A method for testing and analyzing corn plant resistance to lodging

Abstract

The application discloses a method for testing and analyzing the lodging resistance of corn plants, comprising: collecting the aboveground crown layer morphology sequence of the corn plants under high wind speed to quantify the dynamic morphology change; extracting the dynamic structural mechanical response representing the crown layer posture and stem bending; obtaining the volume data of the underground root system and soil through the tomography technology, and analyzing to obtain the three-dimensional topological structure of the root system and the biomechanical field data of the root-soil interaction; and fusing the aboveground and underground measurement data to quantitatively evaluate the synergistic wind resistance performance of the plants. The method overcomes the limitations of the static and one-sided analysis of the prior art, and the quantitative indexes provided by the method can be applied to the screening of high-toughness corn varieties, and the data acquisition efficiency is improved through an intelligent monitoring strategy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of material physical property testing and analysis, and particularly relates to a method for testing and analyzing lodging resistance of corn plants. BACKGROUND

[0002] As an important food crop worldwide, the yield stability of corn is of great importance to food security. Lodging is one of the main agricultural disasters that limits the yield and quality of corn, especially in regions where extreme weather events such as typhoons and heavy rains occur frequently. Lodging can cause more than 50% yield loss. The lodging of corn is not caused by the defect of a single organ, but a complex biomechanical instability process. The fundamental reason is that the wind load borne by the aboveground part, such as stem and canopy, exceeds the anchoring moment provided by the underground part, i.e. the complex of root and soil. In-depth and quantitative understanding of the synergistic mechanism of the aboveground and underground parts of corn under wind stress is the key to solving the problem of lodging of corn varieties.

[0003] However, from the perspective of material testing and analysis, the existing technical means for quantitative research on this process has obvious limitations. Most research methods focus on static measurement of plant traits, such as measuring the bending strength of isolated stems in a controlled environment, or analyzing the configuration of roots by digging after lodging occurs. These test methods are mostly destructive tests, and cannot reflect the dynamic mechanical response of the interaction between the aboveground part and the underground root-soil system of the corn plant as a whole in the in-situ state in the field. The existing phenotype analysis technology is difficult to capture the key physical change process under strong wind action (within seconds to minutes) due to the low data collection frequency. Therefore, there is an urgent need in the art for a new method that can in-situ, dynamically and non-destructively measure the physical and mechanical response of the aboveground and underground parts of the corn plant. SUMMARY

[0004] To solve the technical problem that the existing material testing technology cannot in-situ, dynamically and non-destructively analyze the physical and biomechanical properties of living biological composite materials such as corn plants, the present application provides a method for testing and analyzing the lodging resistance of corn plants.

[0005] To achieve the above-mentioned purpose, the present application provides a method for testing and analyzing the lodging resistance of corn plants, comprising the following steps:

[0006] Step S1, collecting the crown shape sequence. The image data of the aboveground crown of the corn plant under high wind speed or simulated wind field conditions at multiple key time points is collected by a multispectral sensor, and a monitoring reference frequency is preset to obtain a crown shape sequence reflecting the change of crown shape over time.

[0007] Step S2, extracting dynamic structural mechanical response. An image segmentation model (such as a convolutional neural network model) is used to process the images in the canopy morphology sequence to segment out the stem and leaf regions; and based on the segmentation results, feature extraction is performed to calculate the leaf angle distribution feature for representing the windward attitude of the canopy and the bending degree index for representing the mechanical response of the stem, thereby obtaining the dynamic structural mechanical response quantifying the morphological adjustment of the aboveground part of the plant. Specifically, a convolutional neural network model with U-Net architecture can be used to perform semantic segmentation on the images in the canopy morphology sequence to distinguish the background, stem and leaf regions. The center lines of the leaves and stems are extracted by a morphological skeletonization algorithm, the angles between the leaves and the main stem are calculated to generate the leaf angle distribution feature, and the curvature integral of the stem center line is calculated as the bending degree index.

[0008] Step S3, obtaining underground root system structure and biomechanical field data. The volume data of the underground root system of the corn plant at the corresponding time point is obtained by tomography technology, such as a portable X-ray CT (Computed Tomography) scanner. The volume data is processed to extract a root system expansion direction sequence representing the three-dimensional topological structure of the root system, which is used to represent the anchoring state change of the root-soil complex under wind load. Specifically, the Marching Cubes algorithm can be used to reconstruct the three-dimensional grid model of the segmented voxel data, and the model is smoothed and skeletonized to obtain the root system expansion direction sequence.

[0009] Further, the method not only analyzes the geometric morphology of the root system, but also uses the same volume data obtained by the tomography technology to introduce the Digital Volume Correlation (DVC) analysis method to analyze the three-dimensional displacement field and strain field of the soil region around the root system, thereby obtaining the biomechanical field data representing the interaction between the root system and the soil.

[0010] Step S4, fusing data and evaluating the cooperative wind resistance performance. A machine learning fusion model is used to fuse the dynamic structural mechanical response, the root system expansion direction sequence, and the biomechanical field data. By analyzing the functional relationship between wind load and overall plant deformation, including the aboveground sway and the small displacement of the root, the cooperative wind resistance performance of the aboveground and underground is evaluated. Preferably, a random forest regression model is used, which includes the statistics of the leaf angle distribution, the bending degree index of the stem, the statistics of the root system expansion direction sequence, and further includes the strain peak value of the key soil region and the root-soil interface slip rate obtained by DVC analysis. The model outputs a standardized cooperative wind resistance index, which is used to quantitatively evaluate the wind resistance ability of the overall plant.

[0011] Step S5, evaluate post-disaster recovery potential. If the occurrence of lodging event is monitored, that is, the bending degree of the stem exceeds the preset instability threshold, the post-disaster recovery potential evaluation is started. Based on the crown morphology data collected after lodging, the ability of the plant to recover photosynthesis is predicted by using a Photosynthetically Active Radiation (PAR) absorption model, and is quantified as a resource utilization efficiency trend, which is used to evaluate the survival and recovery potential of the plant; the analysis perspective of the application is extended from the lodging event itself to the post-disaster recovery stage, which provides a quantitative basis for evaluating the post-disaster recovery ability or resilience of the variety.

[0012] Step S6, dynamically adjust the data acquisition frequency. According to the external meteorological warning (such as typhoon warning) or real-time wind speed monitoring data, and the recovery state evaluated in step S5, the data acquisition frequency is dynamically adjusted. For example, before and during the landing of a typhoon, the acquisition frequency is increased to the highest to capture the mechanical process at the moment of lodging; and during the post-disaster recovery period, the frequency is correspondingly reduced for long-term tracking. Finally, through filtering processing, such as Kalman filtering, the data collected at different frequencies are smoothed to obtain a spatiotemporally continuous data set covering the entire event process; the application proposes an event-driven intelligent monitoring strategy, which can automatically adjust the data acquisition scheme according to the changes of the external environment and the internal state of the plant, so as to accurately capture the key details of the sudden event and the long-term trend of the slow process with the optimal resource consumption.

[0013] Step S7, construct a four-dimensional dynamic model of the whole plant. Using the spatiotemporally continuous data set, a four-dimensional (3D space + time) dynamic model of the whole plant is constructed, which includes biomechanical parameters. The model integrates the aboveground and underground structures, and can be used to simulate and reproduce the instability process of corn under strong wind and the morphological recovery process after the disaster, providing visual decision support for the mechanism research and screening of lodging-resistant varieties.

[0014] The application has the following beneficial effects through the above technical solutions:

[0015] 1. The application overcomes the limitations of the static and one-sided analysis of the prior art, and can simultaneously capture and quantitatively analyze the coordinated physical response of the aboveground crown and the underground root system of corn under strong wind stress, providing a new material dynamic performance testing method.

[0016] 2. The application not only analyzes the morphological changes, but also introduces digital volume correlation (DVC) analysis to further research the biomechanical interaction of the root-soil complex, providing a more quantitative scientific basis for studying the internal physical mechanism of corn lodging resistance.

[0017] 3. Based on the above-mentioned accurate physical test results, the present application provides a new quantitative physical index for measuring the toughness of corn varieties for breeding workers, which has important practical guiding value for breeding excellent varieties that can quickly restore productivity after extreme climate events. BRIEF DESCRIPTION OF DRAWINGS

[0018] Embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0019] Figure 1 is a flowchart of a method for testing and analyzing the lodging resistance of corn plants according to an embodiment of the present application.

[0020] Figure 2 is a structural diagram of a field in-situ collaborative data acquisition system according to an embodiment of the present application.

[0021] Figure 3 is a strategy diagram for dynamically adjusting data acquisition frequency according to an embodiment of the present application.

[0022] Figure 4 is a strain field distribution diagram of a root-soil complex obtained by digital volume correlation (DVC) analysis according to an embodiment of the present application.

[0023] Figure 5 is a comparison diagram of the recovery potential of different corn varieties after lodging according to an embodiment of the present application.

[0024] Figure 6 is a simulation and experimental data comparison and verification diagram of a four-dimensional dynamic model of a whole plant according to an embodiment of the present application. DETAILED DESCRIPTION

[0025] In order to achieve the purpose of the present application, the present application will be described in further detail below with reference to the accompanying drawings and embodiments.

[0026] Embodiment 1

[0027] The present embodiment provides a method for testing and analyzing the lodging resistance of corn plants, which refers to Figure 1 The technical scheme of the method can be divided into the following steps:

[0028] Step S1, collecting the aboveground canopy morphology sequence. The image data of the aboveground canopy of the corn plant under high wind speed or simulated wind field conditions at multiple key time points is collected by a multispectral sensor, and a monitoring reference frequency is preset to obtain a canopy morphology sequence reflecting the change of canopy morphology over time.

[0029] Specifically, the implementation of this step can use a field automated monitoring platform, for example Figure 2As shown, a set of multispectral imaging units 202 are deployed on a movable gantry structure 201. The multispectral imaging units 202 include but are not limited to sensors in visible (RGB) band, near-infrared (NIR) band, and red-edge band. In a preferred implementation scenario, the monitoring subject, corn plants 203, are planted in a specially designed root box 204 that is transparent to X-ray, and the entire setup is placed in the field. Under simulated wind field conditions, large fans apply gradient wind speed to the plants, while the imaging units 202 on the gantry 201 take vertical and oblique angle shots of the canopy at a pre-set reference frequency (e.g., 1 frame per minute under windless or low wind speed conditions). When the wind speed exceeds a pre-set threshold (e.g., 8 meters per second), the acquisition frequency is boosted to a high frequency mode (e.g., 10 to 30 frames per second) to capture the high speed dynamic response of the canopy under wind load. The collected image data are time-stamp aligned to form a canopy morphology sequence.

[0030] Step S2, extracting dynamic structural mechanical response. A convolutional neural network model is used to extract features from the canopy morphology sequence, calculate the leaf angle distribution feature used to represent the windward attitude of the canopy, and the bending degree index used to represent the mechanical response of the stem, thereby obtaining the dynamic structural mechanical response quantifying the morphological adjustment of the aboveground part of the plant.

[0031] Preferably, this step can be decomposed into multiple sub-steps. Firstly, each frame of the collected canopy morphological sequence is pre-processed, including radiation calibration and geometric correction. Secondly, a pre-trained convolutional neural network (CNN) model with U-Net architecture is used to perform semantic segmentation on the pre-processed image, accurately dividing the image pixels into three categories: background, stem, and leaf. The U-Net model is obtained by training on a corn image dataset containing thousands of manually annotated images. Then, morphological post-processing is performed on the segmented stem and leaf regions, such as using a skeletonization algorithm to extract their respective centerlines. By calculating the angles between each leaf centerline and the main stem centerline, a leaf angle distribution histogram describing all leaf poses can be obtained, which is the leaf angle distribution feature. At the same time, the stem centerline is curve-fitted, and the integral of its curvature along the arc length is calculated, which is defined as the bending degree index. The calculation formula of this index B can be expressed as: B = ∫k(s)ds, where s is the arc length along the stem centerline, and k(s) is the curvature at arc length s. The larger the value of this index B, the more severe the bending of the stem. This step converts macroscopic morphological changes into quantitative mechanical response parameters through deep analysis of the canopy image, providing accurate aboveground input data for subsequent collaborative analysis. As a technical alternative, an instance segmentation model (such as Mask R-CNN) can be used, which not only segments the leaf region but also distinguishes each independent leaf, allowing for more detailed single-leaf level dynamic analysis.

[0032] Step S3, obtain the underground root structure and biomechanical field data. Obtain the volume data of the corn plant underground root system at the corresponding time point through tomographic scanning technology, and process the volume data to extract the root extension direction sequence representing the three-dimensional topological structure of the root system. Further, using the volume data, introduce the digital volume correlation (DVC) analysis method to analyze the three-dimensional displacement field and strain field of the soil region around the root system, and obtain the biomechanical field data of the interaction between the root system and the soil.

[0033] In detail, referring to Figure 2While collecting the aboveground data, a portable X-ray Computed Tomography (CT) instrument 205 is deployed around the rootbox 204, with its X-ray source and detector rotating around the rootbox to acquire the 3D voxel data of the root-soil complex. The resolution of the scan is preferably 50-200 microns to clearly resolve the major root structures and soil particles. For the acquired voxel data, first image segmentation is performed to separate the roots, soil, and pores. Then, the Marching Cubes algorithm is used to reconstruct the 3D mesh model for the segmented root voxel data. Skeletonization is performed on the reconstructed model to extract the 3D topological structure of the roots, and the distribution of the main growth directions of the roots is calculated by Principal Component Analysis (PCA) or other methods to form the sequence of root extension directions.

[0034] Further, to reveal the mechanical mechanism of root-soil interaction, the present application introduces DVC analysis. At least two sets of CT volume data are acquired before and after the wind load is applied (or at different wind levels). The DVC algorithm calculates the 3D displacement vector of each voxel by tracking and matching the 3D texture of the soil internal microstructure (such as small particles, pores) in the two sets of data. The 3D displacement field of the entire soil region is obtained, and the strain field can be further calculated. For example, the deformation gradient tensor F is calculated from the displacement field, and the Green-Lagrange strain tensor E is calculated from F, with the formula E = 0.5 *(F^T * F - I), where F is the deformation gradient tensor, F^T is the transpose matrix of F, and I is the unit tensor. By analyzing the strain field, it can be clearly identified which regions of the soil are compressed and which regions are stretched under the action of wind, thereby quantifying the anchoring effect of the roots on the soil. This step directly relates the anchoring ability of the roots to the mechanical response of the surrounding soil under in-situ and non-destructive conditions, providing biomechanical data for understanding the underground mechanism of lodging resistance.

[0035] Step S4, fusion data and evaluate the synergistic wind resistance performance. The dynamic structural mechanical response, the root extension direction sequence, and the biomechanical field data are fused by a machine learning fusion model to evaluate the synergistic wind resistance performance of aboveground and underground.

[0036] Preferably, the present application employs a Random Forest regression model to perform this fusion evaluation task. The input feature vector of this model is composed of multiple dimensions of data, specifically, including: 1) the aboveground dynamic response patterns from step S2, such as the mean and variance of leaf angle distribution, the maximum value of the stem bending degree indicator B; 2) the root geometry features from step S3, such as the number of main roots, total root length, and the distribution proportion of root system at different depth levels; 3) the root-soil interaction biomechanics field data from step S3, such as the average principal strain, maximum shear strain, and slip rate along the root-soil interface in the key regions (e.g. the tensile region on the windward side and the compression region on the leeward side) around the root system. The output of this model is a standardized Synergistic Wind Resistance Index (SWRI), with a value range of 0 to 1. This index is obtained by training on a large number of samples, where the labels (i.e. the true resistance to lodging ability) of the training samples can be determined by whether the plant eventually lodges or the critical wind speed at which it lodges. Through this step, the complex and nonlinear synergistic effect of the aboveground and underground parts in resisting external forces is learned and quantified by the machine learning model, thereby providing a more comprehensive and accurate evaluation index of the overall wind resistance ability of the plant. As an alternative, other machine learning models such as Gradient Boosting Decision Trees (GBDT) or Support Vector Regression (SVR) can also be used to complete this task.

[0037] Step S5, evaluate post-disaster recovery potential. If a lodging event is monitored, the post-disaster recovery potential evaluation is started. Based on the crown morphology data collected after lodging, a Photosynthetically Active Radiation (PAR) absorption model is used to predict the plant's ability to recover photosynthesis and quantify it as a resource utilization efficiency trend.

[0038] Specifically, the criterion for lodging event is that the bending degree index B exceeds a preset instability threshold, for example, the angle between the stem and the vertical direction exceeds 60 degrees continuously. After the event occurs, the data acquisition switches to a low-frequency long-time monitoring mode. Using the canopy morphology sequence collected in several days to several weeks after lodging, first, a three-dimensional canopy model after lodging is generated through three-dimensional reconstruction technology (such as Structure from Motion, SfM from multi-angle images). Then, the three-dimensional model is input into a PAR absorption model based on a ray tracing algorithm. The model simulates the position of the sun at different times of the day, calculates the shading and absorption of light by the canopy, and estimates the total PAR absorbed by the whole plant per day in the lodged state. By tracking the trend of this indicator over time, the recovery potential of the plant can be evaluated. If the total daily PAR absorption shows an upward trend within a few days after lodging, it indicates that the plant is actively recovering its photosynthetic capacity by adjusting leaf orientation, etc., and its recovery potential is higher. Conversely, if it continues to decline, the recovery potential is lower. This step expands the evaluation dimension from lodging resistance to recovery ability after lodging, providing a screening index more directly related to final yield loss for breeding.

[0039] Step S6, dynamically adjusting the data acquisition frequency. According to the external meteorological warning or real-time wind speed monitoring data, and the recovery state evaluated in step S5, the data acquisition frequency is dynamically adjusted. Finally, through filtering processing, the data collected at different frequencies is smoothed to obtain a spatiotemporal continuous data set covering the entire event process.

[0040] Specifically, referring to Figure 3The method comprises an event-driven intelligent monitoring strategy module. The module presets three working modes: 1) a regular monitoring mode 401: when there is no meteorological warning and the real-time wind speed is less than 5 m / s, the multispectral imaging unit collects at a frequency of 1 time per hour, and the CT scan is performed at a frequency of 1 time per day; 2) a high-frequency response mode 402: when a high-risk meteorological warning such as a typhoon is received, or the real-time wind speed exceeds 10 m / s, the system automatically switches to this mode, the multispectral imaging frequency is increased to more than 10 frames per second, and the CT scan frequency is increased to 1 time per 10 minutes; and 3) a post-disaster recovery mode 403: after detecting the lodging event in step S5, the collection frequency is adjusted to 1 time per 6 hours for multispectral imaging and 1 time per 3 days for CT scanning. All collected data with different time resolutions are processed by a Kalman filter before being fed into the final model. The Kalman filter can fuse observation data of different frequencies to optimally estimate key state parameters (such as stem bending degree and root displacement) of the plant and generate a smooth and spatiotemporally continuous trajectory. The technical effect of this strategy lies in that it greatly optimizes the efficiency and resource allocation of data collection, ensures that sufficient details can be captured when a key event occurs, and avoids data redundancy during a stable period, making long-term and economic field high-precision monitoring possible.

[0041] Step S7, constructing a four-dimensional dynamic model of the whole plant. A four-dimensional (3D space + time) dynamic model of the whole plant containing biomechanical parameters is constructed using the spatiotemporally continuous data set.

[0042] Specifically, the model integrates all registered three-dimensional crown layer models above ground and three-dimensional root system models underground at all time points. The model can be interactively displayed in visualization software, and users can replay the complete response process of the corn plant under strong wind action. Furthermore, the model is a parameterized biomechanical model. Each part of the model, such as the stem internode, leaf, main root, and lateral root, is assigned material mechanical properties such as elastic modulus and Poisson's ratio obtained from experimental measurements or literature. The soil part is modeled as an elastoplastic material, and its parameters are calibrated by DVC analysis results (as shown in Figure 4 Using the four-dimensional model, finite element analysis (FEM) can be performed to simulate the mechanical response of the plant under different wind field conditions and different soil moistures, thereby deeply exploring the internal mechanism of lodging or predicting the lodging resistance of a virtual excellent plant type. To verify the accuracy of the four-dimensional dynamic model, the simulation results of the model are compared with the measured data of the field experiment. Referring to Figure 6The figure shows the high consistency between the model-predicted maximum lateral displacement of the stem top and the experimentally measured data obtained by image analysis in a simulated wind field experiment. The coefficient of determination (R²) shown in the figure reaches 0.97, and the root mean square error (RMSE) is only 1.6 cm, which strongly proves that the model constructed in the present application has high precision and high reliability, and can accurately reproduce the complex mechanical response of the plant under dynamic load.

[0043] Embodiment 2

[0044] The present embodiment provides a corn aboveground and underground synergistic phenotype analysis device, the structural design of which aims to perform the method described in embodiment 1. The device comprises:

[0045] A data acquisition module configured to synchronously acquire the aboveground canopy morphology sequence and the underground root system and soil volume data of the corn plant through the multispectral sensor and the tomographic scanning equipment.

[0046] An aboveground analysis module internally integrated with a convolutional neural network model and an image processing algorithm for extracting dynamic structural mechanical response from the canopy morphology sequence, specifically including leaf angle distribution features and stem bending degree indicators.

[0047] An underground analysis module configured with a three-dimensional reconstruction algorithm and a digital volume correlation (DVC) analysis program for extracting the root system expansion direction sequence from the volume data and calculating the three-dimensional displacement field and strain field of the soil around the root system.

[0048] A fusion evaluation module internally deploying a machine learning fusion model (such as a random forest model) for fusing the output data of the aboveground analysis module and the underground analysis module and generating a quantitative synergistic wind resistance index.

[0049] A recovery potential evaluation module configured to evaluate the post-disaster recovery potential of the plant based on a photosynthetically active radiation absorption model after detecting a lodging event.

[0050] A collection control module that dynamically adjusts the working frequency of the data acquisition module according to external environmental data and plant status, and uses a filtering algorithm to smooth the multi-frequency data.

[0051] A modeling module for integrating all processed spatiotemporal data to construct and visualize a full-plant four-dimensional dynamic model.

[0052] In specific implementation, the above-mentioned modules can be realized in the form of software program modules, running on one or more processors and communicating and controlling corresponding hardware devices (such as cameras, CT scanners, servers).

[0053] Embodiment 3

[0054] This embodiment further illustrates the specific process and technical effect of the method in the actual breeding screening application through a specific comparative field experiment. The experiment is carried out in an automated experimental base, and two representative corn varieties are selected for comparative analysis: variety A (a control variety with strong field resistance to lodging) and variety B (a new breeding line whose anti-lodging ability needs to be evaluated).

[0055] At the beginning of the experiment, the seedlings of variety A and variety B were planted in 10 specially designed root boxes transparent to X-rays and placed under the field automated monitoring system. At the tasseling stage of corn, when the plant biomechanical properties are basically stable, the simulated wind disaster experiment is started. 24 hours before the experiment, the data acquisition system is in the normal monitoring mode, collecting multispectral images at a frequency of once an hour and performing CT scanning once a day to establish the baseline state data of the plants. On the day of the experiment, a large array of axial flow fans in the field is started to simulate a strong wind environment with a maximum wind speed of 25 meters per second (equivalent to a 10-level gale), and the wind lasts for 5 minutes. At the moment the fan starts, the signal of the wind speed sensor exceeding the threshold of 15 meters per second is received by the acquisition control module, which automatically triggers the emergency response mode, instantly increases the acquisition frequency of the multispectral imaging unit to 25 frames per second, and dynamically shortens the scanning interval of the portable X-ray CT instrument to every 5 minutes, to capture the whole response process of the plants at a high time resolution.

[0056] Through real-time processing of the high-frequency crown shape sequence collected, the calculation results of the above-ground analysis module show that under the action of strong wind, the variety A with strong resistance to lodging shows specific morphological adaptability. Its leaves quickly curl and cling to the main stem, and the average angle between the leaves and the main stem is reduced by 35% within 30 seconds, effectively reducing the windward area of the crown; its stem shows high toughness, and the calculated stem bending degree index B reaches a maximum of 0.85 at the peak of the wind force and then stabilizes at this level, and quickly recovers to vertical after the wind ends. In contrast, the leaf shape adjustment of the new variety B is relatively slow, resulting in a larger moment it bears in the wind, and its stem bending degree index B continues to rise under the action of strong wind, breaking through 2.1 at the 95th second, and then the plant suffers irreversible severe leaning, i.e. lodging.

[0057] The results of the processing of the CT scan data by the underground analysis module reveal the fundamental difference in the anti-lodging mechanisms of the two varieties. The DVC analysis results for variety A show that its root system is deep and extensive, with a high density of lateral roots in the 20-40 cm soil layer. Under wind load, the root-soil complex exhibits excellent overall anchoring. As can be seen from the strain field data, the peak tensile strain of the soil on the windward side is only 0.2%, and the stress is effectively dispersed by the thick support roots to a deeper soil layer, without local stress concentration. The analysis of variety B shows that its root system is mainly distributed in the shallow soil layer of 0-20 cm, and the DVC analysis clearly reveals that under wind action, the main root and the soil have an interfacial slip of up to 1.2 mm, and the peak tensile strain of the shallow soil on the windward side is as high as 0.9%, indicating that the anchoring system is the primary failure link leading to lodging.

[0058] Subsequently, the aboveground dynamic response data (leaf angle change rate, stem bending peak value) and underground biomechanics data (root-soil interfacial slip rate, soil strain peak value) are input into the pre-trained random forest regression model. The model calculates that the synergistic wind resistance index (SWRI) of anti-lodging variety A is 0.92 (high resistance), while the SWRI of new variety B is only 0.31 (low resistance), which clearly distinguishes the anti-lodging ability of the two varieties. For the lodged variety B, the system automatically switches to the recovery tracking mode, continuously monitoring its post-disaster state at a frequency of once a day. Referring to Figure 5 , by analyzing the changes in the three-dimensional crown model and the PAR absorption amount within one week after lodging, the recovery potential evaluation module calculates that the total absorption efficiency of photosynthetically active radiation has decreased by 70% compared to before lodging, and there is no obvious recovery, and the recovery potential is evaluated as "low".

[0059] The comparative experiment results of the embodiment show that the method provided by the present application can accurately and quantitatively distinguish and evaluate the anti-lodging comprehensive traits of different corn varieties from multiple dimensions such as dynamic adjustment of aboveground morphology, biomechanical mechanism of root-soil interaction, and post-disaster physiological recovery potential, and the evaluation results are highly consistent with the known actual performance in the field, providing a strong technical reference for efficient and precise stress-resistant breeding screening.

[0060] The above merely illustrates the preferred embodiments of the present application, but should not be used to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for testing and analyzing the lodging resistance characteristics of maize plants, characterized in that, Includes the following steps: Step S1: Collect aboveground canopy morphology sequence. Use a multispectral sensor to collect image data of the aboveground canopy of maize plants under wind conditions at multiple time points to obtain a canopy morphology sequence that reflects the changes in canopy morphology over time. Step S2: Extract the dynamic structural mechanical response and use an image segmentation model to process the images in the canopy morphology sequence to segment the stem and leaf regions; Based on the segmentation results, feature extraction was performed to calculate the leaf angle distribution characteristics used to characterize the windward posture of the canopy and the bending degree index used to characterize the mechanical response of the stem, thereby obtaining the dynamic structural mechanical response of the aboveground part of the plant to quantify the morphological adjustment. Step S3: Obtain underground root structure and biomechanical field data. Simultaneously obtain volume data of the underground root system of the corn plant at multiple time points using tomography. The volume data is processed to extract the root expansion direction sequence representing the three-dimensional topology of the root system; and the microstructure inside the soil in the volume data at different time points is tracked and matched using digital volume correlation analysis method to calculate the three-dimensional displacement field and strain field of the soil area around the root system, so as to obtain biomechanical field data characterizing the change of anchoring state of the root-soil complex under wind load. Step S4: Data fusion and evaluation of synergistic wind resistance performance. A machine learning fusion model is used to fuse the dynamic structural mechanical response, the root extension direction sequence, and the biomechanical field data. By analyzing the functional relationship between wind load and overall plant deformation, a synergistic wind resistance index is output to quantitatively evaluate the overall wind resistance capacity of the plant.

2. The method according to claim 1, characterized in that, The steps for extracting dynamic structural mechanical response specifically include: using a convolutional neural network model with U-Net architecture to perform semantic segmentation on the images in the canopy morphology sequence to distinguish the background, stem and leaf regions; extracting the center lines of the leaves and stems through a morphological skeletonization algorithm, calculating the angle between the leaves and the main stem to generate the leaf angle distribution features, and calculating the curvature integral of the stem center line as the bending degree index.

3. The method according to claim 1, characterized in that, The biomechanical field data include the average principal strain, maximum shear strain, and slip ratio along the root-soil interface in the windward tension zone and the leeward compression zone.

4. The method according to claim 1, characterized in that, The machine learning fusion model is a random forest regression model or a gradient boosting decision tree model; the input feature vector of the model includes: the statistics of the leaf angle distribution, the bending degree index of the stem mechanical response, and the peak strain of the key soil region and the root-soil interface slip rate in the biomechanical field data.

5. The method according to claim 1, characterized in that, The method further includes the step of dynamically adjusting the data acquisition frequency of the multispectral sensor and the tomography technology based on external weather warnings or real-time wind speed monitoring data.

6. The method according to claim 5, characterized in that, The step of dynamically adjusting the data acquisition frequency of the multispectral sensor and the tomography technology includes: when a high-risk weather warning is received or the real-time wind speed exceeds a preset wind speed threshold, the acquisition frequency is increased to a high-frequency emergency response mode; when there is no warning and the wind speed is lower than the conventional threshold, a low-frequency conventional monitoring mode is adopted; the method also includes: using Kalman filtering to smooth the data acquired at different frequencies to obtain a spatiotemporal continuous dataset.

7. The method according to claim 1, characterized in that, The method also includes the following steps: if the bending degree index of the stem mechanical response is detected to exceed a preset instability threshold, then a post-disaster recovery potential assessment is initiated; based on the canopy morphology data collected after lodging, the plant's ability to restore photosynthesis is predicted to assess the plant's recovery potential.

8. The method according to claim 7, characterized in that, The steps for predicting the plant's ability to restore photosynthesis include: generating a three-dimensional canopy model of the plant after lodging using the canopy morphology data collected after lodging; inputting the three-dimensional canopy model into a photosynthetically active radiation absorption model, calculating and tracking the change trend of the total photosynthetically active radiation absorbed by the plant daily over time, in order to quantify the recovery potential.

9. The method according to claim 6, characterized in that, The method also includes the step of: using the spatiotemporal continuous dataset to construct a four-dimensional dynamic model of the whole plant containing biomechanical parameters. The four-dimensional dynamic model integrates above-ground and underground structures and can reproduce the instability process of corn under wind fields and the morphological recovery process after disaster.

10. The method according to claim 9, characterized in that, The steps of constructing a four-dimensional dynamic model of the whole plant also include: assigning material mechanical properties to the stem, leaves and roots of the model, including elastic modulus and Poisson's ratio, and modeling the soil part as an elastoplastic material; and performing finite element analysis using the four-dimensional dynamic model to simulate the mechanical response of the plant under different wind conditions.

Images