A method and system for monitoring crop growth by image monitoring

Abstract

The present invention relates to the technical field of crop growth monitoring, and specifically to a method and system for monitoring crop growth through images, comprising the following steps: deploying an unmanned aerial vehicle (UAV) to fly over the target paddy field, collecting image data, and transmitting the image data to the ground station in real time through data transmission to generate a preliminary image dataset; and cropping and correcting the preliminary image dataset. In the present invention, the use of a UAV for image collection avoids the problem of limited viewing angles of fixed ground cameras, and reduces data latency through real-time data transmission, enabling more timely monitoring. During the process of cropping and correcting the image data, angle deviation and background noise are eliminated, improving the accuracy of leaf area index calculation. The introduction of time series analysis enables dynamic monitoring of crop growth trends, making growth assessment no longer limited to a single time point, but forming a comparison of historical data, and improving the prediction ability of growth status.

Description

Technical Field

[0001] The present invention relates to the technical field of crop growth monitoring, and particularly to a method and system for monitoring crop growth through images. Background Art

[0002] The technical field of crop growth monitoring involves using a variety of sensing and data analysis means to evaluate the growth status, health status, and growth trend of crops in real-time or periodically. The method of monitoring crop growth through images belongs to the technical field of crop growth monitoring, mainly using image data to analyze and evaluate the growth status of crops.

[0003] However, in the prior art, the image data is not cropped and geometrically corrected, the background information is complex, which is easy to interfere with the calculation of leaf area index and affects the accuracy of monitoring. Crop growth assessment usually relies on single-point image data without data comparison, resulting in difficult accurate judgment of growth trends and affecting the reliability of agricultural management decisions. Therefore, improvements are needed. Summary of the Invention

[0004] The purpose of the present invention is to solve the drawbacks existing in the prior art, and to propose a method and system for monitoring crop growth through images.

[0005] To achieve the above purpose, the present invention adopts the following technical solutions. A method for monitoring crop growth through images includes the following steps:

[0006] Deploy an unmanned aerial vehicle (UAV) to fly over the target paddy field, collect image data, and send the image data to the ground station in real-time through data transmission to generate a preliminary image dataset; crop and correct the preliminary image dataset to generate a processed image dataset;

[0007] Based on the processed image dataset, identify and measure the leaf area index of rice in the image, identify the health status and growth stage of the crop, and generate original LAI data; perform time series analysis on the original LAI data, and determine the crop growth trend through comparative analysis to generate a time series analysis result;

[0008] Based on the time series analysis result, combine soil humidity and meteorological data, combine the soil humidity and meteorological data with the time series analysis result to generate comprehensive monitoring data; use the comprehensive monitoring data to evaluate the growth status of rice, analyze the environmental factors affecting crop growth, and generate a crop growth assessment result;

[0009] Based on the crop growth assessment result, evaluate the effectiveness of the current irrigation and fertilization plans, determine whether agricultural management measures need to be adjusted, and generate management adjustment suggestions.

[0010] Preferably, the step of obtaining the preliminary image dataset is as follows:

[0011] Deploy a drone to fly over the target rice paddy, collect image data of the rice paddy through a camera, adjust the camera angle and focal length to cover the entire field surface, and obtain the real-time collected rice paddy images;

[0012] Based on the real-time collected rice paddy images, transmit the data to the ground station in real time through wireless transmission to obtain the preliminary image dataset.

[0013] Preferably, the step of obtaining the processed image dataset is as follows:

[0014] Based on the preliminary image dataset, determine the edge transition region through pixel gradient, eliminate isolated noise points, adjust the boundary range according to the shape characteristics of the target crop region, and remove redundant background regions to obtain the image data with adjusted boundaries;

[0015] Based on the image data with adjusted boundaries, calculate the lens distortion parameters, including radial distortion and tangential distortion, and use image interpolation to restore the pixel offset caused by correction to generate the image data after distortion correction;

[0016] Based on the image data after distortion correction, perform color channel normalization processing, and at the same time perform histogram equalization to adjust the brightness distribution to obtain the processed image dataset.

[0017] Preferably, the step of obtaining the original LAI data is as follows:

[0018] Based on the processed image dataset, perform RGB channel separation, extract the pixel values of the red channel, green channel, and blue channel to obtain the channel data of the rice leaf area;

[0019] Based on the channel data of the rice leaf area, calculate the rice leaf area index, and the calculation formula is:

[0020] and ;

[0021] wherein, is the rice leaf area index, is the variable ratio vegetation index based on the drone image, is the digital value of the green channel, is the digital value of the red channel, is the digital value of the blue channel;

[0022] Based on the rice leaf area index, extract the leaf color and texture features, compare the spectral change trends in the growth cycle, and identify the crop health status and growth stage to obtain the original LAI data.

[0023] Preferably, the steps for obtaining the time series analysis result are as follows:

[0024] Based on the original LAI data, construct a time series matrix, extract the rice leaf area index at each observation time point, calculate the change range of the rice leaf area index between adjacent time points, fill the data missing area by interpolation, and perform time series smoothing processing to obtain time series LAI data;

[0025] Based on the time series LAI data, calculate the crop growth trend index, and the calculation formula is:

[0026] ;

[0027] Wherein, is the crop growth trend index, is the rice leaf area index at the th time point, is the rice leaf area index at the previous time point, is the leaf density corresponding to the th time point, is the rice canopy height, is the rice leaf color change rate, is the influence factor of color change on the trend index, is the number of data points in the time series;

[0028] Based on the crop growth trend index, analyze the change pattern of the crop growth trend index in the time dimension, extract the trend characteristics of growth, stability or decline, and obtain the time series analysis result.

[0029] Preferably, the steps for obtaining the comprehensive monitoring data are as follows:

[0030] Based on the time series analysis result, extract the crop growth trend data, call the soil moisture measurement data and meteorological observation data, calculate the soil moisture change rate, air humidity change range, temperature gradient and precipitation accumulation, and obtain the environmental parameter set;

[0031] Based on the environmental parameter set, calculate the comprehensive monitoring index, and the calculation formula is:

[0032] ;

[0033] Wherein, is the comprehensive monitoring index, respectively represent the real-time measurement values of soil moisture content, air humidity and temperature, are the time series smoothing values of the corresponding parameters respectively, respectively represent the fluctuation ranges of each environmental factor, respectively represent the change rates of soil moisture, air humidity, and temperature, which are environmental change impact factors;

[0034] Based on the comprehensive monitoring index, analyze the impact of environmental factors on the crop growth trend, extract the correlation between environmental variables and crop growth status, and obtain comprehensive monitoring data.

[0035] Preferably, the steps for obtaining the crop growth assessment result are as follows:

[0036] Based on the comprehensive monitoring data, calculate the growth adaptability score, and the calculation formula is:

[0037] ;

[0038] where, is the growth adaptability score, is the rice leaf area index, is the crop photosynthetic capacity index, is the soil temperature, is the air humidity, is the cumulative impact value of meteorological factors, is the change rate of soil moisture;

[0039] Based on the growth adaptability score, analyze the growth status of rice in the current environment, and obtain the crop growth assessment result.

[0040] Preferably, the steps for obtaining the management adjustment suggestion are as follows:

[0041] Based on the crop growth assessment result, extract the soil moisture content, soil nitrogen, phosphorus, and potassium content, chlorophyll index, precipitation, and temperature to obtain the basic agricultural management data;

[0042] Based on the basic agricultural management data, calculate the management adjustment index, and the calculation formula is:

[0043] ;

[0044] where, is the management adjustment index, respectively represent the nitrogen, phosphorus, and potassium concentrations in the current soil, is the supply amount of irrigation water, is the total precipitation, is the chlorophyll index, is the crop leaf temperature;

[0045] Based on the management adjustment index, analyze the adaptability of rice to the current irrigation and fertilization, evaluate whether agricultural management measures need to be adjusted, and obtain the management adjustment suggestion.

[0046] The present invention provides a system, comprising:

[0047] A data acquisition module that deploys an unmanned aerial vehicle (UAV) to fly over a target paddy field, collect image data, and transmit it to a ground station in real time to generate a preliminary image dataset;

[0048] An image processing module that crops and corrects the preliminary image dataset to generate a processed image dataset;

[0049] An LAI calculation module that, based on the processed image dataset, identifies and measures the leaf area index of rice in the image, identifies the health status and growth stage of the crop, and generates raw LAI data;

[0050] A data analysis module that performs time series analysis on the raw LAI data, determines the crop growth trend through comparative analysis, and generates a time series analysis result;

[0051] An agricultural management advice module that, based on the time series analysis result, combines soil humidity and meteorological data, combines the soil humidity and meteorological data with the time series analysis result to generate comprehensive monitoring data, uses the comprehensive monitoring data to evaluate the growth status of rice, analyzes the environmental factors affecting crop growth, generates a crop growth evaluation result, and based on the crop growth evaluation result, evaluates the effectiveness of the current irrigation and fertilization plans, determines whether agricultural management measures need to be adjusted, and generates management adjustment suggestions.

[0052] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0053] In the present invention, the use of a UAV for image acquisition avoids the problem of limited perspective of a fixed ground camera, and reduces data lag through real-time data transmission, making the monitoring more timely. During the process of cropping and correcting the image data, angle deviation and background noise are eliminated, improving the accuracy of leaf area index calculation. The introduction of time series analysis enables dynamic monitoring of the crop growth trend, making the growth evaluation no longer limited to a single time point, but forming a comparison of historical data, improving the prediction ability of the growth state. The integration of environmental factors enhances the adaptability of the monitoring system to external influences, enabling key factors such as meteorological changes and soil moisture dynamics to be quantified and directly acting on crop growth evaluation, avoiding the limitations of single crop apparent feature analysis. Based on comprehensive data analysis, the impact of current agricultural management measures on crop growth is evaluated, enabling operations such as irrigation and fertilization to be dynamically adjusted according to the real-time needs of the crop, optimizing resource allocation, and improving the scientificity and implementation efficiency of agricultural management. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 It is a schematic diagram of the steps of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0055] In order to make the objectives, technical solutions and advantages of the present invention more clear and 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.

[0056] Please refer to Figure 1 , the present invention provides a technical solution, a method for monitoring the growth of crops through images, including the following steps:

[0057] Deploy a drone to fly over the target rice field, collect image data, and send the image data to the ground station in real time through data transmission to generate a preliminary image dataset; crop and correct the preliminary image dataset to generate a processed image dataset;

[0058] Based on the processed image dataset, identify and measure the leaf area index of the rice in the image, identify the health status and growth stage of the crop, and generate the original LAI data; perform time series analysis on the original LAI data, and determine the crop growth trend through comparative analysis to generate the time series analysis result;

[0059] Based on the time series analysis result, combine the soil humidity and meteorological data, combine the soil humidity and meteorological data with the time series analysis result to generate comprehensive monitoring data; use the comprehensive monitoring data to evaluate the growth status of the rice, analyze the environmental factors affecting crop growth, and generate the crop growth evaluation result;

[0060] Based on the crop growth evaluation result, evaluate the effectiveness of the current irrigation and fertilization plans, determine whether agricultural management measures need to be adjusted, and generate management adjustment suggestions.

[0061] The steps for obtaining the preliminary image dataset are as follows:

[0062] Deploy a drone to fly over the target rice field, collect the image data of the rice field through a camera, adjust the camera angle and focal length to cover the entire field surface, and obtain the real-time collected image of the rice field;

[0063] Based on the real-time collected image of the rice field, send the data to the ground station in real time through wireless transmission to obtain a preliminary image dataset.

[0064] Specifically, based on the target paddy field area distribution data and geographical coordinate information collected in the early stage, first, the initial flight altitude range of the drone is set between 20 meters and 50 meters. This range is determined according to the early mapping and statistical results combined with empirical methods. For example, when the area of a single paddy field area exceeds 1 hectare, a flight altitude above 35 meters is used, and a threshold H is set through a calculation example. If the collected farmland area S is 1.2 hectares, H can be set as H = 20 + (S - 1.0) × 15, where S is measured in hectares and H is measured in meters. When H exceeds 50 meters or is lower than 20 meters, it is determined as over-limit and a reminder is triggered. Then, a flight corridor is planned according to the established route and the horizontal spacing is ensured to be between 15 meters and 25 meters. According to experience settings, the camera depression angle is adjusted between 45 degrees and 75 degrees. If it is found that the speed of the drone during flight exceeds the predefined upper limit of 5 meters per second, speed adjustment is performed. This speed upper limit is set according to the statistical data of the field size. For example, in a plot with an area less than 0.5 hectares, a low-speed navigation of 2 meters per second can be used. Record the positioning coordinates, flight altitude, camera focal length and inclination angle values during the flight, and compare them with the farmland coordinates. If it is detected that the flight range exceeds the geographical reference boundary, it is regarded as an invalid flight segment. All flight and camera adjustment operations are executed according to the pre-determined configuration strategy and registered after each change. Finally, the shooting results after covering all flight segments are associated with the corresponding coordinate information to obtain real-time collected paddy field images.

[0065] Based on the real-time collected paddy field images, the image data of each captured frame is matched with the previously registered flight timestamps and recorded. For the wireless transmission part, first, a signal strength threshold is set according to the signal coverage range of the ground station around the farmland. For example, when the received signal strength is lower than -75 dBm, it is determined that the transmission is unstable, and it is necessary to rescan the available frequency bands around after the flight in this section or adjust the drone attitude to send the image again. If the signal strength is higher than -75 dBm and always within the range of -60 dBm to -70 dBm, it is regarded as the valid data range. All image data will have time and position tags added after being transmitted to the ground station, and the metadata of each frame of the image is proofread to prevent duplication or omission. When it is detected that the image data in a certain time period is lost, the flight altitude, camera angle, and focal length of the lost period will be marked together for manual comparison to check for recording errors. Then, all the effectively transmitted data is numbered and sorted and checked against the corresponding geographical coordinate reference table. If the check result shows that the deviation amount exceeds the threshold of 0.5 degrees set through previous years' test experience, this part of the data is marked as suspicious and retained in the exception list. Finally, the confirmed valid image data is classified and sorted to obtain a preliminary image dataset.

[0066] The steps to obtain the processed image dataset are as follows:

[0067] Based on the preliminary image data set, the edge transition area is determined by pixel gradient, and isolated noise points are eliminated. The boundary range is adjusted according to the shape characteristics of the target crop area, and the redundant background area is removed to obtain the image data after boundary adjustment.

[0068] Based on the image data after boundary adjustment, the lens distortion parameters, including radial distortion and tangential distortion, are calculated, and the pixel offset caused by the correction is restored by image interpolation to generate the image data after distortion correction;

[0069] Based on the distortion-corrected image data, color channel normalization is performed, and histogram equalization is performed to adjust the brightness distribution to obtain the processed image data set.

[0070] Specifically, based on the preliminary image data set obtained above, we first perform adjacent difference operations on the grayscale values of each pixel and calculate the gradient size. For continuous pixel gradients greater than the threshold value calculated based on the statistical sample mean of previous years plus the correction coefficient k, When , it can be determined that there is a significant transition boundary in the region, where The setting method can be referred to , represents the average gradient value of the same type of images in previous years. k can be taken as 1.2 as an example. If the pixel gradient value in a certain area is generally less than Then it is classified as a non-transition area, and then the isolated noise points are cleaned up, for example, it is determined that the brightness fluctuation of the local pixel in the 3×3 neighborhood is less than the threshold determined empirically. If the noise point appears in all five frames of continuous images, it will be deleted after confirmation. Then, the detected edges will be fitted and compared according to the shape features of the target crop area recorded in advance, such as approximate rectangles or irregular multi-borders. If the fitting degree deviation exceeds a comparison coefficient obtained from the reference example, The gradient of the deviated part is recalculated and the ownership is confirmed again. Finally, the background partitions that do not conform to the shape of the target crop are listed as redundant areas and deleted. After checking each suspicious area one by one, the remaining areas that conform to the shape characteristics are retained and registered, and finally the image data with boundary adjustment is obtained.

[0071] Based on the image data after boundary adjustment obtained above, first read the known coordinate points and corresponding pixel positions in the reference calibration information, and convert the radial distortion coefficient and tangential distortion coefficient As the parameters to be estimated, the initial values of these parameters can be obtained from the experimental data of the same type of lens in previous years and iteratively estimated by the least squares method. For each corresponding point, the coordinate difference before and after distortion is calculated and a residual vector is generated. When the residual is higher than the threshold given in an empirical way, When the fit is insufficient, it means that the coefficients need to be corrected again. It can be given by adding a correction amount to the average residual obtained from the calibration experiment in previous years. For example, if the average residual in previous years is 0.2 pixels and the correction amount is 0.05, then , during the iteration process, each coefficient is dynamically adjusted according to the change in the residual distribution after each iteration. When the overall residual curve is lower than , the pixel coordinates are interpolated and restored, and the corrected coordinates are redistributed in a bilinear or cubic interpolation manner. Finally, the image data after distortion correction is generated.

[0072] Based on the image data after distortion correction obtained previously, first extract the red channel, green channel, and blue channel in the image respectively and calculate the maximum and minimum pixel values for each. Map them to the range of 0 to 255 to unify the relative intensities of each channel. Then, give an equalization threshold according to the brightness extreme value distribution recorded in the shooting environment in previous years , for example, setting the threshold to 128 represents the median brightness. Normalize the pixel histogram of each channel and compare it with . Stretch the pixel segment lower than to balance the brightness distribution. If it exceeds , it is regarded as a section with too high brightness and its pixel density is compressed accordingly. The value of 30 is deduced from the statistical results of the high-brightness area in previous years. After completing the normalization and brightness distribution adjustment of each channel, recombine the three-channel data. Finally, the processed image data set is obtained.

[0073] The steps to obtain the original LAI data are as follows:

[0074] Based on the processed image data set, perform RGB channel separation, extract the pixel values of the red channel, green channel, and blue channel, and obtain the channel data of the rice leaf area;

[0075] Based on the channel data of the rice leaf area, calculate the rice leaf area index. The calculation formula is:

[0076] and ;

[0077] Among them, is the rice leaf area index, is the variable ratio vegetation index based on UAV images, is the digital value of the green channel, is the digital value of the red channel, is the digital value of the blue channel;

[0078] Based on the rice leaf area index, extract the leaf color and texture features, compare the spectral change trends during the growth cycle, identify the crop health status and growth stage, and obtain the original LAI data.

[0079] Specifically, based on the processed image dataset obtained previously, first read the color components corresponding to each pixel and number them, establish a one-to-one correspondence between the numbers and the pixel position information, obtain the RGB value range by comparing the calibration data of the same model of shooting equipment in previous years, so as to identify the distribution of normal pixels in the range of 0 to 255. Then, according to the historical monitoring records of the illuminance in the actual farmland shooting environment, set a brightness threshold for distinguishing dark and bright areas , for example, analyze the all-weather shooting image samples with a sunshine duration of no less than 10 hours in previous years and find that the average value of the brightness distribution is about 130, and then select according to the deviation statistics as the adjustment interval, so as to determine to be in the range of 100 to 160. For pixel segments below this range, as the dark area, it is necessary to additionally check whether this area is located in the shadow-covered position when extracting. The rice row spacing pattern recorded in the previous calibration information can be used as a reference again through the shape contour. If the row spacing characteristics do not match, it is regarded as an invalid pixel and cleaned. Pixel segments higher than 160 are marked as bright areas and the corresponding red, green, and blue components are registered. During the cleaning process, the red component of each pixel is counted , the green component and the blue component . If it is found that the proportion of a certain component continuously deviates within 15% of the preset value and the other components soar abnormally, it is listed as a suspicious noise point and its performance in adjacent frame images is checked again. If it is continuously abnormal in multiple frames, this pixel component is marked as abnormal. The normal pixel values that have passed the brightness area discrimination are finally summarized and classified as the rice leaf area channel data

[0080] The benefit of the formula is to combine the differences of the red channel, green channel, and blue channel, quantify the color characteristics of the leaves into a variable ratio vegetation index, and amplify the sensitivity to small color differences in the form of an index, so as to more accurately map the actual distribution of the rice leaf area in subsequent analysis

[0081] The steps for obtaining the parameter are represents the digital value of the green channel, and the range is usually from 0 to 255. It is extracted through the green component histogram of each pixel in the image taken by the drone. If the green values of some pixels deviate significantly from the normal statistical interval, it is necessary to check whether there are strong reflection points or noise areas. Then, form a calibration coefficient with the comparison value of the external measured color card to obtain the final , for example, collect 10,000 green pixel values in the same plot and record them as , compare the average value of these pixel values with the average value of the green channel of the standard card. If the difference is 12, linearly correct the captured result to align it with the standard reference, and use the corrected pixel green value as the official recorded result.

[0082] The steps for obtaining the parameter are as follows. represents the digital value of the red channel, ranging from 0 to 255. The acquisition method is similar to First, read the red component pixel by pixel and screen out the high-brightness saturated areas. Obtain the correction coefficient by comparing with a calibration plate with known RGB values on-site. Then, record the unified corrected red channel value as , for example, continuously capture 200 frames in a farmland in Henan. Extract 5000 effective red pixel values from each frame, obtaining a total of 1 million records. Compare the average value of these records with the red average value of the standard calibration plate. If the difference is 8, when correcting, increase the red component of each pixel by the corresponding correction amount of 8, and finally obtain the corrected .

[0083] The steps for obtaining the parameter are as follows. represents the digital value of the blue channel, also in the range of 0 to 255. It is obtained by statistically counting the blue component of the captured image pixel by pixel. For the phenomenon of over-brightening in local areas caused by excessive sky reflection, those extreme high values need to be listed separately according to the actual situation and then proofread, and then error correction is carried out under the guidance of the visible light standard color card and finally recorded as , for example, in a sample area, about 800,000 blue pixel values are recorded during the whole-day shooting, and their average value is compared with the reference card. If the average value difference is 15, the whole shooting data needs to be subtracted by 15 and then updated and recorded as .

[0084] The steps for obtaining the coefficient of 1.4527 are as follows. 1.4527 is a value obtained by fitting after large-scale field measurements. By measuring the leaf coverage rate of different plants in the same plot during the whole season and performing regression analysis with the calculated UAV-based VARI_{RGB}, several sample points can be expanded on the coordinate axis and a least squares method is used to fit an exponential curve. The intercept part fits out a stable value of about 1.45. In order to obtain similar results in different regions, an environmental correction is carried out on this 1.45 to obtain 1.4527. For example, in a centralized experiment, 1000 groups of marked leaf area index and their corresponding UAV-based VARI_{RGB} values are statistically analyzed, segmented and averaged to obtain about 1.45. When comparing with other 50 groups of data in the same region, a deviation of 0.0027 is generated, and finally this deviation is accumulated and determined as 1.4527.

[0085] The steps for obtaining the coefficient of 7.8932 are as follows. 7.8932 is the amplification coefficient of the exponential part in the same regression process, and its obtaining method is similar to that of 1.4527. First, the leaf area index of all samples and UAV-based VARIRGB are segmented and split, and then the exponential form is selected for fitting to obtain and a correction of 0.0032 is supplemented after fine analysis in the local area to form the final value of 7.8932. For example, in the fitting results of the same plot, the distribution of the vast majority of samples indicates being between 7.8 and 8.0. After combined statistics and averaging, the value is stable at around 7.89. Then, a second comparison is made in combination with the lighting conditions during the actual shooting period to obtain 7.8932.

[0086] Calculation process:

[0087] First, calculate , if the value obtained by extracting a certain pixel , then:

[0088]

[0089] Substitute this result into the LAI formula:

[0090]

[0091] This result indicates that the leaf area index corresponding to this pixel area is approximately 2.2569. If the LAI values of different pixels are statistically analyzed as a whole, the average LAI value of a field can be obtained. When this average value is between 1 and 3, it can be regarded as a relatively stable interval for mid-term growing rice. Exceeding 5 indicates a very high degree of leaf coverage, and it is necessary to combine the actual planting density to judge whether there is excessive closure. Less than 0.5 indicates limited leaf coverage, and it should be checked whether it is in the initial stage of rice or there are missing plants during shooting.

[0092] Based on the previously obtained rice leaf area index, first read the LAI values recorded at each location and match them with the corresponding spectral information of samples at different growth stages. To identify the leaf color, it is necessary to extract the reflection curves of the three major categories of red, green, and yellow in the visible light range. Comparing with the spectral distribution patterns for the green-reverting stage, jointing stage, and heading stage in the previous-year shooting library, if it is found that the proportion of the color characteristics of a certain pixel between 600 nm and 700 nm is relatively high, it will be classified into the region from reddish to orange-red, and combined with the texture direction to determine whether it has the characteristics of curling or being completely flat. When the texture direction is continuous and banded, it indicates that the leaf is in a normal growth mode, while when the texture direction is scattered and there are local fracture patterns, it is recorded as a possible damaged part. Then record these texture data for comparison with the spectral information together. Comparing with the leaf samples with multiple growth stages and known health evaluations, the parts with the largest differences in the current leaf texture distribution can be classified. For example, when setting a texture difference threshold When setting a texture difference threshold , first, based on the previous-year samples, calculate the average value of various texture characteristics, and then weight a fluctuation coefficient. If the set texture difference threshold , when the comparison result shows a deviation greater than 0.15, it is marked as abnormal. And in multiple growth cycles, if the spectral color gradually changes from green to yellowish-brown and the texture difference increases significantly, it can be inferred that it has entered the late maturity or lesion stage. Make a phased division of the growth status at this time period and conduct color spectral comparison. Finally, after processing all the leaf data, an overall growth status recognition conclusion can be formed to obtain the original LAI data.

[0093] The steps to obtain the time series analysis results are as follows:

[0094] Based on the original LAI data, construct a time series matrix, extract the rice leaf area index at each observation time point, calculate the change amplitude of the rice leaf area index between adjacent time points, fill in the data missing areas through interpolation, and perform time series smoothing processing to obtain the time series LAI data;

[0095] Based on the time series LAI data, calculate the crop growth trend index, and the calculation formula is:

[0096] ;

[0097] Among them, is the crop growth trend index, is the rice leaf area index at the th time point, is the rice leaf area index at the previous time point, is the leaf density corresponding to the th time point, is the rice canopy height, is the rate of change of rice leaf color, is the influence factor of color change on the trend index, is the number of data points in the time series;

[0098] Based on the crop growth trend index, analyze the change pattern of the crop growth trend index in the time dimension, extract the trend characteristics of growth, stability or decline, and obtain the time series analysis results.

[0099] Specifically, based on the original LAI data obtained previously, first summarize the rice leaf area index corresponding to each observation time point and map it to a time axis. Arrange all records in matrix form in chronological order and compare the records with obvious errors with the established rice growth curve of previous years. If the deviation is found to exceed the threshold set after comparing with previous samples, for example, take the mean value of the previous year's rice LAI plus or minus 0.8 as the reference judgment range. When any record exceeds the mean value by 1.6, it is considered that the deviation is too large. Then place these records with too large deviations in the list to be inspected and combine with manual inspection to check whether there are abnormal acquisitions or data missing. For the observation periods that are confirmed to be completely missing, they are filled in by interpolation. For example, when there are 3 days of data completely missing in a continuous 30-day observation sequence, extract the LAI values before and after the missing period and record them as and respectively. Divide the intermediate period into equidistant steps by linear interpolation and insert the intermediate LAI values day by day, and then complete the weighted average through a smoothing coefficient selected empirically in the range of 0.2 to 0.3. If the smoothing coefficient is set to 0.25, perform adjacent term weighted operations on the interpolated sequence in turn. After excluding anomalies, continuously check the continuity of the overall data in the time dimension. When all interpolation and smoothing processes are completed, a column of LAI time series without omissions and with low noise is counted to obtain the time series LAI data.

[0100] The benefit of the formula is that it couples the change amount of the rice leaf area index between adjacent time points with leaf density, canopy height and the rate of change of leaf color, enabling the growth trend to not only reflect the increase or decrease of the index value, but also reflect the fluctuations brought by the change amplitude of leaf morphology and color, thus more precisely reflecting the overall growth dynamics of rice.

[0101] The acquisition steps of represent the rice leaf area index measured at the th time point.

[0102] The acquisition steps of It is necessary to confirm whether there is a very significant increase or data anomaly. It can be determined by the average maximum change of previous year's samples. For example, through a historical record of 500 time periods, if the maximum daily change of LAI is 2.0, then can be set to 2.0 and the new observed data can be checked. Once the difference exceeds 2.0, it must be listed as a suspicious point. The corresponding difference is calculated at each time point and then summarized into a difference sequence, and then linked with FSj for calculation.

[0103] The acquisition steps of represent the leaf density corresponding to the th time point, which can be measured by reading the number of leaves per unit area in the captured image. Generally, it fluctuates between 500 plants per square meter and 800 plants per square meter, and this index needs to be corrected by considering the plant gap comprehensively. The monitoring method is through visual recognition or field measurement of the number of plants in a fixed quadrat. If there is local seedling sparseness, it will lead to a lower actual FSj. To quantify this value, multiple 1-meter-square grids are sampled in an area for statistics. The average number of leaves is obtained by dividing the total number of leaves by the number of grids, and then this average value is recorded as the basic amount of FSj. If the actual measurement in a certain place shows that the average number of leaves is 650 plants, then directly set FSj = 650.

[0104] The acquisition steps of are as follows. This square root term means amplifying the ratio of leaf density to LAIj + 1. The ratio of FSj to LAIj + 1 can range from 0 to hundreds. If FSj is too large, it will cause the ratio to increase significantly. To ensure the stability of the value, 1 is usually added to the denominator, and then the square root is used to reduce the excessive dispersion. Both FSj and LAIj + 1 come from previous actual measurements and calculation processes. Then, direct division is performed, 1 is added, and then the square root is taken. For example, at the observation time point j, if FSj = 600 and LAIj = 2.2, then the internal ratio is After adding 1 and taking the square root, we get

[0105] The acquisition steps of represent the rice canopy height at the th time point, generally in the range of 10 cm to 120 cm. The measurement method usually randomly selects representative positions in the field and uses a ruler to measure. When the observed value deviates significantly, it is necessary to conduct a secondary verification by combining the plant height records of previous years. After recording all height values, they are sorted according to the time point. If the height at a certain point is greater than the maximum height of previous years, it is necessary to check whether it belongs to a measurement error. Finally, the confirmed value is set as , for example, the height measured on the 20th day after transplanting is about 25 cm, about 55 cm on the 40th day, and about 85 cm on the 60th day, and thus can be serially recorded as etc., and input them into the denominator for summary calculation.

[0106] The acquisition steps of represent the rice leaf color change rate, which is quantified by analyzing the degree of evolution of the leaf color system towards yellowish or brownish in adjacent periods. Usually, first extract the main wavelength band position of the leaf color in the captured image and compare it with the wavelength band position at the previous time point. The change amount at the nm level is converted into a ratio from 0 to 1. If the color is yellowish, it will cause the peak of the wavelength band to shift towards a longer wavelength. If the shift amount is 6 nm, then divide this value by an average visible light wavelength span, such as from 400 nm to 700 nm taking 300 nm, to obtain 0.02 as a color offset rate. If there are greater fluctuations in subsequent periods, this ratio will be correspondingly increased. Finally, in a time series, mark all the leaf color change rates as , for example, if the wavelength band peak position of the color system measured in the j-th period is 560 nm and that in the j+1 period is 558 nm, then the color moves 2 nm and is calculated with 300 nm to obtain 0.0067, denoted as .

[0107] The acquisition steps of represent the influence factor of color change on the trend index. It is necessary to perform piecewise regression on the different degrees of influence observed from multiple experiments when the leaf changes from green to yellowish. Statistically, comparing the FDj of each period with the actual growth trend of the crop observed can obtain a set of samples, and then use the least squares method to fit out The value of , usually in the range of 0.5 to 2.0. For example, in multiple batches of rice experiments in a certain place, it is confirmed that is most consistent with the actual growth conditions observed in the field. Thus, is used as a fixed value for subsequent calculations.

[0108] The acquisition steps of M: m is the number of data points in the time series, directly from the observation frequency and the number of consecutive days set during the observation process. For example, observing once every other day for 30 days gives 15 time points, or observing twice a day for 14 days gives 28 time points. Here, m only needs to be consistent with the actual observation frequency. For example, if a certain field is observed once a day for 60 consecutive days, then m = 60.

[0109] Calculation process:

[0110] In the record of m = 5 time points, first substitute the LAI and FSj, F Hj, FDj of each period into the numerator and denominator. Let , , , , , let , substitute it into the numerator part:

[0111]

[0112] where is defined as the end value of the previous period, which can be recorded as 1.0, and calculate term by term:

[0113] For j = 1 term: ;

[0114] For j = 2 terms: ;

[0115] For j = 3 terms: ;

[0116] For j = 4 terms: ;

[0117] For j = 5 terms: ;

[0118] Total of the numerator: ;

[0119] Denominator part:

[0120]

[0121] Calculate in sequence:

[0122] For j = 1 term: ;

[0123] For j = 2 terms: ;

[0124] For j = 3 terms: ;

[0125] For j = 4 terms: ;

[0126] For j = 5 terms: ;

[0127] Total of the denominator: ;

[0128] Divide the numerator by the denominator:

[0129]

[0130] The results show that under the values of these 5 observation periods, the crop growth trend index T is approximately 0.42. The larger the value, the more significant the growth variation is. When T is greater than 3, it often corresponds to the rapid growth stage, while when T is lower than 1, it indicates that the growth change is relatively gentle.

[0131] Based on the previously obtained crop growth trend index, first list all the observation points in chronological order and sort the index values for each period. Compare with the multiple index distribution intervals summarized in the statistical data of previous years. If it is found that there are index values in the range below 1 or above 4, mark that there may be significant fluctuations in the corresponding period. According to the observation frequency, the overall time dimension can be divided into three segments: the early stage, the middle stage, and the late stage. Conduct a classified statistics on the index distribution of each stage to see if it shows continuous growth in the range of 1 to 2 or stage stability in the range of 3 to 4. Then, combine the additional parameters of leaf density and canopy height to analyze the trend trajectory of the index over time. If the index rises sharply in some periods of the middle stage and exceeds the threshold of 3.5 set in the actual measurement comparison of previous years in this region, mark that period as the rapid change period. Conduct a more detailed analysis of the time interval for these rapid change periods and compare with the measured values of the color change rate. When the index is continuously below 1 for multiple days, record it as the decline mode and list it in the special attention list. Finally, extract the corresponding growth, stability, or decline characteristics based on the distribution pattern of the index values of all periods on the time axis to obtain the time series analysis result.

[0132] The steps for obtaining the comprehensive monitoring data are as follows:

[0133] Based on the time series analysis result, extract the crop growth trend data, call the soil moisture measurement data and meteorological observation data, calculate the soil moisture change rate, air humidity change range, temperature gradient, and precipitation accumulation to obtain the environmental parameter set;

[0134] Based on the environmental parameter set, calculate the comprehensive monitoring index. The calculation formula is:

[0135]

[0136] Among them, is the comprehensive monitoring index, respectively represent the real-time measurement values of soil moisture content, air humidity, and temperature, are the time series smoothed values of the corresponding parameters respectively, respectively represent the fluctuation ranges of each environmental factor, respectively represent the change rates of soil moisture, air humidity, and temperature, is the environmental change impact factor;

[0137] Based on the comprehensive monitoring index, analyze the impact of environmental factors on the growth trend of crops, extract the correlation between environmental variables and crop growth status, and obtain comprehensive monitoring data.

[0138] Specifically, based on the time series analysis results obtained previously, first extract the crop growth trend data from it and correspond them to the corresponding observation time periods one by one. After obtaining the trend data, refer to the recorded soil moisture measurement information and local meteorological observation information. Number the soil moisture values sampled multiple times recently by day and compare them with the humidity mean of the same season in history. When the difference between the single-day measurement data and the mean value of previous years exceeds 2 percentage points of the humidity fluctuation threshold obtained through continuous field measurement, it is marked as a significant change, and record the soil moisture change rate during this period. At the same time, compare the air humidity measurement value with the established relative humidity reference range. If the deviation from the range is greater than the 3% floating boundary obtained through cumulative statistics of previous years, it is recorded as an obvious fluctuation, and record the change range of air humidity during this period. Then summarize the temperature observation content hourly and calculate the difference between the highest temperature and the lowest temperature within a day, which is regarded as the single-day temperature gradient. When the temperature gradient difference is greater than the upper limit of the daily average difference of 5°C observed in the local area in the past three years in summer, it is additionally recorded as extreme temperature gradient data. The precipitation is superimposed by accumulating all rainfall periods on the same day. When the cumulative amount exceeds the precipitation statistical range of 40 mm provided by the hydrological department after continuous observation for many years, it is marked as medium to high level precipitation. Arrange the soil moisture change rate, air humidity change range, temperature gradient, and precipitation cumulative amount recorded respectively in time series and correspond them to the crop growth trend data one by one. Comprehensively compare the differences between the changes in each time period and the similar time periods in previous years. After confirming that there are no obvious abnormal error records, unify and organize the four environmental factor values to form an environmental parameter set.

[0139] The advantage of the formula lies in combining the differences between the real-time measurement values and the smoothed values of the three elements of soil water content, air humidity, and temperature in a dynamic environment, and multiplying an additional quantization term for the fluctuation range in the numerator. At the same time, it is associated with the change rate through the exponential correction factor in the denominator, so as to more comprehensively evaluate the impact of the environment on the crop growth trend.

[0140] The steps for obtaining the parameters are as follows: Represents the real-time measurement value of soil water content, generally in the range of 0% to 50%. When monitoring, dig probe measuring points with a depth of about 10 cm in the field plot and record the soil moisture content at a frequency of every 6 hours. When it is found that a single measurement result is significantly higher than the dry-wet boundary value of 45% determined through many years of actual measurement, it is necessary to conduct a re-verification. The finally confirmed measurement data is recorded as , for example, the soil moisture content in a certain place for the last 4 days is 40%, 42%, 47%, and 46% in sequence. After summarizing the measured values of each day in their respective time periods, they can be called into the formula for calculation.

[0141] The steps for obtaining the parameters are as follows. represents the real-time measured value of air humidity. The common values are between 20% and 100%. Data is collected by a humidity recording device deployed 2 meters above the farmland at a frequency of once per hour. If multiple readings are continuously observed to exceed 80%, it is necessary to compare with the regional reference data. After confirming that the sampled value has no drift, it is recorded as , and corresponding to the humidity distribution range observed in the past. If the humidity in multiple hourly periods recorded on a certain day is between 60% and 70%, the measured values in this period can be regarded as conforming to the local common level. Finally, the humidity measured value of the day is used as the input to the formula for the corresponding time period. For example, at 8 o'clock on the first day, 65% is monitored, at 14 o'clock, 70% is monitored, and at 20 o'clock, 68% is monitored. Finally, it is recorded as a group sequence.

[0142] The steps for obtaining the parameters are as follows. represents the real-time measured value of temperature, with the unit of degree Celsius. It may fluctuate greatly in the range of 0°C to 40°C. The monitoring method is to obtain the temperature record at the same height in the field weather station and track it hourly. When the difference between the temperature reading of the day and the multi-year average temperature exceeds 6°C, it will be marked for extra attention in the record, and then it can be officially registered as after being confirmed as correct during subsequent statistics. For example, during a summer monitoring, the daily maximum temperature can reach 35°C and the minimum temperature is 25°C. After recording at multiple times throughout the day, several values are obtained and all are input into the corresponding time period in the formula.

[0143] The steps for obtaining the parameters are as follows. represents the time series smoothed value of the soil moisture content. The soil moisture content data measured over multiple days is arranged in a column and a relatively smooth curve is obtained by using the moving average method or the exponential smoothing method. The specific operation is to select a moving window length of 3 days and calculate the mean value of the soil humidity within each window, and then arrange all the mean values to form a new sequence and record it as , when it is found that the humidity change in a certain time period is relatively large when tracing back, the weighted moving average method can be used to suppress the sudden increase data. For example, the humidity measured within 5 days is 40%, 39%, 44%, 43%, 41%. Taking the mean value of adjacent 3 days as the smoothed output, we get approximately 40.3%, 42%, 42.6%, etc., and then form continuous values.

[0144] The steps to obtain the parameters are: Represents the time series smoothed value of air humidity, according to and A similar approach is used to make a weighted average of the air humidity recorded multiple times a day, amortize the errors of the extremely fluctuating hourly measurements, and then output the smoothed result and record it as , the range can be between 20% and 100%. If the humidity in a certain section changes dramatically, the length of the smoothing window can be shortened to ensure that the overall trend is accurately reflected. For example, the humidity in a certain area is 65% during the day, 80% at night, and 85% in the early morning. Then, the moving average is used to output smooth results such as 72%, 78%, etc., and finally concatenated into Used for subsequent formula calculations.

[0145] The steps to obtain the parameters are: The time series smoothing value of temperature is represented by marking individual observations that are too high or too low as noise in multiple temperature records monitored daily, and then calculating the mean in the form of a moving window and stringing them into a smooth sequence. If the temperature is collected once an hour in the summer, 24 data will be obtained. The mean of these data will be output as a moving window of 6 hours, and 4 smoother temperature values will be obtained. When they are continuously aggregated for multiple days, a set of If the temperature is only 18℃ at night on a certain day and is above 25℃ in most other periods, you need to check the fluctuations when calculating the mean value and confirm that the actual measurement is correct before writing it. .

[0146] The steps to obtain the parameters are: Indicates the fluctuation range of soil moisture content. The change in moisture content is measured by analyzing the soil moisture range at multiple times of the day. If the difference between the maximum and minimum values exceeds the 5% upper and lower thresholds accumulated over many years, it is classified as a large fluctuation. When recording, the difference is entered into the formula as the fluctuation range. If the soil moisture measured during the day changes from 36% to 42%, the fluctuation range can be determined as 6%. After combining the local threshold test to confirm that it is within the safe range, 6% is recorded as .

[0147] The steps to obtain the parameters are: Indicates the fluctuation range of air humidity. The range is confirmed by comparing the difference between the highest humidity and the lowest humidity within a day or several days. If the difference exceeds the previous experience threshold of 15%, it is considered to have a significant fluctuation and the difference is recorded as In terms of calculation, the maximum and minimum differences can be found in multiple observation periods every day, and then the thresholds can be compared after accumulation or average. If the sum of the highest and lowest differences in air humidity within three days is 45%, and the average difference is 15%, then it is marked 。

[0148] The steps for obtaining the parameter are as follows: represents the fluctuation range of temperature, which is obtained by recording the difference between the highest temperature and the lowest temperature within a day. If the span is close to 10°C for consecutive days, it indicates a large fluctuation. When compared with the local multi-year statistical average of 9°C, it can be regarded as a high fluctuation and registered. If the difference between the highest temperature of 35°C and the lowest temperature of 25°C on the same day is 10°C, then 10 is recorded as 。

[0149] The steps for obtaining the parameter are as follows: represents the change rate of soil moisture, which is obtained by dividing the difference in soil moisture between adjacent observation periods by the time interval. The change rates of each adjacent period are sorted out as a sequence. If the soil humidity rises from 30% to 32% in one day, then 2% is divided by 24 hours to get approximately 0.083% per hour. The daily average change rate is obtained by averaging the rates obtained from consecutive multiple observations and is rechecked when it exceeds the past statistical range. The finally confirmed result can be recorded within the range of 0.01% per hour to 1% per hour as 。

[0150] The steps for obtaining the parameter are as follows: represents the change rate of air humidity, which is obtained by taking the difference of humidity measurement records at hourly or shorter time granularity. When the value rises or falls significantly, it indicates a rapid change in humidity. If it rises from 50% to 60% within 2 hours, the change amplitude is 10%, which is converted to 5% per hour. If a similar trend occurs for two consecutive days, the daily average change rate of that day is set to 5% per hour and marked in the database. The finally obtained value is used as the denominator part in the formula.

[0151] The steps for obtaining the parameter are as follows: represents the change rate of temperature, which is obtained by dividing the difference in temperature between multiple observation periods by the time interval. For example, if the temperature measured in the early morning is 20°C and at noon is 30°C, then it changes by 10°C within 4 hours, and the rate is 2.5°C per hour. Then check the night period to see if there is an equal amplitude of temperature drop. Deriving a daily average change rate by taking the mean or peak value of the data for all periods can be recorded as , and when it is detected that this value exceeds the regional statistical extreme of 8°C per hour for consecutive days, manual review will be carried out. After finally confirming that it is correct, it is recorded in the formula as 。

[0152] The steps for obtaining the parameter are as follows: It represents the environmental change impact factor, which is obtained through comprehensive comparative statistics of the impact degrees of the changes in soil moisture, air humidity, and temperature collected from multiple locations on crop growth. Generally, it ranges from 0.5 to 2.0. The specific method is to collect the actually measured change rates and crop growth indicators in different farmland experiments, perform linear or non-linear regression on the two, and extract the best fitting points to determine When it is found that the error is the smallest, this value is fixed for subsequent unified use. For example, when comparing 30 fields in a certain area, after calculating the coupling degree between the color change rate and the environmental rate, it is found that can have the highest degree of coincidence with the field observations, that is, is determined.

[0153] Calculation process:

[0154] Select a calculation example scenario and specify , , , , and substitute these values into the formula:

[0155]

[0156] Then add the above three items:

[0157]

[0158] Finally, take the square root:

[0159]

[0160] The result shows that the comprehensive environmental monitoring index at this time is approximately 4.4938. When the value exceeds 5, it is related to large fluctuations in dryness and humidity and high-temperature environments. When the value is below 1, it indicates that the environmental factors are basically stable.

[0161] Based on the comprehensive monitoring index obtained previously, first, arrange the index values for multiple observation periods and classify them according to date and period. Then, read the crop growth trend data observed during the same period and directly compare it with the corresponding index values. When it is found that the index is within the range of 2 to 3 in the local historical statistics, the crop growth trend data in the farmland can be compared to see if it is also in a stable growth state. If the index is within the range of 4 to 5, check whether the soil moisture content during the observation period significantly exceeds the 3% daily average change threshold obtained from the previous statistics, and whether the measured values of air humidity or temperature deviate significantly from the previous year's average by more than 5%. When it is confirmed that these external environments have indeed fluctuated significantly, record the corresponding relationship between the environmental parameters and the crop growth trend during this period, and then establish time-series curves one by one for comparative observation. Superimpose the rise and fall of the index and the change in crop LAI. For some consecutive periods, combine the high index with the high increase in leaf area index to check whether it conforms to the historical statistical linear trend. If it is observed that the index increases significantly while the increase in leaf area index is small, add this period to the detailed list for subsequent in-depth inspection. Finally, compare the index of all periods with the crop growth status one by one and write it into the result comparison table to obtain the comprehensive monitoring data.

[0162] The steps to obtain the crop growth assessment result are as follows:

[0163] Based on the comprehensive monitoring data, calculate the growth adaptability score. The calculation formula is:

[0164]

[0165] Where, is the growth adaptability score, is the rice leaf area index, is the crop photosynthetic capacity index, is the soil temperature, is the air humidity, is the cumulative impact value of meteorological factors, is the soil moisture change rate;

[0166] Based on the growth adaptability score, analyze the growth state of rice in the current environment to obtain the crop growth assessment result.

[0167] Specifically, the advantage of the formula lies in comprehensively considering the correlations among the rice leaf area index, the crop photosynthetic capacity index, the soil temperature, the air humidity, as well as the cumulative impact value of meteorological factors and the soil moisture change rate. It measures the gap between crop physiological indicators and external temperature and humidity with the molecular part, and then corrects it through the cumulative impact value and humidity change rate in the denominator, enabling the calculation result to more precisely quantify the growth adaptation degree of rice under different environments and growth conditions.

[0168] The steps to obtain the LAI parameters are as follows: LAI represents the rice leaf area index, which has been obtained previously.

[0169] The steps to obtain the P parameter are:

[0170] P stands for the crop photosynthetic capacity index, which is generally used to describe the level of rice's absorption and conversion of sunlight energy. This index can range from 0 to 2 and is obtained by combining an actual field photometer with a plant physiological meter. The specific steps are to record the correlation between light intensity and leaf photosynthetic rate during the noon period of rice, and then convert the average maximum net photosynthetic rate calculated from multiple samples into a P score. In a certain place, 30 representative rice plants can be selected in the middle of the growth period, and the photosynthetic rate can be recorded and the average value calculated within the 3 hours of the strongest sunlight. For example, the net photosynthetic rate per hour is measured to be 25 micromoles of carbon dioxide per square meter second, and after comparing it with the average rate measured in previous years, it is concluded that P is about 1.7.

[0171] The steps to obtain the T parameter are:

[0172] T represents soil temperature in degrees Celsius, which is generally obtained from a probe 5 to 10 cm below the soil surface. During the monitoring process, temperature sensors are placed in the field and the values are recorded at hourly or half-hourly intervals. To ensure accuracy, the sensors need to be calibrated periodically and the temperatures collected at multiple points are weighted averaged. The soil temperature in a certain area can be monitored at 6 a.m., 12 noon, and 6 p.m., for example, after measuring 24°C, 28°C, and 26°C, an average value is weighted and finally registered as T=25°C. When there is extreme weather that may cause the value to exceed the local historical average by ±5°C, this part of the data must be recalibrated before it can be officially recorded as T.

[0173] The steps to obtain the YH parameters are:

[0174] YH represents air humidity, and its dimension is usually between 0 and 1. It uses decimal form to represent relative humidity, and can also be expressed as a percentage value between 20% and 100%. The specific quantification method is to collect data through a humidity sensor installed about 2 meters above the ground, and then regularly compare it with the comparison table provided by the National Meteorological Observatory to eliminate sensor offset. Field monitoring can record relative humidity multiple times a day, and then combine it into a weighted average as a representative of YH. If the humidity recorded in a day is 70%, 80%, and 76%, the average value can be set to 0.75 (i.e. 75%), and it can be recorded in the formula to form the input item of YH=0.75.

[0175] The steps to obtain the YS parameters are:

[0176] YS represents the cumulative impact value of meteorological factors, which is used to reflect the cumulative impact effect of meteorological changes on crop growth over a long period. This value usually has no fixed upper limit and is accumulated through experimental statistics or continuous observations. Specifically, it is obtained by first integrating the meteorological data of the crop over a period of time, including the number of extreme temperature events, the duration of excessive humidity, precipitation, and abnormal sunshine periods, etc., and assigning an impact score to each meteorological event, and then adding up the scores of multiple events to obtain YS. In a certain area, the daily maximum temperature exceeding 35°C within a one-month observation period can be regarded as an impact event with a score of 2, the relative humidity below 40% with a score of 3, and extreme rainstorms with a score of 1, etc. After adding up the total scores, it is recorded as YS. If the final cumulative score is 2.0, then substitute YS = 2.0 into the formula.

[0177] The steps to obtain the YD parameter are as follows:

[0178] YD represents the rate of change of soil moisture, which can measure the dynamic degree of soil moisture change over time. The field method is to first collect soil moisture values hourly in a fixed monitoring area, and define the ratio of the difference between adjacent time periods to the time interval as the instantaneous rate of change. Then, taking the average of one or several observation days can obtain the representative value of YD. In a certain place, the process of monitoring the soil moisture rising from 30% to 32% within 24 hours takes 6 hours. The difference of 2% divided by 6 hours is equal to 0.33% per hour. After multiple observations and taking the average to form 0.4% per hour, and then converting it to 0.004, that is, 0.4% is substituted into the formula. At this time, YD = 0.004.

[0179] Calculation process:

[0180] Now select an example and let .

[0181] First calculate the numerator:

[0182]

[0183] Then calculate the denominator:

[0184]

[0185] Divide the numerator by the denominator:

[0186]

[0187] The result shows that the growth adaptability score is close to 6, indicating that under the conditions of soil temperature of 25°C, air humidity of 0.8, and LAI = 3.2, etc., the growth of rice matches the environment in the upper-middle range. Generally, a value greater than 8 indicates extremely high adaptability, and a value below 2 may indicate growth limitation and requires additional inspection.

[0188] Based on the growth adaptability scores obtained previously, first record the environmental conditions and rice monitoring data corresponding to each score at different time nodes one by one. By organizing the soil temperature, air humidity, and LAI values for all monitoring periods, a comparison table is formed, and information such as the photosynthetic capacity index P and the cumulative impact value YS of meteorological factors is added to the comparison table. Then, arrange the scores for each day or hour within a continuous observation period and check for large fluctuations. Combine the soil moisture change rate YD to jointly evaluate the source of score changes. When the continuous curve of the score is lower than the empirical threshold of 2 in certain periods, it is necessary to re-check whether the air humidity within that period deviates significantly from the range of 60% ± 5% of the average daily value observed in this region over the years, and check whether the soil temperature exceeds the range of 15°C to 35°C established according to local statistics. If there are obvious deviations, record that period as a potential adverse growth interval and mark the actual situation, such as rainfall and the depth of waterlogging in the field plot. Then, observe whether the score rebounds in the following days. If the score continuously exceeds 8, it indicates that the environment and photosynthetic conditions are highly compatible. Record it in the high adaptability segment and compare it with the actual growth performance of the crop. When all periods are matched, conduct a time series statistics on these score partitions, output the division results including the low and medium intervals and the high interval, providing basic support for subsequent targeted determination of the rice growth status. Finally, at the summary stage, compare all partitions and trends one by one and check them against the latest on-site observations of the farmland to obtain the crop growth assessment result.

[0189] The steps to obtain management adjustment suggestions are as follows:

[0190] Based on the crop growth assessment result, extract the soil moisture content, soil nitrogen, phosphorus, and potassium content, chlorophyll index, precipitation, and temperature to obtain the basic agricultural management data;

[0191] Based on the basic agricultural management data, calculate the management adjustment index. The calculation formula is:

[0192]

[0193] where, is the management adjustment index, respectively represent the nitrogen, phosphorus, and potassium concentrations in the current soil, is the supply amount of irrigation water, is the total precipitation, is the chlorophyll index, is the crop leaf temperature;

[0194] Based on the management adjustment index, analyze the adaptability of rice to the current irrigation and fertilization, evaluate whether agricultural management measures need to be adjusted, and obtain management adjustment suggestions.

[0195] Specifically, the advantage of the formula lies in integrating several key factors, namely the concentrations of nitrogen, phosphorus, and potassium in the soil, the irrigation water supply, the total precipitation, the chlorophyll index, and the crop leaf temperature. By introducing a non-linear expression of the impact on chlorophyll and leaf temperature in the denominator, it can more intuitively reflect the balance degree between the fertilizer-water supply and the crop physiological state.

[0196] The steps for obtaining the parameters are as follows:

[0197] The concentration value representing the soil nitrogen content is generally expressed in milligrams per kilogram of soil (mg / kg) or as a percentage. To ensure accuracy, first randomly select several sampling points from different plots and obtain soil samples within the 0-20 cm plough layer. Detect the contents of ammonium nitrogen, nitrate nitrogen, etc. in them through laboratory chemical analysis methods and sum them up. If a certain area hopes to quantify the total nitrogen, 200 grams of soil can be taken from each of 10 sampling points within one day, totaling 2 kilograms, and then use methods such as distillation with alkali and colorimetric determination to measure the nitrogen content. If it is found through measurement that the nitrogen content per kilogram of soil is between 120 mg and 300 mg, organize this value into an average and record it as In one operation, if the nitrogen content per kilogram of soil is observed to be 200 mg, then mg / kg can be incorporated into the formula.

[0198] The steps for obtaining the parameters are as follows:

[0199] Represents the soil phosphorus concentration, mostly using available phosphorus (Olsen-P or available phosphorus) as an indicator, with the unit of mg / kg. The detection process usually involves collecting soil from the field plough layer, extracting it with sodium bicarbonate, and then using a spectrophotometer to measure the phosphorus concentration in the extract, and finally converting it back to the soil content to form value. If the average available phosphorus content in the plough layer of a region is around 20 mg / kg, and for high-phosphorus fertilizer application areas it may reach 40 to 50 mg / kg, unified laboratory tests need to be carried out after sampling. If the average value of three tests is 30 mg / kg, then mg / kg is brought into the formula.

[0200] The steps for obtaining the parameters are as follows:

[0201] Represents the soil potassium concentration, usually presented in the form of available potassium, and the unit can also be mg / kg. When obtaining it, the soil sample needs to be dried and sieved, then extracted with ammonium acetate solution and measured with a flame photometer. If the detection result of the available potassium in the plough layer of a certain area shows an interval of 50 to 150 mg / kg, then the representative average value or the weighted result can be recorded as For example, after taking multi-point sampling and analysis of a piece of soil, if the average available potassium content is measured to be 80 mg / kg, it can be marked as .

[0202] The steps for obtaining the parameter are as follows:

[0203] represents the supply amount of irrigation water, which is often recorded in tons per hectare or cubic meters per hectare. When obtaining it, first count the actual amount of water input into the field by the irrigation channel, or directly read it from the flowmeter installed at the irrigation outlet, and convert it in combination with the irrigation duration and the field area. If a total of 50 tons of water is injected into 1 hectare of paddy field in a certain area within a week, it can be recorded as tons / ha.

[0204] The steps for obtaining the parameter are as follows:

[0205] represents the total precipitation, which is usually recorded in millimeters or cubic meters per hectare. When monitoring, a rain gauge or an automatic weather station can be used. The official weather station and the field rain gauge are compared with each other and registered according to the daily cumulative value. The precipitation within several days or a week is combined as the total amount. If the total precipitation depth recorded within three days is 30 millimeters, it means that about 300 cubic meters of rainwater is input per hectare. If this value is recorded as mm.

[0206] The steps for obtaining the parameter are as follows:

[0207] represents the chlorophyll index, which is used to reflect the chlorophyll content level in crop leaves. This value can range from 0 to 100, or relative values can also be used. The specific obtaining method is to use a handheld chlorophyll meter (SPAD meter) or a multi-spectral drone to take pictures in the field, extract the spectral reflection characteristics in the leaf area and convert them into the chlorophyll index. If 10 representative plants are selected in a paddy field, 3 leaves are measured for each plant and averaged, and then a set of data is recorded by the instrument. If the average reading reaches 45, the chlorophyll index can be obtained by referring to the instrument calibration table for this value .

[0208] The steps for obtaining the parameter are as follows:

[0209] Indicates the crop leaf temperature, in degrees Celsius, usually obtained by means of an infrared thermometer or a thermal imager. To reflect the temperature distribution of the main leaf surface of the crop, temperature collection can be carried out on the top and middle and lower leaves of several rice plants during sunny periods, and then the results of multiple measurements are statistically unified and averaged. If the temperature of the top leaves is about 28°C and the temperature of the middle and lower leaves is about 26°C on a certain day and is synthesized to 27°C after multiple measurements in a day, then . When it is detected that the plant temperature significantly exceeds the local historical range, it is necessary to check whether drought or poor ventilation has occurred before recording the final temperature value.

[0210] Calculation process:

[0211] Let mg / kg, mg / kg, mg / kg, tons per hectare, millimeters, , , after uniformly differentiating the units and making reasonable conversions, substitute them into the formula:

[0212]

[0213] Numerator:

[0214]

[0215] Denominator:

[0216]

[0217] Overall:

[0218]

[0219] This result shows that when the soil nitrogen, phosphorus, and potassium contents are in the above range and the irrigation water volume, precipitation, chlorophyll index, and leaf temperature are all maintained at the corresponding levels, the management adjustment index can reach about 60,000 magnitude. If it is found in the actual operation of this example that the value is too large, it is necessary to study and judge the impact of nitrogen, phosphorus, and potassium concentrations, irrigation volume, or precipitation on the overall management. Usually, for better explanation, it is also necessary to combine the actual farmland area or further adjust the unit dimension to make the value have a reference meaning closer to the operation habits of growers.

[0220] Based on the management adjustment index obtained earlier, the actual measured nitrogen, phosphorus and potassium concentrations of each rice field, as well as the irrigation amount, precipitation in recent days and leaf temperature records are first summarized, and the chlorophyll index is used as an additional reference item to check whether it matches the planting process. If it is found that the leaf temperature in a certain period of time obviously exceeds the highest critical value of 35°C obtained from local statistics for many years, the soil moisture in the soil moisture monitoring results is compared to see whether it is lower than the reference humidity range of 25% to 40%. If high leaf temperature and low soil moisture are met at the same time, it may cause an imbalance in the transpiration rate. Therefore, the management adjustment index and chlorophyll readings of the period are recorded as related parameters, and then the fertilization and irrigation arrangements in the short term are compared to see whether they are obviously too little, and whether the supplementary amount of nitrogen, phosphorus and potassium is far lower than the average range of 30 to 50 grams per mu per week in previous years. This information is combined with the recent cumulative precipitation value. It is necessary to combine the nitrogen, phosphorus and potassium concentration records to observe whether the lack of nitrogen leads to the stagnation of crop growth, or whether the phosphorus and potassium concentrations have already exceeded the regional average and unnecessary excessive fertilization is still caused. Then, from the perspective of temperature and humidity, it is checked whether the leaf temperature continues to rise or the chlorophyll index is difficult to increase. After confirming the actual situation, fine-tune the fertilizer and water management details. For example, the irrigation amount can be increased by 5% to 10% or the time period for nitrogen fertilizer application can be adjusted, and the application time point can be close to the concentrated precipitation period to penetrate the soil more evenly. Finally, after comparing the index calculations of multiple observation periods with the field observation information, the irrigation and fertilization are evaluated in combination with the changes before and after, and the conclusion on whether to adjust the agricultural management measures is output to obtain management adjustment suggestions.

[0221] The present invention provides a system, comprising:

[0222] The data acquisition module deploys drones to fly over the target rice fields, collects image data, and sends them to the ground station in real time to generate a preliminary image data set;

[0223] The image processing module crops and corrects the preliminary image data set to generate a processed image data set;

[0224] The LAI calculation module, based on the processed image dataset, identifies and measures the rice leaf area index in the image, identifies the health status and growth stage of the crop, and generates raw LAI data;

[0225] The data analysis module performs time series analysis on the original LAI data, determines the crop growth trend through comparative analysis, and generates time series analysis results;

[0226] The agricultural management suggestion module, based on the results of time series analysis, combines soil humidity and meteorological data, combines the soil humidity and meteorological data with the results of time series analysis to generate comprehensive monitoring data, uses the comprehensive monitoring data to evaluate the growth status of rice, analyzes the environmental factors affecting crop growth, generates crop growth evaluation results, based on the crop growth evaluation results, evaluates the effectiveness of the current irrigation and fertilization plans, determines whether agricultural management measures need to be adjusted, and generates management adjustment suggestions.

[0227] 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 they do not depart from the technical content of the technical solution of the present invention, any simple modifications, equivalent changes and modifications 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 method for monitoring the growth of crops through imaging, characterized in that, Including the following steps: Deploy drones to fly over the target rice paddy, collect image data, and send the image data to the ground station in real time through data transmission to generate a preliminary image dataset; crop and correct the preliminary image dataset to generate a processed image dataset; Based on the processed image dataset, identify and measure the rice leaf area index in the image, identify the health status and growth stage of the crop, and generate raw LAI data; perform time series analysis on the raw LAI data, and determine the crop growth trend through comparative analysis to generate time series analysis results; Based on the time series analysis results, combine soil moisture and meteorological data, combine soil moisture and meteorological data with the time series analysis results to generate comprehensive monitoring data; use the comprehensive monitoring data to evaluate the growth status of rice, analyze the environmental factors affecting crop growth, and generate crop growth evaluation results; Based on the crop growth evaluation results, evaluate the effectiveness of the current irrigation and fertilization plans, determine whether agricultural management measures need to be adjusted, and generate management adjustment suggestions; The steps for obtaining the raw LAI data are as follows: Based on the processed image dataset, perform RGB channel separation, extract the pixel values of the red, green, and blue channels to obtain rice leaf area channel data; Based on the rice leaf area channel data, calculate the rice leaf area index, and the calculation formula is: ； and ； Among them, is the rice leaf area index, is the variable ratio vegetation index based on UAV images, is the digital value of the green channel, is the digital value of the red channel, is the digital value of the blue channel; Based on the rice leaf area index, extract leaf color and texture features, compare the spectral change trends during the growth cycle, identify the health status and growth stage of the crop, and obtain raw LAI data.

2. The method for monitoring the growth of crops through image monitoring according to claim 1, characterized in that, The steps for obtaining the preliminary image dataset are as follows: Deploy drones to fly over the target rice paddy, collect image data of the rice paddy through a camera, adjust the camera angle and focal length to cover the entire field surface, and obtain real-time collected rice paddy images; Based on the real-time collected rice paddy images, send the data to the ground station in real time through wireless transmission to obtain a preliminary image dataset.

3. The method for monitoring the growth of crops through image monitoring according to claim 1, wherein The steps for obtaining the processed image dataset are as follows: Based on the preliminary image dataset, determine the edge transition area through pixel gradients, eliminate isolated noise points, adjust the boundary range according to the shape characteristics of the target crop area, and remove redundant background areas to obtain image data with adjusted boundaries; Based on the image data with adjusted boundaries, calculate lens distortion parameters, including radial distortion and tangential distortion, and use image interpolation to restore pixel offsets caused by correction to generate distortion-corrected image data; Based on the distortion-corrected image data, perform color channel normalization processing, and at the same time perform histogram equalization to adjust the brightness distribution to obtain a processed image dataset.

4. The method for monitoring the growth of crops through image monitoring according to claim 1, characterized in that, The steps for obtaining the time series analysis results are as follows: Based on the raw LAI data, construct a time series matrix, extract the rice leaf area index at each observation time point, calculate the change amplitude of the rice leaf area index between adjacent time points, fill in the data missing area through interpolation, and perform time series smoothing processing to obtain time series LAI data; Based on the time - series LAI data, calculate the crop growth trend index, and the calculation formula is as follows: ； Among them, is the crop growth trend index, is the rice leaf area index at the -th time point, is the rice leaf area index at the previous time point, is the leaf density corresponding to the -th time point, is the rice canopy height, is the rice leaf color change rate, is the influence factor of color change on the trend index, is the number of data points in the time series; Based on the crop growth trend index, analyze the change pattern of the crop growth trend index in the time dimension, extract the trend characteristics of growth, stability or decline, and obtain the time - series analysis result.

5. The method for monitoring the growth of crops through image monitoring according to claim 1, characterized in that, The steps for obtaining the comprehensive monitoring data are as follows: Based on the time - series analysis result, extract the crop growth trend data, call the soil moisture measurement data and meteorological observation data, calculate the soil moisture change rate, air humidity change range, temperature gradient and precipitation accumulation, and obtain the environmental parameter set. Based on the environmental parameter set, calculate the comprehensive monitoring index, and the calculation formula is as follows: ； Among them, is the comprehensive monitoring index, respectively represent the real-time measurement values of soil moisture content, air humidity, and temperature, are the time series smoothed values of the corresponding parameters, respectively represent the fluctuation amplitudes of each environmental factor, respectively represent the change rates of soil moisture, air humidity, and temperature, is the environmental change impact factor; Based on the comprehensive monitoring index, analyze the impact of environmental factors on the crop growth trend, extract the correlation between environmental variables and crop growth status, and obtain the comprehensive monitoring data.

6. The method for monitoring crop growth by image as claimed in claim 1, wherein, The steps for obtaining the crop growth assessment result are as follows: Based on the comprehensive monitoring data, calculate the growth adaptability score, and the calculation formula is as follows: ； Among them, is the growth adaptability score, is the rice leaf area index, is the crop photosynthetic capacity index, is the soil temperature, is the air humidity, is the cumulative influence value of meteorological factors, is the soil moisture change rate; Based on the growth adaptability score, analyze the growth status of rice in the current environment, and obtain the crop growth assessment result.

7. The method for monitoring crop growth by image as claimed in claim 1, wherein The steps for obtaining the management adjustment suggestion are as follows: Based on the crop growth assessment result, extract the soil moisture content, soil nitrogen, phosphorus and potassium content, chlorophyll index, precipitation and temperature, and obtain the basic agricultural management data. Based on the basic agricultural management data, calculate the management adjustment index, and the calculation formula is as follows: ； wherein, is the management adjustment index, respectively represent the nitrogen, phosphorus, and potassium concentrations in the current soil, is the supply amount of irrigation water, is the total precipitation, is the chlorophyll index, is the crop leaf temperature; Based on the management adjustment index, analyze the adaptability of rice to the current irrigation and fertilization, evaluate whether agricultural management measures need to be adjusted, and obtain the management adjustment suggestion.

8. The system for the method of monitoring crop growth by image monitoring according to any one of claims 1-7, characterized in that, Including: A data acquisition module, which deploys an unmanned aerial vehicle to fly over the target rice field, collect image data, and send it to the ground station in real - time to generate a preliminary image data set. An image processing module, which crops and corrects the preliminary image data set to generate a processed image data set. An LAI calculation module, which based on the processed image data set, identifies and measures the rice leaf area index in the image, identifies the health status and growth stage of the crop, and generates the original LAI data. A data analysis module, which conducts time - series analysis on the original LAI data, determines the crop growth trend through comparative analysis, and generates the time - series analysis result. An agricultural management suggestion module, which based on the time - series analysis result, combines the soil moisture and meteorological data, combines the soil moisture and meteorological data with the time - series analysis result to generate comprehensive monitoring data, uses the comprehensive monitoring data to evaluate the growth status of rice, analyzes the environmental factors affecting crop growth, generates the crop growth assessment result, and based on the crop growth assessment result, evaluates the effect of the current irrigation and fertilization plan, determines whether agricultural management measures need to be adjusted, and generates the management adjustment suggestion.

Images