Rice field water level multi-scale acquisition method

Abstract

The invention discloses a rice field water level multi-scale acquisition method which comprises the following steps: acquiring image data, telemetry data, weather data, soil entropy data and radar water level data of a rice field, and performing relation extraction on the rice field based on the image data to obtain a rice field map; performing Gaussian initialization and optimization by using the rice field map through the crop coverage data and the rice field area to obtain a Gaussian map; obtaining a first water level area and evaporation capacity according to a Gaussian map and a telemetering water level, determining a crop growth state through weather data and crop coverage data, determining crop actual water volume according to the crop growth state and soil entropy condition data, and correcting radar water level data through a Bayesian network by using the crop actual water volume and the evaporation capacity to obtain radar water level data; obtaining the rice field actual water level. According to the method, plot-level and area-level water level monitoring is coordinated, probabilistic reasoning is combined, and the rice field water level monitoring precision and the anti-jamming capability in a complex scene are improved.

Description

Technical Field

[0001] The present invention belongs to the field of paddy field water level monitoring, and in particular relates to a multi-scale acquisition method for paddy field water level. Background Art

[0002] Paddy field water level monitoring is a core link in agricultural production and ecological management. It regulates and improves the moisture conditions of paddy fields and regional water conservancy conditions, enhances the ability to resist natural disasters, promotes a virtuous cycle of the ecological environment, and makes it conducive to crop production. Among the many paddy field water conservancy measures, water level monitoring is extremely important to prevent excessive irrigation water from affecting the normal growth of crops and to ensure the health and high yield of crops through scientific monitoring methods. Existing technologies only provide a single water level monitoring function, without integrating other related measurement scales such as length and depth measurement. Spatially, the regional average water level is the main feature, and it is difficult to accurately identify water level anomalies of 1 to 5 mu at the plot level, such as local drought or waterlogging. This makes it impossible for hydrological data to effectively reflect the production of paddy field crops, resulting in extensive irrigation decisions and the inability of monitoring data to effectively support decision-making.

[0003] In order to overcome the shortcomings of traditional methods and meet the needs of multi-scale water level information acquisition in modern rice fields, through coordinated plot-level and regional-level water level monitoring, it can not only meet the precise irrigation of a single plot, but also support regional water resource optimization. It integrates image, telemetry, weather, soil moisture, irrigation and other data to solve data silos, and combines Bayesian probabilistic reasoning to improve monitoring accuracy and anti-interference capabilities in complex scenarios. Summary of the Invention

[0004] The purpose of the present invention is to provide a multi-scale acquisition method for paddy field water level.

[0005] To achieve the above object, the present invention is implemented according to the following technical solutions: A first aspect of the present invention provides a multi-scale method for obtaining paddy field water level, comprising: Collecting rice field data and radar water level data for preprocessing, wherein the rice field data includes image data, telemetry data, weather data, and soil entropy data; Divide the paddy field into regions based on the image data, extract paddy field area, remote sensing water level and crop coverage data according to the paddy field regions, and extract relationships between the paddy fields to obtain a paddy field map; Using the rice field atlas to perform Gaussian initialization and optimization through crop height, crop density, crop paddy field coverage and paddy field area of ​​crop cover data to obtain a Gaussian map; Obtaining a first water level area according to the Gaussian map and the remotely measured water level of the corresponding rice field, and obtaining evaporation based on the first water level area using weather data and crop cover data; Correcting the first water level area using the evaporation amount and radar water level data to obtain a second water level area, and determining the crop growth status based on the second water level area using weather data and crop cover data; The actual crop water content is determined based on crop growth status and soil entropy data. The radar water level data is corrected using the actual crop water content and evaporation through a Bayesian network to obtain the actual water level in the rice field.

[0006] As a further method, the pretreatment method comprises: Multispectral images and infrared spectral data of rice fields are collected by drones, and the multispectral images are converted to grayscale and median filtered to obtain image data; telemetry information is collected, and the missing data of the telemetry information is supplemented by time series interpolation method, abnormal data is eliminated, and the data format is unified to obtain telemetry data; temperature, humidity, rainfall, wind speed, and air pressure data are collected and timestamps are aligned to obtain weather data; radar water level data are collected, and the high-frequency interference noise in the radar water level data is removed using the wavelet noise reduction algorithm; single-point soil volume moisture content and irrigation data are collected to obtain soil entropy data.

[0007] As a further method, the method for obtaining the rice field atlas comprises: Superpixel segmentation is performed based on image data to divide the rice field image into small areas with similar colors and textures. The small areas are used as initial markers, and gradient calculation is performed on the rice field image to obtain a gradient image. According to the gradient image and the initial mark, the watershed algorithm is used to merge the internal area of ​​the rice field with low gradient values ​​and separate the boundary area with high gradient values. The contours of the separated areas are extracted and converted into polygonal vector data format to obtain polygonal rice field vector areas. Based on the polygonal rice field vector area, the vegetation index is extracted through the reflectance of the infrared band. The crop coverage of the rice field is calculated based on the vegetation index. The crop coverage is combined with the geographic coordinates to obtain the rice field area. The crop density is obtained based on the crop coverage and rice field area. The crop height and remote sensing water level are extracted from the rice field telemetry data. The crop height, crop density, and crop coverage are used as crop cover data. The rice field plots are used as nodes, the rice field area, telemetered water level, crop cover data, and geographic coordinates are used as node features, and the spatial adjacency relationship between adjacent rice fields and the connection relationship between irrigation and drainage networks are used as edges. Graph database modeling is performed based on nodes, node features, and edges to obtain a rice field map that contains spatial topological relationships and attribute associations.

[0008] As a further method, the Gaussian initialization and optimization method includes: Based on the node and spatial adjacency relationships of the rice field map, GIS topological analysis was used to extract the rice field area. The crop height, crop density, and crop coverage of the rice field plots were extracted based on the rice field area, and the mean and variance were calculated. The crop height, crop density, and crop coverage were Z-score standardized using the mean and variance to obtain a three-dimensional feature vector. A three-dimensional covariance matrix is ​​constructed based on the variance of each feature in the three-dimensional feature vector. The crop height and crop density in the three-dimensional feature vector are mapped to the vertical axis and horizontal axis of the two-dimensional space respectively. The two-dimensional covariance matrix is ​​obtained based on 0.2 times the crop density on the vertical axis and 0.1 times the crop height on the horizontal axis in the two-dimensional space. The ratio of the area of ​​the rice field plot to the total area of ​​the rice field region is used as the plot weight, and the two-dimensional mean, two-dimensional covariance matrix and plot weight are used as two-dimensional Gaussian components. The optimized two-dimensional Gaussian components are obtained by using the Gaussian parameter optimization formula. The Gaussian parameter optimization formula is: in is the optimized two-dimensional Gaussian component, including the mean , covariance and weights , is the number of observation data points, which is obtained by the number of pixels after rasterization of the rice field area. is the number of rice fields, is the plot index, is the plot index, For the The weight of the plot, is the mean of the two-dimensional features of the plot, is the covariance matrix, Adjacent plots obtained from the rice field map.

[0009] As a further method, a method for obtaining the Gaussian map includes: Update the Gaussian map parameters based on the optimized two-dimensional Gaussian distribution through the Gaussian map parameter optimization formula; The paddy fields are sorted in descending order according to the paddy area. The crop coverage of the paddy fields is used as the opacity. The Gaussian map parameters and opacity are optimized in descending order through the transparency blending formula to generate the Gaussian map. The transparency blending formula is: in For the pixels within the rice field area The Gaussian probability value of is the rice field area index, is the pixel coordinate in two-dimensional space, For the The opacity of each rice field plot, For the The weight of the rice field plots, For the The two-dimensional Gaussian distribution after optimization of the rice field plots, is the number of rice fields, Indicates before Treated rice fields.

[0010] As a further method, the method for optimizing the parameters of the Gaussian map includes: The Gaussian map parameter optimization formula is: in, The optimized Gaussian map parameters in the current field of view are minimized Photometric error, iterative adjustment , so that the generated image Better match observation images , Adam optimizer is used to efficiently solve the problem and update Gaussian parameters. is the field of view index, To predict the image, is the optimized two-dimensional Gaussian component in the current field of view, It is a weighted transparent combination.

[0011] As a further method, the method for obtaining the evaporation amount includes: Based on the Gaussian map, the parts of the image with Gaussian probability values ​​greater than 0.6 are taken as crop characteristic plots. Crop coverage data and telemetered water levels are extracted according to the rice field atlas. The areas belonging to the crop characteristic plots and with telemetered water levels 15% lower than the average water level are taken as the first water level areas. Based on the first water level areas, the Penman-Montes evaporation model is used to obtain the plot-level evaporation according to the weather data, crop coverage data and telemetered water levels of the rice field plots. The regional evaporation is then obtained by aggregating the plot-level evaporation using the plot weights.

[0012] As a further method, the method for determining the growth status of the crop includes: Based on the first water level area, radar water level data and plot-level evaporation data were extracted for each point. Points with plot-level evaporation greater than 2 mm within 6 hours were marked based on the radar water level data. The rice fields at the marked points were aggregated using the Gaussian map plot weights to obtain the second water level area. Based on the radar water level data, weather data and crop cover data of the second water level area, the water stress index of the rice field plots was determined by the water stress index model. The plots with a water stress index greater than 70% were classified as healthy, the plots with a water stress index less than 70% and greater than 40% were classified as mildly stressed, and the plots with a water stress index less than 40% were classified as severely stressed.

[0013] As a further method, a method for correcting radar water level data using the actual crop water content and evaporation through a Bayesian network includes: Extract soil entropy data based on the second water level area, and obtain the point coordinates of the soil entropy data The water stress index of crop growth status uses the product of the water stress index and the corresponding plot-level evaporation as the actual evaporation. Based on the actual evaporation, the soil entropy data within the same time period is extracted. The actual crop water content is obtained by adding the sum of the single-point soil volume water content change in the soil entropy data and the change in the irrigation data plus the rainfall data minus the actual evaporation. Based on the actual crop water volume, the corresponding rice field plot is used as a plot particle. The radar water level data is obtained from the plot particle as time series data. The time series data of the plot particle at the current moment is used as the observation node. The actual crop water volume, actual evaporation, irrigation data, rainfall data and time series data of the previous moment are used as state nodes according to the observation node. A Bayesian network is obtained based on the observation node and the state node. The Bayesian network is used to obtain the actual water level of the rice field through the rice field water level correction formula. The rice field water level correction formula is: in is the corrected radar water level data, is the total number of particles in the plot, For the moment Radar water level data, for Moment The predicted water level value of each particle, The standard deviation of the radar measurement error, For the The irrigation amount of each particle, 0 means no irrigation, obtained from the rice field atlas, For the The evaporation of particles, For the The actual amount of water for each particle of crop, For the The rainfall amount of each particle, 0 means no rainfall, obtained from weather data.

[0014] A second aspect of the present invention provides a multi-scale acquisition system for paddy field water level, comprising: Data acquisition module: used to collect rice field data and radar water level data for preprocessing. The rice field data includes image data, telemetry data, weather data, and soil entropy data; Paddy field atlas module: used to divide paddy fields into regions based on the image data, extract paddy field area, telemetry water level and crop coverage data according to the paddy field regions, and extract relationships between paddy fields to obtain paddy field atlas; Gaussian map module: used to use the rice field atlas to perform Gaussian initialization and optimization through crop height, crop density, crop paddy field coverage and paddy field area of ​​crop cover data to obtain a Gaussian map; A first water level area module is configured to obtain a first water level area based on the Gaussian map and the telemetered water level of the corresponding rice field, and obtain evaporation based on the first water level area using weather data and crop cover data; A second water level area module is configured to use the evaporation rate and radar water level data to correct the first water level area to obtain a second water level area, and determine the crop growth status based on the second water level area using weather data and crop cover data; Actual water level acquisition module: used to determine the actual water volume of crops based on crop growth status and soil entropy data, and use the actual water volume and evaporation of crops to correct the radar water level data through the Bayesian network to obtain the actual water level of the rice field.

[0015] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: The present invention coordinates plot-level and regional-level water level monitoring to meet the needs of precise irrigation of a single plot and support regional water resource optimization, thereby solving the problem of scale fragmentation, achieving multi-scale water level acquisition, and deepening multi-source data fusion. It integrates image crop coverage, remotely measured water level time series, weather evaporation and precipitation dynamics, soil moisture storage, irrigation human intervention and other data to solve the problems of data siloing and missing physical mechanisms of the model. It combines physical water balance with the probabilistic reasoning of Bayesian networks to improve the monitoring accuracy and anti-interference ability of radar water levels in complex scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a flowchart of the steps of a multi-scale acquisition method of rice field water level in an embodiment of the present invention. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0018] Reference Figure 1 As shown, the present invention provides a multi-scale acquisition method for paddy field water level, comprising: Paddy field data and radar water level data are collected for preprocessing, wherein the paddy field data includes image data, telemetry data, weather data, and soil entropy data.

[0019] In the actual evaluation, a drone was used to fly at an altitude of 100 meters during the tillering stage of rice in May and the booting stage in July to collect blue, green, and red multispectral images with a resolution of 0.15 meters. During flight, infrared 650nm and near-infrared 840nm channel data and L1 LiDAR data with a point cloud density of 200 points per square meter were simultaneously collected. The multispectral images were converted to grayscale using ENVI software and median filtered using Matlab with a 3×3 kernel to generate 512×512 pixel image data covering 1,000 mu of rice fields. Missing data in the LiDAR data were supplemented using time series interpolation, abnormal data were removed, and the data format was unified to generate a surface model (DSM) and a terrain model (DEM) to obtain telemetry data. Temperature, humidity, rainfall, wind speed, and air pressure in the rice field area were collected every 5 minutes and aligned with the telemetry data using timestamps. Sentinel-1 C Radar water level data with a band and a resolution of 10 meters is subjected to wavelet noise reduction to generate a radar water level grid in the WGS84 standard. Decagon 5TM sensors are buried at a depth of 15 cm in every 100 mu area to collect soil volumetric moisture content every hour. The irrigation amount and irrigation points are recorded through the irrigation system, and the irrigation points are associated with the soil moisture data to form soil moisture data.

[0020] The paddy field regions are divided based on the image data, the paddy field area, remote sensing water level and crop coverage data are extracted according to the paddy field regions, and the relationship between the paddy fields is extracted to obtain a paddy field map.

[0021] In the actual evaluation, the SLIC algorithm was used to segment the image data to obtain superpixel blocks with similar textures and an average area of ​​0.25m². The image data was gradient calculated using the Sobel operator with a 3×3 kernel. The watershed algorithm was used to merge low-gradient areas with a gradient less than 10 as the interior of the rice field, and to separate high-gradient boundaries with a gradient greater than 30 as the ridges. Polygonal vectors containing area and geographic coordinates were extracted using QGIS based on the gradient areas. The vegetation index NDVI was calculated using the near-infrared data (NIR) and infrared data (R) collected from the infrared spectral data as (NIR-R) / (NIR+R). When NDVI was greater than 0.4, it was considered effective coverage, and the crop area ratio was obtained as 85%. The crop height was obtained by subtracting the DSM and DEM data in the telemetry data with an accuracy of ±5cm. The crop height was calibrated using ground stakes to ensure that the crop area was within the range of 100%. The NDWI index is calculated using the DEM data of telemetry data and green spectrum data G and near-infrared data NIR for water body reflection, where the NDWI index is (G-NIR) / (G+NIR). The telemetry water level is obtained, and the paddy field area is obtained using crop coverage combined with geographic coordinates. The crop density is obtained based on crop coverage and paddy field area. The paddy field plots are used as nodes, and the paddy field area, telemetry water level, crop coverage data, and geographic coordinates are used as node features. The spatial adjacency relationship between adjacent paddy fields and the connection relationship of the irrigation and drainage network are used as edges. A graph database model is built based on nodes, node features, and edges to form a paddy field map containing 120 plots.

[0022] The paddy field atlas is used to perform Gaussian initialization and optimization through crop height, crop density, crop paddy field coverage and paddy field area of ​​crop cover data to obtain a Gaussian map.

[0023] In the actual evaluation, paddy field areas were extracted through GIS topological analysis based on the nodes and spatial adjacency of the paddy field map. The crop height, crop density, and crop coverage of the plots in the paddy field map were extracted according to the paddy field area and Z-score normalized. The mean crop height was 80 cm, the standard deviation was 10 cm, the mean crop density was 47,000 plants per mu, the standard deviation was 5,000, and the mean crop coverage was 0.85, the standard deviation was 0.1, and a three-dimensional feature vector was obtained. A three-dimensional covariance matrix was constructed based on the variance of each feature in the three-dimensional feature vector. The crop height and crop density in the three-dimensional feature vector were mapped to the vertical axis and horizontal axis of the two-dimensional space, respectively. The two-dimensional covariance matrix was obtained based on 0.2 times the crop density of the vertical coordinate and 0.1 times the crop height of the horizontal coordinate in the two-dimensional space. The ratio of the area of ​​the rice field plot to the total area of ​​the rice field area is used as the plot weight, and the two-dimensional mean, two-dimensional covariance matrix and plot weight are used as two-dimensional Gaussian components. The two-dimensional Gaussian components are used to optimize the Gaussian parameters using the Adam optimizer, with the learning rate set to 0.001 and 200 iterations to obtain the optimized two-dimensional Gaussian components; based on the optimized two-dimensional Gaussian distribution, the Gaussian map parameters are updated using the Gaussian map parameter optimization formula, and the areas of the rice field plots are sorted in descending order according to the rice field area, with the largest plot being 10 mu and the smallest being 2 mu. The crop coverage of the rice field plot is 0.85 as the opacity, and the optimized Gaussian map parameters and opacity are mixed in descending order to generate a Gaussian map using the transparency mixing formula to generate a Gaussian map with a resolution of 10 meters. The high-probability crop area with a Gaussian probability value of 0.65 corresponds to a dense rice plot, which accounts for 75%.

[0024] A first water level area is obtained according to the Gaussian map and the remotely measured water level of the corresponding rice field, and evaporation is obtained based on the first water level area through weather data and crop cover data.

[0025] In the actual evaluation, based on the Gaussian map, the part of the image with a Gaussian probability value greater than 0.6 was taken as the crop characteristic plot. Crop coverage data and telemetered water level were extracted according to the rice field atlas. The area belonging to the crop characteristic plot and with a telemetered water level 15% lower than the average water level was taken as the first water level area. Among them, the area with a telemetered water level 10.2 cm lower than the average water level during the tillering period accounted for 15%, or 150 mu. Based on the first water level area, the plot-level evaporation was obtained through the Penman-Montes evaporation model according to the weather data, crop coverage data and telemetered water level of the rice field plots. For example, the plot-level evaporation of an area during the tillering period was 3.2 mm per day, and 4.5 mm per day during the booting period. The plot-level evaporation was aggregated according to the plot weight to obtain the evaporation at the first water level area level.

[0026] The first water level area is corrected using the evaporation amount and radar water level data to obtain a second water level area, and the crop growth status is determined based on the second water level area using weather data and crop cover data.

[0027] In the actual assessment, radar water level data and plot-level evaporation were extracted from each point in the first water level area. Points with plot-level evaporation greater than 2 mm within 6 hours were marked based on the radar water level data. For example, in a first water level area, points with plot-level evaporation greater than 2 mm within 6 hours accounted for 20% of the first water level area, or 30 mu. The rice fields at the marked points were aggregated using the Gaussian map plot weights to obtain a second water level area, totaling 50 mu. Based on the radar water level data, weather data and crop cover data of the second water level area, the water stress index (WSI) of the rice field plots was determined using the water stress index model FAO-56. Plots with a water stress index greater than 70% were classified as healthy, plots with a water stress index less than 70% and greater than 40% were classified as mildly stressed, and plots with a water stress index less than 40% were classified as severely stressed.

[0028] The actual crop water content is determined based on crop growth status and soil entropy data. The radar water level data is corrected using the actual crop water content and evaporation through a Bayesian network to obtain the actual water level in the rice field.

[0029] In the actual evaluation, soil entropy data is extracted based on the second water level area, and the water stress index of the crop growth status is obtained according to the point coordinates of the soil entropy data. The product of the water stress index and the corresponding plot-level evaporation is used as the actual evaporation. The soil entropy data in the same time period is extracted based on the actual evaporation. The sum of the single-point soil volume water content change in the soil entropy data and the change in the irrigation data is added to the rainfall data minus the actual evaporation to obtain the actual crop water content. For example, within 6 hours, the soil volume water content increases by 1.5 mm, the irrigation amount increases by 5 mm, and light rain falls by 2 mm. The actual evaporation is 3 mm, and the actual crop water volume is 5.5 mm. Based on the actual crop water volume, the corresponding rice field plot is used as a plot particle. The radar water level data is obtained from the plot particle as time series data. The water level data of 11 cm at the current moment of the plot particle is used as the observation node. The actual crop water volume, actual evaporation, irrigation data, rainfall data and time series data at the previous moment are used as state nodes according to the observation node. A Bayesian network is obtained based on the observation nodes and state nodes. The actual water level of the rice field is 12.3 cm using the Bayesian network and the rice field water level correction formula.

[0030] In this embodiment, the preprocessing method includes: Multispectral images and infrared spectral data of rice fields are collected by drones, and the multispectral images are converted to grayscale and median filtered to obtain image data; telemetry information is collected, and the missing data of the telemetry information is supplemented by time series interpolation method, abnormal data is eliminated, and the data format is unified to obtain telemetry data; temperature, humidity, rainfall, wind speed, and air pressure data are collected and timestamps are aligned to obtain weather data; radar water level data are collected, and the high-frequency interference noise in the radar water level data is removed using the wavelet noise reduction algorithm; single-point soil volume moisture content and irrigation data are collected to obtain soil entropy data.

[0031] In this embodiment, the method for obtaining the rice field atlas includes: Superpixel segmentation is performed based on image data to divide the rice field image into small areas with similar colors and textures. The small areas are used as initial markers, and gradient calculation is performed on the rice field image to obtain a gradient image. According to the gradient image and the initial mark, the watershed algorithm is used to merge the internal area of ​​the rice field with low gradient values ​​and separate the boundary area with high gradient values. The contours of the separated areas are extracted and converted into polygonal vector data format to obtain polygonal rice field vector areas. Based on the polygonal rice field vector area, the vegetation index is extracted through the reflectance of the infrared band. The crop coverage of the rice field is calculated based on the vegetation index. The crop coverage is combined with the geographic coordinates to obtain the rice field area. The crop density is obtained based on the crop coverage and rice field area. The crop height and remote sensing water level are extracted from the rice field telemetry data. The crop height, crop density, and crop coverage are used as crop cover data. The rice field plots are used as nodes, the rice field area, telemetered water level, crop cover data, and geographic coordinates are used as node features, and the spatial adjacency relationship between adjacent rice fields and the connection relationship between irrigation and drainage networks are used as edges. Graph database modeling is performed based on nodes, node features, and edges to obtain a rice field map that contains spatial topological relationships and attribute associations.

[0032] In this embodiment, the Gaussian initialization and optimization method includes: Based on the node and spatial adjacency relationships of the rice field map, GIS topological analysis was used to extract the rice field area. The crop height, crop density, and crop coverage of the rice field plots were extracted based on the rice field area, and the mean and variance were calculated. The crop height, crop density, and crop coverage were Z-score standardized using the mean and variance to obtain a three-dimensional feature vector. A three-dimensional covariance matrix is ​​constructed based on the variance of each feature in the three-dimensional feature vector. The crop height and crop density in the three-dimensional feature vector are mapped to the vertical axis and horizontal axis of the two-dimensional space respectively. The two-dimensional covariance matrix is ​​obtained based on 0.2 times the crop density on the vertical axis and 0.1 times the crop height on the horizontal axis in the two-dimensional space. The ratio of the area of ​​the rice field plot to the total area of ​​the rice field region is used as the plot weight, and the two-dimensional mean, two-dimensional covariance matrix and plot weight are used as two-dimensional Gaussian components. The optimized two-dimensional Gaussian components are obtained by using the Gaussian parameter optimization formula. The Gaussian parameter optimization formula is: in is the optimized two-dimensional Gaussian component, including the mean , covariance and weights , is the number of observation data points, which is obtained by the number of pixels after rasterization of the rice field area. is the number of rice fields, is the plot index, is the plot index, For the The weight of the plot, is the mean of the two-dimensional features of the plot, is the covariance matrix, Adjacent plots obtained from the rice field map.

[0033] In this embodiment, the method for obtaining the Gaussian map includes: Update the Gaussian map parameters based on the optimized two-dimensional Gaussian distribution through the Gaussian map parameter optimization formula; The paddy fields are sorted in descending order according to the paddy area. The crop coverage of the paddy fields is used as the opacity. The Gaussian map parameters and opacity are optimized in descending order through the transparency blending formula to generate the Gaussian map. The transparency blending formula is: in For the pixels within the rice field area The Gaussian probability value of is the rice field area index, is the pixel coordinate in two-dimensional space, For the The opacity of each rice field plot, For the The weight of the rice field plots, For the The two-dimensional Gaussian distribution after optimization of the rice field plots, is the number of rice fields, Indicates before Treated rice fields.

[0034] In this embodiment, the method for optimizing the parameters of the Gaussian map includes: The Gaussian map parameter optimization formula is: in, The optimized Gaussian map parameters in the current field of view are minimized Photometric error, iterative adjustment , so that the generated image Better match observation images , Adam optimizer is used to efficiently solve the problem and update Gaussian parameters. is the field of view index, To predict the image, is the optimized two-dimensional Gaussian component in the current field of view, It is a weighted transparent combination.

[0035] In this embodiment, the method for obtaining the evaporation amount includes: Based on the Gaussian map, the parts of the image with Gaussian probability values ​​greater than 0.6 are taken as crop characteristic plots. Crop coverage data and telemetered water levels are extracted according to the rice field atlas. The areas belonging to the crop characteristic plots and with telemetered water levels 15% lower than the average water level are taken as the first water level areas. Based on the first water level areas, the Penman-Montes evaporation model is used to obtain the plot-level evaporation according to the weather data, crop coverage data and telemetered water levels of the rice field plots. The regional evaporation is then obtained by aggregating the plot-level evaporation using the plot weights.

[0036] In this embodiment, the method for determining the crop growth status includes: Based on the first water level area, radar water level data and plot-level evaporation data were extracted for each point. Points with plot-level evaporation greater than 2 mm within 6 hours were marked based on the radar water level data. The rice fields at the marked points were aggregated using the Gaussian map plot weights to obtain the second water level area. Based on the radar water level data, weather data and crop cover data of the second water level area, the water stress index of the rice field plots was determined by the water stress index model. The plots with a water stress index greater than 70% were classified as healthy, the plots with a water stress index less than 70% and greater than 40% were classified as mildly stressed, and the plots with a water stress index less than 40% were classified as severely stressed.

[0037] In this embodiment, the method for correcting radar water level data using the actual crop water content and evaporation through a Bayesian network includes: Extract soil entropy data based on the second water level area, and obtain the point coordinates of the soil entropy data The water stress index of crop growth status uses the product of the water stress index and the corresponding plot-level evaporation as the actual evaporation. Based on the actual evaporation, the soil entropy data within the same time period is extracted. The actual crop water content is obtained by adding the sum of the single-point soil volume water content change in the soil entropy data and the change in the irrigation data plus the rainfall data minus the actual evaporation. Based on the actual crop water volume, the corresponding rice field plot is used as a plot particle. The radar water level data is obtained from the plot particle as time series data. The time series data of the plot particle at the current moment is used as the observation node. The actual crop water volume, actual evaporation, irrigation data, rainfall data and time series data of the previous moment are used as state nodes according to the observation node. A Bayesian network is obtained based on the observation node and the state node. The Bayesian network is used to obtain the actual water level of the rice field through the rice field water level correction formula. The rice field water level correction formula is: in is the corrected radar water level data, is the total number of particles in the plot, For the moment Radar water level data, for Moment The predicted water level value of each particle, The standard deviation of the radar measurement error, For the The irrigation amount of each particle, 0 means no irrigation, obtained from the rice field atlas, For the The evaporation of particles, For the The actual amount of water for each particle of crop, For the The rainfall amount of each particle, 0 means no rainfall, obtained from weather data.

[0038] The second aspect of the present invention further provides a multi-scale acquisition system for paddy field water level, comprising: Data acquisition module: used to collect rice field data and radar water level data for preprocessing. The rice field data includes image data, telemetry data, weather data, and soil entropy data; Paddy field atlas module: used to divide paddy fields into regions based on the image data, extract paddy field area, telemetry water level and crop coverage data according to the paddy field regions, and extract relationships between paddy fields to obtain paddy field atlas; Gaussian map module: used to use the rice field atlas to perform Gaussian initialization and optimization through crop height, crop density, crop paddy field coverage and paddy field area of ​​crop cover data to obtain a Gaussian map; A first water level area module is configured to obtain a first water level area based on the Gaussian map and the telemetered water level of the corresponding rice field, and obtain evaporation based on the first water level area using weather data and crop cover data; A second water level area module is configured to use the evaporation rate and radar water level data to correct the first water level area to obtain a second water level area, and determine the crop growth status based on the second water level area using weather data and crop cover data; Actual water level acquisition module: used to determine the actual water volume of crops based on crop growth status and soil entropy data, and use the actual water volume and evaporation of crops to correct the radar water level data through the Bayesian network to obtain the actual water level of the rice field.

[0039] The above content is merely an example and explanation of the structure of the present invention. Those skilled in the art may make various modifications or additions to the described specific embodiments or replace them in a similar manner. As long as they do not deviate from the structure of the invention or exceed the scope defined by the claims, they should all fall within the scope of protection of the present invention.

Claims

1. A multi-scale acquisition method for paddy field water level, characterized in that: The following steps are involved: Collecting rice field data and radar water level data for preprocessing, wherein the rice field data includes image data, telemetry data, weather data, and soil entropy data; Divide the paddy field into regions based on the image data, extract paddy field area, remote sensing water level and crop coverage data according to the paddy field regions, and extract relationships between the paddy fields to obtain a paddy field map; Using the rice field atlas to perform Gaussian initialization and optimization through crop height, crop density, crop paddy field coverage and paddy field area of ​​crop cover data to obtain a Gaussian map; Obtaining a first water level area according to the Gaussian map and the remotely measured water level of the corresponding rice field, and obtaining evaporation based on the first water level area using weather data and crop cover data; Correcting the first water level area using the evaporation amount and radar water level data to obtain a second water level area, and determining the crop growth status based on the second water level area using weather data and crop cover data; The actual crop water content is determined based on crop growth status and soil entropy data. The radar water level data is corrected using the actual crop water content and evaporation through a Bayesian network to obtain the actual water level in the rice field.

2. A multi-scale acquisition method for paddy field water level according to claim 1, characterized in that: The pretreatment method comprises: Multispectral images and infrared spectral data of rice fields are collected by drones, and the multispectral images are converted to grayscale and median filtered to obtain image data; telemetry information is collected, and the missing data of the telemetry information is supplemented by time series interpolation method, abnormal data is eliminated, and the data format is unified to obtain telemetry data; temperature, humidity, rainfall, wind speed, and air pressure data are collected and timestamps are aligned to obtain weather data; radar water level data are collected, and the high-frequency interference noise in the radar water level data is removed using the wavelet noise reduction algorithm; single-point soil volume moisture content and irrigation data are collected to obtain soil entropy data.

3. The multi-scale acquisition method for paddy field water level according to claim 1, characterized in that: The method for obtaining the rice field atlas comprises: Superpixel segmentation is performed based on image data to divide the rice field image into small areas with similar colors and textures. The small areas are used as initial markers, and gradient calculation is performed on the rice field image to obtain a gradient image. According to the gradient image and the initial mark, the watershed algorithm is used to merge the internal area of ​​the rice field with low gradient values ​​and separate the boundary area with high gradient values. The contours of the separated areas are extracted and converted into polygonal vector data format to obtain polygonal rice field vector areas. Based on the polygonal rice field vector area, the vegetation index is extracted through the reflectance of the infrared band. The crop coverage of the rice field is calculated based on the vegetation index. The crop coverage is combined with the geographic coordinates to obtain the rice field area. The crop density is obtained based on the crop coverage and rice field area. The crop height and remote sensing water level are extracted from the rice field telemetry data. The crop height, crop density, and crop coverage are used as crop cover data. The rice field plots are used as nodes, the rice field area, telemetered water level, crop cover data, and geographic coordinates are used as node features, and the spatial adjacency relationship between adjacent rice fields and the connection relationship between irrigation and drainage networks are used as edges. Graph database modeling is performed based on nodes, node features, and edges to obtain a rice field map that contains spatial topological relationships and attribute associations.

4. The multi-scale acquisition method for paddy field water level according to claim 1, characterized in that: The Gaussian initialization and optimization method comprises: Based on the node and spatial adjacency relationships of the rice field map, GIS topological analysis was used to extract the rice field area. The crop height, crop density, and crop coverage of the rice field plots were extracted based on the rice field area, and the mean and variance were calculated. The crop height, crop density, and crop coverage were Z-score standardized using the mean and variance to obtain a three-dimensional feature vector. A three-dimensional covariance matrix is ​​constructed based on the variance of each feature in the three-dimensional feature vector. The crop height and crop density in the three-dimensional feature vector are mapped to the vertical axis and horizontal axis of the two-dimensional space respectively. The two-dimensional covariance matrix is ​​obtained based on 0.2 times the crop density on the vertical axis and 0.1 times the crop height on the horizontal axis in the two-dimensional space. The ratio of the area of ​​the rice field plot to the total area of ​​the rice field region is used as the plot weight, and the two-dimensional mean, two-dimensional covariance matrix and plot weight are used as two-dimensional Gaussian components. The optimized two-dimensional Gaussian components are obtained by using the Gaussian parameter optimization formula. The Gaussian parameter optimization formula is: in is the optimized two-dimensional Gaussian component, including the mean , covariance and weights , is the number of observation data points, which is obtained by the number of pixels after rasterization of the rice field area. is the number of rice fields, is the plot index, is the plot index, For the The weight of the plot, is the mean of the two-dimensional features of the plot, is the covariance matrix, Adjacent plots obtained from the rice field map.

5. The multi-scale acquisition method for paddy field water level according to claim 1, characterized in that: The method for obtaining the Gaussian map comprises: Update the Gaussian map parameters based on the optimized two-dimensional Gaussian distribution through the Gaussian map parameter optimization formula; The paddy fields are sorted in descending order according to the paddy area. The crop coverage of the paddy fields is used as the opacity. The Gaussian map parameters and opacity are optimized in descending order through the transparency blending formula to generate the Gaussian map. The transparency blending formula is: in For the pixels within the rice field area The Gaussian probability value of is the rice field area index, is the pixel coordinate in two-dimensional space, For the The opacity of each rice field plot, For the The weight of the rice field plots, For the The two-dimensional Gaussian distribution after optimization of the rice field plots, is the number of rice fields, Indicates before Treated rice fields.

6. A multi-scale acquisition method for paddy field water level according to claim 5, characterized in that: The method for optimizing the parameters of the Gaussian map comprises: The Gaussian map parameter optimization formula is: in, The optimized Gaussian map parameters in the current field of view are minimized Photometric error, iterative adjustment , so that the generated image Better match observation images , Adam optimizer is used to efficiently solve the problem and update Gaussian parameters. is the field of view index, To predict the image, is the optimized two-dimensional Gaussian component in the current field of view, It is a weighted transparent combination.

7. The multi-scale acquisition method for paddy field water level according to claim 1, characterized in that: The method for obtaining the evaporation amount comprises: Based on the Gaussian map, the parts of the image with Gaussian probability values ​​greater than 0.6 are taken as crop characteristic plots. Crop coverage data and telemetered water levels are extracted according to the rice field atlas. The areas belonging to the crop characteristic plots and with telemetered water levels 15% lower than the average water level are taken as the first water level areas. Based on the first water level areas, the Penman-Montes evaporation model is used to obtain the plot-level evaporation according to the weather data, crop coverage data and telemetered water levels of the rice field plots. The regional evaporation is then obtained by aggregating the plot-level evaporation using the plot weights.

8. The multi-scale acquisition method for paddy field water level according to claim 1, characterized in that: The method for determining the growth status of the crop comprises: Based on the first water level area, radar water level data and plot-level evaporation data were extracted for each point. Points with plot-level evaporation greater than 2 mm within 6 hours were marked based on the radar water level data. The rice fields at the marked points were aggregated using the Gaussian map plot weights to obtain the second water level area. Based on the radar water level data, weather data and crop cover data of the second water level area, the water stress index of the rice field plots was determined by the water stress index model. The plots with a water stress index greater than 70% were classified as healthy, the plots with a water stress index less than 70% and greater than 40% were classified as mildly stressed, and the plots with a water stress index less than 40% were classified as severely stressed.

9. The multi-scale acquisition method for paddy field water level according to claim 1, characterized in that: The method for correcting radar water level data using the actual crop water content and evaporation through a Bayesian network comprises: Extract soil entropy data based on the second water level area, and obtain the point coordinates of the soil entropy data The water stress index of crop growth status uses the product of the water stress index and the corresponding plot-level evaporation as the actual evaporation. Based on the actual evaporation, the soil entropy data within the same time period is extracted. The actual crop water content is obtained by adding the sum of the single-point soil volume water content change in the soil entropy data and the change in the irrigation data plus the rainfall data minus the actual evaporation. Based on the actual crop water volume, the corresponding rice field plot is used as a plot particle. The radar water level data is obtained from the plot particle as time series data. The time series data of the plot particle at the current moment is used as the observation node. The actual crop water volume, actual evaporation, irrigation data, rainfall data and time series data of the previous moment are used as state nodes according to the observation node. A Bayesian network is obtained based on the observation node and the state node. The Bayesian network is used to obtain the actual water level of the rice field through the rice field water level correction formula. The rice field water level correction formula is: in is the corrected radar water level data, is the total number of particles in the plot, For the moment Radar water level data, for Moment The predicted water level of each particle, The standard deviation of the radar measurement error, For the The irrigation amount of each particle, 0 means no irrigation, obtained from the rice field atlas, For the The evaporation of particles, For the The actual amount of water for each particle of crop, For the The rainfall amount of each particle, 0 means no rainfall, obtained from weather data.

10. A multi-scale acquisition system for paddy field water level, used to execute the multi-scale acquisition method for paddy field water level according to any one of claims 1 to 9, characterized in that: The system comprises: Data acquisition module: used to collect rice field data and radar water level data for preprocessing. The rice field data includes image data, telemetry data, weather data, and soil entropy data; Paddy field atlas module: used to divide paddy fields into regions based on the image data, extract paddy field area, telemetry water level and crop coverage data according to the paddy field regions, and extract relationships between paddy fields to obtain paddy field atlas; Gaussian map module: used to use the rice field atlas to perform Gaussian initialization and optimization through crop height, crop density, crop paddy field coverage and paddy field area of ​​crop cover data to obtain a Gaussian map; A first water level area module is configured to obtain a first water level area based on the Gaussian map and the telemetered water level of the corresponding rice field, and obtain evaporation based on the first water level area using weather data and crop cover data; A second water level area module is configured to use the evaporation rate and radar water level data to correct the first water level area to obtain a second water level area, and determine the crop growth status based on the second water level area using weather data and crop cover data; Actual water level acquisition module: used to determine the actual water volume of crops based on crop growth status and soil entropy data, and use the actual water volume and evaporation of crops to correct the radar water level data through the Bayesian network to obtain the actual water level of the rice field.

Images