A cloud computing-based intelligent agricultural planting control method and system

Abstract

The application belongs to the technical field of agricultural planting, and provides a smart agricultural planting control method and system based on cloud computing, which comprises the following steps: extracting actual planting result data under historical cloud planting instructions, dividing a field into a plurality of management units according to the same planting area, analyzing the fruit malformation rate of each management unit under the current cloud planting instruction, and evaluating whether cloud algorithm reconstruction is needed; after determining that cloud algorithm reconstruction is needed, extracting environment parameters related to crop growth, analyzing and determining the correlation degree between the environment parameters and the fruit malformation rate, and identifying core correlation parameters that have a great influence on the fruit malformation rate of crops; the application divides homogeneous regions through space-time clustering, customizes special planting instruction models and environment suitable intervals for different regions, solves the problem that traditional global unified instructions cannot adapt to the soil and environment differences in the field, and can effectively reduce the fruit malformation rate.

Description

Technical Field

[0001] This invention belongs to the field of agricultural planting technology, specifically a smart agricultural planting control method and system based on cloud computing. Background Technology

[0002] With the deepening of agricultural digital transformation, cloud-based smart agricultural planting systems have gradually replaced traditional experience-based planting, becoming a core technological means to activate agricultural commercial value, improve agricultural production efficiency, and reduce resource consumption. Existing cloud-based planting systems are typically based on globally unified planting algorithms, issuing standardized irrigation, fertilization, and temperature and humidity control instructions to the entire field to achieve large-scale planting management. In large-scale commercial planting scenarios, this has improved the standardization of planting to a certain extent.

[0003] However, significant spatial heterogeneity exists within actual planting fields: soil water and fertilizer retention capacity, topographic relief, and local microclimates vary naturally in different regions. A globally unified planting instruction cannot adapt to the actual needs of each region, leading to over-irrigation in some areas and insufficient water and fertilizer in others. This results in growth imbalances, increased fruit deformity rates, and other problems, severely impacting crop yield and quality. According to statistics from the Ministry of Agriculture and Rural Affairs, under the current unified management model, the average fruit deformity rate of greenhouse fruits and vegetables remains high, and the improvement of high-quality fruit rate has encountered bottlenecks, directly restricting the profitability of commercial planting.

[0004] Meanwhile, existing cloud-based planting algorithms mostly have fixed update cycles, which cannot adapt to the dynamic changes in the field environment: for areas with drastic environmental fluctuations, the fixed update cycle is too long, causing the algorithm to quickly fail and fail to adapt to environmental changes in a timely manner; for areas with stable environments, the fixed update cycle is too short, causing unnecessary waste of computing resources, failing to balance algorithm accuracy and operating efficiency, and also increasing the computing cost for commercial deployment.

[0005] The aforementioned shortcomings result in poor fruit deformity control, low resource utilization, and poor adaptability of existing smart agriculture planting control methods, failing to meet the demands of high-quality commercial agricultural planting. Therefore, this invention provides a cloud-based smart agriculture planting control method and system. Summary of the Invention

[0006] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0007] In a first aspect, the present invention provides a cloud computing-based smart agriculture planting control method, comprising: Extract the actual planting results data under historical cloud planting instructions, divide the fields into several management units with the same planting area, analyze the fruit deformity rate of each management unit under the current cloud planting instructions, and assess whether cloud algorithm reconstruction is needed. After determining that cloud-based algorithm reconstruction is necessary, environmental parameters related to crop growth are extracted, the correlation between environmental parameters and fruit deformity rate is analyzed and determined, and core correlation parameters that have a significant impact on crop fruit deformity rate are identified. By quantifying the environmental fluctuation frequency of each management unit and combining the benchmark parameters with the deviation of the fruit deformity rate, the reconstruction cycle of the cloud algorithm is determined. The reconstruction cycle is the update interval of the cloud algorithm. Based on the reconstruction cycle, a spatiotemporal clustering algorithm is used to divide the planting plots into several homogeneous zones according to environmental characteristic data, determine the suitable range of core correlation parameters of the homogeneous zones, and construct a dynamic planting instruction model for the homogeneous zones.

[0008] As a further aspect of the present invention: the process of evaluating whether cloud-based algorithm reconstruction is necessary is as follows: Extract historical cloud-based planting instructions from the cloud platform logs, including irrigation duration, fertilizer application amount, temperature, and humidity of the planted crops; AI visual recognition technology is used to automatically identify deformed and normal fruits. The number of deformed fruits harvested by each management unit under historical cloud planting instructions is counted against the total number of fruits. The ratio of the number of deformed fruits harvested by the management unit to the total number of fruits is calculated to obtain the fruit deformity rate of the management unit. By analyzing the fruit deformity rate of each management unit under the current cloud planting instructions, the coefficient of variation of the fruit deformity rate is determined. If the coefficient of variation of the fruit deformity rate is greater than or equal to the coefficient of variation threshold, then cloud-based algorithm reconstruction is required; otherwise, cloud-based algorithm reconstruction is not required.

[0009] As a further aspect of the present invention: the process for determining the coefficient of variation of the fruit deformity rate is as follows: The fruit deformity rates of all management units are integrated into a fruit deformity rate sequence. All data in the fruit deformity rate sequence are extracted and processed into an arithmetic mean to obtain the mean fruit deformity rate. The standard deviation formula is used to calculate the standard deviation of all data in the fruit deformity rate sequence to obtain the standard deviation of the fruit deformity rate. The coefficient of variation formula is used to calculate the ratio of the standard deviation of the fruit deformity rate to the mean fruit deformity rate to obtain the coefficient of variation of the fruit deformity rate.

[0010] As a further aspect of the present invention: the identification process of the core associated parameters is as follows: Environmental parameters include air temperature, air humidity, soil moisture, light intensity, and wind speed; By using sensors deployed in each management unit, historical environmental parameter data of crop growth period are collected. Pearson correlation analysis method is used to match the environmental parameter data of each management unit with the fruit deformity rate data to construct a correlation dataset between environmental parameters and fruit deformity rate. For each type of environmental parameter, the Pearson correlation coefficient between the environmental parameter and the fruit deformity rate is calculated and denoted as the environmental correlation value. If the environmental correlation value is greater than or equal to the environmental correlation threshold, the corresponding environmental parameter is recorded as the core correlation parameter.

[0011] As a further aspect of the present invention: the process for determining the reconstruction cycle of the cloud algorithm is as follows: Screen the core correlation parameters that have a significant impact on crop fruit deformity rate, and for each management unit, identify the effective fluctuations of the core correlation parameters based on the hourly granular sensor historical data of the management unit; Analyze and process the effective fluctuations of the core correlation parameters to determine the basic periodic value; If the base period value is less than the shortest reconstruction period, then the shortest reconstruction period will be used as the reconstruction period for the cloud algorithm. If the base period value is greater than the longest reconstruction period, then the longest reconstruction period will be used as the reconstruction period for the cloud algorithm. If the base period value is between the shortest and longest reconstruction period, then the base period value is used as the reconstruction period of the cloud algorithm.

[0012] As a further aspect of the present invention: the process for identifying the effective fluctuations of the core correlation parameter is as follows: For each hourly node, calculate the absolute value of the difference between the current hour and the previous hour's core related parameter monitoring value to obtain the change range of the core related parameter in the current hour; If the change in the core correlation parameter in the current hour is greater than or equal to the change threshold, then the fluctuation of the core correlation parameter in that hour is considered large and is recorded as a valid fluctuation of the core correlation parameter.

[0013] As a further aspect of the present invention: the process for determining the basic periodic value is as follows: For each management unit, the number of valid fluctuations of all core related parameters within a preset historical period is counted, the percentage of hours in which at least one valid fluctuation of a core parameter occurred is calculated, and then the ratio is calculated with the total number of hours in the preset historical period to obtain the fluctuation frequency of the management unit. If the environmental fluctuation frequency of a management unit is greater than or equal to the reference fluctuation frequency, the corresponding management unit is recorded as a large fluctuation unit. Iterate through all the environmental fluctuation frequencies corresponding to the large fluctuation units, and select the maximum value among them as the maximum fluctuation frequency; The environmental fluctuation adaptation coefficient is obtained by comparing the benchmark fluctuation frequency with the maximum fluctuation frequency. The algorithm adaptation deviation coefficient is obtained by comparing the coefficient of variation threshold with the coefficient of variation of the current fruit deformity rate. The preset benchmark reconstruction period is multiplied by the environmental fluctuation adaptation coefficient, and then multiplied by the algorithm adaptation deviation coefficient. Finally, the product is multiplied by the reproductive option weight to obtain the basic period value.

[0014] As a further aspect of the present invention: the process of dividing the homogeneous region is as follows: The core correlation parameter monitoring data of each management unit at the hourly granularity are used as environmental characteristic data. Among them, the NDVI growth index is collected by multispectral UAV inspection, and the plant height and fruit setting rate are collected by field inspection humanoid robots. The GIS coordinates of each management unit in the planting field are used as spatiotemporal location characteristic data. The spatiotemporal location data is used to characterize the spatial location relationship of the management units and avoid management units that are too far apart in space from being clustered into the same homogeneous area. For each management unit, environmental feature data and spatiotemporal location feature data are integrated to construct a multidimensional spatiotemporal feature vector. The K-means++ algorithm is used to select initial cluster centers. According to the set distance metric, the spatiotemporal distance between the feature vector of each management unit and each cluster center is calculated, and each management unit is assigned to the nearest cluster. Then, the average value of the feature vectors of all management units in the cluster is taken to update the center vector of each cluster. The above assignment and update process is repeated until the convergence condition is met. Each cluster after clustering is defined as a homogeneous region. The list of management units contained in each homogeneous region, the GIS boundary range of the homogeneous region, and the environmental characteristic data of each homogeneous region are output to complete the division of homogeneous regions.

[0015] As a further aspect of the present invention: the construction process of the dynamic planting instruction model is as follows: For each homogeneous zone, an independent dynamic planting instruction model is constructed. Based on any target homogeneous zone, the core correlation parameter monitoring data, fruit deformity rate data, and planting instruction execution effect data within the most recent reconstruction cycle are extracted. Multi-source data alignment is completed by hourly timestamps and used as training datasets. The training datasets are sorted from smallest to largest according to the final fruit deformity rate of the management unit, and the top 20% of samples with the lowest fruit deformity rate are selected as the optimal working condition sample set. Based on the optimal working condition sample set obtained through screening, the maximum value of the core correlation parameter in the optimal working condition sample set is extracted as the upper limit and the minimum value is extracted as the lower limit to obtain the suitable range of the core correlation parameter in the homogeneous area. Based on the suitable range of the core correlation parameter, the mapping relationship between the core correlation parameter and the optimal instruction adjustment amount of all management units in the homogeneous area is fitted by univariate linear regression to construct a dynamic planting instruction model.

[0016] Secondly, the present invention also provides a cloud computing-based smart agricultural planting control system, the system comprising: Reconstruction evaluation module: Extract actual planting result data under historical cloud planting instructions, divide the field into several management units according to the same planting area, analyze the fruit deformity rate of each management unit under the current cloud planting instructions, and evaluate whether cloud algorithm reconstruction is needed; Core correlation parameter identification module: After determining that cloud algorithm reconstruction is required, extract environmental parameters related to crop growth, analyze and determine the degree of correlation between environmental parameters and fruit deformity rate, and identify core correlation parameters that have a significant impact on crop fruit deformity rate; Reconstruction cycle determination module: By quantifying the environmental fluctuation frequency of each management unit and combining the baseline parameters with the deviation of the fruit deformity rate, the reconstruction cycle of the cloud algorithm is determined; Dynamic planting instruction model construction module: Based on the reconstruction cycle, a spatiotemporal clustering algorithm is used to divide the planting field into several homogeneous areas according to environmental characteristic data, determine the appropriate range of core correlation parameters of the homogeneous areas, and construct a dynamic planting instruction model for the homogeneous areas.

[0017] The beneficial effects of this invention are as follows: This invention divides homogeneous regions into spatiotemporal clusters and customizes exclusive planting instruction models and suitable environmental ranges for different regions, solving the problem that traditional globally unified instructions cannot adapt to the differences in soil and environment within the field, and can effectively reduce the fruit deformity rate; it dynamically adapts to time changes, balancing efficiency and accuracy: by dynamically calculating the reconstruction cycle, the algorithm updates more frequently in regions with rapid environmental changes and more slowly in regions with stable environments, ensuring the accuracy of the algorithm while avoiding unnecessary waste of computing resources; through improvements such as correcting statistical logic, supplementing nonlinear correlation analysis, and adaptive thresholds, the logical loopholes of the original solution are solved, and the solution can be adapted to planting scenarios of different crops and different fields, greatly improving robustness and versatility. Attached Figure Description

[0018] The invention will now be further described with reference to the accompanying drawings.

[0019] Figure 1 This is a flowchart illustrating the steps of a cloud-based smart agriculture planting control method according to an embodiment of the present invention. Figure 2 This is a system block diagram of a cloud computing-based smart agricultural planting control system according to an embodiment of the present invention. Detailed Implementation

[0020] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0021] Example 1 Please see Figure 1 As shown in the figure, a cloud computing-based smart agriculture planting control method according to an embodiment of the present invention includes the following steps: Step 1: Extract the actual planting results data under historical cloud planting instructions, divide the field into several management units with the same planting area, analyze the fruit deformity rate of each management unit under the current cloud planting instructions, and assess whether cloud algorithm reconstruction is needed. Extract historical cloud-based planting instructions from the cloud platform logs, including but not limited to irrigation duration, fertilizer application amount, temperature and humidity of the planted crops; AI visual recognition technology is used to automatically identify deformed and normal fruits. The number of deformed fruits harvested by each management unit under historical cloud planting instructions is counted against the total number of fruits. The ratio of the number of deformed fruits harvested by the management unit to the total number of fruits is calculated to obtain the fruit deformity rate of the management unit. The fruit deformity rates of all management units are integrated into a fruit deformity rate sequence. All data in the fruit deformity rate sequence are extracted and processed into an arithmetic mean to obtain the mean fruit deformity rate. The standard deviation formula is used to calculate the standard deviation of all data in the fruit deformity rate sequence to obtain the standard deviation of the fruit deformity rate. The coefficient of variation formula is used to calculate the ratio of the standard deviation of the fruit deformity rate to the mean fruit deformity rate to obtain the coefficient of variation of the fruit deformity rate. If the coefficient of variation of the fruit deformity rate is greater than or equal to the coefficient of variation threshold, it indicates that cloud algorithm reconstruction is required. If the coefficient of variation of the fruit deformity rate is less than the coefficient of variation threshold, it means that cloud algorithm reconstruction is not required. It should be noted that the reason for determining whether to reconstruct the cloud algorithm based on the coefficient of variation of the fruit deformity rate is that if the cloud algorithm is accurate, the fruit deformity rate of all management units should be very similar under the historical cloud planting instructions; if the fruit deformity rate of each management unit under the cloud planting instructions has a large dispersion, the environment of each management unit is different, and the current cloud planting instructions are no longer applicable to the crop planted in the current field. Step 2: After determining that cloud algorithm reconstruction is required, extract environmental parameters related to crop growth, analyze and determine the degree of correlation between environmental parameters and fruit deformity rate, and identify the core correlation parameters that have a significant impact on crop fruit deformity rate. Environmental parameters include, but are not limited to, air temperature, air humidity, soil moisture, light intensity, and wind speed; By using sensors deployed in each management unit, historical environmental parameter data of crop growth period are collected. Pearson correlation analysis method is used to match the environmental parameter data of each management unit with the fruit deformity rate data to construct a correlation dataset between environmental parameters and fruit deformity rate. For each type of environmental parameter, the Pearson correlation coefficient between the environmental parameter and the fruit deformity rate is calculated and denoted as the environmental correlation value. If the environmental correlation value is greater than or equal to the environmental correlation threshold, the corresponding environmental parameter is recorded as the core correlation parameter. If the environmental correlation value is less than the environmental correlation threshold, the corresponding environmental parameter is recorded as a non-core correlation parameter. Step 3: By quantifying the environmental fluctuation frequency of each management unit and combining the baseline parameters with the deviation of the fruit deformity rate, the reconstruction cycle of the cloud algorithm is determined. The reconstruction cycle is the update interval of the cloud algorithm. The core correlation parameters that have a significant impact on crop fruit deformity rate are screened. For each management unit, based on the hourly granular sensor historical data of the management unit, the absolute value of the difference between the monitoring value of the core correlation parameter in the current hour and the previous hour is calculated for each hourly node to obtain the change range of the core correlation parameter in the current hour. If the change in the core correlation parameter in the current hour is greater than or equal to the change threshold, then the fluctuation of the core correlation parameter in that hour is determined to be large, and it is recorded as the effective fluctuation of the core correlation parameter. If the change in the core correlation parameter in the current hour is less than the change threshold, then the fluctuation of the core correlation parameter in that hour is determined to be small and recorded as an invalid fluctuation of the core correlation parameter. For example, taking the core related parameter air temperature as an example: extract the hourly granular air temperature sensor historical data of the management unit, and record them in chronological order as T1, T2, T3, ..., T Where T1 is the air temperature monitoring value for the first hour, T2 is the air temperature monitoring value for the second hour, and so on, with the time interval between two adjacent data points strictly one hour; for each hourly node, the absolute value of the difference between the current hour and the previous hour's air temperature monitoring value is calculated, which is the air temperature change amplitude ΔT in hour t (t≥2). =|T -T -1 |; For each management unit, the number of valid fluctuations of all core related parameters within a preset historical period is counted, the percentage of hours in which at least one valid fluctuation of a core parameter occurred is calculated, and then the ratio is calculated with the total number of hours in the preset historical period to obtain the fluctuation frequency of the management unit. If the environmental fluctuation frequency of a management unit is greater than or equal to the reference fluctuation frequency, the corresponding management unit is recorded as a large fluctuation unit. If the environmental fluctuation frequency of a management unit is less than the reference fluctuation frequency, the corresponding management unit is recorded as a small fluctuation unit. It should be noted that the reference fluctuation frequency was set by those skilled in the art based on historical experience; Iterate through all the environmental fluctuation frequencies corresponding to the large fluctuation units, and select the maximum value among them as the maximum fluctuation frequency; It should be noted that the environmental fluctuation speed of the management unit corresponding to the maximum fluctuation frequency is the fastest, which directly determines the failure speed of the cloud algorithm. Therefore, the maximum fluctuation frequency is used as the core basis for the reconstruction cycle calculation. The environmental fluctuation adaptation coefficient is obtained by comparing the benchmark fluctuation frequency with the maximum fluctuation frequency. The algorithm adaptation deviation coefficient is obtained by comparing the coefficient of variation threshold with the coefficient of variation of the current fruit deformity rate. The preset benchmark reconstruction cycle is multiplied by the environmental fluctuation adaptation coefficient, and then multiplied by the algorithm adaptation deviation coefficient. Finally, the result is multiplied by the reproductive option weight to obtain the basic cycle value. It should be noted that the fertility option weight is adaptively adjusted based on the environmental sensitivity of the crop at its current growth stage. Crops are extremely sensitive to fluctuations in parameters such as temperature, humidity, water, and fertilizer. Even short-term deviations from optimal environmental values ​​can lead to poor pollination, abnormal flower bud differentiation, and deformed fruit development, and most of these effects are irreversible. Setting a fertility option weight of less than 1 shortens the reconstruction cycle, allowing the algorithm to update more frequently and ensuring that planting instructions can quickly adapt to environmental changes, thus avoiding irreversible quality damage. The baseline reconstruction cycle is a pre-determined general baseline parameter based on field test data, applicable to the calculation of the reconstruction cycle of cloud-based agricultural planting algorithms. Its preset basis is that when the field environment is completely stable, this value is the longest update cycle of the cloud algorithm. Even if there is no significant change in the field environment, updating the algorithm based on the baseline reconstruction cycle can effectively offset the impact of sensor drift, slow seasonal changes, and other factors on the accuracy of the algorithm, ensuring that the algorithm maintains its adaptability in the long term. If the base period value is less than the shortest reconstruction period, then the shortest reconstruction period will be used as the reconstruction period for the cloud algorithm. If the base period value is greater than the longest reconstruction period, then the longest reconstruction period will be used as the reconstruction period for the cloud algorithm. If the base period value is between the shortest and longest reconstruction period, then the base period value is used as the reconstruction period of the cloud algorithm. It should be noted that the shortest and longest reconstruction cycles are set by those skilled in the art based on historical experience. The reconstruction cycle can dynamically adapt to the environmental fluctuation characteristics of each management unit in the field, while taking into account the sensitivity of crop growth period and the accuracy requirements of the algorithm, so as to realize the dynamic update of the cloud algorithm, ensure that the cloud planting instructions are always adapted to the actual planting environment in the field, and reduce the crop fruit deformity rate. Step 4: Based on the reconstruction cycle, a spatiotemporal clustering algorithm is used to divide the planting plots into several homogeneous zones according to environmental characteristic data, determine the suitable range of core correlation parameters of the homogeneous zones, and construct a dynamic planting instruction model for the homogeneous zones. The core correlation parameter monitoring data of each management unit at the hourly granularity are used as environmental characteristic data. Among them, the NDVI growth index is collected by multispectral UAV inspection, and the plant height and fruit setting rate are collected by field inspection humanoid robots. The GIS coordinates of each management unit in the planting field are used as spatiotemporal location characteristic data. The spatiotemporal location data is used to characterize the spatial location relationship of the management units and avoid management units that are too far apart in space from being clustered into the same homogeneous area. For each management unit, environmental feature data and spatiotemporal location feature data are integrated to construct a multidimensional spatiotemporal feature vector. The K-means++ algorithm is used to select initial cluster centers. According to the set distance metric, the spatiotemporal distance between the feature vector of each management unit and each cluster center is calculated, and each management unit is assigned to the nearest cluster. Then, the average value of the feature vectors of all management units in the cluster is taken to update the center vector of each cluster. The above assignment and update process is repeated until the convergence condition is met. The process of calculating the spatiotemporal distance between the feature vector of each management unit and each cluster center is as follows: Spatial and temporal weights are defined, where the spatial weight represents the influence of the spatial location of the management unit, and the temporal weight represents the influence of the temporal evolution of environmental parameters within the reconstruction cycle. A weighted Euclidean distance is used, expressed by the formula: Calculate the spatiotemporal distance between the i-th management unit and the j-th management unit. In the formula, This represents the spatial weight, with a value of 0.3, used to characterize the influence of the spatial location relationship of management units on the clustering results. The time weight, with a value of 0.7, is used to characterize the influence of the temporal evolution of environmental parameters on the clustering results within the reconstruction period. It is the spatial location feature data of the i-th management unit. It is the spatial location feature of the j-th management unit. Let k be the time series feature parameter of the environmental feature data of the i-th management unit. Let m be the k-th time series feature parameter of the environmental feature data of the j-th management unit, and m be the total number of dimensions of the multidimensional spatiotemporal feature vector of the management unit. Each cluster after clustering is defined as a homogeneous region. The list of management units contained in each homogeneous region, the GIS boundary range of the homogeneous region, and the environmental characteristic data of each homogeneous region are output to complete the division of homogeneous regions. The construction process of the dynamic planting instruction model is as follows: For each homogeneous zone, an independent dynamic planting instruction model is constructed. Based on any target homogeneous zone, the core correlation parameter monitoring data, fruit deformity rate data, and planting instruction execution effect data within the most recent reconstruction cycle are extracted. Multi-source data alignment is completed by hourly timestamps and used as training datasets. The training datasets are sorted from smallest to largest according to the final fruit deformity rate of the management unit, and the top 20% of samples with the lowest fruit deformity rate are selected as the optimal working condition sample set. It should be noted that the purpose of selecting the top 20% of samples with the lowest fruit deformity rate as the optimal working condition sample set is to: remove abnormal working condition samples with a high fruit deformity rate, retain only the state samples with a low fruit deformity rate in the homogeneous area, and use this as the basis to extract the optimal planting parameters for the area, so as to avoid abnormal samples interfering with the accuracy of the model. Based on the optimal working condition sample set obtained through screening, the maximum value of the core correlation parameter in the optimal working condition sample set is extracted as the upper limit and the minimum value is extracted as the lower limit to obtain the suitable range of the core correlation parameter in the homogeneous area. Based on the suitable range of the core correlation parameter, the mapping relationship between the core correlation parameter and the optimal instruction adjustment amount of all management units in the homogeneous area is fitted by univariate linear regression to construct a dynamic planting instruction model. It should be noted that soil water retention, evaporation, and environmental fluctuation characteristics vary significantly across different homogeneous zones, resulting in inherent differences in the optimal environmental range for crop growth. The suitable range of the core correlation parameters can accurately match the optimal growth state of this zone. The cloud algorithm uses the suitable range of the core correlation parameters as the core control target. When the environmental parameters of this zone are within the suitable range of the core correlation parameters, the current planting instructions remain unchanged. When the environmental parameters exceed the range, corresponding adjustments such as ventilation, irrigation, and supplemental heating are automatically triggered to ensure that the environment of this zone is always in the optimal state with a low fruit deformity rate, thereby minimizing the impact of environmental heterogeneity on crop growth. The technical solution of this invention is as follows: Extract actual planting result data under historical cloud-based planting instructions; divide the field into several management units with the same planting area; analyze the fruit deformity rate of each management unit under the current cloud-based planting instructions; assess whether cloud-based algorithm reconstruction is necessary; after determining that cloud-based algorithm reconstruction is necessary, extract environmental parameters related to crop growth; analyze and determine the correlation between environmental parameters and fruit deformity rate; identify core correlation parameters that have a significant impact on crop fruit deformity rate; determine the reconstruction cycle of the cloud-based algorithm by quantifying the environmental fluctuation frequency of each management unit and combining the deviation between the baseline parameters and the fruit deformity rate; based on the reconstruction cycle, use a spatiotemporal clustering algorithm to divide the planting field into several homogeneous regions according to environmental characteristic data, and determine the core correlation of the homogeneous regions. By constructing a dynamic planting instruction model for homogeneous areas within a suitable parameter range, this invention divides homogeneous areas using spatiotemporal clustering and customizes exclusive planting instruction models and suitable environmental ranges for different regions. This solves the problem that traditional globally unified instructions cannot adapt to differences in soil and environment within a field, effectively reducing fruit deformity rates. Dynamic adaptation to time changes balances efficiency and accuracy: by dynamically calculating the reconstruction cycle, the algorithm updates more frequently in areas with rapid environmental changes and more slowly in areas with stable environments, ensuring algorithm accuracy while avoiding unnecessary waste of computational resources. Improvements such as correcting statistical logic, supplementing nonlinear correlation analysis, and adaptive thresholds address logical flaws in the original solution, making the solution adaptable to different crops and different field planting scenarios, significantly improving robustness and versatility.

[0022] Example 2 Based on the same inventive concept as the cloud computing-based smart agriculture planting control method in the foregoing embodiments, such as... Figure 2 As shown, this application provides a cloud computing-based smart agricultural planting control system, wherein the system specifically includes: Reconstruction evaluation module: Extract actual planting result data under historical cloud planting instructions, divide the field into several management units according to the same planting area, analyze the fruit deformity rate of each management unit under the current cloud planting instructions, and evaluate whether cloud algorithm reconstruction is needed; Core correlation parameter identification module: After determining that cloud algorithm reconstruction is required, extract environmental parameters related to crop growth, analyze and determine the degree of correlation between environmental parameters and fruit deformity rate, and identify core correlation parameters that have a significant impact on crop fruit deformity rate; Reconstruction cycle determination module: By quantifying the environmental fluctuation frequency of each management unit and combining the baseline parameters with the deviation of the fruit deformity rate, the reconstruction cycle of the cloud algorithm is determined; Dynamic planting instruction model construction module: Based on the reconstruction cycle, a spatiotemporal clustering algorithm is used to divide the planting plots into several homogeneous zones according to environmental characteristic data, determine the appropriate range of core correlation parameters of the homogeneous zones, and construct a dynamic planting instruction model for the homogeneous zones.

[0023] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A cloud computing-based smart agriculture planting control method, characterized in that: include: Extract the actual planting results data under historical cloud planting instructions, divide the fields into several management units with the same planting area, analyze the fruit deformity rate of each management unit under the current cloud planting instructions, and assess whether cloud algorithm reconstruction is needed. After determining that cloud-based algorithm reconstruction is necessary, environmental parameters related to crop growth are extracted, the correlation between environmental parameters and fruit deformity rate is analyzed and determined, and core correlation parameters that have a significant impact on crop fruit deformity rate are identified. By quantifying the environmental fluctuation frequency of each management unit and combining the benchmark parameters with the deviation of the fruit deformity rate, the reconstruction cycle of the cloud algorithm is determined. The reconstruction cycle is the update interval of the cloud algorithm. Based on the reconstruction cycle, a spatiotemporal clustering algorithm is used to divide the planting plots into several homogeneous zones according to environmental characteristic data, determine the suitable range of core correlation parameters of the homogeneous zones, and construct a dynamic planting instruction model for the homogeneous zones.

2. The smart agricultural planting control method based on cloud computing according to claim 1, characterized in that: The process of assessing whether cloud-based algorithm reconstruction is necessary is as follows: Extract historical cloud-based planting instructions from the cloud platform logs, including irrigation duration, fertilizer application amount, temperature, and humidity of the planted crops; AI visual recognition technology is used to automatically identify deformed and normal fruits. The number of deformed fruits harvested by each management unit under historical cloud planting instructions is counted against the total number of fruits. The ratio of the number of deformed fruits harvested by the management unit to the total number of fruits is calculated to obtain the fruit deformity rate of the management unit. By analyzing the fruit deformity rate of each management unit under the current cloud planting instructions, the coefficient of variation of the fruit deformity rate is determined. If the coefficient of variation of the fruit deformity rate is greater than or equal to the coefficient of variation threshold, then cloud-based algorithm reconstruction is required; otherwise, cloud-based algorithm reconstruction is not required.

3. The smart agricultural planting control method based on cloud computing according to claim 2, characterized in that: The process for determining the coefficient of variation of the fruit deformity rate is as follows: The fruit deformity rates of all management units are integrated into a fruit deformity rate sequence. All data in the fruit deformity rate sequence are extracted and processed into an arithmetic mean to obtain the mean fruit deformity rate. The standard deviation formula is used to calculate the standard deviation of all data in the fruit deformity rate sequence to obtain the standard deviation of the fruit deformity rate. The coefficient of variation formula is used to calculate the ratio of the standard deviation of the fruit deformity rate to the mean fruit deformity rate to obtain the coefficient of variation of the fruit deformity rate.

4. The cloud computing-based smart agriculture planting control method according to claim 3, characterized in that: The identification process for the core associated parameters is as follows: Environmental parameters include air temperature, air humidity, soil moisture, light intensity, and wind speed; By using sensors deployed in each management unit, historical environmental parameter data of crop growth period are collected. Pearson correlation analysis method is used to match the environmental parameter data of each management unit with the fruit deformity rate data to construct a correlation dataset between environmental parameters and fruit deformity rate. For each type of environmental parameter, the Pearson correlation coefficient between the environmental parameter and the fruit deformity rate is calculated and denoted as the environmental correlation value. If the environmental correlation value is greater than or equal to the environmental correlation threshold, the corresponding environmental parameter is recorded as the core correlation parameter.

5. The cloud computing-based smart agriculture planting control method according to claim 4, characterized in that: The process for determining the reconstruction cycle of the cloud algorithm is as follows: Screen the core correlation parameters that have a significant impact on crop fruit deformity rate, and for each management unit, identify the effective fluctuations of the core correlation parameters based on the hourly granular sensor historical data of the management unit; Analyze and process the effective fluctuations of the core correlation parameters to determine the basic periodic value; If the base period value is less than the shortest reconstruction period, then the shortest reconstruction period will be used as the reconstruction period for the cloud algorithm. If the base period value is greater than the longest reconstruction period, then the longest reconstruction period will be used as the reconstruction period for the cloud algorithm. If the base period value is between the shortest and longest reconstruction period, then the base period value is used as the reconstruction period of the cloud algorithm.

6. The cloud computing-based smart agriculture planting control method according to claim 5, characterized in that: The process for identifying valid fluctuations in the core correlation parameters is as follows: For each hourly node, calculate the absolute value of the difference between the current hour and the previous hour's core related parameter monitoring value to obtain the change range of the core related parameter in the current hour; If the change in the core correlation parameter in the current hour is greater than or equal to the change threshold, then the fluctuation of the core correlation parameter in that hour is considered large and is recorded as a valid fluctuation of the core correlation parameter.

7. The cloud computing-based smart agriculture planting control method according to claim 6, characterized in that: The process for determining the basic period value is as follows: For each management unit, the number of valid fluctuations of all core related parameters within a preset historical period is counted, the percentage of hours in which at least one valid fluctuation of a core parameter occurred is calculated, and then the ratio is calculated with the total number of hours in the preset historical period to obtain the fluctuation frequency of the management unit. If the environmental fluctuation frequency of a management unit is greater than or equal to the reference fluctuation frequency, the corresponding management unit is recorded as a large fluctuation unit. Iterate through all the environmental fluctuation frequencies corresponding to the large fluctuation units, and select the maximum value among them as the maximum fluctuation frequency; The environmental fluctuation adaptation coefficient is obtained by comparing the benchmark fluctuation frequency with the maximum fluctuation frequency. The algorithm adaptation deviation coefficient is obtained by comparing the coefficient of variation threshold with the coefficient of variation of the current fruit deformity rate. The preset benchmark reconstruction period is multiplied by the environmental fluctuation adaptation coefficient, and then multiplied by the algorithm adaptation deviation coefficient. Finally, the product is multiplied by the reproductive option weight to obtain the basic period value.

8. The cloud computing-based smart agriculture planting control method according to claim 7, characterized in that: The process of dividing the homogeneous region is as follows: The core correlation parameter monitoring data of each management unit at the hourly granularity are used as environmental characteristic data. Among them, the NDVI growth index is collected by multispectral UAV inspection, and the plant height and fruit setting rate are collected by field inspection humanoid robots. The GIS coordinates of each management unit in the planting field are used as spatiotemporal location characteristic data. The spatiotemporal location data is used to characterize the spatial location relationship of the management units and avoid management units that are too far apart in space from being clustered into the same homogeneous area. For each management unit, environmental feature data and spatiotemporal location feature data are integrated to construct a multidimensional spatiotemporal feature vector. The K-means++ algorithm is used to select initial cluster centers. According to the set distance metric, the spatiotemporal distance between the feature vector of each management unit and each cluster center is calculated, and each management unit is assigned to the nearest cluster. Then, the average value of the feature vectors of all management units in the cluster is taken to update the center vector of each cluster. The above assignment and update process is repeated until the convergence condition is met. Each cluster after clustering is defined as a homogeneous region. The list of management units contained in each homogeneous region, the GIS boundary range of the homogeneous region, and the environmental characteristic data of each homogeneous region are output to complete the division of homogeneous regions.

9. The cloud computing-based smart agriculture planting control method according to claim 8, characterized in that: The construction process of the dynamic planting instruction model is as follows: For each homogeneous zone, an independent dynamic planting instruction model is constructed. Based on any target homogeneous zone, the core correlation parameter monitoring data, fruit deformity rate data, and planting instruction execution effect data within the most recent reconstruction cycle are extracted. Multi-source data alignment is completed by hourly timestamps and used as training datasets. The training datasets are sorted from smallest to largest according to the final fruit deformity rate of the management unit, and the top 20% of samples with the lowest fruit deformity rate are selected as the optimal working condition sample set. Based on the optimal working condition sample set obtained through screening, the maximum value of the core correlation parameter in the optimal working condition sample set is extracted as the upper limit and the minimum value is extracted as the lower limit to obtain the suitable range of the core correlation parameter in the homogeneous area. Based on the suitable range of the core correlation parameter, the mapping relationship between the core correlation parameter and the optimal instruction adjustment amount of all management units in the homogeneous area is fitted by univariate linear regression to construct a dynamic planting instruction model.

10. A cloud computing-based smart agricultural planting control system, characterized in that, The system is used to perform the method according to any one of claims 1-9, the system comprising: Reconstruction evaluation module: Extract actual planting result data under historical cloud planting instructions, divide the field into several management units according to the same planting area, analyze the fruit deformity rate of each management unit under the current cloud planting instructions, and evaluate whether cloud algorithm reconstruction is needed; Core correlation parameter identification module: After determining that cloud algorithm reconstruction is required, extract environmental parameters related to crop growth, analyze and determine the degree of correlation between environmental parameters and fruit deformity rate, and identify core correlation parameters that have a significant impact on crop fruit deformity rate; Reconstruction cycle determination module: By quantifying the environmental fluctuation frequency of each management unit and combining the baseline parameters with the deviation of the fruit deformity rate, the reconstruction cycle of the cloud algorithm is determined; Dynamic planting instruction model construction module: Based on the reconstruction cycle, a spatiotemporal clustering algorithm is used to divide the planting plots into several homogeneous zones according to environmental characteristic data, determine the appropriate range of core correlation parameters of the homogeneous zones, and construct a dynamic planting instruction model for the homogeneous zones.

Images