Crop growth detection system and method based on machine vision

Abstract

The invention relates to the technical field of agricultural intelligent monitoring, in particular to a crop growth detection system and method based on machine vision, and the system comprises an image collection module, a morphological feature capture module, a growth trend judgment module, an abnormal region marking module and a state information output module. According to the method, crop images are collected through a multi-angle camera, leaf contours, stem bending and plant spacing are extracted, multi-dimensional modeling of morphology is realized, structural change identification is enhanced, time sequence comparison of key morphological characteristics is realized, identification precision and time efficiency are improved, leaf and stem change trends are continuously analyzed, offset is quantified, and growth abnormity is early warned in advance; health degradation identification is combined with fluctuation area marking, dynamic monitoring is achieved, high-risk positioning is carried out on a time overlapping area, time-space locking is enhanced, precise management is assisted, the process is from image analysis to abnormal focusing, an information reasoning chain is established, and the monitoring precision and response efficiency are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of agricultural intelligent monitoring, and particularly relates to a crop growth detection system and method based on machine vision. Background Art

[0002] The technical field of agricultural intelligent monitoring includes using advanced automation technologies and informatization means to achieve data collection, analysis, and processing in agricultural production, so as to achieve real-time monitoring and precise management of the crop growth environment and growth status. The core contents of this technical field include agricultural environment perception, crop growth status detection, pest and disease prediction and control, and intelligent agricultural production decision support. Agricultural intelligent monitoring technology realizes the monitoring of factors such as farmland environment changes, soil quality, crop growth, and climate through the application of devices such as sensors, drones, and satellite remote sensing, combined with data processing and analysis platforms, promoting the development of agriculture towards precision, high efficiency, and sustainable development.

[0003] Among them, a crop growth detection system refers to a system that uses means such as sensors, image processing technology, and Internet of Things technology to monitor the growth status of crops in real time. Through monitoring devices set in the field, data such as soil humidity, temperature, light intensity, climate conditions, and crop growth images are collected to continuously track the growth of crops. The design of the system considers data collection and transmission. Combining image recognition technology, it can accurately judge the growth progress and health status of crops. The system adopts automated control technology to analyze and process data in real time, thereby providing decision support for agricultural management.

[0004] The existing technologies mainly rely on fixed sensors to collect environmental parameters and image data. The processing path is mostly based on static analysis after data collection, lacking the ability to identify the multi-dimensional spatial structure of the crop status in the image. It only relies on the change of leaf color or a single image to infer the growth status, making it difficult to comprehensively reflect the true morphological evolution of the crop. In addition, the current system has a lag in processing the time relationship of data, lacking real-time capture and marking of key growth nodes, unable to accurately track growth transitions on the time axis, and easily missing early signals of growth abnormalities. The assessment of crop trends generally uses average values or threshold judgments, unable to form quantitative quantification of the growth change trends in continuous stages, resulting in fuzzy discrimination of growth degradation and easy loss of accuracy in management intervention. In terms of spatial recognition, common systems mostly use the regional mean method for problem areas in farmland, lacking dynamic anomaly focusing based on morphological trend evolution, and it is difficult to lock in local risks in a high-density crop environment. For example, in a certain crop field, when there is a growth abnormality caused by local light shading, the traditional system is difficult to accurately mark according to local structural changes, easily causing abnormal omissions or misjudgments. The existing technologies are insufficient in multi-dimensional data recognition, time-series trend tracking, and risk focusing, limiting their practicability and response ability in complex agricultural environments. Summary of the Invention

[0005] The object of the present invention is to solve the disadvantages existing in the prior art, and to propose a crop growth detection system and method based on machine vision.

[0006] To achieve the above object, the present invention adopts the following technical solutions: A crop growth detection system based on machine vision includes:

[0007] The image acquisition module obtains crop canopy images through multi-angle cameras set in the field, extracts the changes in the leaf edge contour, the degree of stem bending, and the plant spacing distribution, and performs consistency matching with the overall growth state of the crop according to the spatial distribution characteristics of the three parameters to generate a crop morphological feature sequence;

[0008] The morphological feature capture module reads the crop morphological feature sequence, detects whether the corresponding node positions in the time axis are consistent with the critical points of crop growth state changes. If there is a time coincidence, the corresponding morphological features are marked as key features to generate a list of crop morphological key nodes;

[0009] The growth trend determination module counts the crop image data of two consecutive periods through the list of crop morphological key nodes, compares the leaf expansion area and the stem elongation length in each period respectively, evaluates the change range between the two in the periods. If the change direction turns, it is compared with the original growth benchmark to obtain crop growth trend deviation information;

[0010] The abnormal area marking module detects whether there is degradation in the crop health state during the corresponding time period according to the crop growth trend deviation information, marks the abnormal area, and generates a set of crop growth fluctuation areas.

[0011] As a further solution of the present invention, the crop morphological feature sequence includes morphological distribution characteristics, consistency matching frequency, and growth state stability. The list of crop morphological key nodes includes key feature names, critical points of growth state changes, and time node mapping relationships. The crop growth trend deviation information includes change range, turning direction, and benchmark deviation state. The set of crop growth fluctuation areas includes turning nodes, health degradation marked areas, and key areas of fluctuations within the period.

[0012] As a further solution of the present invention, the image acquisition module includes:

[0013] The viewing angle adjustment sub-module obtains crop canopy images through multi-angle cameras set in the field, adjusts the pitch angle and focal length of the camera, extracts the leaf edge contour, the degree of stem bending, and the plant spacing distribution, and generates a sequence of crop spatial distribution images;

[0014] The feature extraction sub-module analyzes the changing trends of the leaf edge contour, the degree of stem bending, and the plant spacing distribution over continuous time based on the crop spatial distribution image sequence, compares the changing directions of the three with the overall growth state of the crop, calculates the frequency of occurrence of consistency, and obtains a frequency sequence of consistent morphology and growth state;

[0015] The sequence screening sub-module selects the morphological parameters with the optimal frequency of occurrence according to the frequency sequence of consistent morphology and growth state, identifies the corresponding original parameters, and establishes a crop morphological feature sequence.

[0016] As a further solution of the present invention, the morphological feature capture module includes:

[0017] The node positioning sub-module obtains the time points corresponding to the morphological parameters in the crop morphological feature sequence, extracts the extreme value moments of the parameters as representative nodes, sets a periodic time axis, takes three adjacent periods as a sliding window, identifies the local extreme value points in each sequence and marks the key nodes, and obtains a set of key morphological time nodes;

[0018] The coincidence detection sub-module extracts the growth state values corresponding to the time nodes based on the set of key morphological time nodes, determines that the change sign reversal between the previous and the next moments exceeds the fluctuation threshold as the critical point, records the critical point time, analyzes the correlation degree of the coincidence area between the morphological features and the critical point time, and identifies the morphological features with a correlation degree higher than the threshold to obtain a set of matching morphological features;

[0019] The feature registration sub-module marks the features and morphological parameters at the coincidence time points according to the set of matching morphological features, calculates the key feature values, records the corresponding time nodes, and integrates the indexes, morphological names, time point numbers, and growth state fluctuation direction values of the key features to establish a crop morphological key node list.

[0020] As a further solution of the present invention, the growth trend determination module includes:

[0021] The area comparison sub-module obtains the crop image data of two consecutive periods through the crop morphological key node list, classifies each period according to the leaf expansion area and the stem elongation length, compares the measured values of the time nodes of each category, identifies the change values of the leaf area and the stem elongation and arranges them according to the period to generate a periodic growth change sequence;

[0022] The direction evaluation sub-module evaluates the change direction turn of the current and the previous period data according to the periodic growth change sequence, identifies the turning phenomenon and records it, analyzes the corresponding deviation amplitude between the leaf and the stem, and obtains the growth trend deviation degree;

[0023] The benchmark comparison sub-module extracts the growth change data of the original continuous cycles of the crop according to the growth trend offset degree, identifies the benchmark range of growth changes, compares the offset degree of the current cycle with the benchmark range of growth changes, and if it exceeds the benchmark range, marks the corresponding cycle as an abnormal section of growth fluctuation, and obtains the crop growth trend offset information.

[0024] As a further solution of the present invention, the abnormal area marking module includes:

[0025] The steering node extraction sub-module extracts the time identifier of each steering node based on the marked direction steering node time points in the crop growth trend offset information, records them in chronological order, and after determining the steering behavior, regards it as a valid steering and relocates the time to establish a steering node time series;

[0026] The health degradation detection sub-module identifies the continuous data of the crop health state in the adjacent three cycles according to the time points corresponding to each node in the steering node time series, performs a sliding analysis, extracts the health state values within the time points, and compares whether there is degradation. If there is, it is determined as a health degradation state and a health degradation trend sequence is generated;

[0027] The fluctuation area screening sub-module extracts the time points that exist in both sequences according to the health degradation trend sequence, combines the index positions corresponding to the intersection of the two, identifies the key cycle areas that meet the conditions of synchronous steering and health degradation, encodes them into a set uniformly, and generates a set of crop growth fluctuation areas.

[0028] As a further solution of the present invention, the system further includes:

[0029] The status information output module compares the set of crop growth fluctuation areas with the list of morphological key nodes in terms of time, determines whether there is an overlap in the time windows of adjacent cycles. If there is, it lists the corresponding areas as risk concerns and outputs the crop growth status information;

[0030] The crop growth status information includes risk concerns, the overlap status of cycle time windows, and abnormal trend identifiers.

[0031] As a further solution of the present invention, the status information output module includes:

[0032] The window overlap recognition sub-module extracts the abnormal areas and the start and end times of the operation cycles based on the cycle time nodes in the set of crop growth fluctuation areas and the list of morphological key nodes, combines the standard cycle duration to identify the upper and lower bounds of the standard time window, matches the start and end time periods of the areas, analyzes the cycle overlap relationship, and generates time window overlap range data;

[0033] The overlapping interval determination sub-module extracts the start and end time points of the overlap according to the time window overlap range data, combines the regional activity frequency, duration, and the difference in the cycle start point, calculates the overlap relationship index, and compares it with a threshold. If it exceeds the threshold, it is marked as a valid overlap cycle to obtain a sequence of valid overlap cycles.

[0034] The status information generation sub-module extracts the region index according to the sequence of valid overlap cycles, checks whether it is a member of an abnormal region. If so and it is within the cycle overlap interval, it is marked as a risk region and included in the risk concern set, and the crop growth status information is output.

[0035] The machine vision-based crop growth detection method is executed based on the above-mentioned machine vision-based crop growth detection system, and includes the following steps:

[0036] S1: Obtain crop canopy images through multi-angle cameras set in the field, extract the changes in the leaf edge contour, the degree of stem bending, and the plant spacing distribution, judge the consistency between the changes and the overall growth status of the crop, screen the morphological features, and generate a crop morphological feature sequence.

[0037] S2: According to the time nodes of the crop morphological feature sequence, combine the crop growth status change curve, judge whether the nodes coincide, mark the coincident features as key features and record their positions, and generate a list of crop morphological key nodes.

[0038] S3: According to the list of crop morphological key nodes, obtain the crop image data for consecutive cycles, evaluate the change range of the leaf expansion area and the stem elongation length, and judge whether the turning amplitude exceeds the original benchmark to obtain the crop growth trend deviation information.

[0039] S4: According to the turning time points in the crop growth trend deviation information, judge whether the crop health status degenerates during the corresponding period, mark the qualified time points, and generate a set of crop growth fluctuation regions.

[0040] S5: Combine the set of crop growth fluctuation regions and the list of morphological key nodes, judge whether there is a cycle window overlap region, mark it as a risk concern point, and output the crop growth status information.

[0041] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0042] In the present invention, by setting up multi-angle cameras to capture crop canopy images and extracting spatial features such as leaf contours, stem bending, and plant spacing, multi-dimensional modeling of crop morphology is achieved, no longer limited to single-image or single-point data collection, improving the depth and breadth of the recognition of crop structure changes. By comparing time nodes of the morphological feature sequences to detect the synchronization of growth changes and critical moments, a temporal correlation between morphological features and the evolution of growth states is established, obtaining key morphologies during dynamic evolution, improving the timeliness and accuracy of morphological recognition. By periodically analyzing the changing trends of leaf unfolding and stem elongation within consecutive time periods, the inflection point recognition of crop growth potential and the quantitative evaluation of trend deviation are realized, strengthening the pre-warning ability for abnormal growth patterns. Based on the trend deviation information, the degradation recognition of the health state is carried out, which can mark time periods for crop fluctuation areas, breaking the static judgment mode of the result state, introducing time-periodic fluctuation recognition, and making health monitoring more dynamically sensitive. Then, the time-overlapping areas of crop states are focused on as high-risk nodes to achieve the spatio-temporal locking of local problems in the farmland, promoting the precise allocation of management resources. This processing flow, from image feature analysis to temporal trend evaluation, and then to spatial anomaly focusing, constitutes a bottom-up information reasoning chain, strengthening the response ability to dynamic changes in complex growth processes, comprehensively improving the monitoring accuracy, timeliness of risk recognition, and management intervention efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 is the system flow chart of the present invention;

[0044] Figure 2 is the flow chart of the image acquisition module in the present invention;

[0045] Figure 3 is the flow chart of the morphological feature capture module in the present invention;

[0046] Figure 4 is the flow chart of the growth trend determination module in the present invention;

[0047] Figure 5 is the flow chart of the abnormal area marking module in the present invention;

[0048] Figure 6 is the flow chart of the status information output module in the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0049] In order to make the objectives, technical solutions, and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0050] In the description of the present invention, it should be understood that the orientation or positional relationships indicated by the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. are based on the orientation or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation on the present invention. In addition, in the description of the present invention, the meaning of "a plurality of" is two or more, unless otherwise specifically defined.

[0051] Please refer to Figure 1 , the present invention provides a technical solution: a crop growth detection system based on machine vision includes:

[0052] The image acquisition module obtains crop canopy images through multi-angle cameras set in the field, extracts the changes in the leaf edge contours, the bending degree of the stems, and the distribution of plant spacing, and performs consistency matching with the overall growth state of the crops according to the spatial distribution characteristics of the three parameters to generate a crop morphological feature sequence;

[0053] The morphological feature capture module reads the crop morphological feature sequence, detects whether the corresponding node positions in the time axis are consistent with the critical points of the crop growth state changes. If there is a time coincidence, the corresponding morphological features are marked as key features to generate a list of key nodes of crop morphology;

[0054] The growth trend determination module statistically analyzes the crop image data of two consecutive cycles through the list of key nodes of crop morphology, compares the leaf expansion area and the stem elongation length in each cycle respectively, evaluates the change amplitude between the two cycles. If the change direction turns, it is compared with the original growth benchmark to obtain crop growth trend deviation information;

[0055] The abnormal area marking module detects whether there is a degradation in the crop health state during the corresponding time period according to the crop growth trend deviation information, marks the abnormal areas, and generates a set of crop growth fluctuation areas;

[0056] The status information output module compares the set of crop growth fluctuation areas with the list of key nodes of crop morphology in terms of time to determine whether there is an overlap in the time windows of adjacent cycles. If so, the corresponding areas are listed as risk concerns and the crop growth status information is output.

[0057] The crop morphological feature sequence includes morphological distribution characteristics, consistency matching frequency, and growth state stability. The crop morphological key node list includes key feature names, critical points of growth state change, and time node mapping relationships. The crop growth trend deviation information includes the change amplitude, turning direction, and baseline deviation state. The crop growth fluctuation area set includes turning nodes, health degradation marked areas, and key areas of fluctuation within a cycle. The crop growth state information includes risk concern points, overlapping state of cycle time windows, and abnormal trend identification.

[0058] Please refer to Figure 2 , the image acquisition module includes:

[0059] The perspective adjustment sub-module obtains crop canopy images through multi-angle cameras set in the field, adjusts the pitch angle and focal length of the cameras, extracts the leaf edge contours, stem bending degree, and plant spacing distribution, and generates a sequence of crop spatial distribution images;

[0060] By precisely setting the pitch angle and focal length of the cameras, detailed images of the crops can be obtained from different perspectives. To obtain crop canopy images at different growth levels, the adjustment range of the pitch angle of the cameras is set between -15° and +15°, and image acquisition is performed every 5°, thus ensuring coverage of the upper and lower levels of the crops. The adjustment of the focal length determines the clarity and capture range of the images. The focal length is set between 500mm and 1500mm to capture large-scale crop images from a long distance, and the focal length can also be adjusted to focus on nearby details. After image acquisition, image processing technology needs to be used to analyze the images. First, the edge detection algorithm (such as the Canny operator) is used to extract the leaves in the images, and the edge contours of the leaves are accurately identified. Image recognition and curve fitting technology are applied to analyze the bending degree of the stems, identify the curves of each stem, and calculate the bending angle and bending radius of the stems through the fitting algorithm. For the analysis of plant spacing, the image segmentation technology can be used to obtain each plant of the crops and calculate the distance between adjacent plants to form a distribution map of plant spacing. After combining all the image processing data, a series of crop spatial distribution image sequences are formed, which are convenient for subsequent analysis of the growth state and morphological characteristics of the crops. The implementation of this series of steps can be verified through experiments with different shooting conditions, simulating the shooting effects of the cameras under different weather and lighting conditions to ensure stable acquisition of accurate data in practical applications.

[0061] The feature extraction sub-module analyzes the change trends of the leaf edge contours, stem bending degree, and plant spacing distribution in the crop spatial distribution image sequence over continuous time, compares the change directions of the three with the overall growth state of the crops, calculates the frequency of consistency occurrence, and obtains a sequence of consistency frequencies of morphology and growth state;

[0062] Analyze the changes in the leaf edges by calculating the image difference degree. By comparing the images within consecutive time periods, calculate the leaf edge differences between each image and the previous one. The difference degree calculation uses a normalization method. Calibrate the edge contours of each image to obtain a difference value. If this value is greater than the set threshold, it is considered that the leaves have undergone significant changes. For the analysis of the stem bending degree, first extract the morphological changes of the stem by curve fitting in each frame of the image. A common method is to perform least squares fitting through the edge points in the image to calculate the bending angle of the stem. As the crop grows, the bending degree of the stem will increase or decrease. Continuously monitoring the change trend helps to evaluate the health status of the crop. For the distribution of plant spacing, use density function estimation techniques to calculate the distances between plants at different time points, and by analyzing the change trend of the distances, understand the change in the spatial density of crop growth. Combine the changes in leaf edges, the changes in stem bending, and the change trend of plant spacing distribution to further compare the overall growth state of the crop and determine whether there is a phenomenon where the morphology is consistent with the growth state. The analysis process can help accurately grasp the growth dynamics of the crop. Based on the analysis, calculate the frequency of occurrence of consistency. For example, within a certain period of analysis, if the change direction of the leaf edge is consistent with the change in stem bending, and the plant spacing change tends to be moderate, it indicates that the crop is in a good growth state. Through this process, a frequency sequence of morphological and growth state consistency can be obtained, further providing a basis for crop management and regulation.

[0063] The sequence screening sub-module selects the optimal morphological parameters with the highest frequency of occurrence according to the frequency sequence of morphological and growth state consistency, identifies the corresponding original parameters, and establishes a crop morphological feature sequence;

[0064] First, it is necessary to conduct statistical analysis on the frequency sequence of consistent morphology and growth status, count the frequency of each morphological parameter, and obtain the morphological parameters with higher frequencies. For example, within a certain period, if a morphological parameter (such as the stem bending angle) frequently appears in line with the growth status of the crop, it can be considered that this morphological parameter is relatively stable and highly representative during this period. To more precisely select the optimal morphological parameter, the statistical analysis method of the frequency sequence is used to calculate the mean and standard deviation of the frequency sequence. Usually, the parameter with a frequency exceeding the mean plus one standard deviation is selected as the optimal morphological parameter. For example, if the frequency sequence of morphological parameter A is [4, 7, 10, 15, 12, 8], its mean is calculated to be 9 and the standard deviation is 3, then the frequency greater than 15 is considered the optimal morphological parameter. By screening the morphological parameter with the optimal frequency, the original parameters matching the crop growth status, such as the camera angle, focal length, shooting time, etc., can be effectively identified. The original parameters can provide a key basis for subsequent crop monitoring. By associating the original crop parameters with the growth status of the crop, a crop morphological feature sequence is constructed, and the crop is further dynamically monitored at different growth stages to provide accurate data support for agricultural production management. By dynamically updating the morphological feature sequence, farmers can be helped to grasp the growth status of the crop in real time and take appropriate management measures.

[0065] Please refer to Figure 3 , the morphological feature capture module includes:

[0066] The node positioning sub-module obtains the time series points corresponding to the morphological parameters in the crop morphological feature sequence, extracts the parameter extreme value moments as representative nodes, sets a periodic time axis, takes three adjacent periods as a sliding window, identifies the local extreme points in each sequence segment and marks the key nodes, and obtains the set of morphological key time nodes;

[0067] Node positioning requires processing the morphological characteristics of crops in a time series, associating the changes of each morphological parameter with time to generate a series of time points, which indicate the gradually changing state of crop morphological characteristics on the time axis. Taking the leaf edge contour as an example, if the crop is in the rapid growth stage, the edge of the leaf changes more violently, and the interval between time points is shorter. While in the slow growth stage, the interval between time points is longer. After analyzing the time points, the extreme value moments are selected, and the extreme value moments represent the critical moments of crop morphological changes. For example, in the early stage of crop growth, the edge of the leaf will expand rapidly, and as time goes by, the expansion amplitude gradually decreases. At this time, the extreme value moment of the leaf edge change is a key node. By setting a periodic time axis, for example, each period is 10 days, and taking three adjacent periods as a sliding window, local extreme points can be identified from the time points within each sliding window. The local extreme point refers to the moment when significant changes occur in crop morphology during this time period. In this way, the key nodes within each period can be marked. By continuously sliding the window and extracting local extreme points throughout the crop growth process, a complete set of morphological key time nodes can be finally obtained, and the node set can provide important references for subsequent analysis.

[0068] Based on the set of morphological key time nodes, the coincidence detection sub-module extracts the growth state values corresponding to the time nodes, determines that the sign reversal of the change between the previous and subsequent moments and exceeding the fluctuation threshold is the critical point, records the critical point time, analyzes the correlation degree of the coincidence area between the morphological characteristics and the critical point time, and identifies the morphological characteristics with a correlation degree higher than the threshold to obtain the set of matching morphological characteristics;

[0069] The core of coincidence detection lies in revealing the mutual relationship between them by analyzing the changes of morphological characteristics and the fluctuation law of growth state. For example, assume that the change of the leaf edge at a certain key node experiences a transition from expansion to contraction in a short period of time. If the corresponding growth state value also shows the same change trend at this time, it indicates that the crop is at a growth turning point. When this change trend reverses and the change amplitude exceeds the preset fluctuation threshold, this node is identified as a critical point. The setting of the fluctuation threshold can be based on the statistical analysis of historical data. For example, a change with a fluctuation amplitude exceeding ±5% is regarded as a significant fluctuation. In this process, first, the growth state values at each key node moment need to be extracted, and the values are represented by the changes of parameters such as the height of the plant, the number of leaves, and the thickness of the stem. By comparing the growth state at the current moment with that at the previous and subsequent moments, it is judged whether the sign of the change has reversed. If it has reversed and the amplitude exceeds the threshold, then this moment is marked as a critical point, and the time of the critical point will be recorded together with the corresponding morphological characteristics for further analysis of the changes in morphological characteristics during the critical point period. Through this method, the area where the morphological characteristics coincide with the critical point time can be identified, and further evaluate the importance of morphological characteristics in the crop growth process.

[0070] The feature registration sub-module marks the features and morphological parameters at the overlapping time points according to the matching morphological feature set, using the formula:

[0071] ;

[0072] Calculate the key feature values, record the corresponding time nodes, integrate the indexes, morphological names, time point numbers and growth state fluctuation direction values of the key features, and establish a list of key nodes of crop morphology;

[0073] Among them, represents the key feature value, represents the time point number of the i-th time point, represents the time point number of the i-th time point, represents the morphological feature value of the i-th time point, represents the total number of time points;

[0074] According to the previous coincidence detection results, select those parameters whose morphological features change significantly near the critical point time as the matching morphological features. The morphological features include the expansion amplitude of leaves, the bending angle of the stem, the plant spacing, etc. The change of the features will directly reflect the fluctuation of the crop growth state. When it is confirmed that a certain morphological feature coincides with the growth state change at a specific time point, the feature can be marked as a key feature and the corresponding time node can be recorded. For example, assume that the change of the leaf edge exactly coincides with the fluctuation of the crop growth state at a certain critical point moment. Then the change amplitude of the leaf edge and its corresponding time point can be recorded as the key features of the crop. Integrate the relevant information of the key features, including the morphological parameter name, the time point number and its corresponding growth state fluctuation direction value, and finally form a list of key nodes of crop morphology. This list contains the detailed information of all key nodes, which is convenient for subsequent data analysis and crop growth prediction. In this way, necessary reference data can be provided for the precise management of crops to help optimize the planting process.

[0075] Using the formula:

[0076] ;

[0077] Among them, represents the key feature value, represents the time point number of the i-th time point, represents the time point number of the i-th time point, represents the growth state fluctuation direction value at the \(i\)-th time point, represents the total number of time points;

[0078] The key feature value is a comprehensive index used to represent the combined effect of the morphological characteristics and the growth state fluctuation direction value of the crop during the growth process. By calculating the square root of the sum of the squares of the time differences between different time nodes and the crop morphological characteristics and the growth state fluctuation direction value, it reflects the change trend of the crop during the growth cycle. Specifically, the key feature value can comprehensively consider the morphological changes of the crop at each time point (such as leaf growth, stem development, etc.) and the fluctuations in the growth state (such as the acceleration or deceleration of the growth rate), thereby providing a basis for analyzing the key nodes of crop growth, predicting the growth trend, and formulating reasonable agricultural management measures;

[0079] (Time node number):

[0080] Obtaining method: The time node number is obtained through the monitoring data of the crop growth cycle. Usually, the different growth stages of the crop are monitored by ground sensors (such as temperature and humidity sensors, soil moisture sensors) or remote sensing technologies (such as drones, satellite images). Each monitoring time point of the crop growth stage is calibrated as a time node number (for example, the 1st week, the 2nd week, etc. after the crop is sown);

[0081] Dimensional consistency: The time node number is dimensionless, that is, it does not involve units and directly represents the time sequence as an integer. Therefore, the definition and obtaining method of the time nodes in the data directly determine the format of the time node number;

[0082] (Morphological feature value):

[0083] Obtaining method: The morphological feature value reflects the growth form of the crop, such as leaf length, stem diameter, height, leaf area, etc. It is usually extracted by image processing technologies (such as image recognition based on RGB or near-infrared images) or ground sensors (such as lidar). Deep learning algorithms are used to analyze the image or lidar data to measure the crop morphological characteristics and quantify them into numerical values;

[0084] Dimensional consistency: Various morphological features have different units, such as leaf length (unit: cm), stem diameter (unit: mm), etc. Therefore, the data must be normalized. Usually, each feature value is divided by the maximum value of the feature during the entire monitoring period, so that all morphological feature values are normalized between 0 and 1;

[0085] (Growth state fluctuation direction value):

[0086] Obtaining method: The growth state fluctuation direction value reflects the change of the crop growth trend, which can be obtained from the long-term monitored crop growth data. Using meteorological data, soil data, light data, etc., combined with crop growth models or machine learning algorithms to predict the growth state fluctuation direction of the crop (for example, the growth rate accelerating or slowing down, the fluctuation trend of the growth state);

[0087] Dimensional consistency: The unit of the growth state fluctuation direction value is a dimensionless number. Therefore, it needs to be normalized to a value between 0 and 1. A common approach is to perform standardization processing to make the data within the range of 0 to 1, which can avoid the influence brought by the data units of different fluctuation directions;

[0088] Dimensional consistency process (normalization):

[0089] Morphological feature normalization: Morphological feature values (such as leaf length, stem diameter, etc.) need to be normalized. Assuming that for a specific morphological feature (such as leaf length), its maximum value is then the normalized morphological feature value can be expressed as: where is the original morphological feature value, is the maximum value of this morphological feature. After normalization, all morphological feature values will be adjusted to numerical values between 0 and 1;

[0090] Growth state fluctuation direction value normalization: The growth state fluctuation direction value also needs to be standardized. Assuming that the minimum value of the growth state fluctuation direction value is and the maximum value is then the normalized growth state fluctuation direction value

[0091] can be expressed as: where is the original growth state fluctuation direction value, and are the minimum and maximum values of this value respectively. Through standardization processing, the data range is adjusted to between 0 and 1, eliminating the influence of dimensions;

[0092] Formula calculation and derivation process (taking specific numbers as an example):

[0093] Suppose there are data at 3 time points:

[0094] Time node number ( ): , , ;

[0095] Morphological feature value ( ): , , ;

[0096] Growth state fluctuation direction value ( ): , , ;

[0097] Calculate the time difference ( ):

[0098] ;

[0099] ;

[0100] Calculate the sum of squares of the morphological characteristics and the growth state fluctuation direction value ( ):

[0101] ;

[0102] ;

[0103] ;

[0104] Calculate the square root of the sum of squares: , , ;

[0105] Calculate the weighted value at each time point ( ):

[0106] For : ;

[0107] For : ;

[0108] Sum to obtain the key characteristic value ( ): ;

[0109] The calculated key characteristic value K represents the comprehensive situation of the morphological characteristics and the growth state fluctuation direction value of the crop during the growth process. The result can reveal the variation law of the crop at different growth stages and the fluctuation trend of the growth state, help to evaluate the overall health status of the crop and the stability of its growth process. By continuously monitoring the key characteristic value, it can provide important reference for crop management. For example, by adjusting agricultural management measures to optimize crop growth and ensure that the crop plays its potential in the best growth state. At the same time, the change trend of the key characteristic value can also provide a scientific basis for crop production decision-making, guiding the implementation timing and strategy of agricultural activities such as fertilization and irrigation.

[0110] Please refer to Figure 4 , the growth trend determination module includes:

[0111] The area comparison sub-module obtains the crop image data of two consecutive cycles through the list of key morphological nodes of the crop, classifies each cycle according to the leaf expansion area and the stem elongation length, compares the measured values of each category of time nodes, identifies the change values of the leaf area and the stem elongation and arranges them according to the cycle, and generates a cycle growth change sequence;

[0112] By obtaining the list of key morphological nodes of the crop, the key growth points of the crop in each time period can be determined, such as the expansion area of the leaves and the elongation length of the stem. During the processing of the image data, the image data of each cycle will be classified according to the changes in the leaf area and the stem elongation. For example, within a growth cycle, the expansion area of the leaves will increase over time, and the elongation length of the stem will also vary with the seasons. Therefore, by measuring the change values of these two parameters in each cycle, the growth state of the crop can be clearly judged. For example, in the first cycle, the leaf expansion area is 50 cm² and the stem length is 15 cm, while in the next cycle, the leaf expansion area is 70 cm² and the stem elongates to 20 cm. By comparing the changes in these two values, the growth progress of the crop can be identified, generating a cycle growth change sequence, and the data can be used as the basis for subsequent crop growth trend analysis.

[0113] The direction evaluation sub-module evaluates the change direction turn of the current and the previous cycle data according to the cycle growth change sequence, identifies the turning phenomenon and records it, analyzes the corresponding deviation amplitude between the leaf and the stem, and uses the formula:

[0114] ;

[0115] Obtain the growth trend deviation degree;

[0116] Among them, represents the growth trend deviation degree, represents the leaf length of the jth cycle, represents the th cycle of leaf length, represents the stem diameter of the jth cycle, represents the th cycle of stem diameter, represents the total number of cycles;

[0117] Compare the growth data changes between the current cycle and the previous cycle to identify the direction of change. For example, in a certain cycle, the leaf expansion area increases from 50 cm² to 70 cm², and the stem elongation increases from 15 cm to 20 cm. If the growth directions of these two indicators are consistent, it indicates that the growth state of the crop remains stable. However, if the leaf area increases while the stem length decreases, it indicates that the growth direction has deviated. When evaluating the deviation degree, first calculate the change range between the two cycles, and then determine whether the direction of change has reversed. If it has reversed, record this phenomenon and label it as a turning phenomenon. The calculation of the deviation amplitude is by comparing the change amounts of the leaves and the stems. For example, the change amount of the leaf area is 20 cm², and the change amount of the stem length is -5 cm. By calculating the relative difference between the two change amounts, the growth trend deviation degree can be obtained. If the deviation degree exceeds a certain threshold, it is considered that there is a significant directional change in the growth of the crop. For example, if the set threshold is 10%, then if the deviation degree is greater than 10%, it can be considered that the growth direction of the crop has deviated.

[0118] The growth trend deviation degree is a comprehensive index used to measure the change differences between leaves and stems in different growth cycles of crops. Specifically, it reflects the relative deviation degree between the leaf length and the stem diameter. When the deviation degree is large, it means that the change amplitudes of the leaves and the stems are relatively significant, indicating that the growth of the crop fluctuates greatly in a certain cycle, or the growth state is greatly affected by external factors (such as climate change, soil conditions, etc.). On the contrary, when the deviation degree is small, it indicates that the growth changes of the leaves and the stems are relatively consistent, and the growth of the crop is relatively stable. By analyzing the deviation degree, the growth trend of the crop can be better understood, and important basis can be provided for agricultural production decision-making;

[0119] (Leaf length in the j-th cycle):

[0120] Obtaining method: Leaf length data is collected through field observations, remote sensing images or ground sensors. Remote sensing technologies (such as drone or satellite images) can measure the leaf length through image recognition technology. In addition, laser scanning or image processing technology can also accurately determine the leaf length;

[0121] Quantification processing: To ensure that the data is compared under the same dimension, the leaf length is recorded in centimeters (cm). If the leaf length data comes from image recognition or sensor measurement, it needs to be standardized according to the longest leaf. For example, by measuring the maximum leaf length of the crop, all leaf lengths are divided by the maximum leaf length for normalization processing to keep the data within the range of 0 to 1;

[0122] (Leaf length in the previous cycle):

[0123] Obtaining method: The leaf length data of the previous cycle is collected by the same technical means and recorded at the end of each monitoring cycle. The cycle is defined according to the growth stage of the crop. For example, cycle data is recorded every 7 days or 10 days;

[0124] Quantification process: To ensure the comparability of data, the leaf length of the previous cycle is also subjected to the same normalization process to ensure that its unit is the same as that of the leaf length in the current cycle;

[0125] (Stem diameter in the j-th cycle):

[0126] Obtaining method: The data of the stem diameter is obtained through special measurement tools (such as calipers, digital calipers) or remote sensing technology. In the field, high-precision calipers are used for physical measurement. In remote sensing images, the maximum diameter of the stem can be extracted through image processing algorithms and calculated;

[0127] Quantification process: The stem diameter is recorded in millimeters (mm). After being collected through image recognition or sensors, it will be standardized, and all stem diameter values will be normalized to between 0 and 1. For example, using the maximum value method, each stem diameter is divided by the maximum stem diameter of the crop to normalize the data;

[0128] (Stem diameter of the previous cycle):

[0129] Obtaining method: The stem diameter data of the previous cycle is obtained through the same collection method, which is consistent with the periodic monitoring time points of the leaf length. The definition of the cycle affects the data recording and collection timing of each measurement;

[0130] Quantification process: The normalization process is the same as that of the stem diameter in the current cycle to ensure unit consistency and data comparability;

[0131] Dimensional unification process (normalization):

[0132] Leaf length ( and ) :

[0133] The unit of leaf length is centimeters (cm). Among the leaf length data of different cycles, due to the different growth stages of the crop, directly comparing its original data will be affected by unit and scale differences. Therefore, in practical applications, the leaf length will be normalized. Assuming that the maximum leaf length within the cycle is , then the normalized leaf length is: , and the normalized leaf length data is within the range of 0 to 1, eliminating the differences between units;

[0134] Stem diameter ( and ):

[0135] The unit of stem diameter is millimeters (mm). Since the stem diameter varies at different time points and for different crop varieties, normalization is required. Using the same maximum value normalization method, assuming the maximum stem diameter within the period is , then the normalized stem diameter is: . In this way, all stem diameter data will be normalized to the range of 0 to 1, eliminating the influence of size differences;

[0136] Formula calculation and derivation process (taking specific numbers as an example):

[0137] Assume the original data within the monitoring period is as follows:

[0138] Data for period 1: Leaf length cm, stem diameter mm;

[0139] Data for period 2: Leaf length cm, stem diameter mm;

[0140] Data for period 3: Leaf length cm, stem diameter mm;

[0141] Through normalization:

[0142] Normalized leaf length: Assume the maximum leaf length cm, the normalized leaf length is:

[0143] , , ;

[0144] Normalized stem diameter: Assume the maximum stem diameter mm, the normalized stem diameter is:

[0145] , , ;

[0146] Calculation process:

[0147] Calculate the leaf length difference:

[0148] For period 2, ;

[0149] For period 3, ;

[0150] Calculate the difference in stem diameter:

[0151] For cycle 2, ;

[0152] For cycle 3, ;

[0153] Calculate the deviation ratio of the leaf to the stem:

[0154] For cycle 2, the deviation ratio is: ;

[0155] For cycle 3, the deviation ratio is: ;

[0156] Sum to obtain the degree of deviation:

[0157] Total degree of deviation is: ;

[0158] : The calculated degree of deviation is 1.75, indicating that the variation difference between the leaves and the stem of the crop is relatively large during these cycles, and the variation difference between the leaves and the stem decreases as the cycle progresses.

[0159] The benchmark comparison sub-module extracts the growth change data of the original continuous cycles of the crop according to the growth trend deviation degree, identifies the benchmark range of growth changes, compares the deviation degree of the current cycle with the benchmark range of growth changes. If it exceeds the benchmark range, the corresponding cycle is marked as an abnormal growth fluctuation section, and the growth trend deviation information of the crop is obtained;

[0160] According to historical data, determine the benchmark range of the crop in the normal growth state. For example, the fluctuations of the leaf expansion area and the stem elongation length of the crop are very regular in some cycles. The benchmark range can be set based on the regular data, and the average value ± one standard deviation is selected as the benchmark range. Suppose the standard deviation of the leaf expansion area of a certain crop in the past few cycles is 5 cm², and the benchmark range is ±5 cm². If the change amount of the leaf area in the current cycle is 10 cm², exceeding the threshold of the benchmark range, then this cycle is marked as an abnormal fluctuation section. By comparing the deviation degree of the current cycle with the benchmark range, abnormal fluctuations in the growth process of the crop can be detected in a timely manner. Through this method, abnormal fluctuations occurring in the growth process of the crop can be accurately detected, and corresponding countermeasures can be taken in a timely manner. For example, if it is found that the stem elongation amount is significantly lower than the benchmark range in a certain cycle, it indicates that the crop has encountered environmental stress or diseases in this cycle, and appropriate management measures need to be taken.

[0161] Please refer to Figure 5, the abnormal area marking module includes:

[0162] Based on the time points of the marked direction turning nodes in the crop growth trend offset information, the turning node extraction sub-module extracts the time identifiers of each turning node, records them in chronological order, and after determining the turning behavior, regards it as a valid turn and relocates the time to establish a turning node time series;

[0163] The turning nodes in crop growth are identified through the growth trend offset data. A turning node refers to a time point during the crop growth process when the growth direction of the crop changes significantly. For example, the growth direction of the leaves changes from expansion to contraction, or the elongation direction of the stem changes. By extracting the time identifiers of each turning node and sorting them in chronological order, the occurrence time of the turning behavior can be determined, and further verify whether the turning node is a valid turn. For example, when it is observed that the stem growth suddenly slows down and the leaf expansion amplitude no longer increases, it is considered that the growth direction of the crop has turned, and this time point is recorded as the turning node. To further improve the accuracy, the time of the turning node can be relocated according to the time periods before and after the turn, ensuring that the turning node accurately reflects the true state of crop growth. Finally, a turning node time series can be generated, which details the moments when the crop's growth direction changes each time, providing an important reference for subsequent health status assessment and growth trend analysis.

[0164] Based on the time points corresponding to each node in the turning node time series, the health degradation detection sub-module identifies the continuous data of the crop health status in the adjacent three cycles, conducts a sliding analysis, extracts the health status values within the time points, compares whether there is degradation, and if so, determines it as a health degradation state and generates a health degradation trend sequence;

[0165] Based on the time points corresponding to each turning node, the health status data of the crop in the adjacent three cycles is extracted. The health status value is measured by growth parameters such as the leaf condition, stem thickness, and plant height of the crop. By conducting a sliding analysis on the health status data within these three cycles, the change trend of the crop health status can be obtained. During the analysis process, first compare the changes in the health status within consecutive time periods. If it is found that the health status significantly decreases in a certain cycle and the decrease amplitude exceeds the set threshold (for example, the health status value decreases by more than 10%), it is determined that the crop has a health degradation phenomenon. The determination of the degradation state is achieved through the set threshold. For example, when the health status value of the crop decreases by more than 10% compared to the baseline level within three consecutive cycles, it can be determined as a health degradation state. A health degradation trend sequence is generated, which records the health status fluctuations of the crop in each cycle, marks the time periods when health degradation occurs, and provides a basis for subsequent intervention measures.

[0166] The fluctuation region screening sub-module extracts the time points that exist in both sequences based on the health degradation trend sequence, combines the index positions corresponding to the intersection of the two, identifies the key cycle regions that meet the condition of synchronization between turning and health degradation, encodes them uniformly into a set, and generates a set of crop growth fluctuation regions;

[0167] By comparing the health degradation trend sequence with the turning node time sequence, the time points that appear in both sequences are found. These time points are the critical moments when the crop turns and its health degrades simultaneously. For example, in a certain cycle, when the crop reaches the turning node, there is a significant decline in its health status at the same time, indicating that this cycle is an important time point for crop growth fluctuations. By identifying the synchronization phenomenon, the abnormal fluctuation regions in the crop growth process can be further analyzed. Combining the intersection index positions of the turning node and health degradation, the key cycle regions showing synchronous changes can be accurately located, and then a set of crop growth fluctuation regions is generated. This set will help monitor and identify the abnormal fluctuation sections in the crop growth process, such as the impact of diseases or environmental changes on crop health, and provide data support for precision agriculture management;

[0168] Table 1: Crop health status change data

[0169]

[0170] As shown in Table 1, the health status change data of the crop in three consecutive cycles is listed. The "Health status change" column in the table shows the change in the health status value in each cycle. For example, at the T1 node, the health status value drops from 85 to 80, with a decline of 5 units, indicating that the crop's health status has degraded during this cycle.

[0171] Please refer to Figure 6 , the status information output module includes:

[0172] The window overlap identification sub-module extracts the abnormal regions and the start and end times of the operation cycles based on the crop growth fluctuation region set and the cycle time nodes in the morphological key node list, combines the standard cycle duration to identify the upper and lower bounds of the standard time window, matches the start and end time periods of the regions, analyzes the cycle overlap relationship, and generates the time window overlap range data;

[0173] Classify the periodic time nodes, identify the start and end periods of the cycle, combine the time nodes with the growth fluctuation data of the region, and locate the abnormal region. The abnormal region refers to the region where the fluctuation amplitude during the crop growth process exceeds the normal range, which can be defined by calculating the standard deviation of the growth curve. For example, when the fluctuation exceeds ±2 times the standard deviation, it can be regarded as an abnormal fluctuation. On this basis, extract the start and end time periods of the abnormal region. For example, if the fluctuation in a certain region increases significantly between the 5th day and the 8th day, then this time period can be regarded as the abnormal period, and further verify the time period through the standard cycle duration to ensure that it conforms to the duration of the standard cycle. The setting of the standard cycle duration needs to be determined according to the normal growth cycle of the crop. For example, for a certain crop, its normal growth cycle is 30 days, then the duration of the standard cycle is 30 days. If the abnormal cycle deviates significantly from the duration, it needs to be marked. Combine the information to further identify the upper and lower bounds of the time window that meets the standard cycle, and analyze their overlapping situation by matching the start and end time periods of the region to generate the data of the overlapping range of the time window. The process requires the use of a matching algorithm for time series data, such as dynamic time warping (DTW), to measure the similarity of growth fluctuations in different time periods and generate accurate overlapping range data.

[0174] The overlapping interval determination sub-module extracts the start and end time points of the overlap according to the data of the overlapping range of the time window, combines the regional activity frequency, duration, and the difference in the cycle start point, calculates the overlapping relationship index, and compares it with the threshold. If it exceeds the threshold, it is marked as a valid overlapping cycle to obtain a sequence of valid overlapping cycles; The threshold is a judgment boundary for this overlapping relationship index. Only when the overlapping relationship index exceeds this threshold does it indicate that the overlap of the two cycles is significant and will be recognized as a "valid overlapping cycle";

[0175] Extract the overlapping start and end time points, and further calculate by combining the regional activity frequency, duration, and the difference in the start points of the cycles. The activity frequency can be obtained by calculating the number of frequent fluctuation events of crop growth in each time period. For example, if there are three significant fluctuations in a week, the activity frequency is 3 times / week; the duration refers to the duration of the fluctuation. For instance, if a certain fluctuation lasts from the 5th day to the 8th day, the duration is 3 days. The difference in the start points of the cycles is the difference between the start time of the overlapping cycle and the start time of the standard cycle. The larger the difference, the more the overlapping period deviates from the standard cycle. Through parameters, the overlapping relationship index can be calculated, and the index is a key indicator to measure the degree of overlap between two cycles. If the overlapping relationship index exceeds the set threshold, for example, the set threshold is 0.8, then this cycle is considered a valid overlapping cycle. The setting of the threshold needs to be adjusted according to empirical values and actual situations. By statistically analyzing the normal overlapping situations of crop growth cycles, a reasonable percentage can be set. For example, 0.8 means that 80% of the overlapping periods meet the requirements of the standard cycle. In practical applications, if the overlapping relationship index is higher than the threshold, then this cycle is marked as a valid overlapping cycle and incorporated into the sequence of valid overlapping cycles.

[0176] The status information generation sub-module extracts the regional index according to the sequence of valid overlapping cycles, checks whether it is a member of an abnormal region. If so and it is within the cycle overlapping interval, it is marked as a risk region and included in the set of risk concerns, and the crop growth status information is output.

[0177] Check whether this region is a member of an abnormal region. If it is an abnormal region, further check whether it is within the cycle overlapping interval. If the region is within the overlapping interval, then this region is marked as a risk region and included in the set of risk concerns. The calibration process of the risk region depends on the valid overlapping cycles identified previously. If the overlapping time period of the valid cycle exactly falls within the growth fluctuation interval of the abnormal region and the overlapping relationship index is relatively high, then this region is considered a high-risk region. For example, assume that a certain region is within the overlapping cycle range from the 10th day to the 12th day, and there are significant abnormal fluctuations in the growth cycle of this region. At this time, it will be marked as a risk region. The risk region is included in the set of risk concerns, and the crop growth status information is output. The information provides data support for subsequent crop monitoring and management.

[0178] The crop growth detection method based on machine vision is executed based on the above-mentioned crop growth detection system based on machine vision, and includes the following steps:

[0179] S1: Obtain crop canopy images through multi-angle cameras set in the field, extract the changes in the leaf edge contours, the degree of stem bending, and the distribution of plant spacing, judge the consistency between the changes and the overall growth status of the crops, screen the morphological features, and generate a crop morphological feature sequence.

[0180] S2: Based on the time nodes of the crop morphological feature sequences, combined with the crop growth state change curve, determine whether the nodes coincide, mark the coincident features as key features and record the positions, and generate a list of key crop morphological nodes;

[0181] S3: According to the list of key crop morphological nodes, obtain the crop image data of consecutive periods, evaluate the change ranges of the leaf expansion area and the stem elongation length, and determine whether the turning range exceeds the original benchmark to obtain the crop growth trend deviation information;

[0182] S4: According to the turning time points in the crop growth trend deviation information, determine whether the crop health state degenerates during the corresponding period, mark the qualified time points, and generate a set of crop growth fluctuation regions;

[0183] S5: Combine the set of crop growth fluctuation regions and the list of key morphological nodes, determine whether there is an overlapping region of the periodic window, mark it as a risk concern point, and output the crop growth state information.

[0184] The above are only the preferred embodiments of the present invention, and do not limit the present invention in other forms. Any person skilled in the art may use the disclosed technical content to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A crop growth detection system based on machine vision, characterized in that, The system includes: The image acquisition module obtains crop canopy images through multi-angle cameras set in the field, extracts the changes in the leaf edge contours, the degree of stem bending, and the plant spacing distribution, and performs consistency matching based on the spatial distribution characteristics of the three parameters and the overall growth state of the crop to generate a crop morphological feature sequence. The morphological feature capture module reads the crop morphological feature sequence, detects whether the corresponding node positions in the time axis coincide with the critical points of the crop growth state changes. If there is a time coincidence, the corresponding morphological features are marked as key features to generate a list of crop morphological key nodes. The growth trend determination module, through the list of crop morphological key nodes, statistically analyzes the crop image data for two consecutive periods, compares the leaf expansion area and the stem elongation length in each period respectively, evaluates the change amplitude between the two in different periods. If the change direction turns, it is compared with the original growth benchmark to obtain crop growth trend deviation information. The abnormal area marking module, based on the crop growth trend deviation information, detects whether there is a degradation in the crop health state during the corresponding time period, marks the abnormal areas, and generates a set of crop growth fluctuation areas.

2. The crop growth detection system based on machine vision according to claim 1, wherein, The crop morphological feature sequence includes morphological distribution characteristics, consistency matching frequencies, and growth state stability. The list of crop morphological key nodes includes key feature names, critical points of growth state changes, and time node mapping relationships. The crop growth trend deviation information includes change amplitude, turning direction, and benchmark deviation status. The set of crop growth fluctuation areas includes turning nodes, health degradation marked areas, and key areas of fluctuations within a period.

3. The crop growth detection system based on machine vision according to claim 1, characterized in that, The image acquisition module includes: The perspective adjustment sub-module obtains crop canopy images through multi-angle cameras set in the field, adjusts the camera's pitch angle and focal length, extracts the leaf edge contours, the degree of stem bending, and the plant spacing distribution, and generates a sequence of crop spatial distribution images. The feature extraction sub-module, based on the sequence of crop spatial distribution images, analyzes the change trends of the leaf edge contours, the degree of stem bending, and the plant spacing distribution over consecutive time, compares the change directions of the three with the overall growth state of the crop, and calculates the frequencies of consistency occurrences to obtain a sequence of frequencies of morphological and growth state consistency. The sequence screening sub-module, according to the sequence of frequencies of morphological and growth state consistency, selects the morphological parameters with the optimal occurrence frequencies, identifies the corresponding original parameters, and establishes a crop morphological feature sequence.

4. The crop growth detection system based on machine vision according to claim 3, wherein, The morphological feature capture module includes: The node positioning sub-module obtains the time series points corresponding to the morphological parameters in the crop morphological feature sequence, extracts the moments of parameter extrema as representative nodes, sets a periodic time axis, uses three adjacent periods as a sliding window, identifies the local extrema points in each sequence and marks the key nodes to obtain a set of morphological key time nodes. The coincidence detection sub-module, based on the set of morphological key time nodes, extracts the growth state values corresponding to the time nodes, determines that the sign reversal of the change between the front and back moments and exceeding the fluctuation threshold is the critical point, records the critical point time, analyzes the correlation degree of the coincidence area between the morphological features and the critical point time, and identifies the morphological features with a correlation degree higher than the threshold to obtain a set of matching morphological features. The feature registration sub-module marks the features and morphological parameters at the overlapping time points according to the matching morphological feature set, calculates the key feature values, records the corresponding time nodes, integrates the indexes, morphological names, time point numbers and growth state fluctuation direction values of the key features, and establishes a list of key nodes of crop morphology.

5. The crop growth detection system based on machine vision according to claim 4, characterized in that, The growth trend determination module includes: The area comparison sub-module obtains the crop image data of two consecutive periods through the list of key nodes of crop morphology, classifies each period according to the leaf expansion area and the stem elongation length, compares the measured values of the time nodes of each category, identifies the change values of the leaf area and the stem elongation and arranges them according to the period, and generates a sequence of periodic growth changes; The direction evaluation sub-module evaluates the change direction turn between the current and the previous period data according to the sequence of periodic growth changes, identifies and records the turning phenomenon, analyzes the corresponding deviation amplitude between the leaf and the stem, and obtains the growth trend deviation degree; The reference comparison sub-module extracts the growth change data of the original continuous period of the crop according to the growth trend deviation degree, identifies the growth change reference range, compares the deviation degree of the current period with the growth change reference range, and if it exceeds the reference range, marks the corresponding period as an abnormal section of growth fluctuation and obtains the crop growth trend deviation information.

6. The crop growth detection system based on machine vision according to claim 5, characterized in that, The abnormal area marking module includes: The turning node extraction sub-module extracts the time identifiers of each turning node based on the time points of the marked direction turning nodes in the crop growth trend deviation information, records them in chronological order, and after determining the turning behavior, regards it as a valid turn and relocates the time to establish a time sequence of turning nodes; The health degradation detection sub-module identifies the continuous data of the crop health state in three adjacent periods according to the time points corresponding to each node in the time sequence of turning nodes, performs sliding analysis, extracts the health state values within the time points, compares whether there is degradation, and if so, determines it as a health degradation state and generates a health degradation trend sequence; The fluctuation area screening sub-module extracts the time points that exist in both sequences according to the health degradation trend sequence, combines the index positions corresponding to the intersection of the two, identifies the key period areas that meet the condition of synchronous turning and health degradation, encodes them uniformly into a set, and generates a set of crop growth fluctuation areas.

7. The crop growth detection system based on machine vision according to claim 1, characterized in that The system further includes: The status information output module compares the set of crop growth fluctuation areas with the list of key nodes of morphology in terms of time, determines whether there is an overlap in the time windows of adjacent periods, and if so, lists the corresponding areas as risk concerns and outputs the crop growth status information; The crop growth status information includes risk concerns, the overlap status of the periodic time window, and the abnormal trend identifier.

8. The crop growth detection system based on machine vision according to claim 7, characterized in that The status information output module includes: The window overlap identification sub-module extracts the abnormal areas and the start and end times of the operation periods based on the periodic time nodes in the set of crop growth fluctuation areas and the list of key nodes of morphology, combines the standard cycle duration to identify the upper and lower bounds of the standard time window, matches the start and end time periods of the areas, analyzes the cycle overlap relationship, and generates the time window overlap range data; The overlapping interval determination sub-module extracts the start and end time points of the overlap according to the time window overlap range data, combines the regional activity frequency, duration, and the difference in the cycle start point, calculates the overlap relationship index, and compares it with a threshold. If it exceeds the threshold, it is marked as a valid overlap cycle, and a sequence of valid overlap cycles is obtained. The status information generation sub-module extracts the region index according to the sequence of valid overlap cycles, checks whether it is a member of an abnormal region. If so and it is within the cycle overlap interval, it is marked as a risk region and included in the set of risk concern points, and the crop growth status information is output.

9. A crop growth detection method based on machine vision, characterized in that, The method is used to implement the machine vision-based crop growth detection system described in any one of claims 1-8, and includes the following steps: S1: Obtain crop canopy images through multi-angle cameras set in the field, extract the changes in the leaf edge contours, the degree of stem bending, and the plant spacing distribution, judge the consistency between the changes and the overall growth status of the crops, screen the morphological features, and generate a sequence of crop morphological features. S2: According to the time nodes of the crop morphological feature sequence, combine the crop growth status change curve, judge whether the nodes coincide, mark the coincident features as key features and record the positions, and generate a list of crop morphological key nodes. S3: According to the list of crop morphological key nodes, obtain the crop image data of consecutive cycles, evaluate the change range of the leaf expansion area and the stem elongation length, and judge whether the turning amplitude exceeds the original benchmark to obtain the crop growth trend deviation information. S4: According to the turning time points in the crop growth trend deviation information, judge whether the crop health status deteriorates during the corresponding period, mark the qualified time points, and generate a set of crop growth fluctuation regions. S5: Combine the set of crop growth fluctuation regions and the list of morphological key nodes, judge whether there is a cycle window overlap region, mark it as a risk concern point, and output the crop growth status information.

Images