Crop monitoring method based on remote sensing big data

Abstract

The invention provides a crop monitoring method based on remote sensing big data, and relates to the technical field of agricultural remote sensing monitoring. The method comprises the following steps: acquiring a high-resolution remote sensing image; acquiring spectral reflectivity, structural data and growth state data of various crops; inputting the spectral reflectivity and the structural data into a trained pest and disease damage probability prediction model to obtain pest and disease damage probability prediction data; determining whether the various crops meet preset growth conditions or not; if the multiple crops meet the preset growth conditions, determining growth state scores of the multiple crops according to the growth state data; if the multiple crops do not meet the preset growth conditions, the growth state score of the multiple crops is 0; and generating a crop monitoring report. According to the invention, the crop growth state can be monitored according to the integration of the pest and disease damage probability prediction data and the multi-dimensional data, so that the comprehensiveness, accuracy and reliability of crop monitoring are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of agricultural remote sensing monitoring, and in particular to a method for monitoring crops based on remote sensing big data. Background Art

[0002] In the related art, CN116012720A relates to a method, device and medium for monitoring the growth of crops based on high-resolution remote sensing images, belonging to the technical field of agricultural remote sensing monitoring. The method includes: predicting a plurality of monitoring time periods corresponding to each crop to be monitored in the area to be monitored; obtaining high-resolution remote sensing image data of the area to be monitored in the current time period; determining the target crop corresponding to the crop to be monitored in the monitoring time period including the current time period; analyzing the growth of the target crop according to the high-resolution remote sensing image data to obtain the growth information of the target crop; obtaining comparison data information of the target crop corresponding to the current time period; the comparison data information includes historical growth information and preset index information; analyzing the growth information of the target crop and the comparison data information to obtain the monitoring result of the target crop. This solution has the effect of saving remote sensing image monitoring resources.

[0003] CN110287944A discloses a method for monitoring crop pests in multi-spectral remote sensing images based on deep learning, belonging to the technical field of satellite remote sensing image processing and application. The purpose of this solution is to solve the problems of poor timeliness of existing methods for monitoring crop pests, unstable single-index spectral information, and difficult acquisition of hyperspectral UAV data. This solution uses a combination of 10 groups of characteristic band spectra to construct an LSTM long short-term memory network, and trains a model capable of classifying crop pests in remote sensing images through deep learning. This solution can automatically and efficiently identify crop pest disasters from multi-spectral satellite remote sensing images, providing certain technical support for many fields such as agricultural production disaster prediction and prevention, and agricultural insurance claims.

[0004] Therefore, in the related art, although the growth of crops and pest disasters can be monitored, the related art does not monitor the growth state of crops based on the integration of pest probability prediction data and multi-dimensional data, so the comprehensiveness, accuracy and reliability of crop monitoring cannot be improved.

[0005] The information disclosed in the background art of the present application is only intended to deepen the understanding of the general background art of the present application, and should not be regarded as an admission or any form of suggestion that this information constitutes the prior art known to those skilled in the art. Summary of the Invention

[0006] The present invention provides a method for monitoring crops based on remote sensing big data, which can solve the technical problem in the related art that the growth state of crops is not monitored according to the integration of pest probability prediction data and multi-dimensional data.

[0007] According to the first aspect of the present invention, there is provided a method for monitoring crops based on remote sensing big data, including: monitoring the crop growth area through remote sensing technology to obtain high-resolution remote sensing images at multiple moments in the current monitoring period; according to the remote sensing images, obtaining the spectral reflectance, structural data and growth state data of various crops, wherein the spectral reflectance includes the spectral reflectance at multiple wavelengths, the structural data includes the planting area and planting density, and the growth state data includes leaf area index data, average leaf surface temperature data, biomass data and chlorophyll content data; inputting the spectral reflectance and the structural data into a trained pest probability prediction model to obtain pest probability prediction data of various crops at multiple moments in the current monitoring period; determining whether various crops meet the preset growth conditions according to the pest probability prediction data; if various crops meet the preset growth conditions, determining the growth state score of various crops according to the growth state data; if various crops do not meet the preset growth conditions, the growth state score of various crops is 0; generating a crop monitoring report according to the growth state score.

[0008] Further, determining whether various crops meet the preset growth conditions according to the pest probability prediction data includes: obtaining soil iron oxide content data at multiple moments in the current monitoring period; determining whether various crops meet the preset growth conditions according to the soil iron oxide content data and the pest probability prediction data.

[0009] Further, determining whether various crops meet the preset growth conditions according to the soil iron oxide content data and the pest probability prediction data includes: according to the formula and determining the first condition C1 and the second condition C2, wherein, is the pest probability prediction data of the i-th crop at the k-th moment in the current monitoring period, is the pest probability threshold of the i-th crop, F is a piecewise function of comparing the pest probability prediction data of the i-th crop at the k-th moment in the current monitoring period with the pest probability threshold of the i-th crop, is the preset average pest probability of the i-th crop in a monitoring period, is the soil iron oxide content data at the k-th moment in the current monitoring period, is the soil iron oxide content standard value, is the preset average relative difference of the soil iron oxide content data of the i-th crop within a monitoring period, M is the number of moments in the monitoring period, k ≤ M, and both i, k, and M are positive integers. The first condition C1 and the second condition C2 are the preset growth conditions. When the first condition C1 and the second condition C2 are both satisfied, it is determined that the i-th crop meets the preset growth conditions.

[0010] Further, the training steps of the pest and disease probability prediction model include: obtaining historical remote sensing images of the crop growth area at multiple moments in multiple historical monitoring periods; according to the historical remote sensing images, obtaining the historical spectral reflectance and historical structure data of multiple crops at multiple moments in multiple historical monitoring periods, where the historical structure data includes historical planting area and historical planting density; inputting the historical spectral reflectance and historical structure data into the trained pest and disease probability prediction model to obtain historical pest and disease probability prediction data of multiple crops at multiple moments in multiple historical monitoring periods; obtaining historical pest and disease probability data of multiple crops at multiple moments in multiple historical monitoring periods; obtaining historical temperature data and historical humidity data at multiple moments in multiple historical monitoring periods; determining the loss function of the pest and disease probability prediction model according to the structure data, the historical structure data, the historical pest and disease probability prediction data, the historical pest and disease probability data, the historical temperature data, and the historical humidity data; training the pest and disease probability prediction model according to the loss function of the pest and disease probability prediction model to obtain the trained pest and disease probability prediction model.

[0011] Further, determining the loss function of the pest and disease probability prediction model according to the structure data, the historical structure data, the historical pest and disease probability prediction data, the historical pest and disease probability data, the historical temperature data, and the historical humidity data includes: according to the formula Determine the loss function of the pest and disease probability prediction model , where is the planting area of the i-th crop in the current monitoring period, is the historical planting area of the i-th crop in the h-th historical monitoring period, is the planting density of the i-th crop in the current monitoring period, is the historical planting density of the i-th crop in the h-th historical monitoring period, is the historical temperature data of the i-th crop at the k-th moment in the h-th historical monitoring period, is the standard temperature data of the i-th crop, is the historical humidity data of the i-th crop at the k-th moment in the h-th historical monitoring period, is the standard humidity data of the i-th crop, is the historical pest and disease probability data of the i-th crop at the k-th moment in the h-th historical monitoring period, is the historical pest and disease probability prediction data of the i-th crop at the k-th moment in the h-th historical monitoring period. M is the number of moments in the monitoring period, H is the number of historical monitoring periods, h ≤ H, k ≤ M, and i, k, M, h, and H are all positive integers.

[0012] Furthermore, if multiple crops meet the preset growth conditions, then according to the growth state data, determine the growth state scores of the multiple crops, including: obtaining the standard growth state data of the multiple crops, where the standard growth state data includes standard leaf area index data, standard average leaf surface temperature data, standard biomass data, and standard chlorophyll content data; determining the standard growth state vector of the multiple crops according to the standard growth state data; determining the real-time growth state vector of the multiple crops according to the growth state data; and determining the growth state scores of the multiple crops according to the standard growth state vector and the real-time growth state vector.

[0013] Furthermore, determining the growth state scores of the multiple crops according to the standard growth state vector and the real-time growth state vector includes: according to the formula determine the growth state score of the i-th crop , where is the leaf area index data of the i-th crop at the k-th moment in the current monitoring period, is the average leaf surface temperature data of the i-th crop at the k-th moment in the current monitoring period, is the biomass data of the i-th crop at the k-th moment in the current monitoring period, is the chlorophyll content data of the i-th crop at the k-th moment in the current monitoring period, is the standard leaf area index data of the i-th crop at the k-th moment in the current monitoring period, is the standard average leaf surface temperature data of the i-th crop at the k-th moment in the current monitoring period, is the standard biomass data of the i-th crop at the k-th moment in the current monitoring period, is the standard chlorophyll content data of the i-th crop at the k-th moment in the current monitoring period, is the real-time growth state vector of the i-th crop at the k-th moment in the current monitoring period, is the standard growth state vector of the i-th crop at the k-th moment in the current monitoring period. M is the number of moments in the monitoring period, k ≤ M, and i, k, and M are all positive integers.

[0014] Further, according to the growth status score, a crop monitoring report is generated, including: if the growth status score of the i-th crop is greater than or equal to the first growth status score threshold, it is determined that the growth status of the i-th crop is excellent; if the growth status score of the i-th crop is less than the first growth status score threshold and greater than or equal to the second growth status score threshold, it is determined that the growth status of the i-th crop is good; if the growth status score of the i-th crop is less than the second growth status score threshold, it is determined that the growth status of the i-th crop is poor.

[0015] According to a second aspect of the present invention, there is provided a crop monitoring system based on remote sensing big data, including: a remote sensing image module for monitoring the crop growth area through remote sensing technology to obtain high-resolution remote sensing images at multiple moments in the current monitoring period; a data module for obtaining spectral reflectance, structural data, and growth status data of multiple crops according to the remote sensing images, wherein the spectral reflectance includes spectral reflectance at multiple wavelengths, the structural data includes planting area and planting density, and the growth status data includes leaf area index data, average leaf surface temperature data, biomass data, and chlorophyll content data; a pest and disease probability prediction data module for inputting the spectral reflectance and the structural data into a trained pest and disease probability prediction model to obtain pest and disease probability prediction data of multiple crops at multiple moments in the current monitoring period; a judgment module for determining whether multiple crops meet preset growth conditions according to the pest and disease probability prediction data; a growth status score determination module for determining the growth status score of multiple crops according to the growth status data if multiple crops meet the preset growth conditions; a growth status score of 0 module for setting the growth status score of multiple crops to 0 if multiple crops do not meet the preset growth conditions; and a crop monitoring report module for generating a crop monitoring report according to the growth status score.

[0016] Technical effects: According to the present invention, through remote sensing technology, data can be efficiently obtained within a wide range of crop growth areas, saving labor and time costs. Through spectral reflectance, structural data, and growth status data, comprehensive crop information is provided. This multi-dimensional data can more accurately reflect the actual growth conditions and health levels of crops. Through the trained pest and disease probability prediction model, the risk of pests and diseases can be identified in real time, and control measures can be taken in a timely manner to reduce losses and improve crop yield and quality. Based on the integration of pest and disease probability prediction data and multi-dimensional data, the growth status of crops can be monitored, thereby improving the comprehensiveness, accuracy, and reliability of crop monitoring. When determining whether multiple crops meet the preset growth conditions, based on the soil iron oxide content data and the pest and disease probability prediction data of different crops, the first condition and the second condition can be determined. When the first condition and the second condition are both met, it can be determined that the i-th crop meets the preset growth conditions. By quantifying the pest and disease probability prediction data and the soil iron oxide content data, the accuracy of judging the preset growth conditions of crops is improved. Monitoring the soil iron oxide content and the pest and disease probability, and adjusting the planting strategy and management measures in a timely manner can optimize the crop growth environment. When determining the loss function of the pest and disease probability prediction model, during the training process, the influence of historical temperature data and historical humidity data on the pest and disease probability can be used to determine the influence of the above data on the error of the historical pest and disease probability prediction data. Then, based on this influence and the difference between the historical pest and disease probability data and the historical pest and disease probability prediction data, and based on the characteristic that the shorter the time interval from the start time, the higher the accuracy, weights are set, and the closer the structural data of the current monitoring period is to the structural data of the historical control period, the closer the pest and disease probability is, and the greater the reference value of the error of the historical pest and disease probability prediction data. Weights are set accordingly, and the errors output by the pest and disease probability prediction model at each moment of each monitoring period are weighted and summed to obtain the loss function, so as to improve the design accuracy and objectivity of the loss function, thereby improving the training efficiency during the training process and enhancing the accuracy of the pest and disease probability prediction model. When determining the growth status scores of multiple crops, based on the cosine similarity between the real-time growth status vector and the standard growth status vector, the growth status scores of multiple crops can be determined. Based on different growth status data, the growth status of crops can be comprehensively reflected, which helps to timely detect the growth anomalies of crops.

[0017] It should be understood that the above general description and the following detailed description are exemplary and explanatory, rather than limiting the present invention. According to the following detailed description of the exemplary embodiments with reference to the accompanying drawings, other features and aspects of the present invention will become clearer. Brief Description of the Drawings

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other embodiments can be obtained based on these drawings; Figure 1 Exemplarily shown is a flowchart of a crop monitoring method based on remote sensing big data according to an embodiment of the present invention; Figure 2 Exemplarily shown is a block diagram of a crop monitoring system based on remote sensing big data according to an embodiment of the present invention. Detailed implementation manners

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

[0020] The following will detail the technical solutions of the present invention with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments.

[0021] Figure 1A flowchart of a crop monitoring method based on remote sensing big data according to an embodiment of the present invention is exemplarily shown. The method includes: Step S101, monitoring the crop growth area through remote sensing technology to obtain high-resolution remote sensing images at multiple moments in the current monitoring period; Step S102, according to the remote sensing images, obtaining the spectral reflectance, structural data, and growth status data of various crops. Among them, the spectral reflectance includes spectral reflectances at various wavelengths, the structural data includes planting area and planting density, and the growth status data includes leaf area index data, average leaf surface temperature data, biomass data, and chlorophyll content data; Step S103, inputting the spectral reflectance and the structural data into a trained pest and disease probability prediction model to obtain pest and disease probability prediction data of various crops at multiple moments in the current monitoring period; Step S104, determining whether various crops meet the preset growth conditions according to the pest and disease probability prediction data; Step S105, if various crops meet the preset growth conditions, determining the growth status score of various crops according to the growth status data; Step S106, if various crops do not meet the preset growth conditions, the growth status score of various crops is 0; Step S107, generating a crop monitoring report according to the growth status score.

[0022] According to the crop monitoring method based on remote sensing big data of the embodiment of the present invention, through remote sensing technology, data can be efficiently obtained within a wide range of crop growth areas, saving labor and time costs. Through spectral reflectance, structural data, and growth status data, comprehensive crop information is provided. This multi-dimensional data can more accurately reflect the actual growth situation and health level of crops. Through the trained pest and disease probability prediction model, the risk of pests and diseases can be identified in real time, and control measures can be taken in a timely manner to reduce losses and improve crop yield and quality. The growth status of crops can be monitored based on the integration of pest and disease probability prediction data and multi-dimensional data, thereby improving the comprehensiveness, accuracy, and reliability of crop monitoring.

[0023] According to an embodiment of the present invention, in Step S101, each monitoring period can be set to one week, two weeks, etc., and the interval between adjacent moments can be set to 12 hours, 24 hours, etc. The present invention does not limit this. Determining the crop growth area to be monitored includes specific farmland, planting areas, or a larger range of agricultural land. Using remote sensing technology usually involves using devices such as satellites, drones, or other aircraft to obtain high-resolution remote sensing images of the crop growth area. Remote sensing technology can capture changes and features on the earth's surface.

[0024] According to an embodiment of the present invention, in step S102, through the spectral information of the remote sensing image, the spectral reflectance of crops at multiple wavelengths is extracted, which involves different bands (for example, visible light, near-infrared, etc.), and each band corresponds to a specific spectral reflectance value. Through remote sensing image processing software, the remote sensing image is corrected and analyzed to calculate the spectral reflectance of each crop, so as to evaluate the health status and growth of the crops. Using image segmentation and classification techniques in the remote sensing image, the planting areas of crops are identified and quantified, and the total planting area of each crop is calculated. By analyzing the distribution of the crop within the planting area of each crop, the planting density of the crop is obtained, that is, the number of crops per unit area. The leaf area index is calculated through the spectral reflectance of remote sensing, which reflects the leaf coverage of the plant and is usually closely related to the growth status of the crop. The average temperature of the leaves is calculated through the thermal infrared data obtained by remote sensing technology, which helps to evaluate the physiological health of the crops. The biomass of the crops is estimated by combining the spectral reflectance data of remote sensing with the biomass model, that is, the dry weight or total biomass of the crops, which reflects the growth potential of the crops. The chlorophyll content in the crop leaves is estimated using the spectral data in the remote sensing image.

[0025] According to an embodiment of the present invention, in step S103, the collected spectral reflectance and structural data are input into the trained pest and disease probability prediction model, and the input data features can be used to predict the probability of pest and disease occurrence. The pest and disease probability prediction model can be a deep learning neural network model, and the present invention does not limit the specific type of the pest and disease probability prediction model.

[0026] According to an embodiment of the present invention, in step S104, according to the pest and disease probability prediction data, it is determined whether multiple crops meet the preset growth conditions.

[0027] According to an embodiment of the present invention, step S104 includes: at multiple moments in the current monitoring period, obtaining soil iron oxide content data; according to the soil iron oxide content data and the pest and disease probability prediction data, determining whether multiple crops meet the preset growth conditions.

[0028] According to an embodiment of the present invention, at multiple moments in the current monitoring period, soil samples are collected, and chemical analysis methods, such as spectral analysis or chemical titration, are used to determine the iron oxide content in the soil samples. The obtained soil iron oxide content data and pest and disease probability prediction data are compared with the preset growth conditions, that is, it is checked whether the pest and disease probability of each crop in the current monitoring period is lower than the preset risk threshold, and at the same time, whether the soil iron oxide content is within the appropriate range. According to the results of the comparative analysis, it is determined whether each crop meets the preset growth conditions.

[0029] According to an embodiment of the present invention, determining whether various crops meet the preset growth conditions based on the soil iron oxide content data and the pest and disease probability prediction data includes: determining a first condition C1 and a second condition C2 according to formulas (1) and (2). (1), (2), wherein, is the pest and disease probability prediction data of the i-th crop at the k-th moment in the current monitoring period, is the pest and disease probability threshold of the i-th crop, F is a piecewise function obtained by comparing the pest and disease probability prediction data of the i-th crop at the k-th moment in the current monitoring period with the pest and disease probability threshold of the i-th crop, is the preset average pest and disease probability of the i-th crop in a monitoring period, is the soil iron oxide content data at the k-th moment in the current monitoring period, is the soil iron oxide content standard value, is the preset average relative difference of the soil iron oxide content data of the i-th crop in a monitoring period, M is the number of moments in the monitoring period, k ≤ M, and both i, k, and M are positive integers. The first condition C1 and the second condition C2 are the preset growth conditions; when the first condition C1 and the second condition C2 are both satisfied, it is determined that the i-th crop meets the preset growth conditions.

[0030] According to an embodiment of the present invention, in formula (1), means that when the pest and disease probability prediction data of the i-th crop at the k-th moment in the current monitoring period is less than the pest and disease probability threshold of the i-th crop, the value is 1; when the pest and disease probability prediction data of the i-th crop at the k-th moment in the current monitoring period is greater than or equal to the pest and disease probability threshold of the i-th crop, the value is 0. If the pest and disease probability prediction data is relatively large, it indicates that the pest and disease risk of the i-th crop at the k-th moment in the current monitoring period is relatively high, and further indicates that the growth conditions of this crop are poor.

[0031] According to an embodiment of the present invention, in the first condition of formula (2), means that the average value of F at all moments in the current monitoring period is greater than or equal to the preset average pest and disease probability of the i-th crop in a monitoring period. That is, when the pest and disease probability in the monitoring period is lower than the threshold, it indicates that the pest and disease risk is controllable and meets the requirements for crop growth. In the second condition, $D_k$ is the relative difference between the soil iron oxide content data at the $k$-th moment of the current monitoring period and the standard value of the soil iron oxide content. The larger this relative difference is, the greater the deviation degree of the soil iron oxide content from the standard value of the soil iron oxide content, and the greater the impact on the soil texture. For example, iron oxide can combine with other elements in the soil to form complexes, improving the soil fertility. In soils with a relatively high iron oxide content, plant roots may be restricted to a certain extent, resulting in poor root growth or uneven distribution. However, under appropriate conditions, iron oxide can also provide necessary support and fixation for plant roots, improving the soil structure and enhancing the soil's erosion resistance ability. It means that the average value of the above relative difference is less than the preset average relative difference of the soil iron oxide content data of the $i$-th type of crop in a monitoring period. That is, when the soil iron oxide content is relatively stable during the monitoring period, it meets the growth requirements of the $i$-th type of crop.

[0032] In this way, based on the soil iron oxide content data and the predicted data of the pest and disease probabilities of different crops, the first condition and the second condition can be determined. When the first condition and the second condition are both met, it can be determined that the $i$-th type of crop meets the preset growth conditions. By quantifying the predicted data of the pest and disease probabilities and the soil iron oxide content data, the accuracy of judging the preset growth conditions of the crops is improved. Monitoring the soil iron oxide content and the pest and disease probabilities and timely adjusting the planting strategies and management measures can optimize the growth environment of the crops.

[0033] According to an embodiment of the present invention, the above-mentioned pest and disease probability prediction model can be trained before use. The training steps of the pest and disease probability prediction model include: obtaining historical remote sensing images of the crop growth areas at multiple moments in multiple historical monitoring periods; obtaining the historical spectral reflectance and historical structure data of multiple types of crops at multiple moments in multiple historical monitoring periods according to the historical remote sensing images, wherein the historical structure data includes historical planting area and historical planting density; inputting the historical spectral reflectance and historical structure data into the trained pest and disease probability prediction model to obtain the historical pest and disease probability prediction data of multiple types of crops at multiple moments in multiple historical monitoring periods; obtaining the historical pest and disease probability data of multiple types of crops at multiple moments in multiple historical monitoring periods; obtaining the historical temperature data and historical humidity data at multiple moments in multiple historical monitoring periods; determining the loss function of the pest and disease probability prediction model according to the structure data, the historical structure data, the historical pest and disease probability prediction data, the historical pest and disease probability data, the historical temperature data and the historical humidity data; and training the pest and disease probability prediction model according to the loss function of the pest and disease probability prediction model to obtain the trained pest and disease probability prediction model.

[0034] According to an embodiment of the present invention, the historical monitoring period is the actual period, and the data at each moment can be actually collected. The pest and disease probability prediction model can predict the historical pest and disease probability prediction data of crops based on the historical spectral reflectance and historical structure data. The historical pest and disease probability data is based on historical monitoring records and reflects the occurrence of pests and diseases of different crops at different time points. The larger the planting area and the higher the planting density of the crops, the higher the probability of pest and disease occurrence. For example, in areas with large-scale planting, the probability of pest and disease occurrence and spread is likely to increase. Excessively high planting density will lead to poor ventilation and insufficient light among crops. Excessively high planting density of corn will increase the damage of pests such as corn borers and corn aphids. The higher the temperature data and humidity data of the environment, the higher the probability of pest and disease occurrence in crops. For example, a high-temperature and high-humidity environment is conducive to the occurrence and spread of pests and diseases. The pest and disease probability prediction model can predict the historical pest and disease probability prediction data of multiple crops at multiple moments in multiple historical monitoring periods based on the relationship between the above planting area, planting density and pest and disease probability, and based on the historical spectral reflectance and historical structure data of multiple crops at multiple moments in multiple historical monitoring periods. Determine the loss function according to the difference between the historical pest and disease probability prediction data and the historical pest and disease probability data in the historical monitoring records. Obtain the trained pest and disease probability prediction model by performing feedback adjustment on the loss function.

[0035] According to an embodiment of the present invention, determining the loss function of the pest and disease probability prediction model according to the structure data, the historical structure data, the historical pest and disease probability prediction data, the historical pest and disease probability data, the historical temperature data and the historical humidity data includes: determining the loss function of the pest and disease probability prediction model according to formula (3) , (3), where, is the planting area of the i-th crop in the current monitoring period, is the historical planting area of the i-th crop in the h-th historical monitoring period, is the planting density of the i-th crop in the current monitoring period, is the historical planting density of the i-th crop in the h-th historical monitoring period, is the historical temperature data of the i-th crop at the k-th moment in the h-th historical monitoring period, is the standard temperature data of the i-th crop, is the historical humidity data of the i-th crop at the k-th moment in the h-th historical monitoring period, is the standard humidity data of the i-th crop, is the historical pest and disease probability data of the i-th crop at the k-th moment in the h-th historical monitoring period, is the historical pest and disease probability prediction data of the i-th crop at the k-th moment of the h-th historical monitoring period. M is the number of moments in the monitoring period, H is the number of historical monitoring periods, h ≤ H, k ≤ M, and i, k, M, h, and H are all positive integers.

[0036] According to an embodiment of the present invention, in formula (3), is the error magnitude between the historical pest and disease probability data and the historical pest and disease probability prediction data of the i-th crop at the k-th moment of the h-th historical monitoring period. is the ratio of the historical temperature data of the i-th crop at the k-th moment of the h-th historical monitoring period to the standard temperature data of the i-th crop. The larger this ratio is, the larger the historical temperature data is. is the ratio of the historical humidity data of the i-th crop at the k-th moment of the h-th historical monitoring period to the standard humidity data of the i-th crop. The larger this ratio is, the larger the historical humidity data is. Indicates that the historical temperature data and the historical humidity data are positively correlated with the pest and disease probability. For example, when the historical temperature data and the historical humidity data are larger, the hot and humid environment is conducive to the occurrence and spread of pests and diseases, and the probability of pest and disease occurrence of the crop is higher. Therefore, the larger the historical temperature data and the historical humidity data are relative to the standard values, that is, and the larger the values of, the greater the impact on the error of the historical pest and disease probability prediction data. is the weight at the k-th moment, which is used to reasonably weight the relative errors at different moments in the loss function. For the s-th historical monitoring period, the accuracy of the historical pest and disease probability prediction data at the k-th moment output by the pest and disease probability prediction model is usually higher than that of the historical pest and disease probability prediction data at the k + 1-th moment. That is, the longer the time interval between a certain moment in a historical monitoring period and the 1st moment, the less accurate its prediction result. To improve the training efficiency, the higher its weight is set. Conversely, the more accurate the prediction result is, the lower its weight is. Indicates the summation after multiplying the pest and disease probability difference at each moment by the corresponding weight. is the similarity between the planting area of the i-th crop in the current monitoring period and the historical planting area of the i-th crop in the h-th historical monitoring period. is the similarity between the planting density of the i-th crop in the current monitoring period and the historical planting density of the i-th crop in the h-th historical monitoring period. To achieve a similar monitoring effect of pest and disease probability, the larger the planting area and the higher the planting density of the crop, the higher the probability of pest and disease occurrence in the crop. That is, if the structural data of the current monitoring period is closer to the structural data of the historical monitoring period, the historical pest and disease probability data is closer to the pest and disease probability prediction data of the current monitoring period, and its reference value is greater. Therefore, its weight is higher. Multiplying the above three items and taking the average can represent the loss function of the pest and disease probability prediction model.

[0037] According to an embodiment of the present invention, in the process of training the pest and disease probability prediction model, by performing backpropagation on the loss function, some parameters inside the model are adjusted to reduce the value of the loss function of the pest and disease probability prediction model, thereby improving the accuracy of the pest and disease probability prediction model and obtaining the trained pest and disease probability prediction model.

[0038] In this way, during the training process, the influence of historical temperature data and historical humidity data on the pest and disease probability can be used to determine the influence of the above data on the error of the historical pest and disease probability prediction data. Then, based on this influence and the difference between the historical pest and disease probability data and the historical pest and disease probability prediction data, and based on the characteristic that the shorter the time interval from the start time, the higher the accuracy, weights are set, and also based on the characteristic that the closer the structural data of the current monitoring period is to the structural data of the historical control period, the closer the pest and disease probability is, and the greater the reference value of the error of the historical pest and disease probability prediction data, weights are set. Thus, the errors output by the pest and disease probability prediction model at each moment of each monitoring period are weighted and summed to obtain the loss function, so as to improve the design accuracy and objectivity of the loss function, thereby improving the training efficiency during the training process and enhancing the accuracy of the pest and disease probability prediction model.

[0039] According to an embodiment of the present invention, in step S105, if multiple crops meet the preset growth conditions, then according to the growth state data, the growth state scores of the multiple crops are determined.

[0040] According to an embodiment of the present invention, step S105 includes: obtaining the standard growth state data of multiple crops, where the standard growth state data includes standard leaf area index data, standard average leaf surface temperature data, standard biomass data, and standard chlorophyll content data; determining the standard growth state vector of multiple crops according to the standard growth state data; determining the real-time growth state vector of multiple crops according to the growth state data; and determining the growth state scores of multiple crops according to the standard growth state vector and the real-time growth state vector.

[0041] According to an embodiment of the present invention, the standard growth state data of various crops need to be determined according to the growth requirements of different crops. For example, the standard leaf area index of rice is 5, and the standard leaf area index of corn is 4. Combine the respective standard growth state data into a multi-dimensional data structure, that is, the standard growth state vector, which comprehensively reflects the overall growth state of crops under ideal growth conditions. Combine the actually monitored growth state data into a multi-dimensional data structure, that is, the real-time growth state vector, which reflects the current actual growth state of the crops.

[0042] According to an embodiment of the present invention, based on the standard growth state vector and the real-time growth state vector, determine the growth state scores of various crops, including: determining the growth state score of the i-th crop according to formula (4) , (4), where, is the leaf area index data of the i-th crop at the k-th moment of the current monitoring period, is the average leaf surface temperature data of the i-th crop at the k-th moment of the current monitoring period, is the biomass data of the i-th crop at the k-th moment of the current monitoring period, is the chlorophyll content data of the i-th crop at the k-th moment of the current monitoring period, is the standard leaf area index data of the i-th crop at the k-th moment of the current monitoring period, is the standard average leaf surface temperature data of the i-th crop at the k-th moment of the current monitoring period, is the standard biomass data of the i-th crop at the k-th moment of the current monitoring period, is the standard chlorophyll content data of the i-th crop at the k-th moment of the current monitoring period, is the real-time growth state vector at the k-th moment of the current monitoring period, is the standard growth state vector at the k-th moment of the current monitoring period, M is the number of moments in the monitoring period, k ≤ M, and i, k, and M are all positive integers.

[0043] According to an embodiment of the present invention, in formula (4), It is the cosine similarity between the real-time growth state vector at the kth moment of the current monitoring cycle and the standard growth state vector at the kth moment of the current monitoring cycle. The closer the cosine similarity is to 1, the closer the real-time growth state vector at the kth moment of the current monitoring cycle is to the standard growth state vector, and the better the growth state of the i-th crop at the kth moment of the current monitoring cycle. The average value of the cosine similarity at multiple moments of the current monitoring cycle is used as the growth state score of the i-th crop. The larger the growth state score, the worse the growth state of the i-th crop, and measures need to be taken, such as improving the soil, applying pesticides, or adjusting the planting strategy.

[0044] In this way, the growth status scores of various crops can be determined based on the cosine similarity between the real-time growth status vector and the standard growth status vector, and based on different growth status data, the growth status of crops can be fully reflected, which helps to detect growth abnormalities of crops in a timely manner.

[0045] According to one embodiment of the present invention, in step S106, if the multiple crops do not meet the preset growth conditions, the growth status scores of the multiple crops are 0. For example, if corn does not meet the preset growth conditions, that is, does not meet at least one of the first condition C1 and the second condition C2, the growth status score of corn is 0.

[0046] According to an embodiment of the present invention, in step S107, a crop monitoring report is generated according to the growth status score.

[0047] According to one embodiment of the present invention, step S107 includes: if the growth status score of the i-th crop is greater than or equal to a first growth status score threshold, determining that the growth status of the i-th crop is excellent; if the growth status score of the i-th crop is less than the first growth status score threshold and greater than or equal to a second growth status score threshold, determining that the growth status of the i-th crop is good; if the growth status score of the i-th crop is less than the second growth status score threshold, determining that the growth status of the i-th crop is poor.

[0048] According to one embodiment of the present invention, if the growth status of crops is excellent, it means that the growth status of the crops is very good and reaches a higher standard. If the growth status of crops is good, it means that the growth status of the crops is acceptable and requires proper management and maintenance, such as adding organic matter or tilling to improve soil quality. If the growth status of crops is poor, it means that the growth status of the crops is not good and requires corresponding intervention or improvement measures, such as pest and disease control. By setting clear scoring thresholds and making comparisons, the growth status of crops can be effectively judged, thereby providing a scientific basis for agricultural production management and taking timely measures to optimize the growth environment of crops.

[0049] According to the crop monitoring method based on remote sensing big data of an embodiment of the present invention, through remote sensing technology, data can be efficiently acquired within a wide range of crop growth areas, saving labor and time costs. Through spectral reflectance, structural data, and growth status data, comprehensive crop information is provided. This multi-dimensional data can more accurately reflect the actual growth situation and health level of crops. Through the trained pest and disease probability prediction model, the risk of pests and diseases can be identified in real time, and control measures can be taken in a timely manner to reduce losses and improve crop yield and quality. The growth status of crops can be monitored based on the integration of pest and disease probability prediction data and multi-dimensional data, thereby improving the comprehensiveness, accuracy, and reliability of crop monitoring. When determining whether multiple crops meet the preset growth conditions, the first condition and the second condition can be determined based on the soil iron oxide content data and the pest and disease probability prediction data of different crops. When both the first condition and the second condition are met, it can be determined that the i-th crop meets the preset growth conditions. The quantification process of the pest and disease probability prediction data and the soil iron oxide content data improves the accuracy of judging the preset growth conditions of crops. Monitoring the soil iron oxide content and pest and disease probability and adjusting the planting strategy and management measures in a timely manner can optimize the crop growth environment. When determining the loss function of the pest and disease probability prediction model, during the training process, the influence of historical temperature data and historical humidity data on the pest and disease probability can be used to determine the influence of the above data on the error of the historical pest and disease probability prediction data. Then, based on this influence and the difference between the historical pest and disease probability data and the historical pest and disease probability prediction data, and based on the characteristic that the shorter the time interval from the start time, the higher the accuracy, weights are set, and the closer the structural data of the current monitoring period is to the structural data of the historical control period, the closer the pest and disease probability is, and the greater the reference value of the error of the historical pest and disease probability prediction data. Weights are set accordingly, so as to perform weighted summation on the errors output by the pest and disease probability prediction model at each moment of each monitoring period to obtain the loss function, so as to improve the design accuracy and objectivity of the loss function, thereby improving the training efficiency during the training process and enhancing the accuracy of the pest and disease probability prediction model. When determining the growth status scores of multiple crops, the growth status scores of multiple crops can be determined based on the cosine similarity between the real-time growth status vector and the standard growth status vector. Based on different growth status data, the growth status of crops can be comprehensively reflected, which helps to detect the growth anomalies of crops in a timely manner.

[0050] Figure 2The block diagram of a crop monitoring system based on remote sensing big data according to an embodiment of the present invention is exemplarily shown. The system includes: a remote sensing image module for monitoring the crop growth area through remote sensing technology to obtain high-resolution remote sensing images at multiple moments in the current monitoring period; a data module for obtaining the spectral reflectance, structural data, and growth status data of various crops according to the remote sensing images, wherein the spectral reflectance includes the spectral reflectance at multiple wavelengths, the structural data includes the planting area and planting density, and the growth status data includes leaf area index data, average leaf surface temperature data, biomass data, and chlorophyll content data; a pest and disease probability prediction data module for inputting the spectral reflectance and the structural data into a trained pest and disease probability prediction model to obtain the pest and disease probability prediction data of various crops at multiple moments in the current monitoring period; a judgment module for determining whether various crops meet the preset growth conditions according to the pest and disease probability prediction data; a determined growth status scoring module for determining the growth status score of various crops according to the growth status data if various crops meet the preset growth conditions; a growth status score of 0 module for setting the growth status score of various crops to 0 if various crops do not meet the preset growth conditions; and a crop monitoring report module for generating a crop monitoring report according to the growth status score.

[0051] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the accompanying drawings are only examples and do not limit the present invention. The object of the present invention has been fully and effectively achieved. The function and structural principle of the present invention have been shown and described in the embodiments. Without departing from the principle, any deformation or modification can be made to the embodiments of the present invention.

Claims

1. A crop monitoring method based on remote sensing big data, characterized in that: include: Monitor crop growing areas through remote sensing technology and obtain high-resolution remote sensing images at multiple times during the current monitoring period; According to the remote sensing image, the spectral reflectance, structural data and growth status data of multiple crops are obtained, wherein the spectral reflectance includes the spectral reflectance at multiple wavelengths, the structural data includes the planting area and the planting density, and the growth status data includes leaf area index data, average leaf surface temperature data, biomass data and chlorophyll content data; the spectral reflectance and the structural data are input into a trained pest and disease probability prediction model to obtain the pest and disease probability prediction data of multiple crops at multiple moments in the current monitoring period; according to the pest and disease probability prediction data, it is determined whether the multiple crops meet the preset growth conditions; if the multiple crops meet the preset growth conditions, the growth status scores of the multiple crops are determined according to the growth status data; if the multiple crops do not meet the preset growth conditions, the growth status scores of the multiple crops are 0; according to the growth status scores, a crop monitoring report is generated.

2. The crop monitoring method based on remote sensing big data according to claim 1, characterized in that: Determine whether a plurality of crops meet preset growth conditions based on the pest and disease probability prediction data, including: obtaining soil iron oxide content data at multiple moments in the current monitoring cycle; determine whether a plurality of crops meet preset growth conditions based on the soil iron oxide content data and the pest and disease probability prediction data.

3. The crop monitoring method based on remote sensing big data according to claim 2 is characterized in that: According to the soil iron oxide content data and the pest probability prediction data, determining whether a plurality of crops meet the preset growth conditions includes: according to the formula and Determine the first condition C1 and the second condition C2, where: is the probability prediction data of pests and diseases of the i-th crop at the k-th moment in the current monitoring period, is the probability threshold of pests and diseases of the i-th crop, F is the piecewise function that compares the predicted probability data of pests and diseases of the i-th crop at the k-th moment of the current monitoring period with the probability threshold of pests and diseases of the i-th crop, is the preset average probability of pests and diseases for the i-th crop in a monitoring period, is the soil iron oxide content data at the kth moment of the current monitoring period, is the standard value of soil iron oxide content, is the preset average relative difference of the soil iron oxide content data of the i-th crop in a monitoring period, M is the number of monitoring period moments, k≤M, and i, k and M are all positive integers, and the first condition C1 and the second condition C2 are the preset growth conditions; when the first condition C1 and the second condition C2 are met at the same time, it is determined that the i-th crop meets the preset growth conditions.

4. The crop monitoring method based on remote sensing big data according to claim 1, characterized in that: The training steps of the pest and disease probability prediction model include: obtaining historical remote sensing images of crop growth areas at multiple moments in multiple historical monitoring periods; obtaining historical spectral reflectance and historical structural data of multiple crops at multiple moments in multiple historical monitoring periods based on the historical remote sensing images, wherein the historical structural data includes historical planting area and historical planting density; inputting the historical spectral reflectance and historical structural data into the trained pest and disease probability prediction model to obtain historical pest and disease probability prediction data of multiple crops at multiple moments in multiple historical monitoring periods; obtaining historical pest and disease probability data of multiple crops at multiple moments in multiple historical monitoring periods; obtaining historical temperature data and historical humidity data at multiple moments in multiple historical monitoring periods; determining the loss function of the pest and disease probability prediction model based on the structural data, the historical structural data, the historical pest and disease probability prediction data, the historical pest and disease probability data, the historical temperature data and the historical humidity data; training the pest and disease probability prediction model based on the loss function of the pest and disease probability prediction model to obtain a trained pest and disease probability prediction model.

5. The crop monitoring method based on remote sensing big data according to claim 4 is characterized in that: Determine the loss function of the pest probability prediction model according to the structure data, the historical structure data, the historical pest probability prediction data, the historical pest probability data, the historical temperature data and the historical humidity data, including: according to the formula Determine the loss function of the pest probability prediction model ,in, is the planting area of ​​the i-th crop in the current monitoring period, is the historical planting area of ​​the i-th crop in the h-th historical monitoring period, is the planting density of the i-th crop in the current monitoring period, is the historical planting density of the i-th crop in the h-th historical monitoring period, is the historical temperature data of the i-th crop at the k-th moment in the h-th historical monitoring period, is the standard temperature data of the i-th crop, is the historical humidity data of the i-th crop at the k-th moment in the h-th historical monitoring period, is the standard humidity data of the i-th crop, is the historical pest and disease probability data of the i-th crop at the k-th moment in the h-th historical monitoring period, is the historical disease and insect pest probability prediction data of the i-th crop at the k-th moment of the h-th historical monitoring period, M is the number of monitoring period moments, H is the number of historical monitoring periods, h≤H, k≤M, and i, k, M, h and H are all positive integers.

6. The crop monitoring method based on remote sensing big data according to claim 1, characterized in that: If multiple crops meet preset growth conditions, growth status scores of the multiple crops are determined according to the growth status data, including: acquiring standard growth status data of the multiple crops, wherein the standard growth status data includes standard leaf area index data, standard leaf surface average temperature data, standard biomass data and standard chlorophyll content data; determining standard growth state vectors of the multiple crops according to the standard growth status data; determining real-time growth state vectors of the multiple crops according to the growth status data; and determining growth status scores of the multiple crops according to the standard growth state vector and the real-time growth state vector.

7. The crop monitoring method based on remote sensing big data according to claim 6, characterized in that: Determining the growth status scores of multiple crops according to the standard growth status vector and the real-time growth status vector includes: according to the formula Determine the growth status score of the i-th crop ,in, is the leaf area index data of the i-th crop at the k-th moment in the current monitoring period, is the average leaf temperature data of the i-th crop at the k-th moment in the current monitoring period, is the biomass data of the i-th crop at the k-th moment in the current monitoring period, is the chlorophyll content data of the i-th crop at the k-th moment in the current monitoring period, is the standard leaf area index data of the i-th crop at the k-th moment in the current monitoring period, is the standard average leaf temperature data of the i-th crop at the k-th moment in the current monitoring period, is the standard biomass data of the i-th crop at the k-th moment in the current monitoring period, is the standard chlorophyll content data of the i-th crop at the k-th moment in the current monitoring period, is the real-time growth state vector at the kth moment in the current monitoring period, is the standard growth state vector at the kth moment of the current monitoring period, M is the number of moments in the monitoring period, k≤M, and i, k and M are all positive integers.

8. The crop monitoring method based on remote sensing big data according to claim 1, characterized in that: A crop monitoring report is generated according to the growth status score, including: if the growth status score of the i-th crop is greater than or equal to a first growth status score threshold, the growth status of the i-th crop is determined to be excellent; if the growth status score of the i-th crop is less than the first growth status score threshold and greater than or equal to a second growth status score threshold, the growth status of the i-th crop is determined to be good; if the growth status score of the i-th crop is less than the second growth status score threshold, the growth status of the i-th crop is determined to be poor.

9. A crop monitoring system based on remote sensing big data, characterized in that: include: Remote sensing image module, used to monitor crop growing areas through remote sensing technology and obtain high-resolution remote sensing images at multiple moments in the current monitoring cycle; A data module, used to obtain the spectral reflectance, structural data and growth status data of multiple crops according to the remote sensing image, wherein the spectral reflectance includes the spectral reflectance at multiple wavelengths, the structural data includes the planting area and the planting density, and the growth status data includes the leaf area index data, the average leaf temperature data, the biomass data and the chlorophyll content data; a pest and disease probability prediction data module, used to input the spectral reflectance and the structural data into the trained pest and disease probability prediction model to obtain the pest and disease probability prediction data of multiple crops at multiple moments in the current monitoring period; a judgment module, used to determine whether multiple crops meet the preset growth conditions according to the pest and disease probability prediction data; a growth status scoring module, used to determine the growth status scores of multiple crops according to the growth status data if the multiple crops meet the preset growth conditions; a growth status scoring 0 module, used to score the growth status of multiple crops as 0 if the multiple crops do not meet the preset growth conditions; a crop monitoring report module, used to generate a crop monitoring report according to the growth status score.

Images