A deep learning-based automatic monitoring method for crop planting structure

Abstract

The application discloses a kind of based on deep learning's crop planting structure automatic monitoring method, specifically related to data processing technical field, including S1, multi-source data acquisition, using GIS tool, the latitude and longitude boundary of target farmland area is demarcated, generates geographic fence, according to crop growth cycle setting data acquisition frequency;Through multi-source remote sensing data synchronous acquisition protocol, in remote sensing data acquisition layer deployment multi-source sensor fusion framework, construct time-space alignment's time-space reference dataset;The application adopts time-space adaptive acquisition technology, through the acquisition frequency of dynamic scheduling satellite, unmanned aerial vehicle, weather and ground data, improve data processing efficiency, realize the full-course automation from data acquisition to decision generation, reduce manual intervention;The application designs compatible field crop and flexible monitoring framework of facility agriculture, supports the accurate monitoring of multiple crops, adapts to the needs of different planting scales and production modes.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, more particularly, the present application relates to a kind of automatic monitoring method of crop planting structure based on deep learning. BACKGROUND

[0002] With the development of agricultural modernization, accurate monitoring of crop planting structure is of great significance for rational planning of agricultural production, improving the efficiency of agricultural resources, and ensuring food security. Monitoring crop planting structure helps to ensure food security and supply balance and improve resource utilization efficiency. Therefore, it is of great practical significance to develop an automatic monitoring method of crop planting structure based on deep learning.

[0003] Traditional crop planting structure monitoring methods mostly use manual field investigation method, in which professional personnel go deep into planting areas and use their own experience and professional knowledge to observe crop characteristics, determine crop species, and obtain detailed information on planting varieties, planting time, irrigation and fertilization.

[0004] However, it still has some shortcomings in actual use, such as low efficiency, traditional crop monitoring methods rely on manual field investigation, which is low in efficiency, high in cost and difficult to be widely promoted; insufficient recognition accuracy, traditional methods rely on professional experience, only make judgments in a short time, and are difficult to adapt to crop growth timing changes; small application range, traditional crop monitoring methods for manual field investigation only monitor crop conditions in a small area at a certain time, which is difficult to realize real-time monitoring of large areas.

[0005] Therefore, it is urgent to provide an automatic monitoring method of crop planting structure based on deep learning to solve the problems of low efficiency, insufficient recognition accuracy and small application range of existing crop monitoring methods SUMMARY

[0006] In order to overcome the above-mentioned defects of the prior art, embodiments of the present application provide an automatic monitoring method of crop planting structure based on deep learning, which solves the problems raised in the background art by the following scheme.

[0007] To achieve the above-mentioned purpose, the present application provides the following technical scheme: an automatic monitoring method of crop planting structure based on deep learning, comprising:

[0008] S1, multi-source data acquisition, using GIS tool, demarcating the latitude and longitude boundary of target farmland area, generating geographic fence, setting data acquisition frequency according to crop growth cycle; through multi-source remote sensing data synchronous acquisition protocol, deploying multi-source sensor fusion framework in remote sensing data acquisition layer, constructing time-space aligned time-space reference dataset;

[0009] S2, core index calculation, based on S1 to obtain data after standardization processing to establish mathematical model to calculate index, generate satellite multispectral data influence coefficient, unmanned aerial vehicle image data influence coefficient, meteorological data influence coefficient and ground measured data influence coefficient four kinds of crop planting structure core index;

[0010] S3, comprehensive index judgment, through the nonlinear coupling of satellite multispectral data influence coefficient, unmanned aerial vehicle image data influence coefficient, meteorological data influence coefficient and ground measured data influence coefficient obtained by S2, the integrity and reliability of the comprehensive reflection of the crop planting structure are obtained. Crop planting structure influence index;

[0011] S4, automatic monitoring of crop planting structure, based on the crop planting structure influence index obtained by S3, the planting structure grading response is carried out, and the abnormal processing is carried out according to the response result, and the target farmland crop planting structure influence index change curve is generated;

[0012] S5, data interaction, the crop planting structure influence index, the planting structure grading response result, the abnormal processing step, the target farmland crop planting structure influence index change curve and the target farmland data are transported to the user data end, and the reference data for the user to make adjustment measures is provided.

[0013] Preferably, the space-time reference data set includes satellite multispectral data, unmanned aerial vehicle image data, meteorological data and ground measured data.

[0014] Preferably, the satellite multispectral data includes spectral separability, denoted as JM; spectral temporal consistency, denoted as ACF; spectral spatial coverage, denoted as Csat; unmanned aerial vehicle image data includes unmanned aerial vehicle image texture entropy, denoted as Entropy; unmanned aerial vehicle image edge sharpness, denoted as Esharp; unmanned aerial vehicle image resolution factor, denoted as Ruva; meteorological data includes water stress index, denoted as WSI; temperature suitability, denoted as TSI; extreme weather frequency, denoted as f; ground measured data includes sampling density index, denoted as SDI; measurement accuracy evaluation, denoted as MA; space-time representation, denoted as STR.

[0015] Preferably, the satellite multispectral data influence coefficient is constructed by introducing the weighted average of spectral separability, temporal consistency and spatial coverage to construct the comprehensive effectiveness index of satellite data, and the arithmetic average form is adopted to balance the contribution of each item, wherein the spectral separability reflects the spectral differentiation ability of crop type; the spectral temporal consistency reflects the regularity of capturing the growth process, and the spectral spatial coverage is used to quantify the data integrity, which is used to evaluate the reliability of satellite data in crop classification, specifically:

[0016] .

[0017] Preferably, the UAV image data influence coefficient is enhanced by the geometric coupling of texture entropy, edge sharpness and resolution factor, and the normalized product form is used to strengthen the synergistic effect of features, in which the UAV texture entropy represents the complexity of crop morphology, the UAV edge sharpness reflects the boundary clarity, and the UAV resolution factor quantifies the spatial detail retention capability, which is used to determine whether the UAV image quality meets the needs of precision agriculture, specifically:

[0018] .

[0019] Preferably, the meteorological data influence coefficient is based on the Sigmoid function to fuse water stress, temperature suitability and extreme weather frequency, and the exponential term is used to enhance the interaction between water and temperature, the coefficient 10 controls the gradient change rate, and the offset -0.6 adjusts the warning threshold, which is used to real-time warning of meteorological stress risk, specifically:

[0020] .

[0021] Preferably, the ground measured data influence coefficient is obtained by geometrically averaging the sampling density, measurement accuracy and spatio-temporal representativeness, and the square root operation is used to weaken the influence of extreme values, the product form requires balanced development of each item, and the dynamic weight reflects the keyness of the verification data, which is used to calibrate the credibility of remote sensing inversion results, specifically:

[0022] .

[0023] Preferably, the crop planting structure influence index is calculated by the product form to strengthen the synergistic effect between data sources, and the calculation formula is specifically:

[0024] .

[0025] Preferably, the planting structure grading response process is specifically: when the crop planting structure influence index is greater than 0.8, a normal monitoring report is generated, and the existing planting plan is maintained; when 0.6 is less than or equal to 0.8, the drip irrigation system is automatically started, the frequency of unmanned aerial vehicle, weather and ground data acquisition is increased to the next stage, the planting range of the corresponding crop type is reduced to the theoretical crop planting structure influence index of 0.8, the planting range of the crop is reduced according to the theoretical crop planting structure influence index, and the new theoretical crop data is transmitted to S1; when 0.4 is less than or equal to 0.6, an agronomist is dispatched for on-site verification, the planting range of the crop is reduced to the safety threshold of the crop planting structure influence index, and the frequency of unmanned aerial vehicle, weather and ground data acquisition is increased to the early warning stage; when the crop planting structure influence index is less than or equal to 0.4, the whole region is scanned by the unmanned aerial vehicle group networking, the disaster emergency plan is started, the insurance automatic reporting system is triggered, and then the data acquisition is re-performed; wherein the safety threshold is the average of the historical crop planting structure influence index and the target crop planting structure influence index; the frequency of the early warning stage is the minimum value of the historical crop planting structure influence index.

[0026] Preferably, the abnormal processing process is specifically: when the αs reduction rate reaches 0.4 / day, the αu is normal, indicating that the satellite data quality is problematic, and historical data is used for replacement; the αu continuously <0.5+αg normal, indicating that the camera lens is contaminated and needs to be automatically cleaned and calibrated; when all the coefficients decrease synchronously, indicating a major disaster event, an emergency plan needs to be started.

[0027] The technical effects and advantages of the present application are:

[0028] 1. The present application adopts a space-time adaptive acquisition technology, dynamically adjusts the acquisition frequency of unmanned aerial vehicle, weather and ground data to improve data processing efficiency, realizes full automation from data acquisition to decision generation, and reduces manual intervention.

[0029] 2. The present application greatly improves the crop classification accuracy through multi-dimensional feature fusion of spectral feature separability analysis, texture entropy calculation and weather stress modeling, and reduces the boundary recognition error.

[0030] 3. The present application designs a flexible monitoring framework compatible with field crops and facility agriculture, supports precise monitoring of multiple crops, and adapts to the needs of different planting scales and production modes. BRIEF DESCRIPTION OF DRAWINGS

[0031] Figure 1 It is a schematic diagram of the overall structure of the present application.

[0032] Figure 2 It is a schematic diagram of the flow structure of the present application. DETAILED DESCRIPTION

[0033] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of the present application.

[0034] As shown in a deep learning-based automatic monitoring method for crop planting structure, comprising. Figure 1 As shown in a deep learning-based automatic monitoring method for crop planting structure, comprising.

[0035] S1, multi-source data acquisition, using GIS tools, demarcating the latitude and longitude boundary of the target farmland area, generating a geographic fence, setting the data acquisition frequency according to the crop growth cycle; through a multi-source remote sensing data synchronous acquisition protocol, deploying a multi-source sensor fusion framework in the remote sensing data acquisition layer, and constructing a spatio-temporal reference dataset.

[0036] In this embodiment, it needs to be specifically pointed out that the spatio-temporal reference dataset includes satellite multispectral data, unmanned aerial vehicle image data, meteorological data and ground measured data.

[0037] In this embodiment, it needs to be specifically pointed out that the satellite multispectral data includes spectral separability, denoted as JM; spectral temporal consistency, denoted as ACF; spectral spatial coverage, denoted as C sat ; unmanned aerial vehicle image data includes unmanned aerial vehicle image texture entropy, denoted as Entropy; unmanned aerial vehicle image edge sharpness, denoted as E sharp ; unmanned aerial vehicle image resolution factor, denoted as R uva ; meteorological data includes water stress index, denoted as WSI; temperature suitability, denoted as TSI; extreme weather frequency, denoted as f; ground measured data includes sampling density index, denoted as SDI; measurement accuracy evaluation, denoted as MA; spatio-temporal representativeness, denoted as STR.

[0038] In this embodiment, it needs to be specifically pointed out that the spectral separability uses a deep convolutional neural network to automatically learn the spatial-spectral features of multispectral data, and obtains the mean value of each band reflectivity and the covariance matrix of the main crops in the target area through an API L2A level ground reflectivity data, calculates the JM distance according to the crop category, that is, the spectral separability, which is used to quantify the ability of satellite data to distinguish different crops, and specifically:

[0039] ,

[0040] Among them

[0041] ,

[0042] Among them μi and μ j respectively represent the mean vector of crop class i and j in multispectral bands, Σ i and Σ j respectively represent the covariance matrix of crop class i and j; B ij represents the Bhattacharyya distance, which measures the overlap degree of two class distributions.

[0043] In this embodiment, it needs to be specifically pointed out that the spectral timing consistency is extracted by Sentinel-2 time sequence image to obtain near-infrared band reflectivity and red band reflectivity, which is calculated according to pixels, and is used for monitoring abnormal growth, and specifically:

[0044] ,

[0045] Among them

[0046] ,

[0047] Among them, B8 represents near-infrared band reflectivity, B4 represents red band reflectivity, NDVI t represents the NDVI value of the t phase, represents the timing NDVI mean value, and T represents the total phase number.

[0048] In this embodiment, it needs to be specifically pointed out that the spatial coverage is generated by analyzing the QA60 band to generate a cloud mask, and the cloud mask is generated by analyzing the QA60 band, which is used to reflect data availability, and specifically:

[0049] ,

[0050] Among them, N cloud represents the number of cloud coverage pixels, and N total represents the total number of pixels.

[0051] In this embodiment, it needs to be specifically pointed out that the texture entropy of the unmanned aerial vehicle image is monitored by using the unmanned aerial vehicle image of OpenCV to run Laplacian blur, the gray level co-occurrence matrix is calculated by using Pix4Dmapper to generate orthophoto, and the entropy value is calculated according to the matrix, and specifically:

[0052] ,

[0053] Among them, p (i，j) represents the probability value of (i, j) position in the gray level co-occurrence matrix.

[0054] In this embodiment, it needs to be specifically pointed out that the edge sharpness of the unmanned aerial vehicle image is calculated by Canny edge monitoring to calculate the edge gradient mean value, which is used to evaluate the clarity of the farmland boundary, and specifically:

[0055] ,

[0056] wherein ∇I represents the image gradient amplitude, (x i , y i ) represents the coordinate position of the edge pixel, and N represents the total number of edge pixels.

[0057] In the embodiment, it is specifically required to explain that the UAV image resolution factor is calculated by shooting a standard resolution test card, and specifically, the resolution factor is calculated by shooting a standard resolution test card.

[0058] ,

[0059] wherein res represents the resolution.

[0060] In the embodiment, it is specifically required to explain that the water stress index is obtained by a meteorological station, and the reference evapotranspiration is calculated by using a Penman-Monteith model to obtain the water stress index, and specifically, the water stress index is obtained by a meteorological station, and the reference evapotranspiration is calculated by using a Penman-Monteith model to obtain the water stress index.

[0061] ,

[0062] wherein ET0 represents the reference evapotranspiration, and P represents the precipitation.

[0063] In the embodiment, it is specifically required to explain that the temperature suitability degree is replaced by the daily average air temperature, and the crop optimum temperature is calculated to obtain, and specifically, the temperature suitability degree is replaced by the daily average air temperature, and the crop optimum temperature is calculated to obtain.

[0064] ,

[0065] wherein T represents the daily average temperature, T opt represents the crop optimum temperature, and σ represents the crop tolerance range.

[0066] In the embodiment, it is specifically required to explain that the extreme weather frequency is directly obtained by the judgment standard that the daily maximum temperature > 35℃ or the daily precipitation > 50mm, and the duration is more than 3 days.

[0067] In the embodiment, it is specifically required to explain that the sampling density index is recorded by arranging one sampling point per 10 hectares to record the biomass, the plant height, and the chlorophyll content, and the sampling point number per unit area is calculated by RTK measurement of the sampling point coordinates, and specifically, the sampling density index is recorded by arranging one sampling point per 10 hectares to record the biomass, the plant height, and the chlorophyll content, and the sampling point number per unit area is calculated by RTK measurement of the sampling point coordinates.

[0068] ,

[0069] wherein N1 represents the sampling point number, A represents the area, and λ represents the attenuation coefficient, and specifically, the attenuation coefficient is 0.001 by default.

[0070] In this embodiment, it needs to be specifically pointed out that the measurement accuracy evaluation synchronously collects unmanned aerial vehicle image and ground data, calculates relative error, and specifically comprises the following steps:

[0071] ,

[0072] Wherein y k The average of the measured values of the ground biomass, plant height and chlorophyll content of the kth sample, The average of the unmanned aerial vehicle multispectral image ground biomass, plant height and chlorophyll inversion value of the kth sample, K represents the total number of verification samples.

[0073] In this embodiment, it needs to be specifically pointed out that the space-time representation is calculated by recording the sampling time and space distribution, and calculating the coverage ratio, and specifically comprises the following steps:

[0074] ,

[0075] Wherein T1 represents the number of sampling days, T t The number of days in the growth season, wherein S1 represents the sampling area, T t The total area of farmland.

[0076] S2, core index calculation, based on the data obtained after standardization, a mathematical model is established to calculate the index, and four types of crop planting structure core indexes of satellite multispectral data influence coefficient, unmanned aerial vehicle image data influence coefficient, meteorological data influence coefficient and ground measured data influence coefficient are generated.

[0077] In this embodiment, it needs to be specifically pointed out that the satellite multispectral data influence coefficient is constructed by introducing the weighted average of spectral separability, time sequence consistency and spatial coverage rate, and the comprehensive effectiveness index of satellite data is constructed, and the arithmetic mean form is adopted to balance the contribution of each item, wherein the spectral separability reflects the spectral differentiation ability of crop type; The regularity of capturing the growth process is reflected by spectral time sequence consistency, and spectral spatial coverage rate is used to quantify data integrity, which is used to evaluate the reliability of satellite data in crop classification, and specifically comprises the following steps:

[0078] .

[0079] In this embodiment, it needs to be specifically pointed out that the unmanned aerial vehicle image data influence coefficient is coupled by texture entropy, edge sharpness and resolution factor, and the normalized product form is adopted to strengthen the feature synergistic effect, wherein the unmanned aerial vehicle texture entropy represents the complexity of crop morphology, the unmanned aerial vehicle edge sharpness reflects the boundary definition, and the unmanned aerial vehicle resolution factor quantifies the space detail retention ability, which is used to determine whether the unmanned aerial vehicle image quality meets the needs of precision agriculture, and specifically comprises the following steps:

[0080] .

[0081] In this embodiment, it is particularly necessary to explain that the meteorological data influence coefficient is based on the Sigmoid function to fuse water stress, temperature suitability and extreme weather frequency, the exponential term is used to enhance the interaction between water and temperature, the coefficient 10 is used to control the gradient change rate, and the offset -0.6 is used to adjust the early warning threshold, which is used for real-time early warning of meteorological stress risk, specifically:

[0082] .

[0083] In this embodiment, it is particularly necessary to explain that the ground measured data influence coefficient is obtained by using geometric mean to comprehensively consider sampling density, measurement accuracy and space-time representativeness, square root operation is used to weaken the influence of extreme value, product form is used to require balanced development of each part, dynamic weight reflects the keyness of verification data, and is used for calibrating the reliability of remote sensing inversion result, specifically:

[0084] .

[0085] S3, comprehensive index judgment, through the nonlinear coupling of satellite multispectral data influence coefficient, unmanned aerial vehicle image data influence coefficient, meteorological data influence coefficient and ground measured data influence coefficient obtained by S2, the integrity and reliability of the farmland planting structure are comprehensively reflected, and the crop planting structure influence index is obtained.

[0086] In this embodiment, it is particularly necessary to explain that the crop planting structure influence index strengthens the synergistic effect between data sources through product form, and the calculation formula is specifically:

[0087] .

[0088] S4, automatic monitoring of crop planting structure, based on the crop planting structure influence index obtained by S3, the planting structure grading response is carried out, the abnormal processing is carried out according to the response result, and the target farmland area crop planting structure influence index change curve is generated.

[0089] In this embodiment, it is particularly necessary to explain that the planting structure grading response process is specifically:

[0090] When the crop planting structure influence index is greater than 0.8, a normal monitoring report is generated, and the existing planting plan is maintained; when 0.6 is less than or equal to 0.8, the drip irrigation system is automatically started, the frequency of unmanned aerial vehicle, weather and ground data collection is increased to the next stage, the planting range of the corresponding crop type is reduced to the theoretical crop planting structure influence index of 0.8, the planting range of the crop is reduced according to the theoretical crop planting structure influence index, and the new theoretical crop data is transmitted to S1; when 0.4 is less than or equal to 0.6, an agronomist is dispatched for on-site verification, the planting range of the crop is reduced to the safety threshold of the crop planting structure influence index, and the frequency of unmanned aerial vehicle, weather and ground data collection is increased to the frequency of the early warning stage; when the crop planting structure influence index is less than or equal to 0.4, the whole region is scanned by the unmanned aerial vehicle group networking, the disaster emergency plan is started, the insurance automatic reporting system is triggered, and then the data collection is re-performed.

[0091] In the embodiment, it is specifically pointed out that the safety threshold is the average of the historical crop planting structure influence index and the target crop planting structure influence index; and the early warning stage frequency is the minimum value of the historical crop planting structure influence index.

[0092] In the embodiment, it is specifically pointed out that the abnormal processing process specifically includes: when alpha s The reduction rate reaches 0.4 / day, alpha u Normal, indicating a satellite data quality problem, and historical data is used for replacement; alpha g Normal, indicating camera lens contamination, which needs to be automatically cleaned and calibrated; and when all the coefficients decrease synchronously, indicating a major disaster event, an emergency plan needs to be started.

[0093] S5, data interaction, the crop planting structure influence index, the planting structure classification response result, the abnormal processing step, the target farmland crop planting structure influence index change curve and the target farmland data are transmitted to a user data end, and reference data for the user to make adjustment measures is provided.

[0094] Secondly, in the drawings of the disclosed embodiment, only the structures involved in the disclosed embodiment are involved, other structures can be referred to the general design, and in the case of no conflict, the same embodiment and different embodiments of the present application can be combined with each other;

[0095] Finally, the above only describes the preferred embodiments of the present application, and is not used to limit the present application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1.A deep learning-based automatic monitoring method for crop planting structure, characterized in that, The method comprises the following steps: S1, multi-source data acquisition, using GIS tools, demarcating the latitude and longitude boundary of the target farmland area, generating a geographic fence, and setting the data acquisition frequency according to the crop growth cycle; Through a multi-source remote sensing data synchronous acquisition protocol, a multi-source sensor fusion framework is deployed in the remote sensing data acquisition layer to construct a spatio-temporal reference dataset that is spatio-temporally aligned; The spatio-temporal reference dataset includes satellite multispectral data, unmanned aerial vehicle image data, meteorological data, and ground measured data; The satellite multispectral data includes spectral separability, denoted as JM; spectral temporal consistency, denoted as ACF; and spectral spatial coverage, denoted as Csat; the unmanned aerial vehicle image data includes unmanned aerial vehicle image texture entropy, denoted as Entropy; unmanned aerial vehicle image edge sharpness, denoted as Esharp; and unmanned aerial vehicle image resolution factor, denoted as Ruva; the meteorological data includes water stress index, denoted as WSI; temperature suitability, denoted as TSI; and extreme weather frequency, denoted as f; and the ground measured data includes sampling density index, denoted as SDI; measurement accuracy evaluation, denoted as MA; and spatio-temporal representativeness, denoted as STR; S2, core index calculation, after standardizing the data obtained in S1, a mathematical model is established to calculate the indexes, generating four types of crop planting structure core indexes, namely satellite multispectral data influence coefficient, unmanned aerial vehicle image data influence coefficient, meteorological data influence coefficient, and ground measured data influence coefficient; The satellite multispectral data influence coefficient is constructed by introducing the weighted average of spectral separability, temporal consistency, and spatial coverage to balance the contributions of each item, wherein the spectral separability reflects the spectral differentiation ability of crop types; the spectral temporal consistency reflects the regularity of capturing the growth process; and the spectral spatial coverage is used to quantify the data integrity for evaluating the reliability of satellite data in crop classification, specifically: The unmanned aerial vehicle image data influence coefficient is coupled by the texture entropy, edge sharpness, and resolution factor, and adopts a normalized product form to strengthen the feature synergy effect, wherein the unmanned aerial vehicle texture entropy represents the complexity of crop morphology, the unmanned aerial vehicle edge sharpness reflects the boundary clarity, and the unmanned aerial vehicle resolution factor quantifies the spatial detail retention ability, which is used to determine whether the unmanned aerial vehicle image quality meets the needs of precision agriculture, specifically: ； The meteorological data influence coefficient is fused based on the Sigmoid function, including water stress, temperature suitability, and extreme weather frequency, and the interaction between water and temperature is enhanced through an exponential term, wherein the coefficient 10 controls the gradient change rate, and the offset -0.6 adjusts the warning threshold, which is used to real-time warning of meteorological stress risk, specifically: ； The ground measured data influence coefficient is obtained by using geometric mean to comprehensively consider the sampling density, measurement accuracy, and spatio-temporal representativeness, and the square root operation is used to weaken the influence of extreme values, and the product form requires balanced development of each item, and the dynamic weight reflects the keyness of the verification data, which is used to calibrate the credibility of the remote sensing inversion result, specifically: ； ​ ； S3, comprehensive index judgment, through the nonlinear coupling of the satellite multispectral data influence coefficient, the unmanned aerial vehicle image data influence coefficient, the meteorological data influence coefficient and the ground measured data influence coefficient obtained in S2, the integrity and reliability of the farmland planting structure are comprehensively reflected, and a crop planting structure influence index is obtained; The crop planting structure influence index strengthens the synergistic effect between data sources in a product form, and the calculation formula is specifically: ； S4, automatic monitoring of crop planting structure, based on the crop planting structure influence index obtained in S3, a planting structure grading response is performed, and according to the response result, an abnormality is processed, and a target farmland area crop planting structure influence index change curve is generated; S5, data interaction, the crop planting structure influence index, the planting structure grading response result, the abnormality processing step, the target farmland area crop planting structure influence index change curve and the target farmland data are delivered to the user data end, and reference data for the user to make adjustment measures is provided. 2.The deep learning-based automatic monitoring method for crop planting structure according to claim 1, characterized in that: The planting structure grading response process is specifically: When the crop planting structure influence index is greater than 0.8, a normal monitoring report is generated, and the existing planting plan is maintained; when 0.6<crop planting structure influence index≤0.8, the drip irrigation system is automatically started, the unmanned aerial vehicle, meteorological and ground data acquisition frequency is increased to the next stage, the planting range of the corresponding crop is reduced to the theoretical crop planting structure influence index of 0.8, the planting range of the crop is reduced according to the theoretical crop planting structure influence index, and the new theoretical crop data is delivered to S1; when 0.4<crop planting structure influence index≤0.6, an agronomist is dispatched for on-site verification, the planting range of the crop is reduced to the safety threshold of the crop planting structure influence index, the unmanned aerial vehicle, meteorological and ground data acquisition frequency is increased to the early warning stage; when the crop planting structure influence index is less than or equal to 0.4, the whole region is scanned by the unmanned aerial vehicle group network, the disaster emergency plan is started, the insurance automatic reporting system is triggered, and then the data acquisition is performed again; Wherein the safety threshold is the average value of the historical crop planting structure influence index and the target crop planting structure influence index; the early warning stage frequency is the minimum value of the historical crop planting structure influence index. 3.The deep learning-based automatic monitoring method for crop planting structure according to claim 1, characterized in that: The abnormality processing process is specifically: when α s The rate reaches 0.4 / day, α u Normal, indicating that the satellite data quality problem is enabled to replace the historical data; αᵤ continues to be <0.5+α g Normal, indicating that the camera lens is contaminated and needs to be cleaned and calibrated; when all coefficients decrease synchronously, it indicates a major disaster event, and an emergency plan needs to be started.

Images