Methods, systems, apparatuses, media, and products for estimating morphological and biomass phenotypes of crops

Abstract

This application discloses a method, system, device, medium, and product for estimating the morphological and structural phenotypes and biomass phenotypes of crops, relating to the field of intelligent agricultural monitoring technology. The method includes: acquiring point cloud data of the crop to be tested; inputting the point cloud data of the crop to be tested into a phenotypic estimation model to obtain estimated values ​​of the phenotypic parameters of the crop; the phenotypic parameters include: morphological and structural phenotypes and biomass phenotypes, the morphological and structural phenotypes including: plant height and crown width, and the biomass phenotypes including: leaf area, dry weight, and fresh weight; the phenotypic estimation model is obtained by training a multi-task learning network using a training set, the multi-task learning network being constructed based on a multilayer perceptron, an LSTM network, and two regression heads with different numbers of layers. This application improves the efficiency of crop phenotypic estimation.

Description

Technical Field

[0001] This application relates to the field of intelligent agricultural monitoring technology, and in particular to a method, system, device, medium and product for estimating the morphological and structural phenotypes and biomass phenotypes of crops. Background Technology

[0002] Crop phenotypic parameters include plant height, canopy width, leaf area, dry weight, and fresh weight. Plant height and canopy width are important indicators for measuring crop growth rate and growth stage, reflecting the crop's response to environmental conditions and aiding in variety selection and cultivation management in agricultural production. Leaf area is a key indicator for measuring crop photosynthetic capacity and predicting crop yield. Measuring dry weight allows for accurate assessment of crop biomass accumulation and productivity, crucial for understanding crop growth patterns and predicting yield. Fresh weight provides important information about crop growth rate, water status, and biomass allocation, and can be used to optimize agricultural management and irrigation strategy selection. Therefore, obtaining crop phenotypic parameters is extremely important.

[0003] The accurate and high-density information on object shape, structure, and spatial geometry provided by 3D point cloud data can be used to capture subtle surface features of crops and extract local morphological features. These extracted features can then be used to estimate parameters such as crop height, canopy width, leaf area, volume, and biomass. Common methods for estimating crop phenotypic parameters based on point cloud data typically include: 1) preprocessing the original point cloud; 2) point cloud segmentation; 3) constructing 3D models, such as digital surface models, digital elevation models, plot concept models, and skeleton models; and 4) feature selection, fusion, and the use of machine learning algorithms such as random forests, neural networks, and partial least squares algorithms to construct regression models to estimate individual phenotypic parameters.

[0004] While the above-described process typically yields fast and high-precision results, it still has some limitations. First, the point cloud preprocessing and model building processes require manual intervention in feature selection and parameter fine-tuning, including the use of ground control points in 3D model construction and the selection and introduction of visual information during machine learning algorithm development. These additional operations limit the method's portability and automation. Second, the accuracy of each processing step affects the final estimation performance, leading to accumulated errors in the final phenotypic parameter estimation. Finally, when dealing with multiple phenotypic parameter estimation tasks, multiple estimation models are usually constructed. The deep learning network structure is typically estimated by selecting one or more phenotypes as dependent variables from several relevant phenotypic parameters, failing to fully utilize the correlations between phenotypes. Therefore, a model that directly uses the original point cloud as input, can extract feature vectors from multiple phenotypes, and can accurately regress multiple phenotypic parameters simultaneously is crucial for improving the accuracy and efficiency of crop phenotypic analysis. Summary of the Invention

[0005] The purpose of this application is to provide a method, system, device, medium, and product for estimating the morphological and structural phenotypes and biomass phenotypes of crops, in order to solve the problem of low accuracy and efficiency in the estimation of crop phenotypes in related technologies.

[0006] To achieve the above objectives, this application provides the following solution:

[0007] In a first aspect, this application provides a method for estimating the morphological and structural phenotypes and biomass phenotypes of crops, including:

[0008] Acquire point cloud data of the crop to be tested;

[0009] Point cloud data of the crop to be tested is input into the phenotypic estimation model to obtain estimated values ​​of the phenotypic parameters of the crop to be tested. The phenotypic parameters include: morphological structure phenotype and biomass phenotype. The morphological structure phenotype includes: plant height and crown width. The biomass phenotype includes: leaf area, dry weight and fresh weight. The phenotypic estimation model is obtained by training a multi-task learning network using a training set. The multi-task learning network is constructed based on a multilayer perceptron, an LSTM network and two regression heads with different numbers of layers.

[0010] Optionally, point cloud data of the crop to be tested is acquired, including:

[0011] Acquire depth and RGB images of the crop to be tested;

[0012] The point cloud data of the crop under test is determined based on the depth image and RGB image of the crop under test.

[0013] Optionally, based on the depth image and RGB image of the crop to be tested, the point cloud data of the crop to be tested is determined, including:

[0014] Using a coordinate transformation formula, point cloud data is generated based on the depth values ​​of each pixel in the depth image of the crop under test and the pixel coordinates of the corresponding pixel in the RGB image. There is a one-to-one correspondence between pixels in the depth image and pixels in the RGB image; a point cloud is generated by combining the depth value of a pixel in the depth image with the pixel coordinates of the corresponding pixel in the RGB image. The coordinate transformation formula is as follows:

[0015]

[0016] Among them, (x w ,y w ,z w ) represents point cloud data; z c (u,v) represents the depth value of a pixel in the depth image; (u,v) represents the pixel coordinates of a pixel in the RGB image; (u0,v0) represents the center coordinates of the RGB image; and (dx / f,dy / f) represents the camera's intrinsic parameters.

[0017] Optionally, the training process of the phenotypic estimation model includes:

[0018] Obtain the training set; the training set includes point cloud data and measured values ​​of phenotypic parameters for multiple training objects;

[0019] Construct the multi-task learning network;

[0020] The multi-task learning network is trained using point cloud data of each training object as input and measured values ​​of the corresponding phenotypic parameters as output to obtain the phenotypic estimation model.

[0021] Optionally, the multi-task learning network includes: a first sampling and grouping module, a first MLP-MP-C module, a second sampling and grouping module, a second MLP-MP-C module, an MLP module, a max pooling module, an LSTM network, and a fusion module connected in sequence, as well as two regression heads with different numbers of layers; the max pooling module and the two regression heads are respectively connected to the fusion module;

[0022] Both the first MLP-MP-C module and the second MLP-MP-C module include: a multilayer perceptron, a maximum pooling layer, and a stitching layer connected in sequence. The multilayer perceptron and the stitching layer in the first MLP-MP-C module are both connected to the first sampling and grouping module, and the multilayer perceptron and the stitching layer in the second MLP-MP-C module are both connected to the second sampling and grouping module. The MLP module includes: multiple multilayer perceptrons.

[0023] Optionally, after training the multi-task learning network using point cloud data of each training object as input and measured values ​​of the corresponding phenotypic parameters as output to obtain the phenotypic estimation model, the method further includes:

[0024] Obtain a validation set; the validation set includes: point cloud data of multiple validation objects and measured values ​​of phenotypic parameters;

[0025] The point cloud data of each validation object are input into the phenotypic estimation model to obtain the estimated values ​​of the corresponding phenotypic parameters.

[0026] Based on the measured and estimated values ​​of the phenotypic parameters of each validation object in the validation set, the root mean square error and relative root mean square error of the phenotypic estimation model are calculated.

[0027] Secondly, this application provides a system for estimating the morphological and structural phenotypes and biomass phenotypes of crops, including:

[0028] The point cloud data acquisition module is used to acquire point cloud data of the crop to be tested.

[0029] The estimation module is used to input the point cloud data of the crop to be tested into the phenotypic estimation model to obtain the estimated values ​​of the phenotypic parameters of the crop to be tested. The phenotypic parameters include: morphological structure phenotype and biomass phenotype. The morphological structure phenotype includes: plant height and crown width. The biomass phenotype includes: leaf area, dry weight and fresh weight. The phenotypic estimation model is obtained by training a multi-task learning network using a training set. The multi-task learning network is constructed based on a multilayer perceptron, an LSTM network and two regression heads with different numbers of layers.

[0030] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the estimation methods for the morphological and structural phenotypes and biomass phenotypes of crops as described in any of the preceding claims.

[0031] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the estimation methods for the morphological and structural phenotypes and biomass phenotypes of crops as described in any of the preceding claims.

[0032] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the estimation methods for the morphological and structural phenotypes and biomass phenotypes of crops as described in any of the preceding claims.

[0033] According to the specific embodiments provided in this application, the following technical effects are disclosed:

[0034] This application discloses a method, system, device, medium, and product for estimating the morphological and structural phenotypes and biomass phenotypes of crops. First, point cloud data of the crop to be tested is acquired. Then, the point cloud data is input into a phenotypic estimation model to obtain estimated values ​​of the phenotypic parameters of the crop. The phenotypic parameters include morphological and structural phenotypes and biomass phenotypes. The morphological and structural phenotypes include plant height and crown width, while the biomass phenotypes include leaf area, dry weight, and fresh weight. The phenotypic estimation model is trained using a training set on a multi-task learning network, which is constructed based on a multilayer perceptron, an LSTM network, and two regression heads with different numbers of layers. This application directly inputs the point cloud data of the crop to be tested into the phenotypic estimation model constructed based on a multilayer perceptron, an LSTM network, and two regression heads with different numbers of layers, simultaneously obtaining both the morphological and structural phenotypes and the biomass phenotype of the crop, thus improving the accuracy and efficiency of crop phenotypic estimation. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 A schematic diagram of a method for estimating the morphological and structural phenotypes and biomass phenotypes of crops according to an embodiment of this application;

[0037] Figure 2 A schematic diagram of a multi-task learning network structure;

[0038] Figure 3 The diagram shows the structure of the first MLP-MP-C module and the second MLP-MP-C module.

[0039] Figure 4 This is a schematic diagram of the RGB image of lettuce 1 at growth stage 1.

[0040] Figure 5 A schematic diagram of point cloud data for lettuce 1 at growth stage 1;

[0041] Figure 6 This is a schematic diagram of the RGB image of lettuce 1 at growth stage 2;

[0042] Figure 7 A schematic diagram of point cloud data for lettuce 1 at growth stage 2;

[0043] Figure 8 This is a schematic diagram of the RGB image of lettuce 1 at growth stage 3;

[0044] Figure 9 A schematic diagram of point cloud data for lettuce 1 at growth stage 3;

[0045] Figure 10 This is a schematic diagram of the RGB image of lettuce 1 at growth stage 4.

[0046] Figure 11 A schematic diagram of point cloud data for lettuce 1 at growth stage 4;

[0047] Figure 12 This is a schematic diagram of the RGB image of lettuce 2 during growth stage 1.

[0048] Figure 13 A schematic diagram of point cloud data for lettuce 2 during growth stage 1;

[0049] Figure 14 This is a schematic diagram of the RGB image of lettuce 2 at growth stage 2.

[0050] Figure 15 A schematic diagram of point cloud data for lettuce 2 at growth stage 2;

[0051] Figure 16 This is a schematic diagram of the RGB image of lettuce 2 at growth stage 3;

[0052] Figure 17 A schematic diagram of point cloud data for lettuce 2 at growth stage 3;

[0053] Figure 18 This is a schematic diagram of the RGB image of lettuce 2 at growth stage 4;

[0054] Figure 19 A schematic diagram of point cloud data for lettuce 2 at growth stage 4;

[0055] Figure 20 This is a schematic diagram of the RGB image of lettuce 3 at growth stage 1;

[0056] Figure 21 A schematic diagram of point cloud data for lettuce 3 at growth stage 1;

[0057] Figure 22 This is a schematic diagram of the RGB image of lettuce 3 at growth stage 2.

[0058] Figure 23 A schematic diagram of point cloud data for lettuce 3 at growth stage 2;

[0059] Figure 24 This is a schematic diagram of the RGB image of lettuce 3 at growth stage 3.

[0060] Figure 25 A schematic diagram of point cloud data for lettuce 3 at growth stage 3;

[0061] Figure 26 This is a schematic diagram of the RGB image of lettuce 3 at growth stage 4;

[0062] Figure 27 A schematic diagram of point cloud data for lettuce 3 at growth stage 4;

[0063] Figure 28 This is a schematic diagram of the RGB image of lettuce 4 at growth stage 1.

[0064] Figure 29 A schematic diagram of point cloud data for lettuce 4 at growth stage 1;

[0065] Figure 30 This is a schematic diagram of the RGB image of lettuce 4 at growth stage 2.

[0066] Figure 31 A schematic diagram of point cloud data for lettuce 4 at growth stage 2;

[0067] Figure 32 This is a schematic diagram of the RGB image of lettuce 4 at growth stage 3.

[0068] Figure 33 A schematic diagram of point cloud data for lettuce 4 at growth stage 3;

[0069] Figure 34 This is a schematic diagram of the RGB image of lettuce at growth stage 4.

[0070] Figure 35 A schematic diagram of point cloud data for lettuce 4 at growth stage 4;

[0071] Figure 36 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0072] The technical solutions of the embodiments of this application will be clearly and completely described 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 of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0073] The purpose of this application is to provide a method, system, device, medium, and product for estimating the morphological and structural phenotypes and biomass phenotypes of crops, aiming to improve the accuracy and efficiency of crop phenotype estimation.

[0074] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0075] In one exemplary embodiment, such as Figure 1 As shown, the method for estimating the morphological and structural phenotypes and biomass phenotypes of crops in this embodiment includes:

[0076] Step 101: Obtain point cloud data of the crop to be tested.

[0077] As an optional implementation, step 101 includes:

[0078] Step 1011: Obtain the depth image and RGB image of the crop to be tested.

[0079] Step 1012: Determine the point cloud data of the crop to be tested based on the depth image and RGB image of the crop to be tested.

[0080] As an optional implementation, step 1012 includes:

[0081] Using coordinate transformation formulas, point cloud data is generated based on the depth values ​​of each pixel in the depth image of the crop under test and the pixel coordinates of the corresponding pixel in the RGB image. There is a one-to-one correspondence between pixels in the depth image and pixels in the RGB image; the depth value of one pixel in the depth image and the pixel coordinates of the corresponding pixel in the RGB image generate one point cloud data. The coordinate transformation formula is:

[0082]

[0083] Among them, (x w ,y w ,z w ) represents point cloud data; z c (u,v) represents the depth value of a pixel in the depth image; (u,v) represents the pixel coordinates of a pixel in the RGB image; (u0,v0) represents the center coordinates of the RGB image; and (dx / f,dy / f) represents the camera's intrinsic parameters.

[0084] Step 102: Input the point cloud data of the crop to be tested into the phenotypic estimation model to obtain the estimated values ​​of the phenotypic parameters of the crop to be tested.

[0085] The phenotypic parameters include morphological and structural phenotypes and biomass phenotypes. Morphological and structural phenotypes include plant height and crown width, while biomass phenotypes include leaf area, dry weight, and fresh weight. The phenotypic estimation model is obtained by training a multi-task learning network using a training set. The multi-task learning network is constructed based on a multilayer perceptron, an LSTM network, and two regression heads with different numbers of layers.

[0086] As an optional implementation, the training process of the phenotypic estimation model includes:

[0087] Step 1021: Obtain the training set; the training set includes point cloud data and measured values ​​of phenotypic parameters for multiple training objects.

[0088] Step 1022: Construct a multi-task learning network.

[0089] As an optional implementation method, such as Figure 2 As shown, the multi-task learning network includes: a first sampling and grouping module, a first MLP-MP-C module, a second sampling and grouping module, a second MLP-MP-C module, an MLP module, a max pooling module, an LSTM network, a fusion module, and two regression heads with different numbers of layers (corresponding to...). Figure 2 The first and second regression heads in the module; the maximum pooling module and the two regression heads are respectively connected to the fusion module.

[0090] like Figure 3As shown, both the first MLP-MP-C module and the second MLP-MP-C module include: a multilayer perceptron, a maximum pooling layer, and a stitching layer connected in sequence. The multilayer perceptron and the stitching layer in the first MLP-MP-C module are connected to the first sampling and grouping module, and the multilayer perceptron and the stitching layer (Concat) in the second MLP-MP-C module are connected to the second sampling and grouping module. The MLP module includes: multiple multilayer perceptrons.

[0091] Step 1023: Using the point cloud data of each training object as input and the measured values ​​of the corresponding phenotypic parameters as output, train the multi-task learning network to obtain the phenotypic estimation model.

[0092] Specifically, the process of determining the estimated values ​​of crop phenotypic parameters using phenotypic estimation models includes:

[0093] The crop point cloud data with a dimension of 1024×3 is input into the first sampling and grouping module for sampling and grouping, and outputs a first feature with a dimension of 512×3×64. The first feature is processed by the first MLP-MP-C module and outputs a second feature with a dimension of 512×131. The second feature is processed by the second sampling and grouping module and outputs a third feature with a dimension of 128×131×64. The third feature is processed by the second MLP-MP-C module and outputs a fourth feature with a dimension of 128×259. The 128 sub-features of the fourth feature with a dimension of 259 are respectively input into the 128 multilayer perceptrons of the MLP module and output a fifth feature with a dimension of 128×1024. The fifth feature is input into the max pooling module and outputs a sixth feature with a dimension of 1024. The sixth feature is input into the LSTM network and outputs a seventh feature with a dimension of 1024. The sixth and seventh features are fused in the fusion module and output an eighth feature with a dimension of 1024. The eighth feature is input into the first regression head and the second regression head respectively to obtain the morphological structure phenotype and biomass phenotype.

[0094] As an optional implementation, after step 1023, the method further includes:

[0095] Obtain the validation set; the validation set includes: point cloud data of multiple validation objects and measured values ​​of phenotypic parameters.

[0096] The point cloud data of each validation object are input into the phenotypic estimation model to obtain the estimated values ​​of the corresponding phenotypic parameters.

[0097] Based on the measured and estimated values ​​of phenotypic parameters of each validation animal in the validation set, the root mean square error and relative root mean square error of the phenotypic estimation model are calculated.

[0098] Specifically, the formulas for calculating the root mean square error and the relative root mean square error are as follows:

[0099]

[0100]

[0101] Where RMSE is the root mean square error; N is the total number of validation samples in the validation set; Y i Y represents the estimated phenotypic parameter of the i-th validation organism. i * Let be the measured value of the phenotypic parameter of the i-th validation organism; RRMSE is the relative root mean square error. The average of the measured values ​​of the phenotypic parameters of all validation subjects in the validation set.

[0102] Furthermore, to illustrate the method of this application, lettuce was used as an example, and an experiment was conducted using a test set. The experimental content is as follows:

[0103] (1) Obtain the point cloud data of lettuce.

[0104] (1.1) The source of lettuce.

[0105] The test set consisted of four different lettuce varieties at different growth stages. The lettuce varieties included: Aphylion (Aph), Lugano (Lug), Salanova (Sal), and Satine (Sat). Aph and Lugano are green-leaf lettuce, while Sal (Oakleaf) and Sat are red-leaf lettuce. RGB images and point cloud data of some lettuce varieties (Lettuce 1-Lettuce 4) are shown below. Figures 4-35 As shown, the lettuce samples were grown in greenhouses, with different growth conditions artificially controlled and adjusted to enhance the diversity of training and testing samples.

[0106] (1.2) Obtain the RGB image of the lettuce.

[0107] On the image acquisition platform, an RGB-D camera was installed from a top-down view of the lettuce, suspended approximately 0.90m directly above it. The acquired images were in PNG format with an original resolution of 1920×1080 pixels.

[0108] (1.3) Obtain the depth image of the lettuce.

[0109] A computer-driven RGB-D camera was used to acquire images, recording RGB-D top-view images of lettuce at each growth stage from transplanting to harvest. Each lettuce sample corresponds to a set of RGB-D images, with a total of 391 RGB images and depth images in the test set.

[0110] (1.4) Obtain the measured values ​​of phenotypic parameters.

[0111] Plant height and canopy width were measured using a ruler. Plant height was the straight-line distance from the highest point of the canopy leaves to the substrate, and canopy width was the maximum straight-line distance from the leaf tips. Leaf area was measured using a leaf area meter, fresh weight was obtained by weighing using a balance, and dry weight was obtained by weighing using a balance after drying.

[0112] (1.5) Generate point cloud data of lettuce.

[0113] (2) Input the point cloud data of each lettuce into the phenotypic estimation model to obtain the estimated values ​​of the phenotypic parameters of the corresponding lettuce.

[0114] (3) Calculate the root mean square error and relative root mean square error of the phenotypic estimation model.

[0115] (4) Results description.

[0116] To verify the evaluation performance of the phenotypic estimation model in this application, the phenotypic estimation model, the single-task model STR, and the multi-task model MTR, which were obtained after 500 training rounds, were evaluated 50 times on the test set. The average of the RMSE and RRMSE obtained each time is shown in Table 1.

[0117] Table 1. Estimated values ​​of phenotypic parameters and RMSE and RRMSE.

[0118]

[0119] As shown in Table 1, the single-task model STR has a lower estimation error for plant height and crown width than the multi-task model MTR. Nevertheless, in estimating plant height and crown width, the multi-task model can approach or even exceed the accuracy of the single-task prediction model through appropriate feature fusion strategies. The phenotypic estimation model in this application has a lower RRMSE for plant height estimation than the STR model, but a 2.04% higher RRMSE for crown width prediction than the STR model. The single-task model has a higher RRMSE for estimating leaf area, dry weight, and fresh weight compared to the multi-task model.

[0120] Lettuce plant height, crown width, leaf area, dry weight and fresh weight vary at different growth stages. Therefore, it is necessary to evaluate the performance and applicability of the phenotypic estimation model in this application at different growth stages.

[0121] The lettuce in the test set was divided into four periods, SD1 to SD4. The RRMSE values ​​between the estimated values ​​and the actual values ​​of the statistical model were tested to evaluate the robustness of the phenotypic estimation model in this application in estimating the plant height, crown width, leaf area, dry weight, and fresh weight of lettuce at different periods. The results are shown in Table 2.

[0122] Table 2. Estimated values ​​of phenotypic parameters and RMSE and RRMSE for different growth stages.

[0123]

[0124]

[0125] As shown in Table 3, the phenotypic estimation models in this application were affected by the growth stage in their estimation of plant height, crown width, leaf area, dry weight, and fresh weight. The best-performing phenotypic estimation model in this application, in plant height estimation, had a ratio of 7.82% between the estimated and measured values ​​at SD4, compared to 13.48% at SD1, a difference of 5.66%. Regarding crown width, the best-performing model, STR, had an RRMSE of 8.35% at SD1 and 22.92% at SD4, a difference of 14.57%. STR's RRMSEs for lettuce crown width estimation at SD1 and SD4 were 8.19% and 22.68%, respectively. The phenotypic estimation models in this application showed more significant differences in their estimation results for lettuce leaf area, dry weight, and fresh weight at different growth stages. The best-performing model in lettuce dry weight estimation was the phenotypic estimation model in this application, which had an RRMSE of 17.63% in SD4 and 53.42% in SD1, a difference of 35.76%. The best-performing model in lettuce fresh weight estimation was the MTR model, which had an RRMSE of 19.19% in SD4 and 157.47% in SD1, a difference of 138.28%.

[0126] In one exemplary embodiment, a system for estimating the morphological and structural phenotypes and biomass phenotypes of crops is provided, comprising:

[0127] The point cloud data acquisition module is used to acquire point cloud data of the crop to be tested.

[0128] The estimation module is used to input the point cloud data of the crop to be tested into the phenotypic estimation model to obtain the estimated values ​​of the phenotypic parameters of the crop to be tested. The phenotypic parameters include: morphological structure phenotype and biomass phenotype. The morphological structure phenotype includes: plant height and crown width. The biomass phenotype includes: leaf area, dry weight and fresh weight. The phenotypic estimation model is obtained by training a multi-task learning network using a training set. The multi-task learning network is built based on a multilayer perceptron, an LSTM network and two regression heads with different numbers of layers.

[0129] In one exemplary embodiment, a computer device is provided, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a method for estimating the morphological and structural phenotypes and biomass phenotypes of crops.

[0130] In one exemplary embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements methods for estimating the morphological and structural phenotypes and biomass phenotypes of crops.

[0131] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements methods for estimating the morphological and structural phenotypes and biomass phenotypes of crops.

[0132] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 36 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements a method for estimating the morphological and structural phenotypes and biomass phenotypes of a crop.

[0133] Those skilled in the art will understand that Figure 36 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0134] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0135] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0136] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0137] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0138] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method of estimating the morphological and biomass phenotypes of a crop, characterized by, The methods for estimating the morphological and structural phenotypes and biomass phenotypes of the crops include: Acquire point cloud data of the crop to be tested; Point cloud data of the crop to be tested is input into the phenotypic estimation model to obtain estimated values ​​of the phenotypic parameters of the crop to be tested. The phenotypic parameters include: morphological structure phenotype and biomass phenotype. The morphological structure phenotype includes: plant height and crown width. The biomass phenotype includes: leaf area, dry weight and fresh weight. The phenotypic estimation model is obtained by training a multi-task learning network using a training set. The multi-task learning network is constructed based on a multilayer perceptron, an LSTM network and two regression heads with different numbers of layers.

2. The method of estimating the morphological and biomass phenotypes of a crop according to claim 1, characterized in that, Obtain point cloud data of the crop to be tested, including: Acquire depth and RGB images of the crop to be tested; The point cloud data of the crop under test is determined based on the depth image and RGB image of the crop under test.

3. The method of estimating the morphological and biomass phenotypes of a crop according to claim 2, wherein, Based on the depth and RGB images of the crop under test, the point cloud data of the crop under test is determined, including: Using a coordinate transformation formula, point cloud data is generated based on the depth values ​​of each pixel in the depth image of the crop under test and the pixel coordinates of the corresponding pixel in the RGB image. There is a one-to-one correspondence between pixels in the depth image and pixels in the RGB image; a point cloud is generated by combining the depth value of a pixel in the depth image with the pixel coordinates of the corresponding pixel in the RGB image. The coordinate transformation formula is as follows: where (x w ,y w ,z w ) is the point cloud data; z c is the depth value of the pixel on the depth image; (u,v) is the pixel coordinate of the pixel on the RGB image; (u0,v0) is the center coordinate of the RGB image; (dx / f,dy / f) is the intrinsic parameter of the camera.

4. The method of estimating morphological and biomass phenotypes of a crop according to claim 1, wherein, The training process of the phenotypic estimation model includes: Obtain the training set; the training set includes point cloud data and measured values ​​of phenotypic parameters for multiple training objects; Construct the multi-task learning network; The multi-task learning network is trained using point cloud data of each training object as input and measured values ​​of the corresponding phenotypic parameters as output to obtain the phenotypic estimation model.

5. The method of estimating the morphological and biomass phenotypes of a crop according to claim 4, wherein, The multi-task learning network includes: a first sampling and grouping module, a first MLP-MP-C module, a second sampling and grouping module, a second MLP-MP-C module, an MLP module, a max pooling module, an LSTM network, and a fusion module, connected in sequence, as well as two regression heads with different numbers of layers; the max pooling module and the two regression heads are respectively connected to the fusion module; Both the first MLP-MP-C module and the second MLP-MP-C module include: a multilayer perceptron, a maximum pooling layer, and a stitching layer connected in sequence. The multilayer perceptron and the stitching layer in the first MLP-MP-C module are both connected to the first sampling and grouping module, and the multilayer perceptron and the stitching layer in the second MLP-MP-C module are both connected to the second sampling and grouping module. The MLP module includes: multiple multilayer perceptrons.

6. The method of estimating the morphological and biomass phenotypes of a crop according to claim 4, wherein, After training the multi-task learning network using point cloud data of each training object as input and measured values ​​of the corresponding phenotypic parameters as output to obtain the phenotypic estimation model, the method further includes: Obtain a validation set; the validation set includes: point cloud data of multiple validation objects and measured values ​​of phenotypic parameters; The point cloud data of each validation object are input into the phenotypic estimation model to obtain the estimated values ​​of the corresponding phenotypic parameters. Based on the measured and estimated values ​​of the phenotypic parameters of each validation object in the validation set, the root mean square error and relative root mean square error of the phenotypic estimation model are calculated.

7. A system for estimating morphological and biomass phenotypes of a crop, characterized by, The crop morphological and structural phenotype and biomass phenotype estimation system includes: The point cloud data acquisition module is used to acquire point cloud data of the crop to be tested. The estimation module is used to input the point cloud data of the crop to be tested into the phenotypic estimation model to obtain the estimated values ​​of the phenotypic parameters of the crop to be tested. The phenotypic parameters include: morphological structure phenotype and biomass phenotype. The morphological structure phenotype includes: plant height and crown width. The biomass phenotype includes: leaf area, dry weight and fresh weight. The phenotypic estimation model is obtained by training a multi-task learning network using a training set. The multi-task learning network is constructed based on a multilayer perceptron, an LSTM network and two regression heads with different numbers of layers.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the method for estimating the morphological and structural phenotypes and biomass phenotypes of the crop according to any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements a method for estimating the morphological and structural phenotypes and biomass phenotypes of the crop as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements a method for estimating the morphological and structural phenotypes and biomass phenotypes of the crop as described in any one of claims 1-6.

Images