Method and apparatus for extracting rice planting areas based on multi-temporal SAR images

Abstract

This application discloses a method and apparatus for extracting rice planting areas based on multi-temporal SAR imagery. The method includes: acquiring multi-temporal SAR images covering the rice growth cycle and processing each temporal SAR image to obtain scattering and interferometric features; selecting classification feature data based on the scattering and interferometric features, and performing band synthesis based on the classification feature data; selecting feature data for false-color synthesis from the synthesized bands, and performing false-color synthesis based on the feature data to obtain a false-color synthesized image; selecting ground feature samples within the area to be measured based on the false-color synthesized image, and classifying the ground feature samples using a preset classification method to obtain an initial rice planting area identification result; and optimizing the initial rice planting area identification result to obtain an optimized rice planting area identification result. This application solves the technical problem in the prior art where rice planting area extraction does not meet practical needs.

Description

Technical Field

[0001] This application relates to the field of remote sensing image processing technology, and in particular to a method and apparatus for extracting rice planting areas based on multi-temporal SAR images. Background Technology

[0002] Rice is one of my country's most important food crops, holding a vital position in strategic material reserves. Therefore, obtaining information on regional rice planting areas is of great significance for national food security management. Traditional rice identification methods typically rely on collecting ground data and statistical methods, which are time-consuming, labor-intensive, and yield inaccurate results when monitoring large areas. Remote sensing technology, due to its short cycle and wide coverage, is widely used in crop monitoring. In recent years, optical remote sensing has become an important tool for rice identification and monitoring. However, most of my country's rice-growing areas are located in the southeastern region, where frequent rainy weather and high cloud cover make it difficult for optical remote sensing satellites to obtain effective images of rice distribution. Synthetic Aperture Radar (SAR) is an active remote sensing measurement system. Its emitted microwave signals have strong penetrating power through clouds and light rain, enabling all-day and all-weather monitoring. Furthermore, due to the high dielectric constant of rice, it is easier to distinguish rice from other ground features compared to other symbiotic plants. With the continuous improvement of the performance of spaceborne SAR systems, rice identification using SAR remote sensing technology has become an important technical means.

[0003] Currently, domestic and international scholars primarily extract rice information from SAR data based on SAR data from different time phases and with different polarization modes. Identification methods mainly include thresholding, integration with rice growth models, and fusion with multi-source data such as optical images. For example, Ningbo University (e.g., patent document CN114387516B[P]. 2022-08-16) proposed a SAR identification method for single-season rice in small and medium-sized fields under complex terrain. This method involves averaging and time-series filtering of the backscatter intensity time series of SAR data for the entire year in the study area, and using a decision tree algorithm model to identify single-season rice planting areas. The Jiangsu Provincial Institute of Water Resources Science, in conjunction with the Jiangsu Provincial Institute of Surveying and Mapping Engineering (e.g., patent document CN114821362B[P]. 2022-09-23), combined multi-source remote sensing image data and data from the Third National Land Area Survey, and extracted rice distribution and planting area by setting thresholds and spatial cross-operations. The Deqing Institute of Satellite Applications of the Chinese Academy of Sciences, in collaboration with the Institute of Remote Sensing and Digital Earth of the Chinese Academy of Sciences (e.g., CN108766203B[P]. 2020-10-16), converted fully polarimetric SAR data into compact polarimetric SAR data. Based on the attribute characteristics of the collected sample fields, sample data was constructed, and rice mapping results were obtained using the support vector machine classification method.

[0004] However, there are still some problems with the current use of SAR remote sensing technology to extract rice area: (1) Current rice extraction research is mainly based on the intensity information of SAR to analyze its scattering characteristics, without fully exploring the unique phase information of SAR; (2) The fusion application with optical data can improve the extraction accuracy of rice, but in actual applications, in some cloudy and rainy areas, it is difficult to obtain effective optical images during the critical growth period of rice, which limits the fusion application of SAR and optical images in rice extraction; (3) Compared with single and dual polarization data, fully polarized data can provide more complete ground feature information and has certain advantages in rice extraction, but there are still few SAR satellites with fully polarized acquisition methods, and the cost of fully polarized commercial SAR images is also higher than that of single and dual polarization data, which is still some distance from engineering application. Summary of the Invention

[0005] The technical problem solved by this application is that the existing technology does not meet the actual needs for extracting rice planting areas. This application provides a method and apparatus for extracting rice planting areas based on multi-temporal SAR images. In the solution provided by the embodiments of this application, the rice planting area is extracted based on multi-temporal SAR images, utilizing the backscattering characteristics of multi-temporal SAR images and combining interferometric coherence. This explores the application potential of the unique phase information of SAR images in rice extraction, and takes into account both the requirements of engineering application and extraction accuracy.

[0006] In a first aspect, embodiments of this application provide a method for extracting rice planting areas based on multi-temporal SAR imagery. The method includes: acquiring multi-temporal SAR images covering the rice growth cycle; performing multi-temporal processing on each temporal SAR image to obtain scattering features and interferometric processing to obtain interferometric features; selecting classification feature data based on the scattering features and interferometric features, and performing band synthesis based on the classification feature data; selecting feature data from the synthesized bands for false-color synthesis, and performing false-color synthesis based on the feature data to obtain a false-color synthesized image; selecting ground feature samples within the area to be measured based on the false-color synthesized image, classifying the ground feature samples using a preset classification method to obtain an initial rice planting area identification result; optimizing the initial rice planting area identification result to obtain an optimized rice planting area identification result, and calculating the area of ​​the rice planting area based on the optimized rice planting area identification result.

[0007] Optionally, multi-temporal processing is performed on each temporal SAR image to obtain scattering features and interferometric processing to obtain interferometric features, including: registering, multi-temporal speckle filtering, geocoding, and radiometric calibration of the SAR images to obtain the scattering features, wherein the scattering features include time-series backscattering coefficients, shadows, and overlay mask data; baseline estimation, interferogram generation, adaptive filtering, and coherence calculation are performed on each temporal SAR image to obtain the interferometric features, wherein the interferometric features include interferometric coherence coefficients.

[0008] Optionally, the scattering features are obtained by registering, multi-temporal speckle filtering, geocoding, and radiometric calibration of the various temporal SAR images, including: selecting a first SAR image from the multi-temporal SAR images as a reference, registering each SAR image other than the first SAR image based on the first SAR image to obtain a registered SAR image; removing speckle noise from the registered SAR image using a preset multi-temporal filtering method to obtain a filtered SAR image; acquiring data elevation model (DEM) data, converting the filtered SAR image from slant range or ground distance projection to geographic coordinate projection based on the DEM data, and acquiring the local incident angle to detect shadows and overlay areas to obtain the time-series backscattering coefficients, shadows, and overlay mask data.

[0009] Optionally, the classification feature data is selected based on the scattering characteristics and the interferometric characteristics, including: comparing the multi-temporal SAR image with a preset designated remote sensing image, selecting at least two ground object samples, and analyzing the changes in the time series backscattering coefficient and interferometric coherence coefficient of the at least two ground object samples; and selecting feature data with ground object differences greater than a specified condition as the classification feature data based on the changes.

[0010] Optionally, selecting ground feature samples within the area to be measured based on the false-color composite image includes: analyzing the color characteristics of ground features based on the false-color composite image, and selecting ground feature samples from the area to be measured on the preset designated remote sensing image based on the color characteristics.

[0011] Optionally, it further includes: calculating the JM distance between different land feature samples, and determining the land features whose JM distance meets a preset condition, wherein the preset condition is that the JM distance to all other land feature samples is greater than 1.9.

[0012] Optionally, the initial rice planting area identification result is optimized to obtain an optimized rice planting area identification result. The area of ​​the rice planting area is calculated based on the optimized rice planting area identification result, including: performing spatial filtering on the initial rice planting area identification result to remove small patches to obtain the optimized rice planting area identification result; counting the number of rice pixels and obtaining the area of ​​a single pixel in the SAR image from the optimized rice planting area identification result; and multiplying the number of rice pixels by the area of ​​a single pixel in the SAR image to obtain the area of ​​the rice planting area.

[0013] Secondly, embodiments of this application provide a rice planting area extraction device based on multi-temporal SAR imagery, the device comprising:

[0014] The data preprocessing unit is used to acquire multi-temporal SAR images covering the rice growth cycle, and to perform multi-temporal processing on each temporal SAR image to obtain scattering features and interferometric processing to obtain interferometric features.

[0015] The feature analysis unit is used to filter out classification feature data based on the scattering features and the interference features, and to perform band synthesis based on the classification feature data; and to select feature data for false color synthesis from the synthesized bands, and to perform false color synthesis based on the feature data to obtain a false color synthesized image.

[0016] The image classification unit is used to select land feature samples within the area to be measured based on the false-color synthetic image, classify the land feature samples using a preset classification method to obtain an initial rice planting area identification result; and optimize the initial rice planting area identification result to obtain an optimized rice planting area identification result, and calculate the area of ​​the rice planting area based on the optimized rice planting area identification result.

[0017] Thirdly, this application provides a computer device, the computer device comprising:

[0018] Memory, used to store at least one instruction executed by a processor;

[0019] A processor is configured to execute instructions stored in memory to perform the method described in the first aspect.

[0020] Fourthly, this application provides a computer-readable storage medium storing computer instructions that, when executed on a computer, cause the computer to perform the method described in the first aspect.

[0021] Compared with the prior art, the solution provided in this application has at least the following beneficial effects:

[0022] 1. The solution provided in this application embodiment is based on multi-temporal SAR images. It utilizes the backscattering characteristics of multi-temporal SAR images and combines interferometric coherence to extract rice planting area. This explores the application potential of the unique phase information of SAR images in rice extraction, taking into account both engineering applications and extraction accuracy requirements.

[0023] 2. In the solution provided in this application embodiment, the classification samples do not rely on field surveys. They are mainly based on the backscattering and coherence changes of different land features during the rice growth cycle, and samples are selected in combination with high-resolution Google Earth images, saving manpower and material resources.

[0024] 3. In the solution provided in this application embodiment, the shadow and overlapping areas are extracted based on the header information of DEM data and SAR data during the extraction process, thereby removing the influence of both on rice extraction and ensuring the accuracy of rice extraction. Attached Figure Description

[0025] Figure 1 A flowchart illustrating a method for extracting rice planting areas based on multi-temporal SAR imagery provided in this application embodiment;

[0026] Figure 2 A flowchart illustrating another method for extracting rice planting areas based on multi-temporal SAR images provided in this application embodiment;

[0027] Figure 3 A schematic diagram of a rice planting area extraction device based on multi-temporal SAR images provided in an embodiment of this application;

[0028] Figure 4 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0029] The embodiments described in this application are only a part of the embodiments, not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0030] To better understand the above technical solutions, the technical solutions of this application will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of this application and the specific features in the embodiments are detailed descriptions of the technical solutions of this application, rather than limitations on the technical solutions of this application. In the absence of conflict, the embodiments of this application and the technical features in the embodiments can be combined with each other.

[0031] The following description, in conjunction with the accompanying drawings, provides a more detailed explanation of a method for extracting rice planting areas based on multi-temporal SAR imagery, as provided in this application. The specific implementation of this method may include the following steps (method flow as follows): Figure 1 As shown):

[0032] Step 101: Acquire multi-temporal SAR images covering the rice growth cycle, and perform multi-temporal processing on each temporal SAR image to obtain scattering features and interferometric processing to obtain interferometric features.

[0033] As an example, C-band and cross-polarized SAR images are preferred for multi-temporal SAR imagery. Multi-temporal processing of each SAR image includes: registration, multi-temporal speckle filtering, geocoding, and radiometric calibration to obtain time-series backscattering coefficients, shadowing, and overlay masking data. Registration requires slant range geometry, selecting one SAR image as the reference and the remaining images as those to be registered. For example, the earliest captured image is selected as the reference image, and images from other time phases are used as reference images. Another example is acquiring Digital Elevation Model (DEM) data, specifically topographic product data obtained from the 30m resolution Shuttle Radar Topography Mission (SRTM). Based on the acquired DEM data, cross-correlation techniques are used to automatically register multiple radar images covering the same area, achieving sub-pixel registration accuracy. Multi-temporal speckle filtering is based on the registration results, using multi-temporal filtering methods to remove speckle noise from the SAR images. Optionally, a De Grandi filter is used to remove multi-temporal noise.

[0034] Furthermore, geocoding and radiometric calibration process the results obtained from the previous registration and filtering. Combined with the acquired DEM data, the SAR data of each time phase are converted from slant range or ground distance projection to geographic coordinate projection. At the same time, shadow and overlay areas are detected by combining the local incident angle to obtain the backscattering coefficient in dB, as well as shadow and overlay data.

[0035] Optionally, geocoding employs a method based on the Range-Doppler (RD) localization model. Shadow and overlay detection utilizes an indirect localization method based on the RD model to simulate SAR images, obtaining the local incidence angle. When generating SAR simulated images, the system can determine whether each DEM resolution cell belongs to an overlay or shadow area based on the conditions that generate overlays and shadows, thus generating corresponding mask images. Optionally, the aforementioned radiometric calibration is used to correct errors caused by the performance degradation of various internal components of the SAR, suppress noise in the image, and eliminate or reduce signal clutter caused by disturbances in the imaging environment.

[0036] Specifically, formula (1) is the ellipsoid equation that the ground feature point must satisfy, formula (2) is the distance relationship that the satellite position vector and the target vector must satisfy, and formula (3) is the Doppler frequency shift equation. The coherence parameters of the RD positioning model can be extracted from the SAR data header file obtained above.

[0037]

[0038] R 2 =(XX) S ) 2 +(YY S ) 2 +(ZZ S ) 2 (2)

[0039]

[0040] Where (X, Y, Z) represents the coordinates of a ground point in the geocentric inertial coordinate system; (X S Y S Z S ) represents the satellite's coordinates in the geocentric inertial coordinate system; h represents the elevation of the ground point relative to the reference ellipsoid; R a R b Let f be the minor and major semi-axis of the reference ellipsoid, respectively; R is the distance from the satellite to the ground point; D R is the Doppler center frequency corresponding to the ground point; λ is the radar wavelength; S and V S These are the satellite's position and velocity vectors, respectively; V S and V T These are the position and velocity vectors of the point on the ground, respectively.

[0041] In the solution provided in this application embodiment, the shadow and overlapping areas are extracted based on DEM data and SAR data header information during the extraction process, thereby removing the influence of both on rice extraction and ensuring the accuracy of rice extraction.

[0042] Optionally, the detection of shadow and overlay areas adopts the SAR image simulation based on the indirect positioning method of RD model to obtain the local incident angle. The calculation method is shown in formula (4). When generating SAR simulation images, it is possible to judge whether each DEM resolution unit belongs to the overlay or shadow area according to the conditions for generating overlay and shadow, so as to generate the corresponding mask image.

[0043]

[0044] In the above formula, η is the local incident angle, and R ts The radar incident direction vector, It is the normal vector of the surface where the current feature point is located.

[0045] Optionally, the above radiometric calibration is used to correct errors caused by the performance degradation of the internal components of the SAR, suppress noise in the image, and eliminate or reduce signal clutter caused by disturbances in the imaging environment. Equation (5) represents the pixel value (DN) and backscattering coefficient (σ) in the SAR image. 0 ), radar brightness (β) 0 The relationship between ).

[0046]

[0047] In the formula, K is the absolute calibration coefficient, which is determined by the polarization mode, data mode and incident angle; α is the incident angle of the incident wave relative to the flat ground.

[0048] The average backscattering coefficient, expressed in dB, is:

[0049] σ 0 [dB]=10lg σ 0 (6)

[0050] As another example, interferometric synthetic aperture radar (InSAR) is used to perform interferometric processing on various temporal SAR images to obtain interferometric features. InSAR interferometric processing includes baseline estimation, interferogram generation, adaptive filtering, and coherence calculation to obtain the interferometric coherence coefficient. Baseline estimation is used to evaluate the quality of the interferometric image pair. Optionally, the earliest acquired image is used as the master image, and the remaining images as slave images. The baseline length is calculated separately for each image, and the baseline length must be less than 1 / 3 of the critical baseline. Interferogram generation involves processing the master and slave images to obtain an interferogram of the two scenes after registration and multi-view processing, as well as intensity maps of the master and slave images. Adaptive filtering and coherence calculation involve filtering the flattened interferogram to remove phase noise caused by flat-ground interference, and simultaneously generating the interferometric coherence map.

[0051] Step 102: Based on the scattering features and the interference features, select classification feature data for false color synthesis, and perform band synthesis based on the feature data to obtain a false color composite image.

[0052] As an example, the multi-temporal SAR image is compared with a preset designated remote sensing image. At least two ground object samples are selected, and the changes in the time-series backscattering coefficient and interferometric coherence coefficient of the at least two ground object samples are analyzed. Based on the changes, feature data with ground object differences greater than a specified condition are selected as the classification feature data. For example, by comparing a preset designated remote sensing image (Google Earth high-resolution image) and SAR image, a small number of samples of different ground objects are selected, and the changes in the backscattering coefficient and coherence coefficient of different ground object samples are analyzed. Features with large ground object differences are selected as classification features for band synthesis, and three feature data (such as RGB feature data) are selected for false-color synthesis.

[0053] The scattering characteristics are analyzed as follows: As rice grows and develops, the scattering effect of radar beams on rice exhibits characteristics of water body information as the primary component, coexistence of water body and crop information, with crop information being the primary component. In the early stages of rice sowing and transplanting, when the paddy field is filled with water, the backscattering is mainly water-based, resulting in a low backscattering coefficient. As the rice continues to grow, the canopy and leaf layers increase, and backscattering gradually becomes dominated by the volume scattering of the rice itself and the canopy layer scattering, thus increasing the backscattering coefficient with a fluctuating trend. Until the rice reaches the heading stage, the backscattering coefficient of rice generally reaches its maximum value during the heading to milk stage. When the rice is in the waxy-ripe to fully ripe stage, the canopy and leaf layers continue to increase to a certain density, and the panicles are full, with the rice itself becoming the primary scatterer. However, the attenuation effect also increases, so the backscattering coefficient shows a slight decreasing trend and tends to stabilize. After the rice is harvested, only a small amount of stubble remains in the paddy field, the volume scattering weakens, and the backscattering coefficient begins to decrease again. In addition, the main echo mode of man-made structures is secondary scattering, and the backscattering coefficient has been at a high value for a long time; water bodies mainly have surface scattering, and the backscattering coefficient has been at a low value for a long time; the echo of bare land is mainly composed of diffuse scattering, and the backscattering coefficient has been in a relatively stable range for a long time; while other crops and vegetation have no obvious trend in backscattering coefficient due to their complex species. However, because water is involved in the growth of rice, the backscattering at some time phases will be different from that of other crops and vegetation.

[0054] Coherence analysis, including multiple coherence coefficient maps obtained through interferometric processing, analyzes the changes in coherence of various land features over time. For example, artificial structures show relatively small changes over time, exhibiting high correlation coefficients, while water surfaces show low interferometric coherence due to time variations and low signal-to-noise ratios. These two types are clearly distinguishable from other land features. The coherence coefficient of rice plants shows a wave-like trend, and in different test areas, combined with local planting structures, it differs from other vegetation at certain time phases.

[0055] Feature selection and band combination include selecting features with significant differences in land cover as classification feature data, and selecting three feature data (such as RGB feature data) for false color synthesis. Different land cover will show different colors, highlighting the area of ​​rice and other land cover, which facilitates the selection of samples later.

[0056] In the scheme provided in this application embodiment, the classification samples do not rely on field surveys. They are mainly based on the backscattering and coherence changes of different land features during the rice growth cycle, and samples are selected in combination with high-resolution Google Earth images, saving manpower and material resources.

[0057] Step 103: Based on the false-color synthetic image, select land cover samples within the area to be tested, and classify the land cover samples using a preset classification method to obtain the initial rice planting area identification result.

[0058] In the scheme provided in this application embodiment, sample points are selected by combining separability analysis, and land cover classification is performed by combining mask data and using supervised classification method to obtain rice planting areas. Among them, the sample selection is based on the above-mentioned false color synthetic image, analyzing the color characteristics of land cover including rice and combining Google Earth high-resolution image, selecting sample points of different land cover within the area to be tested, and calculating the Jeffries-Matusita (JM) distance separability between different land cover samples. The calculation formula is shown in the following formula (7), ensuring that the JM distance between rice and other land cover samples is above 1.9.

[0059]

[0060] In the formula, p(x / ω) i ) represents the conditional probability density, i.e., the probability density function of the i-th pixel belonging to ω. i The probability of the class, J ij This represents the separability distance between samples, ranging from 0 to 2. A larger value indicates a higher degree of separability between samples.

[0061] Random forest classification, which involves combining shadow and overlay data, removing the effects of shadow and overlay, and using the above training samples to classify land cover using the random forest classification algorithm, yields preliminary results for rice planting areas.

[0062] Step 104: Optimize the initial rice planting area identification results to obtain optimized rice planting area identification results, and calculate the area of ​​the rice planting area based on the optimized rice planting area identification results.

[0063] For example, see Figure 2 The results of rice planting area identification are post-processed and optimized to extract the results, and the rice planting area is calculated, including:

[0064] 1. Spatial filtering, including spatial filtering of classification results to remove small patches and isolated spots;

[0065] 2. Area statistics, including the number of rice pixels in the statistical classification results, multiplied by the area of ​​a single pixel in the SAR image to obtain the area of ​​the rice planting area.

[0066] The solution provided in this application embodiment is based on multi-temporal SAR images. It utilizes the backscattering characteristics of multi-temporal SAR images and combines interferometric coherence to extract rice planting area. This approach explores the application potential of the unique phase information of SAR images in rice extraction, taking into account both engineering applications and extraction accuracy requirements.

[0067] Furthermore, as a response to the above Figure 1 , Figure 2 The implementation of the method shown in this application provides a rice planting area extraction device based on multi-temporal SAR imagery. The aforementioned method is applied to this device. This device embodiment will not repeat the details of the aforementioned method embodiments, but it should be clear that the device in this embodiment can correspondingly implement all the contents of the aforementioned method embodiments. This device is applied to rice planting area extraction based on multi-temporal SAR imagery, specifically as follows... Figure 3 As shown, the device includes:

[0068] Data preprocessing unit 31 is used to acquire multi-temporal SAR images covering the rice growth cycle, and to perform multi-temporal processing on each temporal SAR image to obtain scattering features and interferometric processing to obtain interferometric features.

[0069] The feature analysis unit 32 is used to filter out classification feature data based on the scattering features and the interference features, and to perform band synthesis based on the classification feature data; and to select feature data for false color synthesis from the synthesized bands, and to perform false color synthesis based on the feature data to obtain a false color synthesized image.

[0070] The image classification unit 33 is used to select land feature samples within the area to be tested based on the false-color synthetic image, classify the land feature samples using a preset classification method to obtain an initial rice planting area identification result; and optimize the initial rice planting area identification result to obtain an optimized rice planting area identification result, and calculate the area of ​​the rice planting area based on the optimized rice planting area identification result.

[0071] Specifically, the data preprocessing unit 31 is used to perform multi-temporal processing and InSAR processing on SAR data to obtain time-series backscattering coefficients, shadow and overlay mask data, and interferometric coherence coefficients.

[0072] Feature analysis unit 32 is used to analyze time series backscattering characteristics and InSAR coherence, screen feature data and perform band synthesis;

[0073] Image classification unit 33 is used to select sample points, calculate sample separability, combine mask data, use supervised classification method to classify land features, obtain rice planting areas, post-process the rice planting area identification results to optimize the extraction results, and calculate the rice planting area.

[0074] Furthermore, the data preprocessing unit 31 includes:

[0075] The multi-temporal SAR data processing module 311 is used to perform registration, multi-temporal speckle filtering, geocoding and radiometric calibration on multi-temporal SAR data to obtain time series backscattering coefficients, shadow and overlay mask data.

[0076] InSAR interferometric processing module 312 is used for baseline estimation, interferogram generation, adaptive filtering and coherence calculation to obtain the interferometric coherence coefficient.

[0077] Furthermore, the feature analysis unit 32 includes:

[0078] The characteristic curve plotting module 321 is used to analyze the changes in the backscattering coefficient and coherence coefficient of different ground objects by comparing high-resolution Google Earth images and SAR images, selecting a small number of samples of different ground objects.

[0079] The band synthesis module 322 is used to select features with significant differences in ground features as classification features for band synthesis, and to select three feature data for false-color synthesis.

[0080] Furthermore, the image classification unit 33 includes:

[0081] The sample selection module 331 is used to select samples of different land features by comparing high-resolution Google Earth images and SAR images;

[0082] The sample separability calculation module 332 is used to calculate the JM distance separability between different land cover samples;

[0083] The supervised classification module 333 is used to classify land cover based on the above training samples and mask data using the random forest classification algorithm to obtain preliminary results of rice planting areas.

[0084] The classification post-processing module 334 performs post-processing optimization on the rice planting area identification results to extract the results and calculate the rice planting area.

[0085] Furthermore, the classification post-processing module 334 includes:

[0086] Spatial filtering submodule 3341 is used to perform spatial filtering on the classification results to remove small patches;

[0087] The data statistics submodule 3342 is used to count the number of rice pixels in the classification results and multiply it by the area of ​​a single pixel in the SAR image to obtain the area of ​​the rice planting area.

[0088] See Figure 4 This application provides a computer device, the computer device comprising:

[0089] Memory 401 is used to store at least one instruction executed by a processor;

[0090] Processor 402 is used to execute instructions stored in memory. Figure 1 The method described.

[0091] This application provides a computer-readable storage medium storing computer instructions that, when executed on a computer, cause the computer to perform... Figure 1 The method described.

[0092] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0093] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0094] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0095] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0096] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for extracting rice planting areas based on multi-temporal SAR imagery, characterized in that, include: Acquire multi-temporal SAR images covering the rice growth cycle, and perform multi-temporal processing on each temporal SAR image to obtain scattering features and interferometric processing to obtain interferometric features; Classification feature data is selected based on the scattering and interference features, and band synthesis is performed based on the classification feature data; feature data for false color synthesis is selected from the synthesized bands, and false color synthesis is performed based on the feature data to obtain a false color synthesized image; Based on the false-color synthetic image, land cover samples are selected within the area to be tested, and the land cover samples are classified using a preset classification method to obtain the initial rice planting area identification result. The initial rice planting area identification results are optimized to obtain optimized rice planting area identification results, and the area of ​​the rice planting area is calculated based on the optimized rice planting area identification results. Multi-temporal processing of SAR images from various time phases yields scattering features, and interferometric processing yields interferometric features, including: The scattering features are obtained by registering, multi-temporal speckle filtering, geocoding and radiometric calibration of the various temporal SAR images, wherein the scattering features include time-series backscattering coefficients, shadow and overlay mask data; The interferometric features are obtained by performing baseline estimation, interferogram generation, adaptive filtering, and coherence calculation on each of the temporal SAR images, wherein the interferometric features include the interferometric coherence coefficient; The scattering features are obtained by registering, multi-temporal speckle filtering, geocoding, and radiometric calibration of the various temporal SAR images, including: A first SAR image is selected from the multi-temporal SAR images as a reference, and each SAR image other than the first SAR image is registered based on the first SAR image to obtain a registered SAR image. A preset multi-temporal filtering method is used to remove speckle noise from the registered SAR image to obtain a filtered SAR image; Data Elevation Model (DEM) data is acquired, and the filtered SAR image is converted from slant range or ground distance projection to geographic coordinate projection based on the DEM data; and the local incident angle is acquired to detect shadows and overlay areas to obtain the time series backscattering coefficient, shadow and overlay mask data. Based on the scattering characteristics and the interference characteristics, classification feature data is selected, including: The multi-temporal SAR images are compared with preset designated remote sensing images. At least two ground object samples are selected, and the changes in the time series backscattering coefficient and interferometric coherence coefficient of at least two ground object samples are analyzed. Based on the changes, feature data with differences in ground features greater than specified conditions are selected as the classification feature data; Features with significant differences in ground features were selected as classification features for band synthesis, and three feature data were selected for false-color synthesis. Combine shadow and overlay masking data to remove the effects of shadow and overlay areas.

2. The method as described in claim 1, characterized in that, Based on the false-color composite image, ground feature samples are selected within the area to be measured, including: Based on the analysis of the color characteristics of ground features in the false-color composite image, ground feature samples are selected from the area to be measured in the preset designated remote sensing image based on the color characteristics.

3. The method as described in claim 2, characterized in that, Also includes: Calculate the JM distance between different land cover samples and identify land cover samples whose JM distance meets a preset condition, wherein the preset condition is that the JM distance to all other land cover samples is greater than 1.

9.

4. The method according to any one of claims 1-3, characterized in that, The initial rice planting area identification results are optimized to obtain optimized rice planting area identification results. The area of ​​the rice planting area is then calculated based on the optimized rice planting area identification results, including: The initial rice planting area identification result is spatially filtered to remove small patches, resulting in the optimized rice planting area identification result. The number of rice pixels and the area of ​​a single pixel in the SAR image are counted from the optimized rice planting area identification results. The area of ​​the rice planting area is obtained by multiplying the number of rice pixels by the area of ​​a single pixel in the SAR image.

5. A device for extracting rice planting areas based on multi-temporal SAR imagery, characterized in that, The data preprocessing unit is used to acquire multi-temporal SAR images covering the rice growth cycle, and to perform multi-temporal processing on each temporal SAR image to obtain scattering features and interferometric processing to obtain interferometric features. The feature analysis unit is used to filter out classification feature data based on the scattering features and the interference features, and to perform band synthesis based on the classification feature data; and to select feature data for false color synthesis from the synthesized bands, and to perform false color synthesis based on the feature data to obtain a false color synthesized image. The image classification unit is used to select land feature samples within the area to be measured based on the false-color synthetic image, classify the land feature samples using a preset classification method to obtain an initial rice planting area identification result; and optimize the initial rice planting area identification result to obtain an optimized rice planting area identification result, and calculate the area of ​​the rice planting area based on the optimized rice planting area identification result. Multi-temporal processing of SAR images from various time phases yields scattering features, and interferometric processing yields interferometric features, including: The scattering features are obtained by registering, multi-temporal speckle filtering, geocoding and radiometric calibration of the various temporal SAR images, wherein the scattering features include time-series backscattering coefficients, shadow and overlay mask data; The interferometric features are obtained by performing baseline estimation, interferogram generation, adaptive filtering, and coherence calculation on each of the temporal SAR images, wherein the interferometric features include the interferometric coherence coefficient; The scattering features are obtained by registering, multi-temporal speckle filtering, geocoding, and radiometric calibration of the various temporal SAR images, including: A first SAR image is selected from the multi-temporal SAR images as a reference, and each SAR image other than the first SAR image is registered based on the first SAR image to obtain a registered SAR image. A preset multi-temporal filtering method is used to remove speckle noise from the registered SAR image to obtain a filtered SAR image; Data Elevation Model (DEM) data is acquired, and the filtered SAR image is converted from slant range or ground distance projection to geographic coordinate projection based on the DEM data; and the local incident angle is acquired to detect shadows and overlay areas to obtain the time series backscattering coefficient, shadow and overlay mask data. Based on the scattering characteristics and the interference characteristics, classification feature data is selected, including: The multi-temporal SAR images are compared with preset designated remote sensing images. At least two ground object samples are selected, and the changes in the time series backscattering coefficient and interferometric coherence coefficient of at least two ground object samples are analyzed. Based on the changes, feature data with differences in ground features greater than specified conditions are selected as the classification feature data; Features with significant differences in ground features were selected as classification features for band synthesis, and three feature data were selected for false-color synthesis. Combine shadow and overlay masking data to remove the effects of shadow and overlay areas.

Images