Plant growth form intelligent monitoring system and method based on multi-parameter fusion

Abstract

The invention discloses a plant growth form intelligent monitoring system and method based on multi-parameter fusion, and belongs to the technical field of intelligent agriculture. The system comprises an imaging module used for acquiring a plant form picture; the data processing module is used for identifying and classifying the obtained plant form picture and segmenting a leaf part and a stem part in the plant form picture; the plant morphological parameter extraction module is used for carrying out image processing on the segmented images of the leaf parts and the stem parts, obtaining morphological parameters of various plants, calculating leaf areas, leaf length-width ratios, color HSV feature values, stem thickness change rates and leaf area daily change rates, carrying out pest and disease damage identification and generating plant parameter information; and the plant growth state evaluation module is used for receiving the plant parameter information, and evaluating the plant growth state according to various plant parameter information for reference to the user, so that the user can take corresponding measures in time, and a proper environment is provided for the plant.

Description

Technical Field

[0001] The present invention relates to the field of smart agricultural technology, and more specifically to a plant growth morphology intelligent monitoring system and method based on multi-parameter fusion. Background Art

[0002] Monitoring plant growth is crucial in agricultural production, horticultural cultivation, and plant science research. Traditional manual measurement methods rely on observation, ruler measurement, or laboratory analysis. These methods suffer from low efficiency, strong subjectivity, and poor real-time performance, making them incapable of meeting the demands of modern precision agriculture.

[0003] In recent years, advances in computer vision and deep learning technologies have provided new solutions for plant growth monitoring. Existing technologies primarily use single parameters or simple image analysis methods to assess plant growth, but several issues remain. Most rely solely on leaf color to assess growth status, lacking the integration of multi-dimensional parameters such as stem growth, pests and diseases, and leaf morphology, resulting in inaccurate assessment results. Traditional image processing methods (such as threshold segmentation and edge detection) struggle to accurately segment leaves and stems in complex backgrounds, impacting the accuracy of subsequent parameter calculations. Existing systems often rely on offline analysis, unable to provide real-time feedback on plant growth status, and lack a multi-parameter intelligent assessment mechanism, making it difficult to provide actionable cultivation recommendations.

[0004] Therefore, how to provide a plant growth morphology intelligent monitoring system and method based on multi-parameter fusion is a problem that those skilled in the art urgently need to solve. Summary of the Invention

[0005] In view of this, the present invention provides a plant growth morphology intelligent monitoring system and method based on multi-parameter fusion, aiming to achieve: 1) high-precision plant morphology parameter extraction, through an improved deep learning model, accurate segmentation of leaves and stems, and calculation of key growth parameters such as leaf area, leaf aspect ratio, stem-leaf ratio, HSV color characteristics, stem diameter change rate, leaf area change rate, and use of a target detection model to analyze whether the plant is diseased or pest-infested; 2) intelligent evaluation of multi-parameter fusion, combining plant physiological characteristics with machine learning algorithms to establish a comprehensive evaluation mechanism to automatically identify plant growth status (such as health, water shortage, disease and pests, etc.); 3) real-time monitoring and decision support: providing dynamic data feedback to assist users in timely adjusting cultivation strategies and improving plant growth management efficiency.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] An intelligent monitoring system for plant growth morphology based on multi-parameter fusion, comprising:

[0008] Imaging module, data processing module, plant morphological parameter extraction module, plant growth status evaluation module, among which,

[0009] The imaging module is used to obtain plant morphology images;

[0010] The data processing module is used to identify and classify the acquired plant morphology pictures, and segment the leaf part and the stem part in the plant morphology pictures;

[0011] The plant morphological parameter extraction module is used to perform image processing on the images of the segmented leaf and stem parts to obtain the morphological parameters of various plants, calculate the leaf area, leaf aspect ratio, color HSV eigenvalue, stem diameter change rate, leaf area daily change rate, and perform pest and disease identification to generate plant parameter information;

[0012] The plant growth status evaluation module is used to receive the plant parameter information generated by the plant morphological parameter extraction module and evaluate the plant growth status based on the various plant parameter information.

[0013] Furthermore, the data processing module includes identifying and classifying the acquired plant morphology images using a trained plant stem and leaf target detection model, and the training method includes:

[0014] Obtaining a first sample dataset, wherein the first sample dataset includes overall morphological images of the detected plants at various growth stages, annotating the acquired images by using rectangular boxes to mark the plant stems and leaves, and then rotating, enlarging, and cropping the annotated images to complete data enhancement operations on the dataset, thereby obtaining a final first sample dataset;

[0015] Based on the first sample data set, the yolov8n target detection model is used to perform model training, optimize the model structure, and obtain the target detection model of the plant stems and leaves.

[0016] Furthermore, the data processing module includes segmenting the leaf part and the stem part in the plant morphology image through a trained plant stem and leaf instance segmentation model, and the training method includes:

[0017] Obtain a second sample dataset, wherein the second sample dataset includes target images detected by the target detection model of plant stems and leaves, the target images including partial images of the stems and leaves, annotating the acquired images, accurately annotating the leaves and stems using irregular polygonal rectangular frames, and then rotating, enlarging, and cropping the annotated images to complete data enhancement operations on the dataset, thereby obtaining a final second sample dataset;

[0018] Based on the second sample data set, the yolov8n-seg instance segmentation model is used to train the model, optimize the model structure, and obtain the instance segmentation model of the plant stems and leaves.

[0019] Furthermore, the plant morphological parameter extraction module detects the category of leaf pests and diseases through a target detection model of plant leaf pests and diseases, and the training method includes:

[0020] Obtaining a third sample data set, wherein the third sample data set includes classified images of various types of pests and diseases on the leaves of the detected plants, annotating the acquired images by using rectangular boxes to mark different types of pests and diseases on the leaves, and then rotating, enlarging, and cropping the annotated images to complete data enhancement operations on the data set to obtain a final third sample data set;

[0021] Based on the third sample data set, the yolov8n target detection model is used to train the model, optimize the model structure, and obtain the target detection model for plant leaf pests and diseases.

[0022] Furthermore, the morphological parameters of the various plants include two-dimensional morphological outlines of leaves and two-dimensional morphological outlines of stems.

[0023] Furthermore, the evaluation strategy of the plant growth status evaluation module is divided into basic status evaluation and abnormality cause evaluation, wherein:

[0024] Analyze two levels, first detect whether the basic parameters are abnormal;

[0025] If not, it indicates that the plant is in good condition; if so, it indicates that the plant's basic condition is abnormal;

[0026] When a basic anomaly is detected, the cause of the anomaly is located through a combination of multiple plant parameter information.

[0027] A method for intelligent monitoring of plant growth morphology based on multi-parameter fusion, comprising:

[0028] S100: Acquire plant morphology pictures;

[0029] S200: Identifying and classifying the acquired plant morphology image, and segmenting the leaf portion and stem portion of the plant morphology image;

[0030] S300: Processing the images of the segmented leaf and stem parts to obtain morphological parameters of various plants, calculating leaf area, leaf aspect ratio, color HSV eigenvalue, stem diameter change rate, leaf area daily change rate, and performing pest and disease identification to generate plant parameter information;

[0031] S400: receiving plant parameter information and judging the plant growth status based on the various plant parameter information.

[0032] It can be seen from the above technical solution that compared with the existing technology, the present invention discloses a plant growth morphology intelligent monitoring system and method based on multi-parameter fusion, which is particularly suitable for continuously monitoring changes in plant morphological and physiological indicators. Plant image data is obtained through an imaging module, and the two-dimensional morphological contours of plant leaves and stems are obtained in real time in combination with a deep learning model. The seven core parameters such as leaf area, leaf aspect ratio, stem-leaf ratio, color HSV eigenvalue, stem thickness change rate, and leaf area change rate are calculated, and a plant disease and pest identification function is provided. The plant status is judged by the obtained parameters, and the judgment results are provided to users for reference, so that users can take corresponding measures in time to provide plants with a suitable environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0034] Figure 1 Schematic diagram of the system structure of the present invention;

[0035] Figure 2 Schematic diagram of the method flow of the present invention;

[0036] Figure 3 Specific implementation process of the multi-parameter fusion plant growth morphology intelligent monitoring system;

[0037] Figure 4 This is a schematic diagram of the imaging module for shooting;

[0038] Figure 5 Example image of leaves identified by the object detection model;

[0039] Figure 6 Example image of leaf segmentation for instance segmentation model;

[0040] Figure 7 The image of the leaf part saved after segmentation;

[0041] Figure 8 The following are examples of the original leaf image and binary image;

[0042] Figure 9 This is an example diagram of the minimum circumscribed rectangle of the blade outline;

[0043] Figure 10An example diagram of the target detection model detecting pests and diseases (early blight). DETAILED DESCRIPTION

[0044] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0045] Example 1:

[0046] See also Figure 1 The embodiment of the present invention discloses a plant growth morphology monitoring system based on multi-parameter fusion. The monitoring system includes: an imaging module, a data processing module, a plant morphology parameter extraction module, and a plant growth status evaluation module. The monitoring system is used to obtain multiple morphological parameters of plants, evaluate the plant growth status based on each parameter, and provide feedback of the plant growth status to the user.

[0047] Imaging module, used to obtain plant morphology images;

[0048] The data processing module is used to identify and classify the acquired plant morphology images and further segment them into leaf and stem parts in the images, and accurately segment the leaf and stem parts of the images;

[0049] The plant morphological parameter extraction module is used to perform image processing operations on the segmented leaf and stem images to obtain various plant morphological parameters: the two-dimensional morphological contours of the leaves and stems, and calculate the leaf area, leaf aspect ratio, color HSV eigenvalue, stem diameter change rate, leaf area daily change rate, and identify plant diseases and pests in plant images;

[0050] The plant growth status evaluation module is used to receive the plant parameter information generated by the plant morphological parameter extraction module and evaluate the plant growth status based on the various plant parameter information.

[0051] Specifically, the data processing module includes a target detection model training method for plant stems and leaves, and the training method includes:

[0052] Obtain a sample dataset; the sample dataset includes overall morphological images of the tested plants at various growth stages. Label the acquired images, using rectangular boxes to mark the plant stems and leaves. Then, perform operations such as rotating, enlarging, and cropping the labeled images to complete data augmentation operations on the dataset and obtain the final dataset.

[0053] Based on the dataset, the yolov8n target detection model is used to train the model, optimize the model structure, and obtain the target detection model of plant stems and leaves;

[0054] Specifically, the plant morphological parameter extraction module includes a target detection model training method for plant leaf pests and diseases, and the training method includes:

[0055] Obtain a sample dataset; the sample dataset includes classified images of various types of pests and diseases on the leaves of the tested plants. Label the acquired images, using rectangular boxes to mark the different types of pests and diseases on the leaves. Then, perform operations such as rotating, enlarging, and cropping the labeled images to complete the data enhancement operation of the dataset and obtain the final dataset.

[0056] Based on the data set, the yolov8n target detection model was used to train the model, optimize the model structure, and obtain the target detection model for plant leaf pests and diseases;

[0057] Specifically, the data processing module includes an instance segmentation model training method for plant stems and leaves, and the training method includes:

[0058] Obtain a sample dataset; the sample dataset includes target images detected by the target detection model for plant stems and leaves (including partial images of stems and leaves), annotate the acquired images, use irregular polygonal rectangular boxes to accurately annotate the leaves and stems, and then perform operations such as rotating, enlarging, and cropping on the annotated images to complete data enhancement operations on the dataset and obtain the final dataset;

[0059] Based on the dataset, the yolov8n-seg instance segmentation model is used to train the model, optimize the model structure, and obtain the instance segmentation model of plant stems and leaves;

[0060] Specifically, the plant morphological parameter extraction module includes:

[0061] The instance segmentation model accurately segments the leaf part of the plant in the image to obtain the leaf's two-dimensional morphological contour.

[0062] The instance segmentation model accurately segments the stem of the plant in the image, thereby obtaining the two-dimensional morphological outline of the stem.

[0063] Leaf area: When the shooting distance is fixed and the image is calibrated by a standard reference object, the leaf area can be calculated based on the number of pixels in the image of the segmented leaf part. The leaf area calculation formula is:

[0064] A l =A r *N l

[0065] Among them, A l is the leaf area, A r is the actual corresponding area of a pixel in the image, N l is the number of pixels in the image;

[0066] A r The calculation formula is:

[0067] A r =A s / N s

[0068] Among them A r A is the actual corresponding area of a pixel in the image, s is the area of the standard reference object, N s The number of pixels that a standard reference object occupies in a photo taken at a fixed distance;

[0069] The aspect ratio of the leaf can be calculated based on the image of the segmented leaf part and the minimum rectangle enclosing the leaf by the number of pixels. The calculation formula is:

[0070] R l =C l / K l

[0071] where R l is the blade aspect ratio, C l K is the number of pixels of the minimum circumscribed rectangle of the leaf. l The number of pixels of the minimum circumscribed rectangle of the leaf;

[0072] The leaf color HSV eigenvalue can be used to extract the main colors from the leaf area based on the segmented leaf part image using the K-means clustering algorithm, and convert the color analysis from the RGB space to the HSV (hue, saturation, value) space, taking into account the effects of saturation (S) and value (V);

[0073] The stem diameter change rate can be calculated by the number of pixels based on the image of the segmented stem part and the minimum circumscribed matrix of the stem. The stem diameter parameters are obtained in different time periods and compared with the previous time. The stem diameter change rate can be calculated using the following formula:

[0074]

[0075] Where V j is the stem diameter change rate, C j1 、C j2 is the stem diameter measured at different moments, t1 and t2 are two different measurement times;

[0076] The leaf area change rate can be calculated based on the image of the segmented leaf part, the leaf area at different times is calculated based on the number of pixels in the image, and then compared and calculated. The leaf area change rate can be calculated. The leaf area change rate calculation formula is:

[0077]

[0078] Among them, V l is the leaf area change rate, A l1 、A l2 is the leaf area measured at different times, t1 and t2 are two different measurement times;

[0079] The stem-to-leaf ratio is the area ratio of the stem to the leaf. The formula for calculating the stem-to-leaf ratio is:

[0080]

[0081] Among them, R jl is the stem-to-leaf ratio, A j is the stem area, A l is the leaf area;

[0082] Leaf pest and disease type: Use the target detection model for plant leaf pests to detect rectangular images containing plant leaves to determine whether the leaves are infected with pests and diseases and the types of pests and diseases.

[0083] Specifically, the specific evaluation strategy of the plant growth status evaluation module is as follows:

[0084] The evaluation strategy is divided into basic status judgment and abnormal cause;

[0085] Analyze two levels, first detect whether the basic parameters are abnormal;

[0086] If no, it indicates that the plant is in good condition; if yes, it indicates that the plant’s basic condition is abnormal;

[0087] When a basic anomaly is detected, the cause is further located through parameter combination;

[0088] On the other hand, see Figure 2 This embodiment discloses a method for intelligently monitoring plant growth morphology based on multi-parameter fusion, comprising:

[0089] S100: Acquire plant morphology pictures;

[0090] S200: Identifying and classifying the acquired plant morphology image, and segmenting the leaf portion and stem portion of the plant morphology image;

[0091] S300: Processing the images of the segmented leaf and stem parts to obtain morphological parameters of various plants, calculating leaf area, leaf aspect ratio, color HSV eigenvalue, stem diameter change rate, leaf area daily change rate, and performing pest and disease identification to generate plant parameter information;

[0092] S400: receiving plant parameter information and judging the plant growth status based on the various plant parameter information.

[0093] Example 2:

[0094] See also Figure 3 The specific execution process of the plant growth morphology intelligent monitoring system based on multi-parameter fusion provided by the present invention is as follows:

[0095] Process 1: The imaging module uses an RGB camera to shoot plants at a certain angle to obtain plant phenotypic morphological images, and further transmits the images to the data processing module. Figure 4 This is a schematic diagram for shooting.

[0096] RGB cameras can be used for shooting. Industrial-grade cameras with 20 megapixels or higher (such as the Sony IMX477 sensor) are recommended. The higher the camera pixel, the more accurate the results. Different camera pixels will result in different actual physical areas corresponding to the pixels in the obtained plant phenotypic morphology images. Calibrate the actual physical area corresponding to the camera pixels at a fixed distance using standard reference objects in advance.

[0097] Shooting angle: Fixing the RGB camera on a bracket at a certain distance from the canopy allows capturing top and side views of the plant from this angle;

[0098] Shooting time: you can set the time to automatically take a number of top view and side view images at a fixed time every day, for example, automatically take 3 top view and 3 side view images at 9:00 and 21:00 every day;

[0099] Use fill light for shooting. Use an LED ring fill light (color temperature 5500K, light intensity ≥10000 lux) to avoid the influence of natural light fluctuations on color analysis.

[0100] Process 2: The target detection model in the data processing module first selects the leaves and stems in the image in a rectangular shape, and initially locks the object. Figure 5 Example image of leaves identified for an object detection model.

[0101] The target detection model needs to be pre-trained. The training methods include:

[0102] For shooting, you need to take several pictures of the object to be tested. For example, take 1,000 pictures of the top and side of a tomato at different growth stages.

[0103] Label the image. Use the labelimg tool to label the leaves and stems in the image with rectangular boxes.

[0104] Data augmentation: rotating, zooming in, and cropping the labeled images to increase the complexity of the data and improve the robustness of the trained model. After data augmentation, the dataset required for the final training is obtained.

[0105] Target detection model training, using the improved yolov8n for model training, to obtain a target detection model that can detect tomato leaves and stems;

[0106] Process 3: The instance segmentation model in the data processing module further accurately segments the targets detected by the target detection model, and separates the leaves and stems within the rectangular frame of the image. Figure 6 The instance segmentation model segments the leaf example, and then transmits the segmented part of the image to the plant morphological parameter extraction module. Figure 7 The image of the leaf part is saved after segmentation.

[0107] The instance segmentation model needs to be pre-trained. The training methods include:

[0108] The initial dataset comes from the object images identified by the object detection model in process 2. For example, after the object detection model identifies the tomato leaves and stems, it will select the objects with a rectangular box and save the contents of the rectangular box. The saved content is the initial dataset required for instance segmentation model training;

[0109] Label the images in the initial dataset and use the labelme annotation tool to accurately label the leaves and stems in the images with irregular polygons;

[0110] Data augmentation: rotating, zooming in, and cropping labeled images to increase the complexity of the data and improve the robustness of the trained model.

[0111] Instance segmentation model training, using the improved yolov8n-seg model to train the model, and obtain an instance segmentation model that can accurately detect and segment tomato leaves and stems;

[0112] Process 4: The plant morphological parameter extraction module performs a series of image processing operations on the received pictures to obtain a variety of plant phenotypic parameters, and further transmits the obtained parameters to the plant growth status evaluation module.

[0113] The parameters obtained include: two-dimensional morphological outline of leaves and stems, leaf area, leaf aspect ratio, stem-to-leaf ratio, leaf color HSV characteristic value, stem diameter change rate, leaf area daily change rate, whether there are pests and diseases and the type of pests and diseases;

[0114] The two-dimensional morphological contours of the leaves and stems are obtained from the instance segmentation model described in process 3. The instance segmentation model accurately identifies and segments the tomato leaves and stems and saves them as binary images. Figure 8 The original image of the leaf and the binary image are examples. Specifically, the instance segmentation model detects each mask of the image, marks the mask containing the object as 1, and marks the mask without the object as 0. Then the image is output as a binary image. When the mask is 1, the pixel value of the binary image is 1, and when the mask is 0, the pixel value of the binary image is 0. This binary image is the two-dimensional morphological outline of the leaf and stem;

[0115] Leaf area: When the shooting distance and camera pixel are fixed, the actual physical area corresponding to each pixel in the image is the same. To obtain the leaf area, it is necessary to calculate the actual physical area corresponding to the camera pixel at the shooting distance in advance using a standard reference object. Then, the leaf area can be calculated by the number of pixels with a pixel value of 1. The leaf area calculation formula is:

[0116] A l =A r *N l

[0117] Among them, A l is the leaf area, A r is the actual corresponding area of a pixel in the image, N l is the number of pixels in the image;

[0118] A r The calculation formula is:

[0119] A r =A s / N s

[0120] Among them A r A is the actual corresponding area of a pixel in the image, s is the area of the standard reference object, N s The number of pixels that a standard reference object occupies in a photo taken at a fixed distance;

[0121] The aspect ratio of the leaf is obtained by enclosing the leaf outline into a minimum rectangle through image processing. Figure 9 This is an example diagram of the minimum circumscribed rectangle of the leaf outline, where the leaf outline comes from the two-dimensional morphological outline of the leaf and stem. The leaf aspect ratio can be obtained by calculating the aspect ratio of the rectangle. The calculation formula is:

[0122] Rl =C l / K l

[0123] where R l is the blade aspect ratio, C l K is the number of pixels of the minimum circumscribed rectangle of the leaf. l The number of pixels of the minimum circumscribed rectangle of the leaf;

[0124] The stem-to-leaf ratio is the ratio of the area of the stem to the area of the leaf. The leaf area parameter is derived from the leaf area. The method for obtaining the stem area is the same as that for obtaining the leaf area. The formula for calculating the stem-to-leaf ratio is:

[0125]

[0126] Among them, R jl is the stem-to-leaf ratio, A j is the stem area, A l is the leaf area;

[0127] Leaf color HSV eigenvalues. Use the K-means clustering algorithm to extract the main colors from the leaf area image and convert the image color analysis from RGB space to HSV:

[0128] Among them, the HSV feature value, hue (Hue, H) reflects the dominant color of the leaves, saturation (S) represents the color vividness, and value (V) represents the reflective ability of the leaves;

[0129] K-means is an unsupervised clustering algorithm that iteratively divides data into K clusters, ensuring the highest similarity within a cluster and the greatest differences between clusters. It is used to extract the HSV values of the dominant colors of leaves and stems in plant monitoring.

[0130] The leaf region image is obtained from the detection result of the instance segmentation model;

[0131] The stem diameter change rate is calculated by image processing, where a minimum rectangle is drawn around the stem outline, where the stem outline is derived from the two-dimensional morphological outline of the leaf and stem. The stem diameter is obtained by calculating the width of the rectangle. The stem diameter change rate is then calculated by comparing the difference in stem diameter at different times. The calculation formula is:

[0132]

[0133] Where V j is the stem diameter change rate, C j1 、C j2 is the stem diameter measured at different moments, t1 and t2 are two different measurement times;

[0134] Leaf area change rate can be calculated by comparing the difference in leaf area on different dates. The leaf area change rate calculation formula is:

[0135]

[0136] Among them, V l is the leaf area change rate, A l1 、A l2 is the leaf area measured at different times, t1 and t2 are two different measurement times;

[0137] Whether there are pests and diseases and the types of pests and diseases, by training a target detection model for tomato pest and disease identification, the trained model is used to detect the leaf part segmented in process 3. If there is no identification target, it means there are no pests and diseases. If pests and diseases are detected, it means there are pests and diseases. The pests and diseases are the types of pests and diseases detected by the model. Figure 10 This is an example of a target detection model detecting pests and diseases (early blight). The target detection model for pests and diseases requires pre-training. The training steps include:

[0138] The initial dataset is from the leaf images identified and segmented by the instance segmentation model in process 3, from which images of leaves with pests and diseases are selected;

[0139] Label the images in the initial dataset and use the labelimg annotation tool to mark the patient parts in the images with rectangles;

[0140] Data augmentation: rotating, zooming in, and cropping labeled images to increase the complexity of the data and improve the robustness of the trained model.

[0141] Target detection model training, using the improved yolov8n for model training, to obtain a target detection model that can identify leaf pests and diseases;

[0142] Process 5: The plant growth status evaluation module receives the acquired plant phenotypic parameters, gives a detailed evaluation based on the evaluation strategy, and provides real-time feedback on the plant growth status to the user.

[0143] The evaluation strategy is divided into two levels: basic status determination and abnormal cause analysis. First, it detects whether the basic status is abnormal. When a basic abnormality is detected, the cause is further located through parameter combination. The specific example is as follows:

[0144] Example 1, basic status determination:

[0145] Parameter information: HSV characteristics of leaves (yellowing / brown spots); Evaluation result: severe disease / senescence; Physiological significance: severe chlorophyll degradation (diseases and insect pests, nutritional deficiencies or natural senescence).

[0146] Analysis of abnormal causes:

[0147] Parameter information: HSV characteristics (H value ↑ (yellowing) + S value ↓ (fading)) + normal stem diameter change rate + pests and diseases (early blight); Evaluation result: The plant may have an abnormality, abnormality type: leaf yellowing; possible causes: magnesium / iron deficiency (yellowing of old leaves), early blight;

[0148] Example 2, basic status determination:

[0149] Parameter information: Leaf area change rate ≤ 1% (lasting for 5 days); Evaluation result: Metabolic inhibition; Physiological significance: Photosynthesis is blocked (insufficient light, CO2 deficiency or water stress).

[0150] Analysis of abnormal causes:

[0151] Parameter information: Leaf area change rate ↓ + stem diameter change rate ↓ + HSV (dark green); Abnormal type: Growth retardation; Possible cause: Low temperature stress;

[0152] Compared with the prior art, the present invention has the comprehensive evaluation capability of multi-parameter fusion. For the first time, the present invention integrates and analyzes 7 core parameters including leaf morphological parameters (area, aspect ratio), stem growth parameters (stem diameter change rate), physiological dynamic parameters (change rate, stem-to-leaf ratio), color features (HSV spatial analysis) and pest and disease identification, which significantly improves the comprehensiveness and accuracy of the evaluation. High-precision segmentation and anti-interference ability: the present invention adopts a target detection algorithm to identify whether a plant is diseased or pest-infected, which greatly improves the detection accuracy, detection speed and timeliness. An instance segmentation algorithm is used to identify plant leaves and stems, and its segmentation accuracy and anti-interference ability are significantly improved compared to traditional image processing methods (such as threshold segmentation and edge detection). Real-time dynamic monitoring and early warning: the system supports continuous monitoring and performs full parameter analysis on plants, which ensures the real-time nature of the information and can provide timely warnings.

[0153] Specifically, by collecting plant data sets and performing data enhancement on the data sets, the target detection model has higher anti-interference ability and accuracy, thereby improving the accuracy of the final evaluation results.

[0154] Specifically, by improving the target detection model structure, the target detection model can occupy smaller storage and computing resources while maintaining the original recognition accuracy, thereby reducing the cost of deploying the model.

[0155] Specifically, an instance segmentation model was used to segment leaves and stems. The high-precision segmentation made the subsequent evaluation results more accurate.

[0156] Specifically, a variety of plant phenotypic parameters can be obtained contactlessly from plant images alone.

[0157] Specifically, HSV color space analysis was used to convert color analysis from RGB space to HSV (hue, saturation, value) space, taking into account the effects of saturation (S) and value (V) on plant leaf color, thereby reducing the interference of light on the assessment of plant growth status.

[0158] Specifically, through comprehensive evaluation based on multi-parameter fusion, the evaluation of plant growth status is more accurate, significantly improving the comprehensiveness and accuracy of the evaluation.

[0159] Specifically, through real-time dynamic monitoring of plants, the real-time nature of information on plant growth status is guaranteed, and timely warnings can be issued in the early stages when plant growth is poor, thus avoiding greater losses later.

[0160] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0161] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An intelligent monitoring system for plant growth morphology based on multi-parameter fusion, characterized in that: include: Imaging module, data processing module, plant morphological parameter extraction module, plant growth status evaluation module, among which, The imaging module is used to obtain plant morphology images; The data processing module is used to identify and classify the acquired plant morphology pictures, and segment the leaf part and the stem part in the plant morphology pictures; The plant morphological parameter extraction module is used to perform image processing on the images of the segmented leaf and stem parts to obtain the morphological parameters of various plants, calculate the leaf area, leaf aspect ratio, color HSV eigenvalue, stem diameter change rate, leaf area daily change rate, and perform pest and disease identification to generate plant parameter information; The plant growth status evaluation module is used to receive the plant parameter information generated by the plant morphological parameter extraction module and evaluate the plant growth status based on the various plant parameter information.

2. The plant growth morphology intelligent monitoring system based on multi-parameter fusion according to claim 1 is characterized in that: The data processing module includes identifying and classifying the acquired plant morphology images using a trained plant stem and leaf target detection model, and the training method includes: Obtaining a first sample dataset, wherein the first sample dataset includes overall morphological images of the detected plants at various growth stages, annotating the acquired images by using rectangular boxes to mark the plant stems and leaves, and then rotating, enlarging, and cropping the annotated images to complete data enhancement operations on the dataset, thereby obtaining a final first sample dataset; Based on the first sample data set, the yolov8n target detection model is used to perform model training, optimize the model structure, and obtain the target detection model of the plant stems and leaves.

3. The plant growth morphology intelligent monitoring system based on multi-parameter fusion according to claim 1 is characterized in that: The data processing module includes segmenting the leaf part and the stem part in the plant morphology image using a trained plant stem and leaf instance segmentation model, and the training method includes: Obtain a second sample dataset, wherein the second sample dataset includes target images detected by the target detection model of plant stems and leaves, the target images including partial images of the stems and leaves, annotating the acquired images, accurately annotating the leaves and stems using irregular polygonal rectangular frames, and then rotating, enlarging, and cropping the annotated images to complete data enhancement operations on the dataset, thereby obtaining a final second sample dataset; Based on the second sample data set, the yolov8n-seg instance segmentation model is used to train the model, optimize the model structure, and obtain the instance segmentation model of the plant stems and leaves.

4. The plant growth morphology intelligent monitoring system based on multi-parameter fusion according to claim 1, characterized in that: The plant morphological parameter extraction module detects the category of leaf pests and diseases through a target detection model of plant leaf pests and diseases, and the training method includes: Obtaining a third sample data set, wherein the third sample data set includes classified images of various types of pests and diseases on the leaves of the detected plants, annotating the acquired images by using rectangular boxes to mark different types of pests and diseases on the leaves, and then rotating, enlarging, and cropping the annotated images to complete data enhancement operations on the data set to obtain a final third sample data set; Based on the third sample data set, the yolov8n target detection model is used to train the model, optimize the model structure, and obtain the target detection model for plant leaf pests and diseases.

5. The plant growth morphology intelligent monitoring system based on multi-parameter fusion according to claim 1 is characterized in that: The morphological parameters of the various plants include two-dimensional morphological outlines of leaves and two-dimensional morphological outlines of stems.

6. The plant growth morphology intelligent monitoring system based on multi-parameter fusion according to claim 1, characterized in that: The evaluation strategy of the plant growth status evaluation module is divided into basic status evaluation and abnormality cause evaluation, wherein: Analyze two levels, first detect whether the basic parameters are abnormal; If not, it indicates that the plant is in good condition; if so, it indicates that the plant's basic condition is abnormal; When a basic anomaly is detected, the cause of the anomaly is located through a combination of multiple plant parameter information.

7. A method for intelligent monitoring of plant growth morphology based on multi-parameter fusion, characterized in that: include: S100: Acquire plant morphology pictures; S200: Identifying and classifying the acquired plant morphology image, and segmenting the leaf portion and stem portion of the plant morphology image; S300: Processing the images of the segmented leaf and stem parts to obtain morphological parameters of various plants, calculating leaf area, leaf aspect ratio, color HSV eigenvalue, stem diameter change rate, leaf area daily change rate, and performing pest and disease identification to generate plant parameter information; S400: receiving plant parameter information and judging the plant growth status based on the various plant parameter information.

Images