A smart deployment system for IoT devices in smart farmland

Abstract

This invention relates to the field of intelligent deployment of farmland equipment, specifically to an intelligent deployment system for smart farmland IoT devices. The system includes: a farmland environment and crop coupling module for establishing a mapping relationship between crops and the environment; a monitoring module for quantitatively assessing the sensitivity of different areas of farmland to environmental changes; a region and equipment preprocessing module for finding the optimal equipment layout scheme based on the RANSAC algorithm and incorporating an optimization search mechanism; and a deployment module for converting the optimal equipment layout scheme into an executable equipment deployment task list. This invention accelerates the algorithm's convergence speed by optimizing the sampling and iteration logic of the RANSAC algorithm and combining it with precise quantification of sensing coverage using device vector fields, ensuring that the output layout scheme achieves an optimal balance between coverage effect and resource allocation.

Description

Technical Field

[0001] This invention relates to the field of intelligent deployment of farmland equipment, and more specifically to an intelligent deployment system for smart farmland Internet of Things (IoT) equipment. Background Technology

[0002] In smart farmland, the effectiveness of IoT device deployment directly determines the accuracy of environmental monitoring, the quality of crop growth data collection, and the efficiency of resource utilization.

[0003] Most existing agricultural IoT device deployment schemes lack a deep understanding of the relationship between crop growth and environmental factors. They rely solely on uniform distribution or human experience for layout, failing to take into account the different sensitivity of different areas of farmland to environmental changes such as temperature, humidity, and light. This results in an imbalance where monitoring is insufficient in sensitive areas and there is redundant equipment in non-sensitive areas.

[0004] Furthermore, traditional deployment algorithms often employ a simple random sampling and iteration mode without considering regional monitoring priorities. This results in slow algorithm convergence, making it difficult to quickly output the optimal layout adapted to the actual farmland conditions. In addition, existing solutions lack precision in quantifying the sensing coverage of devices, failing to fully consider the sensing characteristics of different types of devices and the coverage effect after multiple devices are superimposed. This can easily lead to monitoring blind spots or duplicate coverage, affecting the integrity of monitoring data and wasting deployment costs. These issues make it difficult for existing deployment solutions to meet the needs of refined monitoring and efficient resource allocation in smart farmland. Summary of the Invention

[0005] This invention addresses the technical problems existing in the prior art by providing an intelligent deployment system for smart farmland IoT devices.

[0006] The technical solution of this invention to solve the above-mentioned technical problems is as follows: an intelligent deployment system for smart farmland Internet of Things (IoT) devices, the system comprising:

[0007] Farmland environment and crop coupling module; used to establish the mapping relationship between crops and the environment;

[0008] Monitoring module; used to quantitatively assess the sensitivity of different areas of farmland to environmental changes, generate microclimate sensitivity index, and identify sensitive areas of different levels;

[0009] The region and equipment preprocessing module is used to introduce an optimization search mechanism based on the RANSAC algorithm, so that the RANSAC algorithm incorporates the microclimate sensitivity index of each region in each sampling and iteration, and establishes a vector field for each device, and finds the optimal device layout scheme through repeated iterations of the RANSAC algorithm.

[0010] The deployment module transforms the optimal device layout scheme into an executable list of device deployment tasks. After the devices are online, it collects actual system operation data, compares it with the established mapping relationship, and quantifies the actual effect of the deployment scheme.

[0011] In a preferred embodiment, the farmland environment and crop coupling module obtains historical observation data of local temperature, humidity, and light in the farmland through a third party, as well as yield data corresponding to each type of observation data, to obtain a yield sequence corresponding to the observation data and yield data;

[0012] Calculate the linear trend of the production series to obtain the trend value, such as the slope. Subtract the trend value from each production data value in the production series, such as the increase in production due to global warming.

[0013] Calculate the average of each observation data and detrended yield, and calculate the deviation, covariance, and standard deviation of each observation data. Divide the covariance of each observation data by the sedimentation of two standard deviations to obtain the correlation coefficient used to map the relationship between crops and the environment.

[0014] The deviation is calculated as the difference between the environmental variable value and each observed data, and the difference between the detrended output value and the detrended output average. The covariance is the deviation of each observed data multiplied by the deviation of the detrended output, and then summed to obtain the covariance. The standard deviation of each observed data and the detrended output are calculated separately, which is the square root of the average of the squared deviations. The covariance is divided by the product of the two standard deviations to obtain the correlation coefficient.

[0015] In a preferred embodiment, the monitoring module divides the farmland into several grid units at equal intervals. Specifically, it automatically divides the grid using GIS software combined with a digital elevation model, outputs grid boundary files in the WGS-84 coordinate system, and extracts at least seven days' worth of temperature, humidity, and light intensity sequences for each grid unit in each area of ​​the farmland as the basic data for the microclimate sensitivity index, specifically including:

[0016] Satellite data obtained from a third-party platform is spatially interpolated to generate a daily average temperature sequence for each grid cell;

[0017] The humidity sequence is obtained by directly mapping the temperature and humidity sensors distributed in each grid cell to the corresponding grid cell;

[0018] The illumination sequence is obtained by directly mapping the light intensity sensors distributed in each grid cell to the corresponding grid cell;

[0019] The monitoring module calculates the probability density of the temperature sequence of each grid cell, then takes the negative logarithm sum to obtain the temperature information entropy. It also calculates the mean of the absolute values ​​of the differences between the humidity values ​​of each grid cell and the humidity values ​​of adjacent grid cells to obtain the humidity spatial gradient. Finally, it calculates the difference between the maximum and minimum values ​​of the light intensity sequence of each grid cell to obtain the day-night light intensity difference.

[0020] Temperature information entropy, humidity spatial gradient, and diurnal light difference are integrated in an adaptive weighting manner, with the weighting coefficients determined by the mapping relationship between crops and the environment, to generate a microclimate sensitivity index.

[0021] The mapping relationship is specifically formed by the dynamic weights of the correlation coefficients of temperature, humidity, and light. The dynamic weight of temperature is specifically the absolute value of the temperature correlation coefficient divided by the sum of the absolute values ​​of the correlation coefficients of temperature, humidity, and light. The same applies to humidity and light. Furthermore, this application considers that in ordinary farmland planting, temperature, humidity, and light have the greatest impact. Therefore, the correlation coefficients of temperature, humidity, and light are established to establish the mapping relationship between crops and the environment. In special farmland, other environmental factors, such as microbial indicators, can also be addressed by adopting the series of schemes proposed in this application to quantify the relationship between crops and the environment.

[0022] After obtaining the microclimate sensitivity index of the grid units corresponding to each area of ​​farmland, the microclimate sensitivity index of each grid unit is normalized to the range of [0,1]. Grid units with microclimate sensitivity index exceeding the first-level preset threshold are designated as sensitive areas, where the preset threshold is determined by historical experience. Grid units with microclimate sensitivity index exceeding the second-level preset threshold are designated as medium-sensitive areas, and grid units with microclimate sensitivity index less than the second-level preset threshold are designated as low-sensitive areas.

[0023] In a preferred embodiment, when the region and device preprocessing module randomly selects several points from the grid cells as initial samples based on the RANSAC algorithm, an optimization search mechanism is introduced to prioritize the selection of key monitored sensitive areas, including:

[0024] Each grid cell is assigned an initial priority value, representing its potential to be selected. The initial priority value is positively correlated with the microclimate sensitivity index of the corresponding grid cell. That is, the initial priority value of a highly sensitive area is greater than that of a medium sensitive area, and the initial priority value of a medium sensitive area is greater than that of a low sensitive area. In short, in practice, a constant greater than zero and suitable for RANSAC algorithm sampling is introduced, and the positive correlation function can be established by multiplying the constant by the microclimate sensitivity index.

[0025] Each grid cell obtains a selection probability based on its initial priority value. The selection probability is determined by the proportion of the initial priority value of a single grid cell to the sum of the initial priority values ​​of all grid cells. In order to avoid the complete exclusion of low- and medium-microclimate sensitivity areas from selection, a lower limit is set for the basic selection probability. The basic selection probability of low-priority grid cells is no less than 0.5%, and that of medium-priority grid cells is no less than 1%, ensuring that even low-sensitivity areas can have a small number of samples participating in the sampling, thus ensuring the overall continuity of the layout.

[0026] The area is divided into high-sensitivity, medium-sensitivity, and low-sensitivity regions, and a sampling ratio is allocated accordingly. In this application, the sampling ratio can be 7:2:1, with the medium-sensitivity region accounting for 2% to ensure coverage of potentially sensitive areas and prevent the medium-sensitivity region from being overlooked due to excessive bias towards high-priority regions. The low-sensitivity region accounts for the proportion to retain a basic sampling volume, controlling sample redundancy in low-value areas and preventing these areas from completely deviating from the layout model, thus ensuring the integrity of the overall farmland monitoring. When the difference in the preset thresholds among the sensitive areas is small, the ratio is adjusted to reduce the difference, ensuring that the area does not deviate from the layout model and that the integrity of the overall farmland monitoring is maintained.

[0027] In a preferred embodiment, after the RANSAC algorithm completes sampling, it performs iterations, including:

[0028] The initial sample sampling of the RANSAC algorithm is completed according to the determined selection probability and sampling ratio. In the initial and subsequent sample iterations of the RANSAC algorithm, if the device layout corresponding to a grid cell can cover the grid cell area, it is determined as an interior point. The RANSAC algorithm adds an additional value to the initial priority value of the grid cell corresponding to the interior point. The additional value is a fixed value, which is determined by the hierarchical division of high-sensitivity, medium-sensitivity, and low-sensitivity regions. For example, high-priority grid cells are increased by 0.3, medium-priority by 0.1, and low-priority by 0.05.

[0029] In each iteration, if a grid cell is continuously not selected, it is determined to be an outlier. The RANSAC algorithm adds a decay value to the initial priority value of the grid cell corresponding to the outlier. The decay value is a fixed proportion, such as 20% in this application, to avoid the algorithm continuously focusing on poorly adapted areas. In the second and subsequent sampling of the RANSAC algorithm, the selection probability is recalculated by the sum of the initial priority value and the additional value of the grid cell corresponding to the inlier, and the selection probability is recalculated by the product of the initial priority value and the decay value of the grid cell corresponding to the outlier. This completes the introduction of the microclimate sensitivity index of each region into the RANSAC algorithm in each sampling and iteration, accelerating the convergence speed of the RANSAC algorithm in the optimal layout.

[0030] In a preferred embodiment, the area and device preprocessing module integrates the sensing coverage of each device to establish a sensing behavior field, including: establishing a coordinate system consistent with the device coordinate allocation; assigning unique coordinates to the grid cells established by the monitoring module, where z is the vertical height; constructing a vector field for a single device based on the vector direction and vector magnitude, where the vector direction is the sensing coverage of a single device in operation. For example, the vector direction of a soil sensor is vertical downwards from the ground, with the z-axis in the negative direction; the vector direction of a camera is outwards along the fan-shaped field of view axis, i.e., from the installation point to the grid cell, with the angle between the camera and the horizontal direction being the installation tilt angle; the vector direction of a micro-weather station is radial from the device installation point to the grid cell, propagating omnidirectionally without a fixed direction, and the vector direction is determined by the line connecting the grid cell and the installation point. The vector magnitude is the sensing intensity of a single device in the corresponding grid cell. Taking a camera as an example, the sensing intensity of a camera decreases quadratically with the field of view distance, and the optical signal attenuation is proportional to the square of the distance. Therefore, with the device installation point as the vertex and the field of view distance as L, the sensing intensity = initial sensing intensity at near distance of the device × 1 / 1 + k L 2 Where k is the attenuation coefficient, which is determined by the camera manufacturer, and the sensing intensity can also be found in the corresponding device manufacturer. Common devices such as soil sensors have a sensing intensity that decreases linearly with horizontal distance, etc., to complete the establishment of the vector field for each device.

[0031] When a grid cell is covered by multiple devices of the same type, the following settings are also included: the vector direction is the vector direction of the device with the highest sensing intensity, and the vector magnitude is the arithmetic sum of the sensing intensities of all devices within the grid cell;

[0032] When a grid cell is covered by different types of devices, the following settings are also included: the vector magnitude is the superposition of the sensing intensities of multiple devices, and the vector direction is the vector direction of the device with the largest sensing intensity;

[0033] When the vector magnitude of a single device and / or multiple devices within a grid cell exceeds the grid cell boundary, it is determined that the device layout covers the grid cell area.

[0034] In a preferred embodiment, after acquiring the vector field of each device, it is input into the RANSAC algorithm. During the iteration process, the interior points are used to simulate the layout scheme, that is, to determine the type, quantity, and installation coordinates of the devices to be installed in the grid cell. According to the preset requirements, the types of devices to be installed are verified through the device vector field to verify the layout scheme, thus completing the consistency check in the RANSAC algorithm, including:

[0035] Based on the device's vector direction, determine whether the device's sensing range can cover the target grid cell;

[0036] Calculate the vector magnitude of the device and determine whether the vector magnitude exceeds the boundary of the mesh cell;

[0037] Given the existing device layout, after integrating the vector field according to the multi-device coverage rules, it is necessary to verify again whether the integrated vector direction and magnitude meet the coverage requirements.

[0038] If all existence verifications pass, the interior point is determined to be a valid interior point. If the existence verification fails, the interior point is determined to be a pending interior point. The selection probability is then recalculated using the initial priority value of the grid cell corresponding to the interior point, and subsequent iterations are performed.

[0039] In a preferred embodiment, after the RANSAC algorithm iteration is completed, the region and device preprocessing module establishes a consensus set for all valid interior points. If the number of valid interior points in the consensus set is greater than or equal to a predetermined threshold, it is determined that the deployment model output by RANSAC after iteration meets the basic monitoring requirements. If the number of valid interior points in the consensus set is less than the predetermined threshold, RANSAC resamples and iterates. The predetermined threshold is the sum of the product of the total number of grid cells in high-sensitivity areas multiplied by its sampling ratio, the product of the total number of grid cells in medium-sensitivity areas multiplied by its sampling ratio, and the product of the total number of grid cells in low-sensitivity areas multiplied by its sampling ratio. This avoids medium and low-sensitivity areas from completely deviating from the layout model. Therefore, the sampling ratio is used as a component of the predetermined threshold. The deployment module finally outputs a consensus set containing all valid interior points as the optimal device layout scheme.

[0040] In a preferred embodiment, the deployment module simulates the layout scheme of all internal points in the consistent set and the unique coordinates assigned to the grid cells established by the region and equipment preprocessing module for the monitoring module to form an executable equipment deployment task list. Each task in the equipment deployment task list corresponds to the deployment of equipment in one grid cell, including equipment type, number of equipment, unique coordinates, minimum value of vector magnitude, and corresponding sensitive area level. The spacing between all equipment of the same type in the equipment deployment task list is calculated based on the vector magnitude. After the equipment deployment task list is deployed, it is verified according to the judgment rule of equipment layout covering the grid cell area.

[0041] In a preferred embodiment, after the equipment is deployed, the farmland environment and crop coupling module obtains new temperature, humidity and light intensity sequences of the grid units corresponding to each area of ​​the farmland. This can be obtained through third-party software or by the type of sensor equipment added during deployment, i.e., the sensors to be installed in the layout scheme simulation, to generate new correlation coefficients. These correlation coefficients are then compared with those before deployment to quantify the actual effect of the deployment scheme.

[0042] The beneficial effects of this invention are: optimizing the sampling and iteration logic of the RANSAC algorithm, combining the precise quantification of sensing coverage by the device vector field, accelerating the convergence speed of the algorithm, ensuring that the output layout scheme achieves the optimal balance between coverage effect and resource allocation, and incorporating the microclimate sensitivity index into the sampling probability calculation, allowing the algorithm to prioritize sampling of high-value areas, reducing the number of invalid iterations in low-priority areas, and shortening the scheme generation cycle. At the same time, the construction of the device vector field fully considers the sensing characteristics of different types of devices and the coverage effect after multiple devices are superimposed, which can accurately determine the actual coverage of each grid cell and avoid monitoring blind spots or duplicate coverage problems caused by inaccurate estimation of device sensing range.

[0043] By delineating sensitive areas of farmland and introducing an optimized search mechanism, highly sensitive areas can obtain targeted monitoring resources while ensuring basic monitoring coverage in medium- and low-sensitivity areas. This effectively avoids monitoring blind spots and resource redundancy. The sensitivity of different areas of farmland to environmental changes is objectively quantified and classified to ensure that the identification of sensitive areas is more in line with the actual situation of farmland. In terms of equipment deployment, highly sensitive areas can obtain higher sampling priority and equipment configuration density, improving the monitoring accuracy of key areas. It also realizes the on-demand allocation of monitoring resources and significantly optimizes the equipment investment cost and monitoring benefit ratio of smart farmland. Attached Figure Description

[0044] Figure 1 This is a schematic diagram of the process of executing the present invention. Detailed Implementation

[0045] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0046] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0047] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0048] This embodiment provides: an intelligent deployment system for smart farmland IoT devices, the system comprising:

[0049] Farmland environment and crop coupling module; used to establish the mapping relationship between crops and the environment;

[0050] Monitoring module; used to quantitatively assess the sensitivity of different areas of farmland to environmental changes, generate microclimate sensitivity index, and identify sensitive areas of different levels;

[0051] The region and equipment preprocessing module is used to introduce an optimization search mechanism based on the RANSAC algorithm, so that the RANSAC algorithm incorporates the microclimate sensitivity index of each region in each sampling and iteration, and establishes a vector field for each device, and finds the optimal device layout scheme through repeated iterations of the RANSAC algorithm.

[0052] The deployment module transforms the optimal device layout scheme into an executable list of device deployment tasks. After the devices are online, it collects actual system operation data, compares it with the established mapping relationship, and quantifies the actual effect of the deployment scheme.

[0053] The specific deployment method of this system in the deployment of smart farmland IoT devices is as follows:

[0054] S101. Establish a mapping relationship between crops and the environment.

[0055] The farmland environment and crop coupling module obtains historical observation data of local temperature, humidity and light in the farmland through a third party, as well as yield data corresponding to each observation data, and obtains the yield sequence corresponding to the observation data and yield data.

[0056] Calculate the linear trend of the production series to obtain the trend value, such as the slope. Subtract the trend value from each production data value in the production series, such as the increase in production due to global warming.

[0057] Calculate the average of each observation data and detrended yield, and calculate the deviation, covariance, and standard deviation of each observation data. Divide the covariance of each observation data by the sedimentation of two standard deviations to obtain the correlation coefficient used to map the relationship between crops and the environment.

[0058] The deviation is calculated as the difference between the environmental variable value and each observed data, and the difference between the detrended output value and the detrended output average. The covariance is the deviation of each observed data multiplied by the deviation of the detrended output, and then summed to obtain the covariance. The standard deviation of each observed data and the detrended output are calculated separately, which is the square root of the average of the squared deviations. The covariance is divided by the product of the two standard deviations to obtain the correlation coefficient.

[0059] S102. Quantitatively assess the sensitivity of different areas of farmland to environmental changes, generate a microclimate sensitivity index, and identify sensitive areas of different levels.

[0060] The monitoring module divides the farmland into several grid units at equal intervals. Specifically, it automatically divides the grid using GIS software combined with a digital elevation model, outputs grid boundary files in the WGS-84 coordinate system, and extracts temperature, humidity, and light intensity sequences for at least seven days for each grid unit in each area of ​​the farmland as the basic data for the microclimate sensitivity index. This data includes:

[0061] Satellite data obtained from a third-party platform is spatially interpolated to generate a daily average temperature sequence for each grid cell;

[0062] The humidity sequence is obtained by directly mapping the temperature and humidity sensors distributed in each grid cell to the corresponding grid cell;

[0063] The illumination sequence is obtained by directly mapping the light intensity sensors distributed in each grid cell to the corresponding grid cell;

[0064] The monitoring module calculates the probability density of the temperature sequence of each grid cell, then takes the negative logarithm sum to obtain the temperature information entropy. It also calculates the mean of the absolute values ​​of the differences between the humidity values ​​of each grid cell and the humidity values ​​of adjacent grid cells to obtain the humidity spatial gradient. Finally, it calculates the difference between the maximum and minimum values ​​of the light intensity sequence of each grid cell to obtain the day-night light intensity difference.

[0065] Temperature information entropy, humidity spatial gradient, and diurnal light difference are integrated in an adaptive weighting manner, with the weighting coefficients determined by the mapping relationship between crops and the environment, to generate a microclimate sensitivity index.

[0066] The mapping relationship is specifically formed by the dynamic weights of the correlation coefficients of temperature, humidity, and light. The dynamic weight of temperature is specifically the absolute value of the temperature correlation coefficient divided by the sum of the absolute values ​​of the correlation coefficients of temperature, humidity, and light. The same applies to humidity and light. Furthermore, this application considers that in ordinary farmland planting, temperature, humidity, and light have the greatest impact. Therefore, the correlation coefficients of temperature, humidity, and light are established to establish the mapping relationship between crops and the environment. In special farmland, other environmental factors, such as microbial indicators, can also be addressed by adopting the series of schemes proposed in this application to quantify the relationship between crops and the environment.

[0067] After obtaining the microclimate sensitivity index of each grid cell in each area of ​​farmland, the microclimate sensitivity index of each grid cell is normalized to the range of [0,1]. Grid cells with microclimate sensitivity index exceeding the first-level preset threshold are designated as sensitive areas, where the preset threshold is determined by historical experience. Grid cells with microclimate sensitivity index exceeding the second-level preset threshold are designated as medium-sensitive areas, and grid cells with microclimate sensitivity index less than the second-level preset threshold are designated as low-sensitive areas.

[0068] Preset thresholds: By analyzing farmland monitoring data from the past 3-5 years, the correlation between environmental change amplitude and yield fluctuation amplitude under different microclimate sensitivity index ranges is statistically analyzed to determine key critical points as thresholds. For example, if historical data shows that when the microclimate sensitivity index > 0.8, a 2°C fluctuation in temperature or a 15% fluctuation in humidity will lead to a wheat yield fluctuation of > 8%, exceeding the acceptable yield fluctuation threshold, then the primary threshold is set to 0.8. When 0.5 ≤ microclimate sensitivity index ≤ 0.8, the same environmental fluctuation only leads to a yield fluctuation of 3%-8%, which is within the acceptable range, then the secondary threshold is set to 0.5.

[0069] Furthermore, in some specific implementations, different crops have vastly different sensitivities to the environment, and the sensitivity of the same crop can also change at different growth stages. Warm-loving crops, such as rice and tomatoes, are sensitive to temperature and have a high temperature correlation coefficient. Therefore, it is necessary to lower the threshold to expand the coverage of highly sensitive areas. For example, the primary threshold can be adjusted to 0.7 and the secondary threshold can be adjusted to 0.4 to ensure that no temperature-sensitive areas are missed.

[0070] As described above, the preset threshold is determined by historical experience. In the specific implementation process, common methods such as vector machines or machine learning can be used to train the system using historical experience data as a training set. This can be applied to this application to achieve a fast and intelligent way of obtaining the preset threshold.

[0071] S103. Based on the RANSAC algorithm, an optimization search mechanism is introduced, which allows the RANSAC algorithm to incorporate the microclimate sensitivity index of each region in each sampling and iteration, and to establish a vector field for each device. The optimal device layout scheme is found through repeated iterations of the RANSAC algorithm.

[0072] When the region and equipment preprocessing module randomly selects several points from the grid cells as initial samples based on the RANSAC algorithm, it introduces an optimization search mechanism to prioritize the selection of key monitoring sensitive areas, including:

[0073] Each grid cell is assigned an initial priority value, representing its potential to be selected. The initial priority value is positively correlated with the microclimate sensitivity index of the corresponding grid cell. That is, the initial priority value of a highly sensitive area is greater than that of a medium sensitive area, and the initial priority value of a medium sensitive area is greater than that of a low sensitive area. In short, in practice, a constant greater than zero and suitable for RANSAC algorithm sampling is introduced, and the positive correlation function can be established by multiplying the constant by the microclimate sensitivity index.

[0074] In some other specific implementations, the initial priority value is positively correlated with the microclimate sensitivity index of the corresponding grid cell. That is, a constant of 10 is selected to adapt to the RANSAC algorithm sampling. 10 is greater than zero and can make the calculation result fall within a reasonable sampling probability range, avoiding values ​​that are too small or too large. The microclimate sensitivity index is normalized to the [0,1] interval and multiplied by 10 to obtain the initial finite value. As mentioned above, temperature information entropy, humidity spatial gradient, and light diurnal difference are integrated in an adaptive weighting manner. The weight coefficient is determined by the mapping relationship between crops and the environment to generate the microclimate sensitivity index. The dynamic weights of temperature, humidity, and light are specifically the absolute value of the correlation coefficient of temperature divided by the sum of the absolute values ​​of the correlation coefficients of temperature, humidity, and light. The specific microclimate sensitivity index is determined by the region, which will not be elaborated here. Therefore, the initial priority values ​​of multiple sensitive regions of the same type are also different. Therefore, this application further refines the initial finite value of sensitive regions of the same type, instead of determining the initial priority by simple threshold classification, thereby improving the accuracy of subsequent sampling.

[0075] Each grid cell obtains a selection probability based on its initial priority value. The selection probability is determined by the proportion of the initial priority value of a single grid cell to the sum of the initial priority values ​​of all grid cells. In order to avoid the complete exclusion of low- and medium-microclimate sensitivity areas from selection, a lower limit is set for the basic selection probability. The basic selection probability of low-priority grid cells is no less than 0.5%, and that of medium-priority grid cells is no less than 1%, ensuring that even low-sensitivity areas can have a small number of samples participating in the sampling, thus ensuring the overall continuity of the layout.

[0076] The area is divided into high-sensitivity, medium-sensitivity, and low-sensitivity regions, and a sampling ratio is allocated accordingly. In this application, the sampling ratio can be 7:2:1, with the medium-sensitivity region accounting for 2% to ensure coverage of potentially sensitive areas and prevent the medium-sensitivity region from being overlooked due to excessive bias towards high-priority regions. The low-sensitivity region accounts for the proportion to retain a basic sampling volume, controlling sample redundancy in low-value areas and preventing these areas from completely deviating from the layout model, thus ensuring the integrity of the overall farmland monitoring. When the difference in the preset thresholds among the sensitive areas is small, the ratio is adjusted to reduce the difference, ensuring that the area does not deviate from the layout model and that the integrity of the overall farmland monitoring is maintained.

[0077] The specific setting of the sampling ratio is determined by the difference between the first-level threshold and the second-level threshold in the preset threshold. For example, if the first-level threshold is 0.6 and the second-level threshold is 0.4, then the range of the first-level threshold is [0.6,1] and the range of the second-level threshold is [0.4,0.6]. Based on the division of the range, the sampling ratio is set to 5:2:3.

[0078] After the RANSAC algorithm completes sampling, iterations are performed, including:

[0079] The initial sample sampling of the RANSAC algorithm is completed according to the determined selection probability and sampling ratio. In the initial and subsequent sample iterations of the RANSAC algorithm, if the device layout corresponding to a grid cell can cover the grid cell area, it is determined as an interior point. The RANSAC algorithm adds an additional value to the initial priority value of the grid cell corresponding to the interior point. The additional value is a fixed value, which is determined by the hierarchical division of high-sensitivity, medium-sensitivity, and low-sensitivity regions. For example, high-priority grid cells are increased by 0.3, medium-priority by 0.1, and low-priority by 0.05.

[0080] In each iteration, if a grid cell is continuously not selected, it is determined to be an outlier. The RANSAC algorithm adds a decay value to the initial priority value of the grid cell corresponding to the outlier. The decay value is a fixed proportion, such as 20% in this application, to avoid the algorithm continuously focusing on poorly adapted areas. In the second and subsequent sampling of the RANSAC algorithm, the selection probability is recalculated by the sum of the initial priority value and the additional value of the grid cell corresponding to the inlier, and the selection probability is recalculated by the product of the initial priority value and the decay value of the grid cell corresponding to the outlier. This completes the introduction of the microclimate sensitivity index of each region into the RANSAC algorithm in each sampling and iteration, accelerating the convergence speed of the RANSAC algorithm in the optimal layout.

[0081] The area and equipment preprocessing module integrates the sensing coverage of each device to establish a sensing behavior field, including: establishing a coordinate system consistent with the device coordinate allocation; assigning unique coordinates to the grid cells established by the monitoring module, where z is the vertical height; constructing a vector field for a single device based on the vector direction and vector magnitude, where the vector direction is the sensing coverage of a single device in operation. For example, the vector direction of a soil sensor is vertical downwards from the ground, with the z-axis in the negative direction; the vector direction of a camera is outwards along the fan-shaped field of view axis, i.e., from the installation point to the grid cell, with the angle between the camera and the horizontal direction being the installation tilt angle; the vector direction of a micro-weather station is radial from the device installation point to the grid cell, propagating omnidirectionally without a fixed direction, and the vector direction is determined by the line connecting the grid cell and the installation point. The vector magnitude is the sensing intensity of a single device in the corresponding grid cell. Taking a camera as an example, the sensing intensity of a camera decreases quadratically with the field of view distance, and the optical signal attenuation is proportional to the square of the distance. Therefore, with the device installation point as the vertex and the field of view distance as L, the sensing intensity = initial sensing intensity at near distance × 1 / 1 + k L 2 Where k is the attenuation coefficient, which is determined by the camera manufacturer. The sensing intensity can also be found in the corresponding equipment manufacturer's information, such as the attenuation coefficient of lens transmittance provided by the manufacturer in optical consistency monitoring, in dB. For more common equipment, such as soil sensors, the sensing intensity decreases linearly with horizontal distance, etc., to complete the establishment of the vector field for each device.

[0082] When a grid cell is covered by multiple devices of the same type, the following settings are also included: the vector direction is the vector direction of the device with the highest sensing intensity, and the vector magnitude is the arithmetic sum of the sensing intensities of all devices within the grid cell;

[0083] When a grid cell is covered by different types of devices, the following settings are also included: the vector magnitude is the superposition of the sensing intensities of multiple devices, and the vector direction is the vector direction of the device with the largest sensing intensity;

[0084] When the vector magnitude of a single device and / or multiple devices within a grid cell exceeds the grid cell boundary, it is determined that the device layout covers the grid cell area.

[0085] After obtaining the vector field for each device, it is input into the RANSAC algorithm. During the iteration process, the interior points are used to simulate the layout scheme, that is, to determine the type, quantity, and installation coordinates of the devices to be installed in the grid cell. According to the preset requirements, the types of devices to be installed are verified through the device vector field to verify the layout scheme, completing the consistency check in the RANSAC algorithm, including:

[0086] Based on the device's vector direction, determine whether the device's sensing range can cover the target grid cell;

[0087] Calculate the vector magnitude of the device and determine whether the vector magnitude exceeds the boundary of the mesh cell;

[0088] Given the existing device layout, after integrating the vector field according to the multi-device coverage rules, it is necessary to verify again whether the integrated vector direction and magnitude meet the coverage requirements.

[0089] If all existence verifications pass, the interior point is determined to be a valid interior point. If the existence verification fails, the interior point is determined to be a pending interior point. The selection probability is then recalculated using the initial priority value of the grid cell corresponding to the interior point, and subsequent iterations are performed.

[0090] After the RANSAC algorithm iteration is completed, the region and device preprocessing module establishes a consensus set for all valid interior points. If the number of valid interior points in the consensus set is greater than or equal to a certain threshold, it is determined that the deployment model output by RANSAC after iteration meets the basic monitoring requirements. If the number of valid interior points in the consensus set is less than the certain threshold, RANSAC resamples and iterates. The certain threshold is the sum of the product of the total number of grid cells in high-sensitivity areas multiplied by its sampling ratio, the product of the total number of grid cells in medium-sensitivity areas multiplied by its sampling ratio, and the product of the total number of grid cells in low-sensitivity areas multiplied by its sampling ratio. This avoids medium and low-sensitivity areas from completely deviating from the layout model. Therefore, the sampling ratio is used as a component of the certain threshold. The deployment module finally outputs a consensus set containing all valid interior points as the optimal device layout scheme.

[0091] S104. Transform the optimal equipment layout scheme into an executable equipment deployment task list. After the equipment is online, collect the actual data of the system operation, compare it with the established mapping relationship, and quantify the actual effect of the deployment scheme.

[0092] The deployment module simulates the layout scheme of all internal points in the consistent set, and the unique coordinates assigned to the grid cells established by the region and equipment preprocessing module for the monitoring module, forming an executable equipment deployment task list. Each task in the equipment deployment task list corresponds to the deployment of equipment in one grid cell, including equipment type, number of equipment, unique coordinates, minimum value of vector magnitude, and corresponding sensitive area level. The spacing between all equipment of the same type in the equipment deployment task list is calculated based on the vector magnitude. After the equipment deployment task list is deployed, it is verified according to the judgment rules for equipment layout covering the grid cell area.

[0093] After the equipment is deployed, the farmland environment and crop coupling module obtains new temperature, humidity, and light intensity sequences for the grid units corresponding to each area of ​​the farmland. This can be obtained through third-party software or by the type of sensor equipment added during deployment, i.e., the sensors to be installed in the layout scheme simulation. New correlation coefficients are generated and compared with the correlation coefficients before deployment. If there is an improvement, it indicates that the environment after the equipment is deployed has led to positive crop growth. More commonly, irrigation equipment overcomes the situation of water shortage or untimely irrigation in the original environment, thus quantifying the actual effect of the deployment scheme.

[0094] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0095] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0096] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0097] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0098] These computer program instructions can also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0099] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0100] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A smart deployment system for smart farmland IoT devices, characterized in that, The system includes: Farmland environment and crop coupling module; used to establish the mapping relationship between crops and the environment; The farmland environment and crop coupling module obtains historical observation data of local temperature, humidity and light in the farmland through a third party, as well as yield data corresponding to each observation data, and obtains the yield sequence corresponding to the observation data and yield data. Calculate the linear trend of the production series to obtain the trend value, and subtract the trend value from each production data value in the production series; Calculate the average of each observation data and detrended yield, and calculate the deviation, covariance, and standard deviation of each observation data. Divide the covariance of each observation data by the product of the two standard deviations to obtain the correlation coefficient used to map the relationship between crops and the environment. Monitoring module; used to quantitatively assess the sensitivity of different areas of farmland to environmental changes, generate microclimate sensitivity index, and identify sensitive areas of different levels; The monitoring module divides the farmland into several grid units at equal intervals and extracts the temperature, humidity, and light intensity sequences for at least seven days for each grid unit in each area of ​​the farmland, specifically including: Satellite data obtained from a third-party platform is spatially interpolated to generate a daily average temperature sequence for each grid cell; The humidity sequence is obtained by directly mapping the temperature and humidity sensors distributed in each grid cell to the corresponding grid cell; The light intensity sequence is obtained by directly mapping the light intensity sensors distributed in each grid cell to the corresponding grid cell. The monitoring module calculates the probability density of the temperature sequence of each grid cell, then takes the negative logarithm sum to obtain the temperature information entropy. It also calculates the mean of the absolute values ​​of the differences between the humidity values ​​of each grid cell and the humidity values ​​of adjacent grid cells to obtain the humidity spatial gradient. Finally, it calculates the difference between the maximum and minimum values ​​of the light intensity sequence of each grid cell to obtain the day-night light intensity difference. Temperature information entropy, humidity spatial gradient, and diurnal light difference are integrated in an adaptive weighting manner, with the weighting coefficients determined by the mapping relationship between crops and the environment, to generate a microclimate sensitivity index. After obtaining the microclimate sensitivity index of the grid units corresponding to each area of ​​farmland, the microclimate sensitivity index of each grid unit is normalized to the range of [0,1]. Grid units with microclimate sensitivity index exceeding the first-level preset threshold are designated as high-sensitivity areas, grid units with microclimate sensitivity index exceeding the second-level preset threshold are designated as medium-sensitivity areas, and grid units with microclimate sensitivity index less than the second-level preset threshold are designated as low-sensitivity areas. The region and equipment preprocessing module is used to introduce an optimization search mechanism based on the RANSAC algorithm, so that the RANSAC algorithm incorporates the microclimate sensitivity index of each region in each sampling and iteration, and establishes a vector field for each device, and finds the optimal device layout scheme through repeated iterations of the RANSAC algorithm. When the region and equipment preprocessing module randomly selects several points from the grid cells as initial samples based on the RANSAC algorithm, it introduces an optimization search mechanism to prioritize the selection of key monitoring sensitive areas, including: Each grid cell is assigned an initial priority value, which is positively correlated with the microclimate sensitivity index of the corresponding grid cell. Each grid cell obtains a selection probability based on its initial priority value. The selection probability is determined by the proportion of the initial priority value of a single grid cell to the sum of the initial priority values ​​of all grid cells. The regions were categorized into high-sensitivity, medium-sensitivity, and low-sensitivity areas, and sampling ratios were allocated accordingly. After the RANSAC algorithm completes sampling, iterations are performed, including: The initial sample sampling of the RANSAC algorithm is completed according to the determined selection probability and sampling ratio. In the sample iteration of the RANSAC algorithm, if the device layout corresponding to a grid cell can cover the grid cell area, it is determined as an interior point. The RANSAC algorithm adds an additional value to the initial priority value of the grid cell corresponding to the interior point. The additional value is a fixed value. In each iteration, if a grid cell is continuously not selected, it is determined to be an outlier. The RANSAC algorithm adds a decay value to the initial priority value of the grid cell corresponding to the outlier. The decay value is a fixed proportion. In the second and subsequent sampling of the RANSAC algorithm, the selection probability is recalculated by the sum of the initial priority value and the additional value of the grid cell corresponding to the inlier, and the selection probability is recalculated by the product of the initial priority value and the decay value of the grid cell corresponding to the outlier. This completes the introduction of the microclimate sensitivity index of each region into the RANSAC algorithm in each sampling and iteration. The region and device preprocessing module integrates the sensing coverage of each device to establish a sensing behavior field, including: establishing a coordinate system consistent with the device coordinate allocation, assigning unique coordinates to the grid cells established by the monitoring module, and constructing a vector field for a single device based on the vector direction and vector magnitude, wherein the vector direction is the sensing coverage of a single device, and the vector magnitude is the sensing intensity of a single device in the corresponding grid cell, thus completing the establishment of the vector field for each device. When a grid cell is covered by multiple devices of the same type, the following settings are also included: the vector direction is the vector direction of the device with the highest sensing intensity, and the vector magnitude is the arithmetic sum of the sensing intensities of all devices within the grid cell; When a grid cell is covered by different types of devices, the following settings are also included: the vector magnitude is the sum of the sensing intensities of multiple devices, and the vector direction is the vector direction of the device with the largest sensing intensity; When the vector magnitude of a single device and / or multiple devices within a grid cell exceeds the grid cell boundary, it is determined that the device layout covers the grid cell area; The deployment module transforms the optimal device layout scheme into an executable list of device deployment tasks. After the devices are online, it collects actual system operation data, compares it with the established mapping relationship, and quantifies the actual effect of the deployment scheme.

2. The intelligent deployment system for smart farmland IoT devices according to claim 1, characterized in that, After obtaining the vector field for each device, it is input into the RANSAC algorithm. During the iteration process, the interior points are used to simulate the layout scheme, including: Based on the device's vector direction, determine whether the device's sensing range can cover the target grid cell; Calculate the vector magnitude of the device and determine whether the vector magnitude exceeds the boundary of the mesh cell; Given the existing device layout, after integrating the vector field according to the multi-device coverage rules, it is necessary to verify again whether the integrated vector direction and magnitude meet the coverage requirements. If all existence verifications pass, the interior point is determined to be a valid interior point. If the existence verification fails, the interior point is determined to be a pending interior point. The selection probability is then recalculated using the initial priority value of the grid cell corresponding to the interior point, and subsequent iterations are performed.

3. The intelligent deployment system for smart farmland IoT devices according to claim 1, characterized in that, After the RANSAC algorithm iteration is completed, the region and device preprocessing module establishes a consensus set for all valid interior points. If the number of valid interior points in the consensus set is greater than or equal to a certain threshold, it is determined that the deployment model output by RANSAC after the iteration meets the basic monitoring requirements. If the number of valid interior points in the consensus set is less than the certain threshold, RANSAC is resampled and iterated. The deployment module finally outputs a consensus set containing all valid interior points as the optimal device layout scheme.

4. The intelligent deployment system for smart farmland IoT devices according to claim 3, characterized in that, The deployment module simulates the layout scheme of all internal points in the consistent set, and the unique coordinates assigned to the grid cells established by the region and equipment preprocessing module for the monitoring module, forming an executable equipment deployment task list. Each task in the equipment deployment task list corresponds to the deployment of equipment in one grid cell, including equipment type, number of equipment, unique coordinates, minimum value of vector magnitude, and corresponding sensitive area level. The spacing between all equipment of the same type in the equipment deployment task list is calculated based on the vector magnitude. After the equipment deployment task list is deployed, it is verified according to the judgment rules for equipment layout covering the grid cell area.

5. The intelligent deployment system for smart farmland IoT devices according to claim 1, characterized in that, After the equipment is deployed, the farmland environment and crop coupling module obtains new temperature, humidity and light intensity sequences of grid units corresponding to each area of ​​the farmland, generates new correlation coefficients, compares them with the correlation coefficients before deployment, and quantifies the actual effect of the deployment plan.

Images