Monitoring method and system for identifying crop drought based on leaf morphology

Abstract

The invention discloses a monitoring method and system for identifying crop drought based on leaf morphology, and relates to the field of agricultural drought monitoring, and the method comprises the following steps: selecting a completely unfolded healthy leaf, cleaning with ultrapure water, blow-drying with nitrogen, and fixing on a standard objective table to obtain a standardized leaf sample; based on the standardized leaf sample, a high-resolution scanner is used for obtaining a front gray image and a back gray image of the leaf, and a binary image with a clear vein network is obtained through image splicing and binarization processing; skeletonizing the binary image with the clear vein network, and calculating the vein fractal dimension by adopting a box counting method to obtain a fractal dimension change rate; the method is in a fixed detection area of a standardized leaf sample. Through innovative leaf vein fractal dimension analysis and contact angle dynamic measurement, a crop drought monitoring system is constructed, and in the aspect of leaf vein analysis, skeletonization processing and a box counting method are adopted to realize accurate quantification of a leaf vein network topological structure.

Description

Technical Field

[0001] This invention relates to the field of agricultural drought monitoring, and in particular to a monitoring method and system for crop drought based on leaf morphology identification. Background Technology

[0002] In the field of agricultural drought monitoring, diagnostic techniques based on plant physiological characteristics have received widespread attention in recent years. Among existing technologies, the degree of water stress is assessed by collecting two-dimensional images of leaves and calculating morphological parameters such as leaf area shrinkage rate and leaf margin curl. This technology uses machine learning algorithms to establish a mapping model between morphological characteristics and soil moisture content, realizing non-destructive monitoring and showing certain practicality in field applications. Similarly, multispectral imaging technology is used to capture leaf water reflectance characteristics and combine them with vegetation indices for drought early warning research. These methods have efficiency advantages in large-scale farmland monitoring.

[0003] Existing technologies have limitations due to their singular monitoring dimensions. The two-dimensional morphological parameters they rely on are insufficient to capture changes in the topology of leaf vein networks. As a key channel for water transport, leaf veins often exhibit early response characteristics before macroscopic morphological changes. The microstructural reorganization of leaf veins in the early stages of drought leads to changes in their fractal properties. However, existing image analysis methods are limited by resolution and algorithms and cannot quantify this sub-millimeter-level morphological evolution. Leaf surface wettability, as an important indicator reflecting the state of the waxy layer, is directly related to stomatal regulation mechanisms but has not been incorporated into existing monitoring systems. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a monitoring method for crop drought based on leaf morphology identification, which solves the monitoring blind spot problem that existing technologies cannot simultaneously capture changes in leaf micro-topology and the evolution of surface physical properties.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a method for monitoring crop drought based on leaf morphology identification, comprising,

[0008] Fully unfolded healthy leaves were selected, washed with ultrapure water and dried with nitrogen, and then fixed on a standard stage to obtain standardized leaf samples.

[0009] Based on standardized leaf samples, grayscale images of the front and back of the leaves were obtained using a high-resolution scanner. After image stitching and binarization, a binarized image with a clear leaf vein network was obtained.

[0010] Skeletonization processing was performed on the binarized image with a clear leaf vein network, and the fractal dimension of the leaf vein was calculated using the box counting method to obtain the rate of change of fractal dimension;

[0011] In a fixed testing area of ​​a standardized blade sample, the advance angle and retreat angle are continuously measured using a contact angle measuring instrument, and the contact angle hysteresis value is calculated to obtain contact angle hysteresis change data.

[0012] The fractal dimension change rate and contact angle hysteresis change data are input into the random forest model to calculate the drought composite index, obtain the drought level results, and generate a visualization report based on the drought level results.

[0013] As a preferred embodiment of the crop drought monitoring method based on leaf morphology identification described in this invention, the method includes the following steps: selecting fully expanded healthy leaves, washing them with ultrapure water and drying them with nitrogen, and then fixing them on a standard stage to obtain standardized leaf samples.

[0014] Using sterile sampling forceps, select fully expanded healthy leaves from the middle of the crop canopy. Place the collected healthy leaves into a pre-cooled constant temperature transport box, and place a temperature and humidity recorder inside the pre-cooled constant temperature transport box to monitor the environment.

[0015] Healthy leaves are immersed in ultrapure water, cleaned with an ultrasonic cleaner to remove surface contaminants, and then dried with high-purity nitrogen to obtain dried healthy leaves.

[0016] The dried, healthy leaves were laid flat on the stage, and the edges were fixed with stress-free transparent fixing rings to obtain standardized leaf samples with clean surfaces, stable shapes, and no stress deformation.

[0017] As a preferred embodiment of the crop drought monitoring method based on leaf morphology identification described in this invention, the method includes: acquiring grayscale images of the front and back of standardized leaf samples using a high-resolution scanner, and then performing image stitching and binarization to obtain a binarized image with a clear leaf vein network, comprising the following steps.

[0018] Standardized leaf samples were placed on the glass platform of a high-resolution scanner, and the parameters of the high-resolution scanner were adjusted to acquire grayscale images of the front and back of the leaf, and the grayscale images of the front and back of the leaf were registered.

[0019] The Laplacian pyramid algorithm is used to fuse the registered grayscale images of the front and back of the leaf into a complete panoramic image of the leaf. The dynamic Otsu algorithm is then used to convert the registered grayscale images of the front and back of the leaf into binary images. Morphological operations using 3×3 circular structuring elements are then performed to remove noise points, resulting in a binarized image with a clear leaf vein network.

[0020] As a preferred embodiment of the crop drought monitoring method based on leaf morphology identification described in this invention, the method includes the following steps: skeletonizing a binarized image with a clear leaf vein network, calculating the fractal dimension of the leaf veins using box counting, and obtaining the rate of change of the fractal dimension.

[0021] The connectivity of the binarized image with a clear leaf vein network is checked. The binarized image with the clear leaf vein network is then processed iteratively using the Zhang-Suen thinning algorithm to generate a leaf vein skeleton image.

[0022] The scale range of the leaf vein skeleton image was analyzed using box counting, the minimum number of grids covering the leaf vein skeleton was counted, and the fractal dimension of the leaf vein was calculated using the fitted linear regression curve.

[0023] By collecting leaf vein fractal dimension data from healthy leaves of the same crop variety under standard irrigation conditions and performing statistical analysis, a baseline value for healthy samples of the same variety was obtained.

[0024] The fractal dimension of leaf veins was compared with the baseline value of healthy samples of the same variety, and the rate of change of fractal dimension was calculated.

[0025] As a preferred embodiment of the crop drought monitoring method based on leaf morphology identification described in this invention, the method includes the following steps: In a fixed detection area of ​​a standardized leaf sample, the advancing angle and receding angle are continuously measured using a contact angle measuring instrument, and the contact angle hysteresis value is calculated to obtain contact angle hysteresis change data.

[0026] At one-third of the distance from the leaf tip, a laser locator was used to mark the detection point on the standardized leaf sample. The marked leaf was placed in a constant temperature and humidity chamber for equilibration. The contact angle measuring instrument was calibrated using a standard PDMS sheet. Ultrapure water was injected at the marked detection point, and the profile of the droplet at the moment of maximum expansion was recorded to obtain the advance angle. The profile of the droplet at the moment of beginning to contract was captured to obtain the retreat angle, and the contact angle hysteresis value was calculated.

[0027] By measuring and statistically analyzing healthy leaves of the same growth period under standard environmental control conditions, a baseline value for healthy samples of the same variety was obtained.

[0028] The contact angle hysteresis value was compared with the baseline value of healthy samples of the same variety to obtain the contact angle hysteresis change data.

[0029] As a preferred embodiment of the crop drought monitoring method based on leaf morphology identification described in this invention, the method includes the following steps: inputting fractal dimension change rate and contact angle hysteresis change data into a random forest model to calculate a comprehensive drought index and obtain drought level results.

[0030] The data on the rate of change of geometric dimension and the hysteresis change of contact angle are combined into a two-dimensional feature vector, and the RandomForestClassifier of Scikit-learn is used for ten-fold cross-validation training to obtain the trained RandomForest model.

[0031] The two-dimensional feature vector is input into a pre-trained random forest model to calculate the drought index.

[0032] Based on ROC curve analysis of field measured data and crop physiological response inflection points, the drought threshold is obtained by forming optimized segmentation points through the Youden index.

[0033] When the drought composite index is less than the drought threshold, the crop moisture status is determined to be within the normal range.

[0034] When the drought composite index equals the drought threshold, the crop water status is determined to be mild drought.

[0035] When the drought composite index is greater than the drought threshold, the crop moisture status is determined to be severe drought.

[0036] As a preferred embodiment of the crop drought monitoring method based on leaf morphology identification described in this invention, the method includes the following steps: generating a visualization report based on drought level results.

[0037] JET color level mapping is performed on a binarized image with a clear leaf vein network based on the fractal dimension change rate to obtain a degradation heatmap with a color scale. The time series curves of the advance angle and the retreat angle are plotted using Matplotlib.

[0038] A vector trend map is generated based on the drought composite index, drought threshold, and irrigation event markers;

[0039] By using Jinja2, you can insert degradation heatmaps with color scales, time-series curves of advance and retreat angles, and vector trend charts into an HTML5 template to generate a visual report.

[0040] Secondly, the present invention provides a crop drought monitoring system based on leaf morphology identification, comprising,

[0041] Standardized module: Select fully expanded healthy leaves, clean them with ultrapure water and dry them with nitrogen, and then fix them on a standard stage to obtain standardized leaf samples.

[0042] The quantization module acquires grayscale images of the front and back of the leaf using a high-resolution scanner based on standardized leaf samples. After image stitching and binarization, a binarized image with a clear leaf vein network is obtained.

[0043] The detection module performs skeletonization processing on the binarized image with a clear leaf vein network, and uses the box counting method to calculate the fractal dimension of the leaf veins to obtain the rate of change of fractal dimension.

[0044] The fusion module continuously measures the advance angle and retreat angle in a fixed detection area of ​​a standardized blade sample using a contact angle measuring instrument, calculates the contact angle hysteresis value, and obtains contact angle hysteresis change data.

[0045] The execution module inputs the fractal dimension change rate and contact angle hysteresis change data into the random forest model, calculates the drought composite index, obtains the drought level results, and generates a visualization report based on the drought level results.

[0046] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the monitoring method for crop drought based on leaf morphology identification as described in the first aspect of the present invention.

[0047] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the crop drought monitoring method based on leaf morphology identification as described in the first aspect of the present invention.

[0048] The beneficial effects of this invention are as follows: This invention constructs a crop drought monitoring system through innovative leaf vein fractal dimension analysis and contact angle dynamic measurement. In terms of leaf vein analysis, the skeletonization process and box counting method are used to achieve accurate quantification of the leaf vein network topology, which can detect early changes in soil moisture content. In terms of surface characteristic detection, high-speed imaging captures dynamic changes in contact angle and combines it with temperature compensation. This forms a technical closed loop with standardized sample preparation, high-resolution scanning, random forest model fusion, and visualization report generation steps, realizing multi-dimensional collaborative monitoring of micromorphology and surface characteristics, and solving the problem of single monitoring dimensions in existing technologies. Attached Figure Description

[0049] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 This is a flowchart of a crop drought monitoring method based on leaf morphology identification.

[0051] Figure 2 This is a schematic diagram of a crop drought monitoring system based on leaf morphology identification.

[0052] Figure 3 This is a schematic diagram of the standardization process for leaf samples.

[0053] Figure 4 Generate schematic diagrams for visual reports. Detailed Implementation

[0054] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0055] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0056] Secondly, the term "one embodiment" or "example" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the invention. The appearance of an embodiment in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that mutually excludes other embodiments.

[0057] Reference Figures 1-4 This is one embodiment of the present invention, which provides a method for monitoring crop drought based on leaf morphology identification, including the following steps:

[0058] S1. Select fully expanded healthy leaves, wash them with ultrapure water and dry them with nitrogen, then fix them on a standard stage to obtain standardized leaf samples.

[0059] S1.1 Use sterile sampling forceps to select fully expanded healthy leaves in the middle of the crop canopy. Place the collected healthy leaves into a pre-cooled constant temperature transport box and place a temperature and humidity recorder inside the pre-cooled constant temperature transport box to monitor the environment.

[0060] Furthermore, using sterile sampling forceps, fully expanded healthy leaves were selected from the third to fifth leaf position in the middle of the crop canopy, avoiding areas affected by pests, diseases, or mechanical damage. The collected healthy leaves were immediately placed in a pre-cooled, constant-temperature transport box with a temperature controlled at 25±1℃ and a relative humidity of 60±5%. A temperature and humidity recorder with an accuracy of ±0.5℃ was placed inside the pre-cooled, constant-temperature transport box to monitor environmental parameters in real time, ensuring that temperature fluctuations did not exceed 1℃ and humidity fluctuations did not exceed 3% during transport. The healthy leaves were kept in the pre-cooled, constant-temperature transport box for no more than 2 hours. Direct sunlight and violent vibrations were avoided during transport. Temperature and humidity records were maintained. The instrument collects data every minute, including temperature, relative humidity, and corresponding timestamps. The data is transmitted to a mobile terminal for storage via Bluetooth. After the healthy leaves arrive at the laboratory, the temperature and humidity recorder data must be checked. Subsequent processing can only proceed after confirming that the environmental parameters during transportation meet the requirements. The pre-cooled constant temperature transport box is disinfected with 75% ethanol before and after each use. After disinfection, it is left to stand for 30 minutes to ensure that the ethanol evaporates completely. The temperature and humidity recorder is calibrated once a week using a standard temperature and humidity source, and the calibration records are kept for at least three years. When collecting healthy leaves, the crop variety, growth stage, collection time, and field location information are recorded simultaneously.

[0061] S1.2 Immerse healthy leaves in ultrapure water, clean them with an ultrasonic cleaner to remove surface contaminants, and use high-purity nitrogen to dry the residual moisture on the leaf surface to obtain dried healthy leaves.

[0062] Furthermore, healthy leaves were completely immersed in ultrapure water with a resistivity ≥18.2 MΩ·cm and cleaned for 30 seconds using an ultrasonic cleaner with a frequency of 40 kHz and a power of 100 W. During the ultrasonic process, the water temperature was maintained at 25±1℃. After cleaning, the healthy leaves were placed on a sterile operating table and purged with high-purity nitrogen (99.999% purity) at a pressure of 0.2 MPa from a distance of 10 cm at a 45-degree angle. The nitrogen flow rate was controlled at 5 L / min during the purging process, and the purging was continued until there was no obvious reflective water film on the leaf surface. After purging, the healthy leaves were left to stand in a constant temperature and humidity environment (25±1℃, 60±5% RH) for 5 minutes. After confirming that there were no visible water stains on the surface, the dried healthy leaves were obtained. The ultrasonic cleaner was rinsed three times with ultrapure water before and after each use. The leaves were sterilized by passing through a 0.22 μm filter before high-purity nitrogen purging. The dried healthy leaves were to be placed in the subsequent fixation steps within 30 minutes to avoid secondary contamination caused by prolonged exposure.

[0063] S1.3. Lay the dried healthy leaves flat on the stage and fix the edges with stress-free transparent fixing rings to obtain standardized leaf samples with clean surfaces, stable shapes and no stress deformation.

[0064] Furthermore, the dried healthy leaf was placed flat in the center of an aluminum alloy stage with a surface flatness of ≤0.01mm. A stress-free transparent fixing ring with a diameter of 50mm and a thickness of 2mm was used to gently press the edge of the leaf. The contact surface between the fixing ring and the leaf was pre-coated with medical-grade petroleum jelly (e.g., 0.1g) to reduce friction. During the fixing process, the leaf was kept in a naturally extended state to avoid artificial stretching or folding. After fixing, the distance between the leaf edge and the fixing ring was measured with calipers to ensure that the distance deviation in each direction was ≤1mm. After fixing, the healthy leaf was inspected under standard lighting conditions (e.g., 500lux white LED light source) to ensure that there were no wrinkles or warping, and standardized leaf samples with clean surfaces, stable shapes and no stress deformation were obtained.

[0065] S2. Based on standardized leaf samples, grayscale images of the front and back of the leaves are obtained using a high-resolution scanner. After image stitching and binarization, a binarized image with a clear leaf vein network is obtained.

[0066] S2.1 Place the standardized leaf sample on the glass platform of the high-resolution scanner, adjust the parameters of the high-resolution scanner, acquire grayscale images of the front and back of the leaf respectively, and register the grayscale images of the front and back of the leaf.

[0067] Furthermore, standardized leaf samples were placed flat in the center of the tempered glass platform of a high-resolution scanner. The surface cleanliness of the glass platform met the ISO Class 5 cleanliness standard. The high-resolution scanner parameters were set to 2400 dpi optical resolution, 48-bit color depth, and uncompressed TIFF format output, with all automatic enhancement functions turned off. The front and back of the standardized leaf samples were scanned separately. White balance calibration was performed using a standard grayscale card before each scan. The grayscale images of the front and back of the leaf obtained from the scan were registered using the SIFT feature point matching algorithm, with no less than 50 pairs of matching points and a registration error controlled within 5 μm. The registered grayscale images of the front and back of the leaf were stored as 16-bit unsigned integer arrays with pixel values ​​ranging from 0 to 65535. The high-resolution scanner was used to verify the optical performance with a dedicated calibration plate after scanning every 50 samples. The glass platform was wiped with a lint-free cloth soaked in analytical grade ethanol before each batch of scans. During the registration process, feature point pairs with a matching error greater than 2 pixels were removed, and the remaining matching point pairs were to be evenly distributed in the midrib and lateral vein areas of the leaf.

[0068] S2.2. Using the Laplacian pyramid algorithm, the grayscale images of the front and back of the registered leaf are fused into a complete panoramic image of the leaf. The dynamic Otsu algorithm is used to convert the grayscale images of the front and back of the registered leaf into binary images. Morphological operations are performed using 3×3 circular structuring elements to remove noise points, resulting in a binary image with a clear leaf vein network.

[0069] Furthermore, the registered grayscale images of the front and back of the leaf are input into the Laplacian pyramid algorithm for multi-scale fusion. The number of pyramid layers is set to 5, the Gaussian kernel size is 5×5 pixels, and the fusion weight coefficient is 0.5. The fused complete leaf panorama is then segmented using the dynamic Otsu algorithm with a window size of 50×50 pixels and a threshold search step size of 5 gray levels. In the segmented binary image, the leaf vein pixel value is 1 and the leaf mesophyll pixel value is 0. Morphological operations are performed on the binary image using 3×3 circular structuring elements. First, a 3×3 circular structuring element erosion operation is performed to remove isolated noise points, and then a 3×3 circular structuring element dilation operation is performed to restore the leaf vein morphology. Finally, a binary image with a clear leaf vein network is obtained.

[0070] S3. The binarized image with a clear leaf vein network is skeletonized, and the fractal dimension of the leaf vein is calculated using the box counting method to obtain the rate of change of fractal dimension.

[0071] S3.1 Check the connectivity of the binarized image with a clear leaf vein network. Then, use the Zhang-Suen thinning algorithm to iteratively process the binarized image with the clear leaf vein network to generate a leaf vein skeleton image.

[0072] Furthermore, an 8-connectivity analysis is performed on the binarized image with a clear vein network to detect the connectivity status of each vein branch and mark the location of broken areas. After the connectivity check is completed, the binarized image with a clear vein network is input into the Zhang-Suen thinning algorithm, which performs two iterative steps alternately: the first iteration deletes boundary pixels that satisfy 2 ≤ neighbor pixel count ≤ 6, have a connectivity of 1, and have at least one east / south / west / north neighbor as background; the second iteration adjusts the deletion conditions to maintain symmetry until the image no longer changes and a vein skeleton image is generated. After each iteration, the topological integrity of the skeleton is verified, and the connection relationship between the main vein and lateral veins of level three or above is preserved. The final output vein skeleton image maintains a width of 1 pixel, and the branch breakage error is controlled within 3 pixels. During the Zhang-Suen thinning algorithm processing, the calculation is terminated when the number of pixels deleted in 5 consecutive iterations is less than 0.1% of the total number of pixels. The vein skeleton image is stored in 1-bit BMP format, with skeleton node coordinates and branch level information.

[0073] S3.2. Use box counting to analyze the scale range of the leaf vein skeleton image, count the minimum number of grids covering the leaf vein skeleton, and use the fitted linear regression curve to calculate the fractal dimension of the leaf vein.

[0074] Specifically, the expression is,

[0075] log N(∈)=-D f ·log∈+C;

[0076] Where N(∈) is the minimum number of grids required to cover the leaf vein skeleton at scale ∈, and D f denoted as the fractal dimension of the leaf veins, C is the intercept of the linear regression, and ∈ is the scale index.

[0077] S3.3. By collecting leaf vein fractal dimension data of healthy leaves of the same crop variety under standard irrigation conditions and performing statistical analysis, a baseline value for healthy samples of the same variety is obtained.

[0078] Furthermore, under standard irrigation conditions (soil moisture content maintained at 70-80% field capacity), at least 30 healthy leaf samples of the same crop variety and at the same growth stage were collected. After each healthy leaf sample was prepared into a standardized sample according to uniform specifications, grayscale images of the front and back of the leaf were acquired using a high-resolution scanner. After image registration, binarization, and skeletonization, the fractal dimension of the leaf veins of each healthy leaf sample was obtained in the range of 0.1-5 mm using box counting. The arithmetic mean of the fractal dimensions of the remaining healthy leaf samples was used as the baseline value for healthy samples of the same variety. The standard deviation was recorded for subsequent significance testing. The crop variety, growth stage, sampling time, and field location information were recorded simultaneously when collecting healthy leaf samples. All leaf vein fractal dimension calculation results were stored in association with metadata. The baseline value for healthy samples of the same variety was updated quarterly. When updating, it was necessary to ensure that the sampling conditions, measurement methods, and data processing procedures were completely consistent with the previous period.

[0079] S3.4. Compare the fractal dimension of leaf veins with the baseline value of healthy samples of the same variety, and calculate the rate of change of fractal dimension.

[0080] Specifically, the expression is,

[0081]

[0082] ΔD f The rate of change of fractal dimension This serves as the baseline value for healthy samples of the same variety.

[0083] S4. In the fixed detection area of ​​the standardized blade sample, the forward angle and backward angle are continuously measured by the contact angle measuring instrument, and the contact angle hysteresis value is calculated to obtain the contact angle hysteresis change data.

[0084] S4.1 Mark the detection point with a laser locator at one-third of the distance from the leaf tip on the standardized leaf sample. Place the marked leaf in a constant temperature and humidity chamber for equilibration. Use a standard PDMS sheet to calibrate the contact angle measuring instrument. Inject ultrapure water at the marked detection point and record the profile of the droplet at the moment of maximum expansion to obtain the advance angle. Capture the profile of the moment the droplet begins to contract to obtain the retreat angle and calculate the contact angle hysteresis value.

[0085] Specifically, the expression is,

[0086] CAH = θ adv -θ rec +α·ΔT;

[0087] Where CAH is the contact angle hysteresis value, θ adv Let θ be the angle of advance. rec Let α be the recoil angle, α be the temperature compensation coefficient, and ΔT be the temperature deviation.

[0088] It should be noted that, in the standardized leaf sample, a laser locator with a wavelength of 635nm and a power of <5mW was used to mark the detection point in an area one-third of the way from the leaf tip, avoiding the midrib, with a marking accuracy of ±0.1mm. The marked standardized leaf samples were then placed in a constant temperature and humidity chamber at 25±0.5℃ and 60±2% relative humidity for 30 minutes to equilibrate. The surface energy was 20.1±0.5mJ / m². 2 The contact angle measuring instrument was calibrated using standard PDMS sheets, with the calibration error controlled within ±0.5°. At the marked detection point, 2μL of ultrapure water with a resistivity ≥18.2MΩ·cm was injected at a rate of 1μL / s. The droplet dynamics were recorded using a 1000fps high-speed camera. When the droplet contact line reached its maximum expansion radius, a contour image was captured, and the advancing angle was obtained by fitting the Young-Laplace equation. The droplet was then re-absorbed at a rate of 0.5μL / s until the contact line began to contract, and the contour image was captured again. The retreat angle was obtained using the same method. Finally, the contact angle hysteresis value was calculated as the advancing angle minus the retreat angle. Each detection point was measured three times, and the average value was taken. The laser positioner's positioning accuracy was checked with a standard grid plate before each use. The environmental parameters of the constant temperature and humidity chamber were recorded every minute. The injection needle was rinsed three times with ultrapure water before and after calibration of the contact angle measuring instrument. During the measurement of the advancing and retreat angles, the ambient temperature fluctuation was controlled within ±0.3℃.

[0089] S4.2. Under standard environmental control conditions, healthy leaves of the same growth period were uniformly and standardized before measurement and statistics were performed to obtain the baseline value of healthy samples of the same variety.

[0090] Furthermore, under standard environmental control conditions of 25±1℃ and 60±5% relative humidity, at least 30 healthy leaf samples of the same crop variety at the same growth stage were collected. All healthy leaf samples were selected from the third to fifth fully expanded leaves in the middle of the canopy. All healthy leaf samples were processed according to a standardized procedure, including collection using sterile sampling forceps, preservation in a pre-cooled constant-temperature transport box, ultrasonic cleaning with ultrapure water, drying with high-purity nitrogen, and stress-free fixation. After processing, grayscale images of the front and back of the healthy leaf samples were obtained using a high-resolution scanner, and then image registration and... After binarization, the fractal dimension of leaf veins was calculated using the box counting method. Simultaneously, the advance angle and retreat angle were measured using a contact angle meter, and the contact angle lag value was calculated. The arithmetic mean of the fractal dimension of leaf veins and the contact angle lag value of the remaining healthy leaf samples was obtained and used as the benchmark values ​​for the fractal dimension of leaf veins and the benchmark values ​​for the contact angle lag value of healthy samples of the same variety, respectively. When collecting healthy leaf samples, crop variety, growth stage, sampling time, and field location information were recorded simultaneously. All benchmark value calculation results were stored in association with metadata. Standard environmental control conditions were monitored in real time using calibrated temperature and humidity sensors.

[0091] S4.3. Compare the contact angle hysteresis value with the baseline value of healthy samples of the same variety to obtain the contact angle hysteresis change data.

[0092] Specifically, the expression is,

[0093]

[0094] Wherein, ΔCAH represents the contact angle hysteresis change data, and CAH represents the baseline value of healthy samples of the same variety.

[0095] S5. Input the fractal dimension change rate and contact angle hysteresis change data into the random forest model, calculate the drought composite index, and obtain the drought level result.

[0096] S5.1 Combine the data of the rate of change of geometric dimension and the hysteresis change of contact angle into a two-dimensional feature vector, and use Scikit-learn's RandomForestClassifier to perform ten-fold cross-validation training to obtain the trained random forest model.

[0097] Furthermore, the fractal dimension change rate and contact angle hysteresis change data were combined into a two-dimensional feature vector with a weight of 0.6:0.4. Using RandomForestClassifier version 1.2.2 of Scikit-learn, the number of decision trees was set to 500, the maximum depth to 5, and the minimum number of leaf samples to 3. Ten-fold cross-validation was performed on the feature vector matrix containing at least 300 sets of samples. During the training process, the grid search method was used to optimize the parameters. The validation set accounted for 20%. Finally, the model with the highest average accuracy was selected as the trained random forest model.

[0098] S5.2 Input the two-dimensional feature vector into the pre-trained random forest model to calculate the drought comprehensive index.

[0099] Specifically, the expression is,

[0100] DSI = 0.6·ΔD fnorm +0.4·ΔCAH norm ;

[0101] Wherein, DSI is the drought composite index, ΔDf norm To standardize the rate of change of fractal dimension, ΔCAH norm This is the standardized contact angle hysteresis rate of change.

[0102] S5.3. Based on ROC curve analysis of field measured data and crop physiological response inflection points, the drought threshold is obtained by forming an optimized segmentation point through the Youden index.

[0103] Furthermore, based on drought composite index data containing at least 500 sets of field measured samples and simultaneously acquired crop physiological parameters (such as stomatal conductance and photosynthetic rate), receiver operating characteristic (ROC) curves were plotted, with the horizontal axis representing 1-specificity and the vertical axis representing sensitivity. Combining the inflection point of crop physiological response, the Youden index corresponding to each candidate drought threshold was calculated (Youden index = sensitivity + specificity - 1). The drought composite index value corresponding to the maximum Youden index was selected as the optimal split point to obtain the drought threshold.

[0104] S5.4 When the drought comprehensive index is less than the drought threshold, the crop moisture status is determined to be within the normal range.

[0105] Furthermore, when the drought composite index is less than the drought threshold (e.g., DSI ≤ 0.7), the current crop moisture status is determined to be within the normal range. At this time, the deviation of leaf physiological indicators (e.g., stomatal conductance, photosynthetic rate) from the baseline value of healthy samples of the same variety does not exceed 15%, the leaf vein fractal dimension change rate is ≤ 5%, and the contact angle hysteresis change rate ΔCAH is ≤ 8%. The original agronomic management measures are maintained unchanged, and the next test is carried out according to the standard monitoring cycle. During the determination process, the consistency of the fractal dimension change rate and the contact angle hysteresis change rate ΔCAH is checked simultaneously. The two change trends are required to be consistent. Otherwise, the data verification process is triggered. The determination result of the normal state is cross-validated with the field measured soil moisture content data to ensure the reliability of the determination.

[0106] S5.5 When the drought composite index equals the drought threshold, the crop moisture status is determined to be mild drought.

[0107] Furthermore, when the drought composite index calculation result equals the drought threshold (e.g., DSI = 0.7), the current crop moisture status is determined to be mild drought. At this time, leaf physiological indicators show that stomatal conductance decreases by 20-30%, photosynthetic rate decreases by 15-25%, leaf vein fractal dimension change rate ΔDf is in the range of 5-15%, and contact angle hysteresis change rate ΔCAH is 8-15%. The system automatically generates a mild drought warning and suggests retesting within 24 hours. At the same time, the fertilization plan is adjusted (e.g., reducing nitrogen fertilizer application by 10-15%). During the determination process, the synergy between the fractal dimension change rate and the contact angle hysteresis change rate is verified. If the data is abnormal, the quality control review process is initiated. The mild drought determination result is cross-validated with field transpiration meter data.

[0108] S5.6 When the drought composite index is greater than the drought threshold, the crop moisture status is determined to be severe drought.

[0109] Furthermore, when the drought composite index calculation result is greater than the drought threshold determined by ROC curve analysis and Youden index optimization (e.g., DSI>0.85), the current crop water status is determined to be severe drought. At this time, leaf physiological indicators show that stomatal conductance decreases by more than 50%, photosynthetic rate decreases by more than 40%, leaf vein fractal dimension change rate is >15%, and contact angle hysteresis change rate is >20%, immediately triggering a severe drought alarm and automatically generating irrigation instructions.

[0110] S6. Generate a visualization report based on the drought level results.

[0111] S6.1. Based on the fractal dimension change rate, perform JET color level mapping on the binarized image with a clear leaf vein network to obtain a degradation heatmap with a color scale, and use Matplotlib to plot the time series curves of the advance angle and the retreat angle.

[0112] Furthermore, the binarized image with a clear leaf vein network and the corresponding fractal dimension change rate data are input into the applyColorMap function of the OpenCV library. The JET color level mode is selected for color mapping to generate a pseudo-color heatmap. In this heatmap, blue (RGB 0,0,255) represents ΔDf≤5%, yellow (RGB 255,255,0) represents 5%<ΔDf≤15%, and red (RGB 255,0,0) represents ΔDf>15%. A standard color scale is added to the heatmap to indicate the correspondence between ΔDf values ​​and colors. At the same time, the pyplot module of the Matplotlib library is used to plot a time-series curve with marked points, using the measurement time points of the advance angle and retreat angle as the x-axis and the angle value as the y-axis. The advance angle is represented by a solid green line, and the retreat angle is represented by a dashed red line. The trend of the contact angle hysteresis value CAH is marked above the curve.

[0113] S6.2 Generate a vector trend map based on the drought composite index, drought threshold, and irrigation event markers.

[0114] Furthermore, based on historical drought composite index time series data, pre-set drought thresholds, and irrigation event logs, a vector-format trend chart is created using the Matplotlib library. A line chart is plotted with time as the horizontal axis (e.g., date scale) and the drought composite index as the vertical axis (range 0-1.0) to represent the trend of the drought composite index. Two horizontal dashed lines are added to the chart to mark the thresholds for mild drought (e.g., orange dashed line y=0.7) and severe drought (e.g., red dashed line y=0.85). Green triangle symbols are used to mark the time points of irrigation events and their corresponding water volumes. Axis labels (e.g., time (days) and drought composite index) and a title (e.g., drought trend during wheat jointing stage) are added to the trend chart. The chart is saved as a Scalable Vector Graphics (SVG) format, with metadata recording crop variety, monitoring period, and chart generation date. During the plotting process, missing data is processed using linear interpolation, and outliers (e.g., DSI>1.0) are excluded after verification. The color scheme of the trend chart is consistent with the heatmap: blue indicates normal conditions, yellow indicates mild drought, and red indicates severe drought.

[0115] S6.3. Using Jinja2, degradation heatmaps with color scales, time-series curves of advance and retreat angles, and vector trend charts are inserted into an HTML5 template to generate a visualization report.

[0116] Furthermore, using the Jinja2 template engine, PNG files of degradation heatmaps with color scales, PNG files of time-series curves with advance and retreat angles, and SVG files of vector trend graphs are inserted into a predefined HTML5 report template. The template includes a title area (displaying crop variety, detection date, and geographical location), a data visualization area (three graphs displayed side by side), and a conclusion area (summarizing drought level and agronomic recommendations). An interactive floating tooltip is added to the degradation heatmap in the HTML5 template to display the precise value of the rate of change of leaf vein fractal dimension at the mouse hover position. A legend is added to the time-series curves to illustrate the correspondence between the advance and retreat angles r and the contact angle hysteresis value CAH. Clickable irrigation event markers are embedded in the vector trend graph, which display the specific irrigation time and water volume when clicked, generating a visualization report.

[0117] This embodiment also provides a crop drought monitoring system based on leaf morphology identification, including:

[0118] Standardized module: Select fully expanded healthy leaves, clean them with ultrapure water and dry them with nitrogen, and then fix them on a standard stage to obtain standardized leaf samples.

[0119] The quantization module acquires grayscale images of the front and back of the leaf using a high-resolution scanner based on standardized leaf samples. After image stitching and binarization, a binarized image with a clear leaf vein network is obtained.

[0120] The detection module performs skeletonization processing on the binarized image with a clear leaf vein network, and uses the box counting method to calculate the fractal dimension of the leaf veins to obtain the rate of change of fractal dimension.

[0121] The fusion module continuously measures the advance angle and retreat angle in a fixed detection area of ​​a standardized blade sample using a contact angle measuring instrument, calculates the contact angle hysteresis value, and obtains contact angle hysteresis change data.

[0122] The execution module inputs the fractal dimension change rate and contact angle hysteresis change data into the random forest model, calculates the drought composite index, obtains the drought level results, and generates a visualization report based on the drought level results.

[0123] This embodiment also provides a computer device applicable to the monitoring method of crop drought based on leaf morphology identification, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the monitoring method of crop drought based on leaf morphology identification as proposed in the above embodiment.

[0124] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0125] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the crop drought monitoring method based on leaf morphology identification as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0126] In summary, this invention constructs a crop drought monitoring system through innovative leaf vein fractal dimension analysis and dynamic contact angle measurement. In terms of leaf vein analysis, the skeletonization process and box counting method are used to accurately quantify the topology of the leaf vein network, which can detect early changes in soil moisture content. In terms of surface characteristic detection, high-speed imaging captures dynamic changes in contact angle and combines it with temperature compensation. This forms a technical closed loop with standardized sample preparation, high-resolution scanning, fusion with random forest models, and visualization report generation steps, realizing multi-dimensional collaborative monitoring of micromorphology and surface characteristics, and solving the problem of single monitoring dimensions in existing technologies.

[0127] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for monitoring crop drought based on leaf morphology identification, characterized in that: include, Fully unfolded healthy leaves were selected, washed with ultrapure water and dried with nitrogen, and then fixed on a standard stage to obtain standardized leaf samples. Based on standardized leaf samples, grayscale images of the front and back of the leaves were obtained using a high-resolution scanner. After image stitching and binarization, a binarized image with a clear leaf vein network was obtained. Skeletonization processing was performed on the binarized image with a clear leaf vein network, and the fractal dimension of the leaf vein was calculated using the box counting method to obtain the rate of change of fractal dimension; In a fixed testing area of ​​a standardized blade sample, the advance angle and retreat angle are continuously measured using a contact angle measuring instrument, and the contact angle hysteresis value is calculated to obtain contact angle hysteresis change data. The fractal dimension change rate and contact angle hysteresis change data are input into the random forest model to calculate the drought composite index, obtain the drought level results, and generate a visualization report based on the drought level results.

2. The crop drought monitoring method based on leaf morphology identification as described in claim 1, characterized in that: Fully expanded, healthy leaves were selected, washed with ultrapure water, dried with nitrogen, and then fixed on a standard stage to obtain standardized leaf samples. This process included the following steps: Using sterile sampling forceps, select fully expanded healthy leaves from the middle of the crop canopy. Place the collected healthy leaves into a pre-cooled constant temperature transport box, and place a temperature and humidity recorder inside the pre-cooled constant temperature transport box to monitor the environment. Healthy leaves are immersed in ultrapure water, cleaned with an ultrasonic cleaner to remove surface contaminants, and then dried with high-purity nitrogen to obtain dried healthy leaves. The dried, healthy leaves were laid flat on the stage, and the edges were fixed with stress-free transparent fixing rings to obtain standardized leaf samples with clean surfaces, stable shapes, and no stress deformation.

3. The crop drought monitoring method based on leaf morphology identification as described in claim 2, characterized in that: Based on standardized leaf samples, grayscale images of the front and back of the leaves are acquired using a high-resolution scanner. These images are then stitched together and binarized to obtain a binarized image with a clear vein network. The process includes the following steps: Standardized leaf samples were placed on the glass platform of a high-resolution scanner, and the parameters of the high-resolution scanner were adjusted to acquire grayscale images of the front and back of the leaf, and the grayscale images of the front and back of the leaf were registered. The Laplacian pyramid algorithm is used to fuse the registered grayscale images of the front and back of the leaf into a complete panoramic image of the leaf. The dynamic Otsu algorithm is then used to convert the registered grayscale images of the front and back of the leaf into binary images. Morphological operations using 3×3 circular structuring elements are then performed to remove noise points, resulting in a binarized image with a clear leaf vein network.

4. The crop drought monitoring method based on leaf morphology identification as described in claim 3, characterized in that: The binarized image with a clear leaf vein network is skeletonized, and the fractal dimension of the leaf veins is calculated using box counting to obtain the rate of change of fractal dimension. Includes the following steps, The connectivity of the binarized image with a clear leaf vein network is checked. The binarized image with the clear leaf vein network is then processed iteratively using the Zhang-Suen thinning algorithm to generate a leaf vein skeleton image. The scale range of the leaf vein skeleton image was analyzed using box counting, the minimum number of grids covering the leaf vein skeleton was counted, and the fractal dimension of the leaf vein was calculated using the fitted linear regression curve. By collecting leaf vein fractal dimension data from healthy leaves of the same crop variety under standard irrigation conditions and performing statistical analysis, a baseline value for healthy samples of the same variety was obtained. The fractal dimension of leaf veins was compared with the baseline value of healthy samples of the same variety, and the rate of change of fractal dimension was calculated.

5. The crop drought monitoring method based on leaf morphology identification as described in claim 4, characterized in that: Within a fixed testing area of ​​a standardized blade sample, the advance angle and retreat angle are continuously measured using a contact angle measuring instrument. The contact angle hysteresis value is then calculated to obtain contact angle hysteresis change data. Includes the following steps, At one-third of the distance from the leaf tip, a laser locator was used to mark the detection point on the standardized leaf sample. The marked leaf was placed in a constant temperature and humidity chamber for equilibration. The contact angle measuring instrument was calibrated using a standard PDMS sheet. Ultrapure water was injected at the marked detection point, and the profile of the droplet at the moment of maximum expansion was recorded to obtain the advance angle. The profile of the droplet at the moment of beginning to contract was captured to obtain the retreat angle, and the contact angle hysteresis value was calculated. By measuring and statistically analyzing healthy leaves of the same growth period under standard environmental control conditions, a baseline value for healthy samples of the same variety was obtained. The contact angle hysteresis value was compared with the baseline value of healthy samples of the same variety to obtain the contact angle hysteresis change data.

6. The crop drought monitoring method based on leaf morphology identification as described in claim 5, characterized in that: By inputting the fractal dimension change rate and contact angle hysteresis change data into a random forest model, a comprehensive drought index is calculated to obtain the drought level results. Includes the following steps, The data on the rate of change of geometric dimension and the hysteresis change of contact angle are combined into a two-dimensional feature vector, and the RandomForestClassifier of Scikit-learn is used for ten-fold cross-validation training to obtain the trained RandomForest model. The two-dimensional feature vector is input into a pre-trained random forest model to calculate the drought index. Based on ROC curve analysis of field measured data and crop physiological response inflection points, the drought threshold is obtained by forming optimized segmentation points through the Youden index. When the drought composite index is less than the drought threshold, the crop moisture status is determined to be within the normal range. When the drought composite index equals the drought threshold, the crop water status is determined to be mild drought. When the drought composite index is greater than the drought threshold, the crop moisture status is determined to be severe drought.

7. The crop drought monitoring method based on leaf morphology identification as described in claim 6, characterized in that: Generate a visualization report based on drought level results, including the following steps: JET color level mapping is performed on a binarized image with a clear leaf vein network based on the fractal dimension change rate to obtain a degradation heatmap with a color scale. The time series curves of the advance angle and the retreat angle are plotted using Matplotlib. A vector trend map is generated based on the drought composite index, drought threshold, and irrigation event markers; By using Jinja2, you can insert degradation heatmaps with color scales, time-series curves of advance and retreat angles, and vector trend charts into an HTML5 template to generate a visual report.

8. A crop drought monitoring system based on leaf morphology identification, based on the crop drought monitoring method based on leaf morphology identification according to any one of claims 1 to 7, characterized in that: include, Standardized module: Select fully expanded healthy leaves, clean them with ultrapure water and dry them with nitrogen, and then fix them on a standard stage to obtain standardized leaf samples. The quantization module acquires grayscale images of the front and back of the leaf using a high-resolution scanner based on standardized leaf samples. After image stitching and binarization, a binarized image with a clear leaf vein network is obtained. The detection module performs skeletonization processing on the binarized image with a clear leaf vein network, and uses the box counting method to calculate the fractal dimension of the leaf veins to obtain the rate of change of fractal dimension. The fusion module continuously measures the advance angle and retreat angle in a fixed detection area of ​​a standardized blade sample using a contact angle measuring instrument, calculates the contact angle hysteresis value, and obtains contact angle hysteresis change data. The execution module inputs the fractal dimension change rate and contact angle hysteresis change data into the random forest model, calculates the drought composite index, obtains the drought level results, and generates a visualization report based on the drought level results.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the crop drought monitoring method based on leaf morphology identification as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the crop drought monitoring method based on leaf morphology identification as described in any one of claims 1 to 7.

Images