Late rice extraction method, system, terminal and medium based on geographical zoning

Abstract

This application provides a method, system, terminal, and medium for extracting late-season rice based on geographical partitioning. The late-season rice extraction method includes: acquiring a geographical dataset of the target area and performing clustering to obtain several geographical partitions; acquiring historical time-series vegetation growth data sets corresponding to each geographical partition based on historical time-series remote sensing data, to train an initial classifier model and obtain corresponding late-season rice extraction models; acquiring current-year time-series remote sensing data of the target area; acquiring current-year time-series vegetation growth data sets corresponding to each geographical partition based on the current-year time-series remote sensing data; and extracting the corresponding late-season rice planting areas for each partition using the late-season rice extraction models to obtain the total late-season rice planting area. By dividing the target area into geographical partitions with consistent late-season rice phenological patterns and training models for each geographical partition, the accuracy of late-season rice planting area identification and extraction is effectively improved.

Description

Technical Field

[0001] This application belongs to the field of remote sensing monitoring and relates to a rice planting monitoring technology, particularly a method, system, terminal, and medium for extracting late rice based on geographical zoning. Background Technology

[0002] Obtaining the area of ​​late-season rice cultivation is of great significance. Currently, the method for obtaining the area of ​​late-season rice cultivation areas is usually based on satellite data identification, with the total cultivated area derived from the identified areas. Generally, deep learning models can be used to capture the phenological patterns of late-season rice growth, identifying and extracting late-season rice cultivation areas based on input time-series remote sensing image data. However, due to the inconsistency in phenology of late-season rice in different regions, the planting time and growth status may vary, making it difficult to capture the different phenological patterns of late-season rice in different regions using a single deep learning model. Especially when identifying and extracting from large-scale areas, the phenological patterns of late-season rice within these areas vary significantly, leading to greater difficulty in identifying late-season rice cultivation areas and lower identification and extraction accuracy.

[0003] Therefore, how to improve the accuracy of extracting the planting area of ​​late rice is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0004] The purpose of this application is to provide a method, system, terminal, and medium for extracting late-season rice based on geographical zoning, in order to solve the problem that the existing technology uses a single deep learning model for identifying late-season rice planting areas, which is difficult to adapt to the different phenological patterns of late-season rice in different regions, thus leading to a lack of accuracy in the identification and extraction of late-season rice.

[0005] In a first aspect, this application provides a method for extracting late-season rice based on geographical partitioning, comprising: acquiring a geographical dataset of a target area; performing clustering and partitioning based on the geographical dataset to obtain at least one geographical partition; wherein the geographical dataset includes a topographic data group and a climate data group; acquiring a historical time-series vegetation growth data group corresponding to each geographical partition based on historical time-series remote sensing data; training an initial classifier model based on the historical time-series vegetation growth data group to obtain each late-season rice extraction model corresponding to each geographical partition; acquiring current-year time-series remote sensing data of the target area; acquiring current-year time-series vegetation growth data group corresponding to each geographical partition based on the current-year time-series remote sensing data; extracting the partitioned late-season rice planting area corresponding to each geographical partition through each late-season rice extraction model based on the current-year partitioned time-series vegetation growth data group, and obtaining the total late-season rice planting area based on the partitioned late-season rice planting area.

[0006] In one embodiment of this application, obtaining a geographic dataset of a target area and performing clustering and partitioning based on the geographic dataset to obtain at least one geographic partition includes: dividing the target area into multiple grids of a preset size; obtaining geographic features corresponding to each grid based on the geographic dataset; performing comprehensive similarity analysis based on the geographic features of each grid to obtain several partition clusters; and obtaining each geographic partition based on each partition cluster.

[0007] In one embodiment of this application, the step of performing comprehensive similarity analysis based on the geographical features of each grid to obtain several partition clusters includes: calculating the geographical difference value between any two grids based on the geographical features of each grid; obtaining each parameter group based on a preset set of difference thresholds and a set of minimum grid numbers; obtaining each initial cluster group based on each geographical difference value and each parameter group; each parameter group includes a difference threshold and a minimum grid number; calculating the corresponding average profile coefficient based on each initial cluster group, and obtaining the initial cluster group with the largest average profile coefficient as the optimal cluster group, and outputting each initial cluster in the optimal cluster group as each partition cluster.

[0008] In one embodiment of this application, for any parameter group, the difference threshold corresponding to the parameter group is used as the current difference threshold, and the corresponding minimum number of grids is used as the current minimum number of grids; the method for obtaining the initial cluster group corresponding to the parameter group includes: dividing each grid into grid groups based on each geographical difference value and the current difference threshold; wherein, the geographical difference value of any two grids in the same grid group is not greater than the current difference threshold; obtaining the number of grids in each grid group, and if the number of grids is not less than the current minimum number of grids, taking the corresponding grid group as an initial cluster; obtaining the initial cluster group based on each initial cluster.

[0009] In one embodiment of this application, for any two grids, the weighted sum of squares of the corresponding parameter differences in the corresponding geographic features is obtained, and the square root of the weighted sum of squares is used as the geographic difference value of the two grids.

[0010] In one embodiment of this application, the terrain data set includes elevation data, the climate data set includes daily average sunshine data, daily average temperature data, and total precipitation data for each year; the geographic dataset also includes a historical data set; the historical data set includes planting area data for each year.

[0011] In one embodiment of this application, the method of obtaining the time-series vegetation growth data set corresponding to each geographic region based on the time-series remote sensing data of each geographic region is the same as the method of obtaining the time-series vegetation growth data set of each geographic region based on the time-series remote sensing data of the current year, including: cropping each image in the time-series remote sensing data of the current year based on each geographic region to obtain the time-series remote sensing data of each geographic region corresponding to each geographic region; extracting the corresponding time-series spectral raster data of the current year based on the time-series remote sensing data of each geographic region, and calculating the corresponding time-series NDVI raster data of the current year based on the time-series spectral raster data of each geographic region; and using the time-series NDVI raster data of the current year corresponding to each geographic region as the time-series vegetation growth data set of the current year corresponding to each geographic region.

[0012] Secondly, this application provides a late-season rice extraction system based on geographical partitioning, including a partitioning module, a training module, and an extraction module. The partitioning module is used to acquire a geographical dataset of a target area, and perform clustering partitioning based on the geographical dataset to obtain at least one geographical partition. The training module is used to acquire historical time-series data corresponding to each geographical partition based on historical time-series remote sensing data, and train an initial classifier model based on the historical time-series data of each geographical partition to obtain late-season rice extraction models corresponding to each geographical partition. The extraction module is used to acquire current-year time-series remote sensing data of the target area; acquire current-year time-series data of each geographical partition based on the current-year time-series remote sensing data; and extract the planting area corresponding to each geographical partition using the late-season rice extraction models based on the current-year time-series data of each geographical partition, and obtain the total planting area based on the planting area of ​​each partition.

[0013] Thirdly, this application provides a terminal, including: a processor and a memory, wherein the memory and the processor are communicatively connected;

[0014] The memory is used to store computer programs, and the processor is used to execute the computer programs stored in the memory to enable the terminal to perform the geographically partitioned late rice extraction method as described above.

[0015] Fourthly, this application provides a computer storage medium storing a computer program that, when executed by a processor, implements the aforementioned method for extracting late-season rice based on geographic partitioning.

[0016] As described above, this application provides a method, system, terminal, and medium for extracting late-season rice based on geographical partitioning. By dividing the target area into various geographical partitions, the phenological patterns of late-season rice within each geographical partition remain consistent. The late-season rice extraction model trained for each geographical partition has high recognition accuracy for the corresponding geographical partition, thereby improving the recognition accuracy of late-season rice in each geographical partition and further improving the recognition accuracy of the total late-season rice planting area within the target area. It has high industrial application value. Attached Figure Description

[0017] Figure 1 The diagram shown is a flowchart illustrating a late-season rice extraction method according to an embodiment of this application.

[0018] Figure 2 The diagram shown is a flowchart illustrating a method for obtaining geographical partitions as described in an embodiment of this application.

[0019] Figure 3 The diagram shown is a flowchart illustrating a comprehensive similarity analysis as described in an embodiment of this application.

[0020] Figure 4 The diagram shown is a flowchart illustrating an initial cluster group acquisition method as described in an embodiment of this application.

[0021] Figure 5 The diagram shown is a schematic diagram of a classifier model training sample as described in an embodiment of this application.

[0022] Figure 6 The diagram shown is a flowchart illustrating a method for acquiring partitioned time-series vegetation growth data groups as described in an embodiment of this application.

[0023] Figure 7 The diagram shown is a structural schematic of a late-season rice extraction system according to an embodiment of this application.

[0024] Figure 8 The diagram shown is a structural schematic of a terminal as described in an embodiment of this application.

[0025] Explanation of reference numerals in the attached figures

[0026] 41: Partitioning module; 42: Training module; 43: Extraction module; 50: Terminal; 51: Processor; 52: Memory; 521: Operating system; 522: Application program; 53: User interface; 54: Network interface; 55: Bus system. Detailed Implementation

[0027] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, unless otherwise specified, the following embodiments and features in the embodiments can be combined with each other.

[0028] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this application. Therefore, the drawings only show the components related to this application and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0029] Existing methods for identifying late-season rice planting areas often rely on deep learning models to capture and learn the phenological patterns of late-season rice, enabling them to identify and extract planting areas from input time-series remote sensing image data. However, due to the significant variations in phenological patterns of late-season rice across different regions, single deep learning models often struggle to adapt to the identification of large-scale late-season rice planting areas, resulting in low accuracy in late-season rice planting area identification and extraction.

[0030] To address the technical problems existing in the prior art, the following embodiments of this application provide a method, system, terminal, and medium for extracting late-season rice based on geographical partitioning. By using a geographical dataset of the target area, the target area is divided into various geographical partitions, thereby ensuring that the phenological patterns of late-season rice within each geographical partition remain consistent. A corresponding late-season rice extraction model is trained for each geographical partition, so that each late-season rice extraction model has high accuracy in identifying the late-season rice planting area within the corresponding geographical partition, and the accuracy of the extracted late-season rice planting area corresponding to each geographical partition is high, thereby effectively improving the accuracy of late-season rice identification in the target area.

[0031] The following embodiments of this application provide a method, system, terminal, and medium for extracting late-season rice based on geographic zoning, including but not limited to extraction of planting area and identification of main distribution location for late-season rice. The following description will take the identification and extraction of planting areas for late-season rice as an example.

[0032] It should be noted that the geographically partitioned late rice extraction method, system, terminal and medium provided in the following embodiments of this application can also be used to identify and extract the planting areas of other crops, including but not limited to early rice, corn or wheat, etc., which are not specifically limited here.

[0033] The technical solutions in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0034] like Figure 1 As shown, this embodiment provides a method for extracting late-season rice based on geographical zoning, including:

[0035] S100: Obtain the geographic dataset of the target region, and perform clustering and partitioning based on the geographic dataset to obtain at least one geographic partition.

[0036] The geographic dataset includes topographic and climatic data sets. Topography and climate reflect the late-season rice planting conditions at different geographical locations within the target area. For example, in higher-altitude, colder locations, late-season rice is sown earlier, grows more slowly, and has a longer overall growth cycle. In lower-altitude, warmer locations, sowing can be later, growth is faster, and the overall growth cycle is shorter. Crop rotation with other crops is common, resulting in a more complex planting system. Based on this, the geographic dataset reflects the late-season rice planting conditions at different geographical locations within the target area, thereby reflecting the phenological characteristics of late-season rice in each location. The target area is then divided into zones based on these geographic datasets, ensuring consistency in the phenological patterns of late-season rice within each zone. This facilitates model learning and capture of the phenological patterns of late-season rice, improving the model's recognition and extraction accuracy.

[0037] Furthermore, each geographical location within the target area is used to represent the smallest unit for analyzing the geographical environment of late-season rice. The geographical environment remains consistent within each unit, and units with similar geographical environments are grouped together to form a geographical partition. Since areas with similar geographical environments typically exhibit similar phenological patterns of late-season rice, the similarity in phenological patterns across different geographical partitions allows the model to be trained based on these partitions. This helps the model capture more accurate phenological patterns of late-season rice, thereby improving the model's recognition and extraction accuracy. For example, the target area is divided into several grids of a preset size to represent each geographical location; for instance, the target area is divided into 1km x 1km grids, with each grid serving as a geographical location.

[0038] For example, the topographic data set includes elevation data, such as a DEM (Digital Elevation Model), to characterize the altitude of each geographical location within the target area; or it includes latitude and longitude coordinate data to characterize the spatial location information of each geographical location within the target area; or it includes slope data to characterize the topographic relief of each geographical location within the target area. The climate data set includes daily average sunshine data, daily average temperature data, and total precipitation data to characterize the climate conditions of each geographical location within the target area. Specifically, daily average sunshine data refers to the average daily sunshine amount for each geographical location during the late rice growing season; daily average temperature data refers to the average daily temperature for each geographical location during the late rice growing season; and total precipitation refers to the average annual total precipitation for each geographical location during the late rice growing season.

[0039] In some alternative implementations, such as Figure 2 As shown, the methods for obtaining each geographic region include S110-S130:

[0040] S110 divides the target area into multiple grids of preset size and obtains the geographic features corresponding to each grid based on the geographic dataset.

[0041] In this context, geographic features are used to characterize the late-season rice planting conditions within the corresponding grid. For example, the geographic dataset includes a topographic data set and a climate data set, and the geographic features include the topographic and climatic conditions within the corresponding grid.

[0042] Each grid maintains consistent geographic features within its corresponding geographical area; therefore, each grid is used as the smallest unit for geographic feature analysis. For example, the target area is divided into several grids of size 1km x 1km.

[0043] To facilitate understanding of the process of acquiring geographic features corresponding to each grid, this paper takes a geographic dataset that includes a terrain data group and a climate data group. The terrain data group includes DEM data, and the climate data group includes daily average sunshine data, daily average temperature data, and total precipitation data as an example to illustrate the specific process of acquiring geographic features.

[0044] Specifically, for any given grid, based on DEM data, the average elevation value within the corresponding actual geographic area is obtained to reflect the grid's elevation; based on slope data, the average slope value within the corresponding actual geographic area is obtained to reflect the grid's topographic relief; based on daily average sunshine data, the total daily sunshine within the corresponding actual geographic area is obtained, which is used as daily average temperature data to obtain the daily average temperature within the corresponding actual geographic area; and based on monthly precipitation data, the total precipitation within the corresponding actual geographic area during the late rice growing season is obtained. The set of average elevation, average slope, total sunshine, average temperature, and total precipitation values ​​is taken as the geographic feature corresponding to that grid.

[0045] S120 performs a comprehensive similarity analysis based on the geographical characteristics of each grid to obtain several partition clusters.

[0046] Specifically, the geographical similarity of each grid is obtained based on geographical features, so that grids with similar afternoon patterns of late rice can be clustered together to facilitate the formation of geographical zones.

[0047] For example, based on the geographical features of each grid, the OPTICS clustering method (Ordering Points To Identify the Clustering Structure, a density-based clustering algorithm) is used to calculate the degree of similarity of geographical features between grids to form clusters for each partition.

[0048] In some alternative implementations, such as Figure 3 As shown, a comprehensive similarity analysis is performed on each grid to obtain the clusters for each partition, including S121-S123:

[0049] S121, based on the geographical characteristics of each grid, calculate the geographical difference value between any two grids.

[0050] The geographic difference value is used to characterize the degree of geographic similarity between two corresponding grids. The larger the geographic difference value, the lower the geographic similarity between the two grids; the smaller the geographic difference value, the higher the geographic similarity between the two grids.

[0051] For example, the weighted sum of squares of the differences of corresponding parameters in the geographic features of any two grids is obtained, and the square root of the weighted sum of squares is used as the geographic difference value between the two grids.

[0052] To facilitate understanding, we will take the set of average height, average slope, total sunshine, average temperature, and total precipitation of each grid as an example to illustrate the specific calculation method for the geographical difference between any two grids.

[0053] Specifically, if two grids are grids and grid :

[0054]

[0055] in, For grid and grid Geographical differences between them; For grid The average height value; For grid The average height value; For example, with a preset height weight, It is 0.1; For grid The average slope value; For grid The average slope value; For example, a preset slope weight is used. It is 0.15; For grid Total amount of light; For grid Total amount of light; For example, the preset illumination weights, It is 0.25; For grid The average temperature value; For grid The average temperature value; For example, with a preset temperature weight, It is 0.25; For grid The total amount of precipitation; For grid The total amount of precipitation; For example, based on the preset precipitation weights, The weight is set to 0.25. It should be noted that those skilled in the art can set the values ​​of height weight, light weight, temperature weight, and precipitation weight according to actual conditions; this embodiment does not impose specific limitations. Furthermore, the average height value, total light intensity, average temperature value, and total precipitation value of each grid are normalized using the Z-score method to avoid the difference in values ​​between different features affecting the accuracy of the final result. Specifically, taking the average height value as an example, the mean and standard deviation of the average height value of each grid in the entire target area are obtained, and the weight corresponding to each grid is set to... ,in, This represents the normalized average height value for each grid cell. This represents the initial average height value for each grid cell. This is the average height value of each grid cell. is the standard deviation of the average height value of each grid.

[0056] S122: Based on the preset set of difference thresholds and the set of minimum grid numbers, obtain each parameter group; based on each geographical difference value, combine each parameter group to obtain the corresponding initial cluster group.

[0057] The preset set of difference thresholds includes various preset difference thresholds; the preset set of minimum grid numbers includes various preset minimum grid numbers. The difference threshold is used to characterize the maximum geographical difference value between any two grids in the same initial cluster; the minimum grid number is used to characterize the minimum number of grids in any initial cluster.

[0058] Each parameter group includes a difference threshold and a minimum number of grids. For example, the difference thresholds in the difference threshold set and the minimum number of grids in the minimum number of grids set are combined to form each parameter group.

[0059] The parameter set is used as the dividing standard for each grid to cluster them. It should be noted that the difference threshold is used to ensure that the grids in the same initial cluster have a high degree of similarity in geographical environment and consistent phenological patterns of late rice. The minimum number of grids is used to avoid a few grids being classified into a cluster due to accidental factors, which would generate a lot of meaningless small clusters, thereby reducing the number of clusters and improving the efficiency of subsequent processing.

[0060] Based on this, for each parameter group, the corresponding initial clusters are obtained and each initial cluster group is formed, so as to facilitate the selection of the optimal clustering method and the acquisition of the optimal cluster group.

[0061] In some alternative implementations, such as Figure 4 As shown, for any parameter set, the method for obtaining its corresponding initial cluster group includes S1221-S1222:

[0062] S1221, based on each geographical difference value and combined with the current difference threshold, divide each grid into grid groups.

[0063] Specifically, the difference threshold in the current parameter group is used as the current difference threshold.

[0064] The geographic difference between any two grids in the same grid group is not greater than the current difference threshold. Specifically, if the geographic difference between any two grids is not greater than the current difference threshold, the two grids are assigned to the same grid group, and the search continues to find other grids whose geographic difference between them is not greater than the current difference threshold, until any grid outside the grid group has a geographic difference between it and at least one grid in the grid group that is greater than the current difference threshold.

[0065] S1222, obtain the number of grids in each grid group. If the number of grids is not less than the current minimum number of grids, take the corresponding grid group as an initial cluster.

[0066] Specifically, a grid group is considered as an initial cluster only when the number of grids in the grid group is greater than or equal to the current minimum number of grids. This is to avoid a small number of grids being assigned to a grid group by chance, forming meaningless small clusters, which would result in an excessive number of clusters and affect the efficiency of subsequent processing.

[0067] Furthermore, the set of each initial cluster obtained is output as the initial cluster group corresponding to the current parameter group.

[0068] S123, calculate the corresponding average profile coefficient based on each initial cluster group, and obtain the initial cluster group with the largest average profile coefficient, so as to output each initial cluster group in the initial cluster group as the partition cluster group.

[0069] The mean silhouette coefficient is used to measure the clustering effect of each initial cluster in the corresponding initial cluster group. The larger the mean silhouette coefficient, the better the clustering effect of each initial cluster in the initial cluster group; the smaller the mean silhouette coefficient, the worse the clustering effect of each initial cluster in the initial cluster group.

[0070] Based on this, the initial cluster group with the largest average contour coefficient is the optimal cluster group with the best clustering effect. By using each initial cluster group in the optimal cluster group as the output of each partition cluster, the optimal clustering effect of each partition cluster can be obtained, thereby effectively improving the effect of geographical partitioning, which is conducive to improving the extraction effect of late rice based on each geographical partition, and thus improving the accuracy of the obtained late rice planting area.

[0071] Furthermore, the average profile coefficient corresponding to each initial cluster group is the average of the profile coefficients corresponding to each grid. Here, the profile coefficient is a normalized measure of the intra-cluster cohesion and extra-cluster separation of the corresponding grid, specifically:

[0072]

[0073]

[0074]

[0075] in, This represents the geographical difference between the two grids. For grid Intra-cluster cohesion; For grid The initial cluster in which it belongs; For the initial cluster Total number of grids in the system; Characterizing the mesh For the initial cluster In addition to the grid Other grids outside; For grid extra-cluster separation; The representation takes the minimum value; Characterize the initial cluster The initial cluster group excluding the initial cluster Other initial clusters; For the initial cluster Total number of grids in the system; Characterizing the mesh For the initial cluster The grid in the middle; For grid The profile coefficient; The maximum value is represented.

[0076] Based on this, the contour coefficients of all grids are obtained, and their average value is used as the average contour coefficient to characterize the clustering effect of the corresponding initial cluster group, thereby selecting the optimal cluster group to obtain the clusters of each partition.

[0077] S130: Obtain each geographic partition based on the clusters of each partition.

[0078] Specifically, the actual geographic spatial range corresponding to each partition cluster is extracted as the actual geographic spatial range of each geographic partition, thereby obtaining each geographic partition.

[0079] In some alternative implementations, to further improve the accuracy of subsequent late-season rice identification and extraction, historical late-season rice planting conditions are also considered when dividing the target area into zones. It should be noted that areas that have historically planted late-season rice are more likely to plant rice in the current year. Therefore, when training within each geographical zone, since each geographical zone incorporates historical late-season rice planting conditions, the model is trained based on these conditions, which helps improve the identification accuracy of the late-season rice extraction model for each geographical zone, thereby increasing the accuracy of the obtained late-season rice planting area.

[0080] Specifically, the geographic dataset also includes a historical data set to characterize historical late-season rice planting conditions. For example, the historical data set includes planting area data for each year, which refers to the late-season rice planting conditions for each geographic location, i.e., each grid. For instance, the historical planting area data includes the late-season rice planting conditions for the two years preceding the current year. For a given grid, if no late-season rice was planted the year before last, and the area planted with late-season rice within its corresponding geographic area last year accounted for 80%, then the historical planting conditions for that grid in the historical planting area data would be (0, 80%).

[0081] Furthermore, in this embodiment, the geographic dataset also includes a historical data group. In step S110, the geographic features corresponding to each grid acquired also include historical planting information. For ease of understanding, let's take a geographic dataset that includes a terrain data group and a climate data group. The terrain data group includes DEM data, the climate data group includes daily average sunshine data, daily average temperature data, and total precipitation data, and the historical data group includes planting area data from previous years, which includes the late rice planting situation of the previous two years. For any grid, based on the DEM data, the average height value within the actual geographic range corresponding to the grid is obtained; based on the daily average sunshine data, the total daily sunshine within the actual geographic range corresponding to the grid is obtained; based on the daily average temperature data, the daily average temperature within the actual geographic range corresponding to the grid is obtained; based on the total precipitation data, the total precipitation within the actual geographic range corresponding to the grid during the late rice growing season is obtained; and based on the planting area data from previous years, the planting situation of the grid from previous years is obtained. The set of average height value, total sunshine, average temperature value, total precipitation, and planting situation from previous years is used as the geographic feature corresponding to the grid. For the specific methods of obtaining the average height, average slope, total sunshine, average temperature, and total precipitation, please refer to the aforementioned step S110. This embodiment will not repeat them here.

[0082] Furthermore, when the geographic dataset also includes historical data groups, in step S121, referring to the aforementioned method for obtaining geographic difference values, historical planting information is added to obtain new geographic difference values ​​between any two grids, specifically:

[0083]

[0084] in, The geographic dataset also includes geographic difference values ​​when historical data groups are included. For grid The percentage of the area where late rice was planted within its corresponding actual geographical range two years ago; For grid The percentage of the area where late rice was planted within its corresponding actual geographical range two years ago; For grid Last year, the area of ​​late rice planted within its corresponding actual geographical range accounted for the proportion of the total area. For grid Last year, the area of ​​late rice planted within its corresponding actual geographical range accounted for the proportion of the total area. For example, by pre-setting historical planting data weights, The value is 0.1. It should be noted that those skilled in the art can set the weight of historical planting conditions according to actual conditions, and this embodiment does not make specific limitations here.

[0085] Alternatively, different weights can be set for the proportion of late rice area in the year before last and the proportion of late rice area in the year before last, with the weight of last year being greater than that of the year before last.

[0086] Based on this, the target area is divided into geographical zones to ensure that the phenological patterns of late rice within each geographical zone remain consistent. This facilitates the training of corresponding late rice extraction models based on each geographical zone, thereby improving the accuracy of late rice planting area identification.

[0087] S200, based on historical time-series remote sensing data, obtains historical time-series vegetation growth data sets for each geographical region, trains the initial classifier model based on the historical time-series vegetation growth data sets for each geographical region, and obtains the extraction model for each late rice variety corresponding to each geographical region.

[0088] Among them, the historical time-series remote sensing data is used to represent the time-series remote sensing data of all years preceding the current year. Time-series remote sensing data refers to remote sensing images of the target area arranged in chronological order. For example, the historical time-series remote sensing data includes time-series remote sensing data of the 10 years preceding the current year, and each time-series remote sensing data includes remote sensing images of the target area from June to November, with at least one remote sensing image for each month.

[0089] The historical zonal temporal vegetation growth data set is based on historical temporal remote sensing data to obtain vegetation growth changes corresponding to each geographic zone. It should be noted that remote sensing imagery is acquired by satellites collecting the spectrum of light reflected from the Earth's surface. Therefore, remote sensing images can reflect the reflected light waves from the Earth's surface, and temporal remote sensing images over a period of time can obtain the spectral changes of reflected light. Since the reflected light spectrum of vegetation changes at different stages of growth, by obtaining the spectral changes within each pixel of the temporal remote sensing image, the corresponding vegetation changes can be analyzed. Combined with the growth patterns of late-season rice, the planting status of late-season rice can be obtained based on the zonal temporal vegetation growth data set. Note that the size of each pixel in the temporal remote sensing image is typically smaller than the size of a single grid cell.

[0090] For example, changes in NDVI (Normalized Difference Vegetation Index) values ​​are obtained based on time-series remote sensing imagery. NDVI values ​​characterize the differences in absorption and reflection of near-infrared and red light bands by the land surface. The near-infrared band is where the internal structure of plant leaves strongly reflects light, while the red band is where chlorophyll strongly absorbs light, reflecting changes in vegetation growth. Based on this, the planting status of late-season rice can be analyzed by whether changes in NDVI values ​​conform to the growth patterns of late-season rice. Optionally, changes in blue, green, red, and near-infrared spectral bands are also obtained based on time-series remote sensing imagery. These changes in light bands can then be used to analyze whether vegetation growth changes conform to the growth patterns of late-season rice, thus enabling the identification and extraction of late-season rice.

[0091] It should be noted that the planting status of other crops, including but not limited to early rice, corn, or wheat, can also be analyzed by checking whether the changes in NDVI values ​​or blue, green, red, and near-infrared spectral bands conform to the growth patterns of other crops. As long as the classifier model extracts the corresponding region based on the growth patterns of the corresponding crops, the extraction and identification of other crops can be achieved. This application does not make any specific limitations here.

[0092] In this embodiment, a classifier model is used to analyze the late-season rice planting situation to extract the late-season rice planting area. For example, the classifier model is an SVM (Support Vector Machine) classifier model. It should be noted that the initial classifier model is a deep learning model to be trained. Through training with a large number of training samples, the initial classification model is trained to capture and learn the spectral changes during the late-season rice planting process. Further validation samples are used to validate the model. Training stops when the accuracy of the initial classifier model reaches a preset accuracy threshold, and the trained model is used as the final classifier model. Specifically, those skilled in the art should know the specific execution steps and principles of model training with training samples and validation with validation samples; this embodiment does not impose specific limitations here. For example, the total number of training samples and validation samples is in the tens of millions, and the ratio of training samples to validation samples is: training samples : validation samples = 7 : 3. The training samples used for model training need to have the late-season rice planting area manually labeled, for example, such as... Figure 5 As shown, the images displayed represent the training samples for each time period. The area within the red box represents the planting area of ​​late rice, which facilitates model learning and enables the identification and extraction of late rice planting areas, thereby obtaining the late rice extraction model corresponding to each geographical region.

[0093] It should be noted that the samples used to train the initial classification model are historical time-series vegetation growth data sets for different geographic regions, obtained from historical time-series remote sensing data. Specifically, each image in the historical time-series remote sensing data is cropped to obtain historical time-series remote sensing data for each geographic region. Each image in the historical time-series remote sensing data for different geographic regions is a raster image to obtain the historical spectral information of each pixel. The spectral information reflects the changes in vegetation growth, thereby obtaining the historical time-series vegetation growth data sets for different geographic regions. The late-season rice planting area is marked on each image in the historical time-series vegetation growth data sets for use in training the initial classifier model.

[0094] To facilitate understanding by those skilled in the art of obtaining partitioned time-series vegetation generation data based on time-series remote sensing data in this application, the following will use the change of NDVI value to reflect vegetation growth as an example to illustrate the process of obtaining partitioned time-series vegetation growth data based on time-series remote sensing data.

[0095] For example, such as Figure 6 As shown, the methods for obtaining the time-series vegetation growth data sets by region include S201-S203:

[0096] S201, based on each geographic region, crop each image in the time-series remote sensing data to obtain the time-series remote sensing data corresponding to each geographic region.

[0097] Among them, the time-series remote sensing data of each region are used to characterize the time-series remote sensing situation within the corresponding geographic region.

[0098] Specifically, the spatial extent corresponding to each geographic region is obtained, and each image in the time-series remote sensing data is cropped. The cropped image blocks are then arranged in chronological order to serve as the time-series remote sensing data for each geographic region.

[0099] S202. Based on the time-series remote sensing data of each region, extract the corresponding time-series spectral raster data, and calculate the corresponding time-series NDVI raster data based on each time-series spectral raster data.

[0100] Temporal spectral raster data is used to characterize the spectral variations of each pixel. Specifically, each image patch in the temporal remote sensing data for each region is also a raster image patch. The reflectance of each band within each pixel is obtained as the spectral information of that pixel. Based on the spectral information of each pixel, spectral raster image patches are obtained for each image patch, and these spectral raster image patches are arranged temporally to obtain temporal spectral raster data. It should be noted that temporal arrangement refers to arranging the spectral raster image patches of the same geographic region in the same year from earliest to latest time to obtain the temporal spectral raster data corresponding to each year for each geographic region.

[0101] Furthermore, the size of a single pixel in a raster image is typically smaller than the size of a single grid cell. For example, if the size of a single grid cell is 1 km x 1 km, the size of a single pixel in a raster image is 2 m x 2 m.

[0102] Temporal NDVI raster data is used to characterize the variation of NDVI values ​​for each pixel. Specifically, based on any spectral raster image patch, the reflectance of each pixel in the near-infrared and red bands is obtained to calculate the NDVI value of each pixel. Based on the NDVI values ​​of each pixel, the corresponding NDVI raster image patch is obtained. NDVI raster image patches corresponding to each spectral raster image patch in the same geographic region and the same year are obtained and arranged from earliest to latest time to obtain the temporal NDVI raster data for each geographic region and each year.

[0103] For example, the NDVI value corresponding to any pixel is calculated as follows:

[0104]

[0105] in, This represents the reflectance of the current pixel in the near-infrared band. This represents the reflectivity of the current pixel in the red light band.

[0106] S203, take the time-series NDVI raster data corresponding to each geographic region as the time-series vegetation growth data group corresponding to each geographic region.

[0107] Based on this, the vegetation growth changes within each geographic region can be reflected by using the time-series NDVI raster data corresponding to each geographic region.

[0108] It should be noted that in step S200, when obtaining the historical time-series vegetation growth data set corresponding to each geographic region based on historical time-series remote sensing data, the time-series remote sensing data in steps S201 to S213 is historical time-series remote sensing data, and the obtained regional time-series vegetation growth data set is the historical regional time-series vegetation growth data set.

[0109] Based on this, historical zonal time-series vegetation growth data sets are obtained, and late-season rice planting areas are marked on each image of the historical zonal time-series vegetation growth data sets to train the initial classifier model. This allows the classifier model to capture and learn the growth patterns of late-season rice, thereby improving the accuracy of the classifier model in extracting and recognizing late-season rice. Those skilled in the art should understand the specific execution methods and principles of training the initial classifier model based on historical zonal time-series vegetation growth data sets; this embodiment does not impose specific limitations on these methods.

[0110] It should be noted that, due to the different phenological patterns of late-season rice in different geographical regions, in order to improve the accuracy of model recognition, for each geographical region, an initial classifier model is trained based on the corresponding historical time-series vegetation growth data to obtain a corresponding late-season rice extraction model. Based on this, in the subsequent identification and extraction of late-season rice planting areas, the data corresponding to each geographical region is identified and extracted based on the late-season rice extraction model corresponding to each geographical region. Since the geographical environment within each geographical region is the same, the phenological patterns of late-season rice within the same geographical region are relatively consistent, resulting in higher accuracy of the trained late-season rice extraction model and thus effectively improving the accuracy of late-season rice identification and extraction.

[0111] S300: Obtain the current year's time-series remote sensing data of the target area; based on the current year's time-series remote sensing data, obtain the current year's time-series vegetation growth data sets for each geographical region; based on the current year's time-series vegetation growth data sets for each geographical region, extract the corresponding regional late rice planting areas for each geographical region through each late rice extraction model, and obtain the total late rice planting area based on the late rice planting areas for each region.

[0112] Among them, the current year's time-series remote sensing data is used to characterize the time-series remote sensing data for the current year, that is, the time-series remote sensing data corresponding to the year in which the late rice planting area needs to be extracted. For example, the current year's time-series remote sensing data is obtained through satellite acquisition. The zonal time-series vegetation growth data group is used to characterize the vegetation growth changes corresponding to each geographic zone obtained based on the current year's time-series remote sensing data.

[0113] Based on the temporal vegetation growth data sets corresponding to each geographical region, the late-season rice planting areas for each geographical region are extracted using the late-season rice extraction model corresponding to each geographical region. The set of late-season rice planting areas for each geographical region is then used as the total late-season rice planting area. Based on this, the extraction and identification of late-season rice planting areas within the target region can be achieved.

[0114] It should be noted that for the specific methods and principles of obtaining the time-series vegetation growth data sets for each geographic region based on the time-series remote sensing data of the current year, please refer to steps S201 to S213. Specifically, when obtaining the time-series vegetation growth data sets for each geographic region, the time-series remote sensing data in steps S201 to S213 is the time-series remote sensing data of the current year, and the obtained time-series vegetation growth data sets for each region are the time-series vegetation growth data sets for each region of the current year.

[0115] like Figure 7 As shown, this embodiment provides a late-season rice extraction system, which includes a partitioning module 41, a training module 42, and an extraction module 43.

[0116] The partitioning module 41 is used to obtain a geographic dataset of the target area and perform clustering partitioning based on the geographic dataset to obtain at least one geographic partition; wherein the geographic dataset includes a terrain data group and a climate data group.

[0117] Training module 42 is used to obtain the time-series vegetation growth data set corresponding to each of the geographical regions based on the time-series remote sensing data of each year, and to train the initial classifier model based on the time-series vegetation growth data set of each of the geographical regions to obtain the late rice extraction model corresponding to each of the geographical regions.

[0118] The extraction module 43 is used to obtain the current year's time-series remote sensing data of the target area; based on the current year's time-series remote sensing data, obtain the current year's time-series vegetation growth data group corresponding to each of the geographical partitions; based on the current year's time-series vegetation growth data group, extract the partitioned late rice planting area corresponding to each of the geographical partitions through each of the late rice extraction models, and obtain the total late rice planting area based on the partitioned late rice planting area.

[0119] Based on the same technical concept, the late rice extraction method provided in this embodiment of the invention can be implemented on the terminal side or the server side.

[0120] like Figure 8 The diagram illustrates an optional hardware structure of a terminal according to an embodiment of the present invention. The terminal 50 can be a mobile phone, computer device, tablet device, personal digital processing device, factory back-end processing device, etc. The terminal 50 includes at least one processor 51, a memory 52, at least one network interface 54, and a user interface 53. The various components in the system are coupled together through a bus system 55. It is understood that the bus system 55 is used to realize the connection and communication between these components. In addition to a data bus, the bus system 55 also includes a power bus, a control bus, and a status signal bus.

[0121] The user interface 53 may include a monitor, keyboard, mouse, trackball, clicker, button, touchpad, or touch screen.

[0122] It is understood that memory 52 can be volatile memory or non-volatile memory, or both. Non-volatile memory can be read-only memory (ROM) or programmable read-only memory (PROM), used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM) and synchronous static random access memory (SSRAM). The memory characterized in the embodiments of the present invention is intended to include, but is not limited to, these and any other suitable categories of memory.

[0123] In this embodiment of the invention, the memory 52 is used to store various types of data to support the operation of the terminal. Examples of this data include: any executable program for operation on the terminal 50, such as the operating system 521 and application programs 522; the operating system 521 contains various system programs, such as the framework layer, core library layer, driver layer, etc., for implementing various basic services and handling hardware-based tasks. Application programs 522 may contain various applications, such as media players, browsers, etc., for implementing various application services. The method for extracting late-season rice provided in this embodiment of the invention can be included in application program 522.

[0124] The methods disclosed in the above embodiments of the present invention can be applied to processor 51, or implemented by processor 51. Processor 51 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in processor 51 or by instructions in the form of software. The processor mentioned above may be a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Processor 51 can implement or execute the methods, steps and logic block diagrams disclosed in the embodiments of the present invention. Processor 51 may be a microprocessor or any conventional processor, etc. The steps of the accessory optimization method provided in the embodiments of the present invention can be directly reflected as being executed by a hardware decoding processor, or being executed by a combination of hardware and software modules in the decoding processor. The software module may be located in a storage medium, which is located in a memory. The processor reads the information in the memory and combines it with its hardware to complete the steps of the aforementioned method.

[0125] In an exemplary embodiment, terminal 50 may be used by one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), or complex programmable logic devices (CPLDs) to execute the aforementioned method.

[0126] This invention also provides a computer-readable storage medium storing a computer program that, when invoked by a processor, implements the late-season rice extraction method provided by this invention.

[0127] Computer-readable storage media can be tangible devices capable of holding and storing instructions used by an instruction execution device. Computer-readable storage media can be, for example, (but not limited to) electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, and mechanical encoding devices.

[0128] The computer-readable program represented herein can be downloaded from a computer-readable storage medium to various computing / processing devices, or downloaded via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network, to an external computer or external storage device. A network adapter card or network interface in each computing / processing device receives computer-readable program instructions from the network and forwards these instructions to the computer-readable storage medium in the respective computing / processing device.

[0129] In summary, this application divides the target area into geographical regions and trains separate late-season rice extraction models for each geographical region. This avoids the problem of low accuracy in late-season rice extraction models due to different phenological patterns of late-season rice growth in different regions, thereby effectively improving the recognition accuracy of the late-season rice extraction model and thus improving the accuracy of the extracted late-season rice planting areas.

[0130] The descriptions of the processes or structures corresponding to the above figures each have their own emphasis. For parts of a process or structure that are not described in detail, please refer to the relevant descriptions of other processes or structures.

[0131] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in this application should still be covered by the claims of this application.

Claims

1. A method for extracting late-season rice based on geographical zoning, characterized in that, include: A geographic dataset of a target region is obtained, and clustering and partitioning are performed based on the geographic dataset to obtain at least one geographic partition. The geographic dataset includes a topographic data group and a climate data group; it also includes a historical data group, which includes planting area data from previous years. The process of obtaining the geographic dataset of the target region and performing clustering and partitioning to obtain at least one geographic partition includes: dividing the target region into multiple grids of a preset size; obtaining geographic features corresponding to each grid based on the geographic dataset; performing a comprehensive similarity analysis based on the geographic features of each grid; for any two grids, obtaining the weighted sum of squares of the differences between corresponding parameters in the geographic features, and using the square root of the weighted sum of squares as the geographic difference value between the two grids, whereby the geographic difference value characterizes the degree of geographic similarity between the two corresponding grids; obtaining the geographic similarity of each grid based on the geographic features, so that grids with similar late-rice phenological patterns form clusters based on the similarity, thereby obtaining several partition clusters; and obtaining each geographic partition based on each partition cluster. Based on historical time-series remote sensing data, historical time-series vegetation growth data sets corresponding to each of the aforementioned geographical regions are obtained. The initial classifier model is trained based on the historical time-series vegetation growth data sets corresponding to each of the aforementioned geographical regions to obtain the late rice extraction model corresponding to each of the aforementioned geographical regions. Obtain the current year's time-series remote sensing data of the target area; based on the current year's time-series remote sensing data, obtain the current year's time-series vegetation growth data group corresponding to each of the geographical partitions; based on the current year's time-series vegetation growth data group, extract the partitioned late rice planting area corresponding to each of the geographical partitions through each of the late rice extraction models, and obtain the total late rice planting area based on each of the partitioned late rice planting areas. Specifically, the temporal NDVI raster data corresponding to each geographic region is used as the temporal vegetation growth data for each geographic region; the temporal NDVI raster data corresponding to each geographic region reflects the vegetation growth changes within each geographic region.

2. The method according to claim 1, characterized in that, The comprehensive similarity analysis based on the geographical features of each grid is used to obtain several partition clusters, including: Based on the geographical characteristics of each grid, calculate the geographical difference value between any two grids; Based on a preset set of difference thresholds and a set of minimum grid numbers, each parameter group is obtained; based on each of the geographic difference values ​​and in combination with each of the parameter groups, each of the corresponding initial cluster groups is obtained; each of the parameter groups includes a difference threshold and a minimum grid number. The average profile coefficient is calculated based on each initial cluster group, and the initial cluster group with the largest average profile coefficient is selected as the optimal cluster group. Each initial cluster group in the optimal cluster group is then output as the partition cluster group.

3. The method according to claim 2, characterized in that, For any of the parameter sets, the difference threshold corresponding to the parameter set is used as the current difference threshold, and the corresponding minimum number of grid cells is used as the current minimum number of grid cells; the method for obtaining the initial cluster group corresponding to the parameter set includes: Based on the aforementioned geographical difference values ​​and in conjunction with the current difference threshold, each of the grids is divided into grid groups; wherein, the geographical difference value between any two grids in the same grid group is not greater than the current difference threshold; Obtain the number of grids in each grid group. If the number of grids is not less than the current minimum number of grids, take the corresponding grid group as an initial cluster. Based on each initial cluster, obtain the initial cluster group.

4. The method according to claim 1, characterized in that, The topographic data set includes elevation data, and the climate data set includes daily average sunshine data, daily average temperature data, and total precipitation data for each year.

5. The method according to claim 1, characterized in that, The method for obtaining the historical time-series vegetation growth data set corresponding to each of the geographic regions based on historical time-series remote sensing data is the same as the method for obtaining the current year's time-series vegetation growth data set based on the current year's time-series remote sensing data for each of the geographic regions, including: Based on each of the geographical partitions, each image in the current year’s time-series remote sensing data is cropped to obtain the current year’s time-series remote sensing data corresponding to each of the geographical partitions. Based on the time-series remote sensing data of each of the current year's partitions, the corresponding time-series spectral raster data of the current year is extracted, and based on the time-series spectral raster data of each of the current year's partitions, the corresponding time-series NDVI raster data of the current year is calculated. The current year's temporal NDVI raster data corresponding to each of the aforementioned geographical partitions are used as the current year's temporal vegetation growth data group corresponding to each of the aforementioned geographical partitions.

6. A late-season rice extraction system based on geographical zoning, characterized in that, It includes a partitioning module, a training module, and an extraction module; The partitioning module is used to acquire a geographic dataset of a target area, and perform clustering partitioning based on the geographic dataset to obtain at least one geographic partition. The geographic dataset includes a terrain data group and a climate data group; it also includes a historical data group, which includes planting area data from previous years. Acquiring the geographic dataset of the target area and performing clustering partitioning based on it to obtain at least one geographic partition includes: dividing the target area into multiple grids of a preset size; acquiring the geographic features corresponding to each grid based on the geographic dataset; performing a comprehensive similarity analysis based on the geographic features of each grid; for any two grids, acquiring the weighted sum of squares of the differences between corresponding parameters in the geographic features, and using the square root of the weighted sum of squares as the geographic difference value between the two grids, whereby the geographic difference value characterizes the degree of geographic similarity between the two corresponding grids; acquiring the geographic similarity of each grid based on the geographic features, so that grids with similar late-rice phenological patterns form clusters based on the similarity, thereby obtaining several partition clusters; and acquiring each geographic partition based on each partition cluster. The training module is used to obtain the time-series vegetation growth data set of each geographical region based on the time-series remote sensing data of each year, and to train the initial classifier model based on the time-series vegetation growth data set of each geographical region to obtain the late rice extraction model of each geographical region. The extraction module is used to acquire the current year's time-series remote sensing data of the target area; based on the current year's time-series remote sensing data, acquire the current year's time-series vegetation growth data group corresponding to each of the geographical partitions; based on the current year's time-series vegetation growth data group, extract the partitioned late rice planting area corresponding to each of the geographical partitions through each of the late rice extraction models, and obtain the total late rice planting area based on the partitioned late rice planting area. Specifically, the temporal NDVI raster data corresponding to each geographic region is used as the temporal vegetation growth data for each geographic region; the temporal NDVI raster data corresponding to each geographic region reflects the vegetation growth changes within each geographic region.

7. A terminal, characterized in that, include: A processor and a memory, wherein the memory and the processor are communicatively connected; The memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory to enable the terminal to perform the geographically partitioned late rice extraction method as described in any one of claims 1 to 5.

8. A computer storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the late-season rice extraction method based on geographic partitioning as described in any one of claims 1 to 5.

Images