectangular regions and output the maximum value for each sub-region. By continuously reducing the spatial size of the data, the number of computational parameters and the computational cost decrease, thus controlling overfitting. They are used to evaluate the performance of fully connected layers.

[0068] The normalization layer is used to perform local subtraction and division normalization in the feature map of sample image 5, forcing adjacent features in the feature map to compete locally, so that features at the same spatial location in different feature maps compete, thereby achieving feature normalization.

[0069] Dropout layers are used to avoid overfitting in CNNs by randomly updating the parameter values ​​in each iteration, thus increasing the network's versatility.

[0070] The Softmax layer is used for multi-class classification. It maps the data output by multiple neurons to probability values ​​in the interval (0, 1), where the class with the highest probability value is the coverage type of the target plant in the sample image 5 input to the CNN. In some specific implementations, the coverage types output by the Softmax layer include: full coverage, partial coverage, and no coverage.

[0071] The structure of the cover classification model used in this specific embodiment can both obtain the cover type of the target plant in sample image 5 and ensure the efficiency of data processing.

[0072] In some specific implementations, the preset latitude and longitude information of the Yunnan-Guizhou Plateau includes at least 100°–111° east longitude and 22°–30° north latitude; the preset altitude information includes at least 2km–4km altitude.

[0073] Furthermore, in some specific implementations, the target image is segmented and sampled based on a preset sample specification to obtain multiple sample images 5. The specifications of each sample image 5 meet the preset sample specifications, for example, the preset sample specifications include 64 pixels × 64 pixels. The preset sample specifications are set according to the channel specifications of the convolutional neural network.

[0074] like Figure 4 As shown, the two convolutional layers include a first convolutional layer and a second convolutional layer. The kernel parameters of the first convolutional layer are 5×5 and the neuron description parameters are 64. The kernel parameters of the second convolutional layer are 3×3 and the neuron description parameters are 128.

[0075] Furthermore, in some specific implementations, the first convolutional layer is followed by the first pooling layer, and the second convolutional layer is followed by the second pooling layer.

[0076] The pooling kernel parameter of the first pooling layer has a value of 2×2 and the step parameter has a value of 2, while the pooling kernel parameter of the second convolutional layer has a value of 3×3 and the step parameter has a value of 2.

[0077] Furthermore, in some specific implementations, the node parameter value of the fully connected layer is 192, the node parameter value of the dropout layer is 192 and the probability value of the probability parameter is 0.73 to 0.87, and the node parameter value of the Softmax layer is 6.

[0078] In this application embodiment, the parameter values ​​of multiple model parameters are determined by grassland type in order to adjust the classification characteristics of the cover classification model so that the cover classification model can adapt to the classification requirements of grassland type in the location of grassland monitoring point.

[0079] Step S102b-3: Based on the preset sample specifications, image sampling is performed on the target image to obtain multiple sample images 5, such as... Figure 5 As shown.

[0080] Since non-target plants are not plants needed for agricultural and pastoral production or daily life, they are irrelevant to the measurement of plant cover within quadrat 3. In some specific implementations, the various plants 4 within quadrat 3 also include non-target plants.

[0081] Accordingly, the target image is sampled based on a preset sample specification to obtain multiple sample images 5, including the following steps:

[0082] S102b-3-1, using a semantic segmentation model to remove images of non-target plants from the target image.

[0083] Semantic segmentation models can label each pixel in an image with its corresponding category, without distinguishing individual pixels. This can be understood as dividing a target image into various semantically interpretable categories. An example is the deeplabV3+ model.

[0084] In the cleaned target image, the pixels occupied by the original non-target plant image are marked. This means the RGB value of the pixel is assigned a specific color value. In other words, in the cleaned target image, the area occupied by the original non-target plant image is a specific color. For example, if the specific color value is zero, i.e., black, then the area occupied by the original non-target plant image in the cleaned target image is black.

[0085] S102b-3-2, Based on the preset sample specifications of the samples, image sampling is performed on the cleaned target image to obtain multiple sample images 5.

[0086] In this specific embodiment, before sampling, images of non-target plants in the target image are removed, and images of target plants and non-plants are saved. This avoids the influence of images of non-target plants on the statistical coverage and ensures the validity of the results.

[0087] Step S102b-4: Apply each sample image 5 to the coverage classification model set by the parameter values ​​of the multiple model parameters to obtain the coverage type of the corresponding sample image 5.

[0088] In some specific implementations, the coverage classification model outputs coverage types including: full coverage, partial coverage, and no coverage.

[0089] In the manual estimation of coverage within sample frame 3, when classifying the areas within the frame according to the preset sample specifications, the coverage types only include full coverage and no coverage. When partial coverage occurs, the designated areas are manually classified as full coverage or no coverage based on the coverage ratio. For example, when 1 / 3 of the designated area is covered, it is classified as no coverage; when 2 / 3 of the designated area is covered, it is classified as full coverage. This method results in significant errors in the calculated coverage. This application classifies sample images 5 with coverage less than or equal to 20% as no coverage, sample images 5 with coverage greater than 20% but less than 80% as partially covered, and sample images 5 with coverage greater than or equal to 80% as no coverage. This reduces the error in coverage and improves the accuracy of the statistics.

[0090] Step S102b-5: Analyze the multiple sample images 5 based on the grassland type and each coverage type of the multiple sample images 5 to obtain the coverage within the quadrat frame 3.

[0091] Coverage refers to the ratio of the area of ​​the vertical projection of the aboveground parts of a plant to the ground surface.

[0092] In some specific embodiments, the step of analyzing the plurality of sample images 5 based on the grassland type and each coverage type of the plurality of sample images 5 to obtain the coverage within the quadratic frame 3 includes the following steps:

[0093] Step S102b-5-1: Perform cluster analysis on the multiple sample images 5 based on each coverage type to obtain the number of sample images 5 for each coverage type.

[0094] This can be understood as counting all sample images 5 according to their coverage type to obtain the number of sample images 5 for each coverage type. For example, if there are a total of 100 sample images 5, after cluster analysis, 45 sample images 5 belong to the full coverage type, 25 sample images 5 belong to the partial coverage type, and 30 sample images 5 belong to the no-coverage type.

[0095] Step S102b-5-2: Obtain preset weight values ​​for each cover type based on the grassland type.

[0096] Because grassland types differ, the plants (4) growing on those grasslands also differ, and therefore, the methods for calculating the projection of plant (4) also differ. For example, the method for calculating thorns in a grassland differs from the method for calculating alfalfa. In this specific embodiment, preset weight values ​​are assigned to each cover type based on the grassland type to improve the accuracy of the calculation.

[0097] Step S102b-5-3: Obtain the coverage within the sample frame 3 based on the number of sample images 5 for each coverage type, the preset weight value, and the preset number of sample images 5.

[0098] Specifically, the formula for calculating the coverage within the box is as follows:

[0099]

[0100] Where C represents coverage, N1 represents the number of sample images 5 of the full coverage type, w1 represents the preset weight value of the full coverage type, N2 represents the number of sample images 5 of the partial coverage type, w2 represents the preset weight value of the partial coverage type, N3 represents the number of sample images 5 of the no-coverage type, w2 represents the preset weight value of the no-coverage type, and M represents the preset number of sample images 5.

[0101] Of course, coverage can be expressed not only by the mean, but also by the standard deviation and / or variance.

[0102] This embodiment uses a near-infrared camera 1 to acquire original images including quadrat frames 3, which better highlights the features of plants 4, facilitating plant 4 identification during image recognition. Combining the preset latitude, longitude, and altitude information of the grassland monitoring point helps determine the grassland type, taking into account the influence of geographical location on the growth cycle of plants 4 and grassland growth. The grassland type is used to determine the parameter values ​​of multiple model parameters in the cover classification model, allowing the classification characteristics of the cover classification model to be adjusted by adjusting the parameter values ​​of each model parameter. This enables the cover classification model to adapt to the classification requirements of the grassland type at the location of the grassland monitoring point, thereby ensuring the accuracy of the calculation. This embodiment uses artificial intelligence to automatically measure the cover within the quadrat frame above the grassland monitoring point, improving monitoring efficiency and making it suitable for large-scale calculations.

[0103] Step S103: Apply the height and the coverage to the biomass model to obtain the biomass within the quadrat 3.

[0104] Biomass refers to the total amount of living organic matter (dry or wet weight) per unit area at a given moment (including the weight of food stored within organisms), measured in kg / m². 2 or t / hm 2 express.

[0105] In some specific embodiments, the biomass model includes at least a wet weight model of plant 4. (The last part, "through methods such as...", appears to be an error and doesn't translate directly.) Figure 6 The diagram showing the fitting of the wet weight model illustrates that the wet weight model is obtained through the following calculation formulas:

[0106] f(x,y)=a×sin(n×PI×x×y)+b×exp(-(k×y) 2 );

[0107] Where f(x, y) represents the quotient of the wet weight in the quadrat 3 and 10000, x represents the quotient of the cover in the quadrat 3 and 100, y represents the quotient of the height of the plant 4 in the quadrat 3 and 100, PI represents pi, and a, b, n, and k are model coefficients.

[0108] Optionally, the model coefficient 'a' can range from 1.143 to 1.236.

[0109] Optionally, the model coefficient b can range from 0.08892 to 0.1463.

[0110] Optionally, the model coefficient n can range from 0.5995 to 0.7279.

[0111] Optionally, the model coefficient k can range from 1.031 to 5.385.

[0112] This embodiment of the application uses a near-infrared camera 1 to acquire original images including quadrat frames 3, which can better highlight the features of the plants 4 within them, facilitating the identification of plants 4 in the image during image recognition. The height of the plants 4 within the quadrat frames 3 and the canopy cover within the quadrat frames 3 are obtained from the original images, and then the height and canopy cover are applied to a biomass model to obtain the biomass within the quadrat frames 3. This embodiment of the application automatically measures the biomass within the quadrat frames above grassland monitoring points using artificial intelligence, improving monitoring efficiency and suitable for the need for large-scale biomass measurement.

[0113] Example 2

[0114] This application also provides apparatus embodiments that follow the above embodiments, for implementing the method steps described in the above embodiments. The interpretation of the same names is the same as that in the above embodiments, and they have the same technical effects as those in the above embodiments, so they will not be repeated here.

[0115] like Figure 7 As shown, this application provides a biomass measurement device 700 within a quadrat frame, comprising:

[0116] The acquisition unit 701 is used to acquire raw images through a near-infrared camera at a grassland monitoring point, wherein the raw images include images of preset sample boxes;

[0117] Analysis unit 702 is used to obtain the height of the plants within the quadrat frame and the canopy coverage within the quadrat frame based on the original image;

[0118] The calculation unit 703 is used to apply the height and the cover to the biomass model to obtain the biomass within the quadrat frame.

[0119] Optionally, the biomass model includes at least a wet weight model of the plant; the wet weight model includes the following calculation formula:

[0120] f(x,y)=a×sin(m×PI×x×y)+b×exp(-(w×y) 2 );

[0121] Where f(x, y) represents the quotient of the wet weight in the quadrat and 10000, x represents the quotient of the cover in the quadrat and 100, y represents the quotient of the height of the plant in the quadrat and 100, PI represents pi, and a, b, m, and w are model coefficients.

[0122] Optionally, the model coefficient a ranges from 1.143 to 1.236; the model coefficient b ranges from 0.08892 to 0.1463; and the model coefficient m ranges from 0.5995 to 0.7279.

[0123] Optionally, the original image also includes the target image within the quadrat frame and the scale image of the height measuring rod set within the quadrat frame, wherein the target image includes images of various plants, including the target plant;

[0124] Accordingly, the analysis unit 702 includes:

[0125] The analysis subunit is used to determine the height of the plants within the quadrat frame based on the positional relationship between the images of the various plants and the scale in the ruler image in the original image.

[0126] Optionally, the analysis unit 702 includes:

[0127] The first acquisition subunit is used to obtain the grassland type within the sample box based on the target image in the original image and the preset latitude and longitude information and preset altitude information of the grassland monitoring point;

[0128] A sub-unit is defined for determining parameter values ​​of multiple model parameters in the cover classification model based on the grassland type, wherein the cover classification model includes a trained convolutional neural network model;

[0129] The second obtaining subunit is used to sample the target image based on a preset sample specification to obtain multiple sample images;

[0130] The third obtaining subunit is used to apply each sample image to a coverage classification model set by the parameter values ​​of the multiple model parameters to obtain the coverage type of the corresponding sample image.

[0131] The fourth subunit is used to analyze the multiple sample images based on the grassland type and each coverage type of the multiple sample images to obtain the coverage within the quadrat.

[0132] Optionally, the coverage classification model includes at least two convolutional layers for input sample images and one fully connected layer after the two convolutional layers; each convolutional layer is followed by a pooling layer, and each pooling layer is followed by a normalization layer; the fully connected layer is followed by a dropout layer, and the dropout layer is followed by a Softmax layer that outputs the classification result.

[0133] Optionally, the original image includes fused information of near-infrared light, green light, and red light.

[0134] This application embodiment uses a near-infrared camera to acquire original images including quadrat frames, which better highlights plant features and facilitates plant identification during image recognition. The height and canopy of the plants within the quadrat frames are obtained from the original images, and then these height and canopy are applied to a biomass model to obtain the biomass within the quadrat frames. This application embodiment automatically measures the biomass within the quadrat frames at grassland monitoring points using artificial intelligence, improving monitoring efficiency and suitable for large-scale biomass measurement needs.

[0135] Example 3

[0136] This embodiment provides an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method steps described in the above embodiment.

[0137] Example 4

[0138] This application provides a non-volatile computer storage medium storing computer-executable instructions that can perform the steps described in the above embodiments.

[0139] Finally, it should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0140] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for measuring biomass within a quadrat frame, characterized in that, include: Raw images are acquired using near-infrared cameras at grassland monitoring points, wherein the raw images include images of pre-defined sample boxes; The height of the plants within the quadrat frame and the canopy coverage within the quadrat frame are obtained based on the original image. The height and the cover are applied to the biomass model to obtain the biomass within the quadrat frame; The preset altitude of the grassland monitoring point is 2800 m; the monitoring point is located in Dali, on the Yunnan-Guizhou Plateau; the obtained grassland type is temperate grassland. The biomass model includes at least a wet weight model of the plant; the wet weight model includes the following calculation formula: ； Where f(x, y) represents the quotient of the wet weight within the quadrat frame to 10000, x represents the quotient of the cover within the quadrat frame to 100, y represents the quotient of the height of the plant within the quadrat frame to 100, PI represents pi, and a, b, n, and k are model coefficients; the value range of model coefficient a is 1.143 to 1.236; the value range of model coefficient b is 0.08892 to 0.1463; the value range of model coefficient n is 0.5995 to 0.7279; and the value range of model coefficient k is 1.031 to 5.

385.

2. The method according to claim 1, characterized in that, The original image also includes the target image within the quadrat frame and the scale image of the height measuring rod set within the quadrat frame. The target image includes images of various plants, including the target plant. Accordingly, obtaining the height of the plants within the quadrat frame based on the original image includes: In the original image, the height of the plants within the quadrat frame is determined based on the positional relationship between the images of the various plants and the scale markings in the ruler image.

3. The method according to claim 2, characterized in that, The step of obtaining the plant cover within the quadrat frame based on the original image includes: The grassland type within the quadrat is obtained by combining the target image in the original image with the preset latitude, longitude and altitude information of the grassland monitoring point; The parameter values ​​of multiple model parameters in the cover classification model are determined based on the grassland type, wherein the cover classification model includes a trained convolutional neural network model; Based on the preset sample specifications of the samples, the target image is sampled to obtain multiple sample images; Each sample image is applied to a coverage classification model defined by the parameter values ​​of the multiple model parameters to obtain the coverage type of the corresponding sample image; The grassland type and the various coverage types of the multiple sample images are analyzed to obtain the coverage within the quadrat frame.

4. The method according to claim 3, characterized in that, The coverage classification model includes at least two convolutional layers for input sample images and one fully connected layer after the two convolutional layers; each convolutional layer is followed by a pooling layer, and each pooling layer is followed by a normalization layer; the fully connected layer is followed by a dropout layer, and the dropout layer is followed by a Softmax layer that outputs the classification result.

5. The method according to claim 3, characterized in that, The original image includes near-infrared light information, green light information, and blue light information.

6. A device for measuring biomass within a quadrat frame, characterized in that, include: The acquisition unit is used to acquire raw images through a near-infrared camera at a grassland monitoring point, wherein the raw images include images of preset sample boxes; Analysis unit, used to obtain the height of the plants within the quadrat frame and the canopy coverage within the quadrat frame based on the original image; A calculation unit is used to apply the height and the cover to a biomass model to obtain the biomass within the quadrat frame; The preset altitude of the grassland monitoring point is 2800 m; the monitoring point is located in Dali, on the Yunnan-Guizhou Plateau; the obtained grassland type is temperate grassland. The biomass model includes at least a wet weight model of the plant; the wet weight model includes the following calculation formula: ； Where f(x, y) represents the quotient of the wet weight within the quadrat frame to 10000, x represents the quotient of the cover within the quadrat frame to 100, y represents the quotient of the height of the plant within the quadrat frame to 100, PI represents pi, and a, b, n, and k are model coefficients; the value range of model coefficient a is 1.143 to 1.236; the value range of model coefficient b is 0.08892 to 0.1463; the value range of model coefficient n is 0.5995 to 0.7279; and the value range of model coefficient k is 1.031 to 5.

385.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 5.

8. An electronic device, characterized in that, include: One or more processors; Storage device for storing one or more programs. Wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1 to 5.

Images