Adaptive de-aliasing method, device and storage medium for isolated echo radial velocity

Abstract

The application discloses an adaptive deblurring method and device for isolated echo radial velocity and a storage medium, and relates to the technical field of radar signal processing. The method first identifies a main echo area and an isolated echo area from a preprocessed radar radial velocity matrix, and establishes a trusted area set. According to the speed distribution characteristics of the isolated echo area, the isolated echo area is divided into two categories and is processed respectively. For the first category area, the starting speed is predicted by using local radial data, and the deblurring is guided to be performed in a direction propagation type pixel by pixel. After the deblurring is completed, the first category area is included in the trusted area set. For the second category area, the anchor speed is determined according to the spatial relationship between the second category area and adjacent pixels in the trusted area set, and the overall sign consistency correction is performed. The application realizes automatic and accurate deblurring of various isolated echoes, solves the failure problem of traditional methods in the isolated target scene, and significantly improves the quality of the radar velocity field.

Description

Technical Field

[0001] This invention belongs to the field of radar signal processing technology, and particularly relates to an adaptive deblurring method, device and storage medium for isolated echo radial velocity. Background Technology

[0002] Pulse Doppler weather radar, through the transmission and reception of electromagnetic waves, can not only acquire reflectivity factors characterizing precipitation intensity, but also extract radial velocity based on the Doppler effect, revealing the motion state of the atmospheric flow field. This parameter is the most direct and crucial physical quantity for three-dimensional wind field inversion, identification of mesoscale eddies (such as tornadoes and mesocyclones), and monitoring of severe weather phenomena such as wind shear and downbursts. The quality of radar data, especially the accuracy of radial velocity, directly determines the timeliness and reliability of short-term forecasts and warnings.

[0003] However, limited by the pulse repetition frequency, the velocity measurement capability of Doppler weather radar has an insurmountable theoretical upper limit, namely the Nyquist velocity (±V). N When the true radial velocity of precipitation particles or turbulent scatterers in the atmosphere exceeds this range, radar observations will experience periodic jumps; this phenomenon is known as "velocity ambiguity." An uncorrected ambiguous velocity field distorts the true airflow structure into an unrecognizable chaotic image, rendering all subsequent advanced products (such as wind field inversion and storm identification) scientifically meaningless. Therefore, developing efficient and robust de-ambiguity algorithms is a core technical challenge that must be overcome in the radar data preprocessing stage.

[0004] To overcome this challenge, the industry has developed various technical approaches, but they all face severe challenges in dealing with the complexities of real-world weather, especially the prevalent "isolated echoes."

[0005] The global reference method relies on a stable velocity reference field provided by a large-scale, continuous, strong echo. This method performs reasonably well when dealing with large weather systems such as typhoons and squall lines. However, its fatal flaw lies in the fact that once the echo field is fragmented and isolated (such as initiation of local thermal convection, isolated cells within the outer spiral rainbands of a typhoon, or scattered precipitation at sea), the algorithm completely fails because it cannot anchor itself to a reliable global reference, and may even trigger erroneous global velocity reversals.

[0006] Adjacent radial propagation method: This method assumes that the wind field changes continuously between adjacent azimuths, propagating correction information through an "infectious" path. Its success heavily relies on the continuous and uninterrupted azimuth of the echoes. In actual observations, terrain obstruction, electromagnetic wave attenuation, or discontinuities in precipitation itself can all cause "discontinuities" in the echo data. Once the propagation chain is broken, echoes located on "islands" become information islands, unable to obtain any corrections. This makes this method extremely vulnerable to processing isolated, weak echoes.

[0007] External wind field constraint method: This method uses external wind fields, such as numerical weather prediction or radiosonde, as prior knowledge for constraint. While this method has some theoretical value, it has fundamental limitations: First, the spatiotemporal resolution of external data and radar observations is severely mismatched, which can easily lead to systematic biases after its introduction; second, relying on external data streams seriously impairs the radar system's own real-time and autonomous processing capabilities, resulting in low operability in operational applications.

[0008] In summary, existing technologies all face a prominent technical contradiction and bottleneck:

[0009] Faced with small-scale, isolated, and discrete echo targets (which are the key targets for monitoring strong convection), existing algorithms generally suffer from "de-blurring blind spots" due to a lack of adaptive local reference establishment capabilities and intelligent region segmentation and coordination mechanisms. Specifically, this manifests as follows: (1) The algorithm cannot autonomously generate reliable local velocity trend predictions for isolated regions without global reference; (2) It cannot effectively identify and utilize the processed credible regions to provide dynamic anchor points for subsequent processing; (3) It ignores the azimuth periodicity of radar scanning, resulting in artificial velocity breaks at the image edges; (4) It lacks effective spatial consistency correction methods for the "velocity islands" that remain after processing and are inconsistent with the physical field.

[0010] Therefore, weather radar operations urgently require a paradigm shift in de-ambiguity solutions. This solution must move away from reliance on large-area continuous echoes or external data, and instead possess "endogenous" intelligent processing capabilities for isolated targets. This means automatically identifying and segmenting isolated regions within complex echo fields, and achieving stable, adaptive velocity field reconstruction based on the inherent physical constraints and spatial relationships within the radar data itself. This is not only a technical requirement for improving data quality, but also an essential requirement for achieving precise and intelligent monitoring of severe weather. Summary of the Invention

[0011] To address the aforementioned deficiencies in existing technologies, the present invention aims to provide an adaptive deblurring method, device, and storage medium for isolated echo radial velocity, thereby solving the technical problems of deblurring failure, correction propagation interruption, velocity sign reversal, and residual "velocity islands" caused by the lack of effective local reference establishment capabilities and adaptive coordination mechanisms in existing deblurring algorithms in isolated, fragmented, weak echo, or azimuth boundary echo scenarios.

[0012] This invention solves the above-mentioned technical problems through the following technical solution: an adaptive deblurring method for isolated echo radial velocity, comprising:

[0013] Obtain the radar radial velocity matrix after pre-de-blurring, and identify the main echo region and at least one isolated echo region based on the radial velocity matrix;

[0014] Initialize a set of trusted regions, and mark all pixels in the main echo region as trusted;

[0015] Based on the statistical distribution characteristics of velocity values ​​within each isolated echo region, they are divided into two categories: Category I and Category II.

[0016] Perform a first deblurring process on the first type of region: For each first type of region, predict an initial velocity based on its local radial data, and use the initial velocity as a guide to complete the pixel-by-pixel deblurring of the velocity value in the region by propagating the correction operation along the azimuth angle to the adjacent radial direction covered by the region; after completion, mark all pixels of the region as trustworthy in the set of trustworthy regions;

[0017] A second deblurring process is performed on the second type of region: based on the updated set of trusted regions, for each second type of region, an anchor velocity is determined according to its spatial proximity relationship with trusted pixels in the set of trusted regions, and the velocity values ​​in the region are corrected for sign consistency according to the sign relationship between the overall velocity statistics of the region and the anchor velocity.

[0018] Existing deblurring algorithms often suffer from blind spots in processing isolated, fragmented, small-scale echoes due to logical flaws. This invention proactively captures and locks onto all discrete targets at the algorithmic level by "identifying and classifying the main echo and isolated echo regions," ensuring that any valid echo is included in the processing flow. This achieves a fundamental shift from "selective processing" dependent on echo continuity to "adaptive full-scene processing" accommodating arbitrary distributions, completely resolving the failure issues of traditional methods caused by target scale and shape.

[0019] To address the industry challenge of reliable references for isolated echoes, this invention designs a core step for the first type of region: predicting the initial velocity based on local radial data. This mechanism abandons the reliance on large-area continuous echoes or external wind field data, instead autonomously extrapolating and generating an initial de-ambiguity reference from extremely short sequences of reliable data upstream of the target echo (near the radar end). This "internal-driven" reference establishment capability enables the algorithm to independently and stably start and complete de-ambiguity processing for the first time under conditions completely lacking strong external references, overcoming the primary technical bottleneck in isolated echo processing.

[0020] This invention constructs a unique data-driven enhancement closed loop by introducing and dynamically maintaining a set of trusted regions. This set starts with only the main echo and expands in real time with the successful processing of each first-class region. This transforms the algorithm's reference benchmark from static to dynamically growing. Subsequent processing of second-class regions can make more accurate judgments based on this richer and more spatially proximate set of trusted regions. This recursive positive feedback mechanism of "processing-labeling-enhancing" enables the overall deblurring accuracy and coverage to improve synchronously with the processing progress, achieving self-optimization of global consistency in the results.

[0021] To address the differences in the internal characteristics of isolated echoes, this invention does not adopt a "one-size-fits-all" strategy. Instead, based on their statistical distribution characteristics, it pioneers a "dual-path differentiated processing" architecture. For the first type of region, which exhibits significant internal variations and contains structural information, a "propagation correction" method that can recover the fine wind field structure is employed. For the second type of region, which is internally homogeneous and lacks self-correction capabilities, a "symbol consistency correction" method that ensures physical uniformity is used. This categorized approach allows all types of echoes to be accurately corrected in the most suitable and efficient manner, thereby significantly improving the overall accuracy and processing efficiency of deblurring.

[0022] Further, identifying the main echo region and at least one isolated echo region based on the radial velocity matrix includes:

[0023] Generate the binary effective mask matrix of the radial velocity matrix;

[0024] The radial velocity matrix and the effective mask matrix are respectively subjected to the same azimuth periodic expansion to generate corresponding expanded velocity matrices and expanded mask matrices; wherein, the azimuth periodic expansion is achieved by adding a row before the first row and after the last row of the original matrix, and copying the last row and the first row of the original matrix to the newly added first row and last row, respectively;

[0025] Based on the extended mask matrix, connected component analysis is performed to identify all connected components. The connected component with the largest area is identified as the main echo region, and the remaining connected components are identified as isolated echo regions.

[0026] This invention overcomes the inherent "azimuth boundary fragmentation" problem in traditional radar data processing due to the polar coordinate matrix data structure by introducing a three-pronged preprocessing workflow of "mask generation, azimuth period expansion, and connected component analysis." This design proactively restores the topological continuity of physical space at the data level, ensuring that any real continuous echo crossing the 0° / 360° scan boundary is completely and unambiguously reconstructed as a single connected entity during the algorithm recognition stage. This provides a unique and physically correct basis for all subsequent partitioning, classification, and deblurring operations, fundamentally eliminating systemic failures in the entire processing chain caused by initial recognition errors, and significantly improving the algorithm's fundamental robustness and result reliability in complex scanning scenarios.

[0027] Furthermore, an initial velocity is predicted based on local radial data from the first type of region, including:

[0028] The radial direction containing the most pixels in the first type of region is determined as the central radial direction;

[0029] In the central radial direction, determine the pixel that belongs to the first type of region and is closest to the radar end, and use its column index as the starting distance gate;

[0030] In the central radial direction, a continuous effective velocity sequence is searched for near the radar end of the starting range gate;

[0031] Based on the effective velocity sequence, the predicted velocity at the starting distance gate is obtained by fitting or statistical extrapolation, and the predicted velocity is used as the starting velocity.

[0032] This invention addresses the challenge of high-standard-deviation isolated echoes lacking external references by designing a local reference self-generation mechanism that "uses the central radial direction as an anchor point, traces back to the radar end, and extrapolates the prediction." This mechanism starts from the radial direction where the target region's structure is most stable, reverses (towards the radar direction) to mine a nearby, de-blurred, reliable velocity sequence, and intelligently extrapolates and predicts a reasonable velocity value at the target region's leading edge using the continuous wind field variation trend implied in this sequence. This value serves as the "starting velocity" for initiating all subsequent correction operations. This process completely eliminates dependence on any external reference or global continuous echoes, endowing the algorithm with the core capability to "create something from nothing" and autonomously establish a reliable correction starting point in isolated scenarios, fundamentally solving the industry bottleneck of the inability to autonomously initiate the de-blurring process for isolated echoes.

[0033] Furthermore, a multi-level degradation strategy is employed to search for a continuous effective velocity sequence, including:

[0034] Search for a first velocity sequence that meets a first quality condition, the first quality condition including the sequence being continuous, its velocity standard deviation being lower than a first standard deviation threshold, and the sequence length being not less than a length threshold; if the search is successful, then fit the first velocity sequence and extrapolate the fitting result to the starting distance gate to obtain the predicted velocity.

[0035] If the first velocity sequence search fails, a second velocity sequence that meets the second quality condition is searched. The second quality condition includes that the sequence is continuous, its velocity standard deviation is lower than the second standard deviation threshold, and the sequence length is not less than the length threshold. The second standard deviation threshold is greater than the first standard deviation threshold. If the search is successful, the second velocity sequence is fitted, and the fitting result is extrapolated to the starting distance gate to obtain the predicted velocity.

[0036] If the second velocity sequence search fails, then starting from the initial distance gate, search for effective velocity values ​​in the direction of decreasing radial column index; collect the first consecutive effective velocity values ​​found that are equal in number to the length threshold, and use the median of the collected velocity values ​​as the predicted velocity.

[0037] This invention employs a three-stage decreasing strategy of "strict fitting - relaxed fitting - conservative anchoring" to ensure that even in extreme cases where the quality of local reference data is unsatisfactory (e.g., low signal-to-noise ratio, insufficient length), it can still adaptively adjust the search and prediction criteria to ultimately output a relatively reliable predicted velocity. This mechanism greatly enhances the robustness and fault tolerance of the initial velocity prediction, avoiding prediction failures or severe deviations caused by poor local data quality in traditional single strategies. This ensures that the entire defuzzification process can be stably and reliably initiated in complex and variable real-world observation data.

[0038] Further, the azimuth-directed radial propagation correction operation covering the adjacent radial region includes:

[0039] In the central radial direction, based on the initial velocity, a pixel velocity that satisfies the proximity condition and the same direction condition is searched as the reference velocity;

[0040] According to the pixel-by-pixel correction rule, all pixels in the central radial direction are sequentially corrected to obtain the reference radial direction. The pixel-by-pixel correction rule is as follows: For the first pixel, the absolute difference between its original velocity and the reference velocity is calculated. Based on the comparison result of the absolute difference with the Nyquist velocity of a preset multiple, the correction amount is determined. After correction, the reference velocity is updated to the corrected velocity of the pixel. For non-first pixels, the absolute difference used for comparison is the larger of the absolute difference between the original velocity of the current pixel and the reference velocity, and the absolute difference between the corrected velocity of the previous pixel and the reference velocity.

[0041] The corrected velocity of each pixel in the reference radial direction is used as the reference velocity, and the correction is propagated recursively to the adjacent radial directions on both sides. For any pixel in the radial direction to be corrected, at least one nearest neighbor reference velocity is dynamically selected from the corrected pixels in its spatial neighborhood, and the pixel-by-pixel correction rule is applied for correction.

[0042] This invention addresses two inherent flaws of traditional deblurring algorithms: the tendency for errors within isolated echoes to accumulate linearly along the radial direction, and the chain-like interruption during azimuth propagation due to irregular shapes. It provides a systematic solution. First, the scheme introduces a "dual-point reference determination" rule in the radial correction of the reference. By simultaneously considering the differences between the current point and the previous point relative to the reference, it actively intercepts and suppresses the propagation of correction errors, ensuring the inherent stability of radial correction. Furthermore, it innovatively employs a "dynamic nearest neighbor matching" mechanism in azimuth propagation, completely eliminating the rigid constraint that the target point and reference point must be aligned. This allows the algorithm to intelligently find the most relevant reliable reference in two-dimensional space for each point to be corrected. The synergistic effect of these two mechanisms enables the correction process to resist the spread of internal errors while closely adhering to and completely covering echo regions of arbitrarily complex shapes. This overcomes the fundamental problems of poor robustness and weak adaptability of traditional methods, achieving highly stable, full-shape adaptive deblurring for isolated echoes with high standard deviations.

[0043] Furthermore, performing the first deblurring process on the first type of region also includes performing a discrete point correction step after completing the pixel-by-pixel deblurring:

[0044] Calculate the median of the corrected velocity for all pixels within the first type of region;

[0045] Traverse all pixels within the first type of region. If the corrected velocity of a pixel has the opposite sign to the median, apply a sign-flipping correction to that pixel so that the sign of the corrected velocity of that pixel is consistent with the sign of the median.

[0046] This invention adds a "discrete point correction based on median benchmark" step. This step uses the statistical median of the corrected velocity within the region as the true benchmark for wind direction, automatically screening and locking a few abnormal pixels that contradict the overall wind direction. By applying precise sign-flipping correction, it can efficiently filter out sporadic artifacts of different signs caused by local noise, weak echo interference, or complex wind shear edge effects. This not only achieves forced normalization of velocity direction within the region at the pixel level, but also significantly improves the quality and physical reliability of input data generated by downstream advanced products such as wind field inversion and vortex identification at key meteorological application levels, ensuring that the final output of the algorithm has structural details while also meeting strict physical consistency requirements.

[0047] Further, an anchoring velocity is determined based on its spatial proximity to trusted pixels in the trusted region set, including:

[0048] Calculate the coordinates of the geometric center of the second type of region;

[0049] Using the geometric center as a reference, find all reliable pixels in the set of reliable regions whose spatial distance from the geometric center is less than a preset distance threshold;

[0050] Calculate the median of the velocity values ​​of all the found reliable pixels, and determine the median as the anchor velocity.

[0051] This invention presents a "space proximity-based anchoring velocity" determination mechanism for isolated echoes (velocity islands) with low standard deviation. This mechanism cleverly utilizes a dynamically growing set of reliable regions built by the algorithm as a global wind field reference library. By calculating the geometric center of the target region and retrieving reliable pixels within its spatial neighborhood, the median velocity of these reliable neighboring points is used as the decision criterion. This design enables weak echoes, which would otherwise be unable to determine their own direction due to a lack of internal information, to "borrow" reliable wind direction information from the most relevant processed regions in the external physical space. This not only provides a unique, objective, and physically meaningful decision anchor point for subsequent symbol correction but also demonstrates the system-level design wisdom of this invention: transforming the initial processing results (reliable set) into an effective resource for solving subsequent, more challenging problems (velocity islands), thereby systematically overcoming the fundamental problem of lacking intrinsic criteria for deblurring in low-variance regions.

[0052] Furthermore, based on the sign relationship between the overall velocity statistics of the region and the anchoring velocity, a sign consistency correction is performed on the velocity values ​​within the region, including:

[0053] Calculate the mean of the original velocities of all pixels within the second type of region;

[0054] Compare the sign of the mean with that of the anchoring speed;

[0055] If the mean value has the opposite sign to the anchor velocity, then a sign flip correction is applied to all pixels in the second type region whose velocity signs are opposite to the anchor velocity; wherein, the sign flip correction is: adding or subtracting a preset correction amount to the original velocity value of the pixel, so that the sign of the corrected velocity of the pixel becomes consistent with the sign of the anchor velocity.

[0056] This invention achieves robust macroscopic diagnosis of systematic wind direction errors in a region by comparing the sign relationship between the overall average velocity of the region and the reliable external anchor velocity, effectively immunizing against interference from sporadic internal noise. Once the diagnosis is confirmed, targeted correction is performed: only pixels within the region that deviate from the anchor wind direction are subject to a preset, uniform sign flip, while pixels with the correct sign are fully preserved. This design ensures that the final wind direction of the entire region is forced to conform to the physical environment while preserving the effective information in the original observation data to the greatest extent possible. Thus, with extremely high efficiency and accuracy, "velocity islands" are seamlessly corrected and integrated into a continuous and reasonable overall wind field.

[0057] Based on the same concept, the present invention also provides an electronic device, including a memory, a processor, and a computer program or instructions stored in the memory, wherein the processor executes the computer program or instructions to implement the adaptive deblurring method for isolated echo radial velocity as described above.

[0058] Based on the same concept, the present invention also provides a computer-readable storage medium having a computer program or instructions stored thereon, which, when executed by a processor, implements the adaptive deblurring method for isolated echo radial velocity as described above.

[0059] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0060] By actively identifying and classifying all isolated echoes, this invention breaks through the dependence of traditional methods on echo continuity, ensuring that isolated targets of any scale and shape can be included in the processing flow, achieving a paradigm shift from selective processing to full-scene adaptive processing. For isolated echoes with high standard deviation, a mechanism of "predicting the initial velocity based on local radial data" is designed. This mechanism autonomously constructs an initial reference from reliable data in the vicinity of the target, completely eliminating dependence on external wind fields or large-area continuous echoes, enabling the algorithm to autonomously start in isolated scenarios. By dynamically maintaining and expanding the "reliable region set," the results of initial processing are transformed into reliable reference sources for subsequent processing, forming a positive feedback loop of "processing-learning-enhancement," significantly improving the overall coverage and consistency of the processing. Based on the statistical characteristics of the echoes, "propagation correction" is implemented for structurally complex high standard deviation regions to restore details, while "sign consistency correction" is implemented for uniform low standard deviation regions to enforce physical consistency. This classification-based approach significantly improves the accuracy and efficiency of the processing.

[0061] Through the above-mentioned systematic innovations, this invention effectively solves the common failure problem of traditional deblurring algorithms in isolated, fragmented, and weak echo scenarios, and significantly improves the quality and reliability of radar radial velocity data. Attached Figure Description

[0062] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only one embodiment of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0063] Figure 1 This is a flowchart of the adaptive defuzzification method for isolated echo radial velocity in an embodiment of the present invention;

[0064] Figure 2 This is a schematic diagram of velocity prediction at the starting distance gate in an embodiment of the present invention;

[0065] Figure 3 These are comparison images of isolated echo radial deblurring before and after in an embodiment of the present invention;

[0066] Figure 4 This is a schematic diagram of isolated echo velocity propagation correction in an embodiment of the present invention;

[0067] Figure 5 This is the original velocity echo PPI view in an embodiment of the present invention;

[0068] Figure 6 This is the PPI view of the speed after the first type of region deblurring is completed in this embodiment of the invention;

[0069] Figure 7 This is the PPI view of the second type of region after deblurring is completed in this embodiment of the invention. Detailed Implementation

[0070] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0071] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0072] Example 1

[0073] This embodiment provides an adaptive deblurring method for the radial velocity of isolated echoes from weather radar, aiming to solve the problem that traditional deblurring algorithms have weak or even complete failure capabilities in processing isolated, fragmented, and small-scale echoes. Figure 1 As shown, the method specifically includes the following steps:

[0074] Step S1: Obtain the radar radial velocity matrix after pre-deblurring, and identify the main echo region and isolated echo region based on the radial velocity matrix.

[0075] Obtain the radar radial velocity matrix after preliminary processing using basic deblurring algorithms such as the traditional adjacent radial propagation method. Its size is M×N, where M represents the number of azimuth angles and N represents the number of range parameters. Region identification is then performed based on the radar radial velocity matrix.

[0076] S1.1 Generating an effective mask matrix: Invalid data (NaN values) in the radial velocity matrix are masked to generate a binary effective mask matrix. Specifically, if the element in the i-th row and j-th column of the radar radial velocity matrix is ​​a valid value, then the element in the i-th row and j-th column of the effective mask matrix is ​​1; otherwise, it is 0.

[0077] S1.2, Azimuth Period Extension: In order to deal with the echo breakage problem at the boundary between the radar scanning azimuth angles of 0° and 360°, the same azimuth period extension operation is performed on the radial velocity matrix and the effective mask matrix.

[0078] Specifically, construct an extended velocity matrix and an extended mask matrix of size (M+2)×N. Copy the last row of the radial velocity matrix to the first row of the extended velocity matrix, and copy the first row of the radial velocity matrix to the last row of the extended velocity matrix. Place the radial velocity matrix in rows 2 to M+1 of the extended velocity matrix. Perform the same operation on the effective mask matrix to obtain the extended mask matrix.

[0079] S1.3 Connectivity Analysis and Primary / Secondary Region Division: The eight-neighbor clustering algorithm is used to perform connectivity analysis on the extended mask matrix to identify all independent echo connectivity sets; the area (i.e., the number of pixels contained) of each echo connectivity is calculated, and the connectivity with the largest area is determined as the primary echo region, while all other connectivity regions are determined as isolated echo regions.

[0080] In image processing, the eight-neighborhood relationship is defined as follows: for a central pixel, pixels above, below, left, right, and along its four diagonals are considered its neighboring pixels. If two pixels are adjacent within the eight-neighborhood and both have valid values ​​(a value of 1 in the extended mask matrix), they are considered to belong to the same connected region. Crucially, thanks to azimuth periodic extension processing, the last row of the effective mask matrix (recording the echo validity state at azimuth ~360°) is copied to the first row of the extended mask matrix, while the first row of the effective mask matrix (recording the echo validity state at azimuth ~0°) itself is located in the second row of the extended mask matrix. Under the eight-neighborhood rule, these two rows of data are spatially adjacent in the matrix structure. Therefore, the continuous echo region that actually spans the radar scan 0° / 360° boundary in physical space (whose corresponding first and last row states in the effective mask matrix are 1) can be correctly identified as a single connected region in the extended mask matrix, thus completely solving the problem of azimuth boundary echo breakage caused by the limitations of the matrix data structure in traditional methods.

[0081] Step S2: Initialize a set of trusted regions, and mark all pixels in the main echo region as trusted.

[0082] Create a reliable region labeling matrix of the same size (M×N) as the radar radial velocity matrix, and initialize all elements to 0. Since the radar radial velocity matrix has undergone pre-deblurring, the main echo region has been deblurred. Mark all pixel positions corresponding to the main echo region as 1 in the reliable region labeling matrix, indicating that the velocity values ​​of these pixels have been deblurred and are reliable. At this point, the initial state of the reliable region set is complete.

[0083] Step S3: Based on the statistical distribution characteristics of velocity values ​​within each isolated echo region, divide them into the first type of region and the second type of region.

[0084] Traverse all isolated echo regions and calculate the standard deviation of the velocity values ​​within each region. Set a first standard deviation threshold (e.g., an empirical value of 9 or 10). Classify isolated echo regions with a standard deviation ≥ the first standard deviation threshold as Class I regions (high standard deviation regions); classify isolated echo regions with a standard deviation < the first standard deviation threshold as Class II regions (low standard deviation regions). Different deblurring strategies will then be applied to these two types of regions.

[0085] Step S4: Perform the first deblurring process on the first type of region.

[0086] This step aims to recover the true velocity field of isolated echoes with distinct internal structural features. For each Class I region, the following sub-steps are performed:

[0087] S4.1 Determine the processing anchor point: Find the radial direction (matrix row) containing the most pixels in the first type of region, and use it as the central radial direction; on the central radial direction, find the pixel that belongs to the first type of region and has the smallest column index (i.e., the closest to the radar end), and record the column index of the pixel as the starting distance gate.

[0088] S4.2. Use a multi-level degradation strategy to predict the initial velocity: In the central radial direction, starting from the initial range gate and moving towards the direction where the column index decreases (near the radar end), search for a continuous effective velocity sequence to predict the velocity.

[0089] In one specific implementation, a three-level degradation strategy is employed to ensure robustness:

[0090] Level 1 (Exact Fit): Search for a continuous first velocity sequence with a velocity standard deviation lower than a first standard deviation threshold. If the length of the first velocity sequence is not less than a length threshold (i.e., the number of fitting points threshold, such as 20), then use the least squares method to perform a first-order polynomial fitting on the first velocity sequence, and extrapolate the fitted line to the starting distance gate position to obtain the predicted velocity.

[0091] Level Two (Relaxed Fit): If the first velocity sequence search fails, the standard deviation threshold is relaxed, i.e., a second standard deviation threshold is preset, which is greater than the first standard deviation threshold. For example, the second standard deviation threshold is 1.2 times the first standard deviation threshold. A second velocity sequence that is continuous and has a velocity standard deviation lower than the second standard deviation threshold is searched again. If the length of the second velocity sequence is not less than a length threshold, a first-order polynomial is fitted to the second velocity sequence using the least squares method, and the fitted line is extrapolated to the starting distance gate position to obtain the predicted velocity.

[0092] Level 3 (Conservative Anchoring): If the first two levels fail, start searching from the initial range gate toward the near-radar end, collect the first consecutive effective velocity values ​​found that are equal in number to the length threshold, and take the median as the predicted velocity.

[0093] Ultimately, the prediction velocity was used as the starting velocity to guide the deblurring of the first type of region.

[0094] S4.3 Determine the initial baseline velocity: Starting from the point with the smallest column index (nearest distance) along the central radial direction of this first-class region, search sequentially in the direction of increasing column index until all elements along the entire central radial direction have been traversed. Within this range, find the first pixel that simultaneously satisfies the following two conditions, and use its velocity value as the initial baseline velocity:

[0095] (a) The velocity value V of the pixel i With initial velocity V start The absolute difference (i.e., the absolute value of the difference between the two) is less than αV. N VN This represents the Nyquist velocity, and α represents a preset multiple (e.g., 0.9).

[0096] (b) The velocity value V of this pixel i With initial velocity V start The signs are the same (i.e., V) i ×V start >0).

[0097] If no point that meets the conditions is found within this search range, the predicted starting velocity will be used directly as the initial reference velocity.

[0098] Figure 2 A schematic diagram of velocity prediction at the initial distance gate is shown. (Example) Figure 2 As shown, at the initial distance gate, the predicted velocity value at that point is obtained by extrapolating along the fitted line.

[0099] S4.4, Pixel-by-Pixel Correction in the Central Radial Direction: Starting from the initial reference velocity, sequentially correct all pixels in the central radial direction from the initial distance gate. The pixel-by-pixel correction rules are as follows:

[0100] For the first pixel (at the starting distance gate in the center radial direction): calculate the absolute difference between its original velocity V1 and the current reference velocity (which is the initial reference velocity at this point). (That is, the absolute value of the difference between the two). According to With Nyquist velocity V N The multiple relationship determines the amount of correction applied. After correction, the base velocity is updated to the corrected velocity of the first pixel.

[0101] Specifically, if If no correction is made, the correction amount is 0; if Then apply ±V N Correction; if Then apply ±2V N The correction is as follows: α and β represent preset multiples; the sign of the correction is determined based on the relationship between the original speed and the reference speed. If the original speed is greater than the reference speed, the sign of the correction is positive; otherwise, it is negative.

[0102] For the j-th (j≥2) pixel: calculate its original velocity V j The absolute difference from the current reference velocity (the current reference velocity is the corrected velocity of the previous pixel) And the absolute difference between the corrected velocity of the previous pixel and the current baseline velocity. ;Pick , The larger of the two values ​​is used as the judgment difference ΔV max .

[0103] According to ΔVmax With Nyquist velocity V N The multiple relationship determines the application of 0, ±V N or ±2V N The correction amount.

[0104] After correction, the base speed is set to the corrected speed of the j-th pixel.

[0105] This rule effectively solves the problem of error accumulation and propagation by introducing a two-point reference decision mechanism. Specifically, for pixels other than the first one, the correction decision not only refers to the original velocity of the current pixel but also to the corrected velocity of the previous pixel, and takes the larger of the differences between the two and the reference velocity as the decision criterion. This design can keenly capture the potential error jump trend formed by the accumulation of continuous small velocity differences, and thus actively intervene and correct it before it develops into an obvious error, ensuring the high stability of the correction process in the radial direction. After completing this step, a corrected reference radial direction with internal consistency and controlled error is obtained.

[0106] S4.5, Regional Radial Propagation Correction: Using the corrected reference radial direction as a "seed," the correction is recursively propagated to adjacent radial directions on both sides (clockwise and counterclockwise). For a valid pixel P on the target radial direction to be corrected:

[0107] Step S4.5 uses the corrected center radial direction (i.e., the reference radial direction) from step S4.4 as the initial reference radial direction to initiate recursive propagation correction across the entire isolated echo region. The propagation process follows the core mechanism below:

[0108] The recursive propagation framework uses the corrected velocity of each pixel on the initial reference radial direction as a reliable reference value for that location. The correction operation first propagates to adjacent radial directions on both sides of this radial direction (i.e., clockwise and counterclockwise, corresponding to the upper and lower adjacent rows of the 2D radial velocity matrix). Once all pixels on a given radial direction have been corrected, that radial direction becomes the new reference radial direction, and propagation continues to its unprocessed adjacent radial directions on both sides. This process is repeated recursively, like ripples spreading, until the entire target area is covered.

[0109] Adaptive Reference Matching Mechanism (Nearest Neighbor Matching): Due to the irregularity of isolated echo shapes, effective pixels in the target radial direction and pixels in the reference radial direction may not be on the same distance gate (column index), i.e., spatially misaligned. To solve this problem, the algorithm uses a nearest neighbor matching mechanism to dynamically determine the baseline speed.

[0110] For a valid pixel P in the radial direction of the target to be corrected, first check whether there is a corrected pixel (i.e., the corrected velocity of the pixel) at the same distance gate position in the reference radial direction.

[0111] If it exists, the corrected speed of that pixel is used as the base speed.

[0112] If not found (i.e., misaligned), then among all pixels already marked as corrected, search for the two corrected pixels that are spatially closest to and second closest to the valid pixel P. The spatial distance is calculated by combining azimuth difference and distance gate difference, where the azimuth difference uses the minimum period difference. ,in Indicates the orientation (row index) of the unambiguous reference pixel (i.e., the corrected pixel). This indicates the orientation of the effective pixel P in the radial direction of the target to be corrected, and M is the row number of the radial velocity matrix.

[0113] The speeds (i.e., the corrected speeds) of the two corrected pixels that are spatially closest and second closest to the effective pixel P are used as the baseline speeds.

[0114] Unified judgment and pixel-by-pixel correction rules: After obtaining the baseline velocity, the subsequent velocity blur judgment and correction logic is as follows:

[0115] Case 1 (Single-point reference): If a baseline velocity is obtained through direct alignment, then the original velocity V of the effective pixel P is directly calculated. P The absolute difference from this reference velocity, based on the ratio of this absolute difference to the Nyquist velocity, determines the application of 0, ±V. N or ±2V N The correction amount.

[0116] Case 2 (Two-point reference - nearest neighbor matching): If two baseline velocities are obtained through nearest neighbor matching, then, following the rule described in step S4.4 "for the j-th (j≥2) pixel", calculate the original velocity V of the effective pixel P. P The absolute difference between each of the two reference speeds is taken, and the larger of the two absolute differences is used as the judgment difference value ΔV. max According to ΔV max With Nyquist velocity V N The multiple relationship determines the application of 0, ±V to the effective pixel P. N or ±2V N The correction amount.

[0117] This design ensures that, whether in the radial or azimuth direction, when multiple reliable references exist, a more conservative decision strategy (taking the larger difference) is adopted to effectively suppress the propagation of errors.

[0118] This mechanism of "recursive propagation + nearest neighbor matching + unified rules" ensures that the correction information can adaptively fill echo regions of arbitrary irregular shapes, while maintaining the robustness and consistency of the correction process. This process continues until all pixels within the isolated echo region of the target have undergone deblurring correction.

[0119] To visually demonstrate the effect of deblurring, Figure 3 A schematic diagram comparing radial velocity before and after deblurring is shown. Figure 3 As shown, the velocity curve before deblurring exhibits a clear step jump, a typical manifestation of velocity folding; the velocity curve after deblurring restores the physically true continuous trend, the jump is eliminated, and the velocity value is corrected to a reasonable range. To clearly demonstrate the different pixel states during the processing, Figure 4 A schematic diagram of velocity propagation correction based on a reference radial direction is shown. (See diagram below.) Figure 4 As shown, the processed pixels are located in the reference radial direction and several adjacent radial directions, indicating that the velocity blurring process has been completed in this area; the pixels to be processed are located in the outer radial direction that has not yet been polled, and their velocity blurring process has not yet been completed.

[0120] S4.6 Discrete Point Correction. After completing the region propagation correction, to further improve the physical consistency of the velocity field within the region, discrete point correction can be performed:

[0121] Calculate the median V of the corrected velocity of all pixels within the current first-class region. median ;

[0122] Traverse all pixels within the current first-class region. If the corrected velocity of a certain pixel is similar to the median V... median If the sign is opposite, then apply a sign-flip correction to the pixel (e.g., apply ±2V to the corrected velocity of the pixel). N (the correction amount), making its sign similar to the median V median The symbols are consistent, thus ensuring the consistency of the velocity field symbols within the first type of region.

[0123] Specifically, if the median V median If the sign is positive, then a +2V is applied to the corrected velocity of that pixel. N The correction amount; if the median V median If the sign is negative, then -2V is applied to the corrected velocity of that pixel. N The correction amount.

[0124] S4.7 Update the set of trusted regions: Mark all pixel positions in the first type of region that has successfully completed the first deblurring process as 1 in the trusted region labeling matrix, indicating that the velocity values ​​of these pixels have been deblurred and are trusted. At this point, the region is included in the updated set of trusted regions.

[0125] Repeat step S4 until all identified first-class regions have been traversed and processed.

[0126] Step S5: Perform a second deblurring process on the second type of region.

[0127] This step aims to correct weak, isolated echoes that lack obvious internal structure and may form velocity islands. For each type II region, perform the following sub-steps:

[0128] S5.1 Determine the spatial anchoring velocity. Use the updated set of reliable regions (which now includes the main echo region and all processed Type I regions) to find a reliable external wind direction reference for the current region.

[0129] Calculate the geometric center coordinates (R) of the current second type region. T G T );

[0130] Using the geometric center as a reference, among all reliable pixels marked by the reliable region marking matrix, find all pixels whose spatial distance to the geometric center is less than a preset distance threshold. Distance calculation considers both azimuth difference and distance threshold, with the azimuth difference using the minimum period difference. ,in Indicates the orientation (i.e., row index) of an unambiguous reference pixel (i.e., a reliable pixel).

[0131] Calculate the median of the velocity values ​​of all found neighboring reliable pixels, and determine this median as the anchoring velocity V. anchor .

[0132] S5.2 Determine and perform symbol consistency correction.

[0133] Calculate the mean of the original velocities of all pixels within the current second-class region, and compare the sign of this mean with the anchor velocity;

[0134] If the signs of the two are opposite, it is considered that there is an overall velocity fold (i.e., an "island") in the current second type region. In this case, the signs of all velocities within the current second type region are compared with the anchor velocity V. anchor For the opposite pixel, a uniform sign-flip correction is applied. The correction method involves adding or subtracting a preset correction amount (typically ±2V) from the pixel's original velocity value. N This makes the velocity sign of the pixel after correction consistent with the anchor velocity.

[0135] If the two signs are the same, the current second-type region is considered to be in the correct direction, and no sign flipping correction is required.

[0136] S5.3 Update the set of reliable regions. Mark all pixels in the second type of region that have completed sign flip correction (or are determined not to need correction) as 1 in the reliable region labeling matrix.

[0137] Repeat step S5 until all identified second-class regions have been traversed and processed. Finally, the ambiguity of all isolated echo regions in the original input radial velocity matrix has been adaptively corrected, and all valid echo points in the reliable region labeling matrix have been marked as reliable. The corrected final velocity field is output for subsequent meteorological analysis and product generation.

[0138] Figure 5 This displays a radar radial velocity PPI view after basic deblurring. Figure 5 It is evident that the velocity blurring in the main echo region (large continuous area) has been largely eliminated, but several isolated small-scale echo regions remain, whose velocity values ​​exhibit obvious folding and jumping (color abrupt change), constituting the main target to be processed.

[0139] Figure 6 The velocity field after performing the first deblurring process on the first type of region is shown. For example... Figure 6 As shown, those isolated echoes with high standard deviations and relatively clear structures, exhibiting drastic internal velocity variations (i.e., the first type of region), have been successfully corrected, restoring the physical continuity and plausibility of their velocity fields. However, Figure 6 Several small echo blocks with uniform internal velocity but abnormal overall direction (their color is out of place with the surrounding wind field) can still be seen, i.e., velocity islands.

[0140] Figure 7 The final velocity field after performing the second deblurring process on the first type of region is shown. For example... Figure 7 As shown, Figure 6 All remaining low-standard-deviation "velocity islands" (Type II regions) have been effectively identified and corrected. Their velocity directions have been uniformly corrected, achieving physical consistency with the main wind field and the surrounding processed areas, ultimately forming a complete velocity field that is continuous and physically reliable across the entire domain.

[0141] pass Figures 5 to 7 The comparison clearly shows that the "classification processing and recursive enhancement" framework proposed in this invention achieves systematic deblurring of various isolated echoes that traditional methods are ineffective against. Traditional methods (such as the adjacent radial propagation method) highly rely on the continuity of the echoes, and fail once the propagation chain is interrupted at an isolated echo; while the global reference method fails because isolated echoes lack reliable internal references. This invention successfully overcomes this industry challenge by establishing an intrinsically driven local reference for high standard deviation regions and a spatially anchored external reference for low standard deviation regions, significantly improving the availability and accuracy of radar wind field data in complex weather scenarios.

[0142] Through the specific implementation methods described above, this invention achieves fully automatic, adaptive, and highly robust deblurring of various isolated echoes in the radial velocity field of weather radar. Its significant advantages include: the ability to handle isolated targets of arbitrary shape and scale, eliminating processing blind spots; no external reference required, possessing bootstrapping capability; employing error accumulation prevention correction and nearest neighbor propagation to ensure a stable and reliable correction process; forming a self-reinforcing processing loop through recursive updating of the trusted region set; effectively eliminating velocity islands and improving data physical consistency; and a clear algorithm structure that meets real-time processing requirements.

[0143] Example 2

[0144] This invention also provides an electronic device, which includes a memory, a processor, and a computer program or instructions stored in the memory. The processor executes the computer program or instructions to implement the adaptive deblurring method for isolated echo radial velocity in this invention.

[0145] Although not shown, the electronic device includes a processor that can perform various appropriate operations and processes based on programs and / or data stored in read-only memory (ROM) or loaded from a storage portion into random access memory (RAM). The processor can be a multi-core processor or may contain multiple processors. In some embodiments, the processor may include a general-purpose main processor and one or more specialized coprocessors, such as a central processing unit, graphics processing unit (GPU), neural network processor (NPU), digital signal processor (DSP), etc. Various programs and data required for device operation are also stored in RAM. The processor, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus.

[0146] The processor and memory described above are used together to execute programs / instructions stored in the memory. When the program / instructions are executed by the computer, they can implement the methods, steps, or functions described in the above embodiments.

[0147] Although not shown, embodiments of the present invention also provide a computer-readable storage medium having a computer program or instructions stored thereon that, when executed by a processor, implements the adaptive deblurring method for isolated echo radial velocity in embodiments of the present invention.

[0148] Readable storage media include both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0149] The above description only discloses specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or modifications that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. An adaptive defuzzification method for the radial velocity of isolated echoes, characterized in that, The method includes: Obtain the radar radial velocity matrix after pre-deblurring, and identify the main echo region and at least one isolated echo region based on the radial velocity matrix; Initialize a set of trusted regions, and mark all pixels in the main echo region as trusted; Based on the statistical distribution characteristics of velocity values ​​within each isolated echo region, they are divided into two categories: Category I and Category II. Perform a first deblurring process on the first type of region: For each first type of region, predict an initial velocity based on its local radial data, and use the initial velocity as a guide to complete the pixel-by-pixel deblurring of the velocity value in the region by propagating the correction operation along the azimuth angle to the adjacent radial direction covered by the region; after completion, mark all pixels of the region as trustworthy in the set of trustworthy regions; A second deblurring process is performed on the second type of region: based on the updated set of trusted regions, for each second type of region, an anchor velocity is determined according to its spatial proximity relationship with trusted pixels in the set of trusted regions, and the velocity values ​​in the region are corrected for sign consistency according to the sign relationship between the overall velocity statistics of the region and the anchor velocity.

2. The adaptive defuzzification method for isolated echo radial velocity according to claim 1, characterized in that, Identifying the main echo region and at least one isolated echo region based on the radial velocity matrix includes: Generate the binary effective mask matrix of the radial velocity matrix; The radial velocity matrix and the effective mask matrix are respectively subjected to the same azimuth periodic expansion to generate corresponding expanded velocity matrices and expanded mask matrices; wherein, the azimuth periodic expansion is achieved by adding a row before the first row and after the last row of the original matrix, and copying the last row and the first row of the original matrix to the newly added first row and last row, respectively; Based on the extended mask matrix, connected component analysis is performed to identify all connected components. The connected component with the largest area is identified as the main echo region, and the remaining connected components are identified as isolated echo regions.

3. The adaptive defuzzification method for isolated echo radial velocity according to claim 1, characterized in that, Predicting an initial velocity based on local radial data from the first type of region includes: The radial direction containing the most pixels in the first type of region is determined as the central radial direction; In the central radial direction, determine the pixel that belongs to the first type of region and is closest to the radar end, and use its column index as the starting distance gate; In the central radial direction, a continuous effective velocity sequence is searched for near the radar end of the starting range gate; Based on the effective velocity sequence, the predicted velocity at the starting distance gate is obtained by fitting or statistical extrapolation, and the predicted velocity is used as the starting velocity.

4. The adaptive defuzzification method for isolated echo radial velocity according to claim 3, characterized in that, A multi-level degradation strategy is employed to search for a continuous effective velocity sequence, including: Search for a first velocity sequence that meets a first quality condition, the first quality condition including the sequence being continuous, its velocity standard deviation being lower than a first standard deviation threshold, and the sequence length being not less than a length threshold; if the search is successful, then fit the first velocity sequence and extrapolate the fitting result to the starting distance gate to obtain the predicted velocity. If the first velocity sequence search fails, a second velocity sequence that meets the second quality condition is searched. The second quality condition includes that the sequence is continuous, its velocity standard deviation is lower than the second standard deviation threshold, and the sequence length is not less than the length threshold. The second standard deviation threshold is greater than the first standard deviation threshold. If the search is successful, the second velocity sequence is fitted, and the fitting result is extrapolated to the starting distance gate to obtain the predicted velocity. If the second velocity sequence search fails, then starting from the initial distance gate, search for effective velocity values ​​in the direction of decreasing radial column index; collect the first consecutive effective velocity values ​​found that are equal in number to the length threshold, and use the median of the collected velocity values ​​as the predicted velocity.

5. The adaptive defuzzification method for isolated echo radial velocity according to claim 1, characterized in that, The azimuth-oriented radial propagation correction operation covering the adjacent radial region includes: In the central radial direction, based on the initial velocity, a pixel velocity that satisfies the proximity condition and the same direction condition is searched as the reference velocity; According to the pixel-by-pixel correction rule, all pixels in the central radial direction are sequentially corrected to obtain the reference radial direction. The pixel-by-pixel correction rule is as follows: For the first pixel, the absolute difference between its original velocity and the reference velocity is calculated. Based on the comparison result of the absolute difference with the Nyquist velocity of a preset multiple, the correction amount is determined. After correction, the reference velocity is updated to the corrected velocity of the pixel. For non-first pixels, the absolute difference used for comparison is the larger of the absolute difference between the original velocity of the current pixel and the reference velocity, and the absolute difference between the corrected velocity of the previous pixel and the reference velocity. The corrected velocity of each pixel in the reference radial direction is used as the reference velocity, and the correction is propagated recursively to the adjacent radial directions on both sides. For any pixel in the radial direction to be corrected, at least one nearest neighbor reference velocity is dynamically selected from the corrected pixels in its spatial neighborhood, and the pixel-by-pixel correction rule is applied for correction.

6. The adaptive defuzzification method for the radial velocity of isolated echoes according to any one of claims 1 to 5, characterized in that, Performing the first deblurring process on the first type of region further includes performing a discrete point correction step after completing the pixel-by-pixel deblurring: Calculate the median of the corrected velocity for all pixels within the first type of region; Traverse all pixels within the first type of region. If the corrected velocity of a pixel has the opposite sign to the median, apply a sign-flipping correction to that pixel so that the sign of the corrected velocity of that pixel is consistent with the sign of the median.

7. The adaptive defuzzification method for isolated echo radial velocity according to claim 1, characterized in that, An anchoring velocity is determined based on its spatial proximity to trusted pixels in the trusted region set, including: Calculate the coordinates of the geometric center of the second type of region; Using the geometric center as a reference, find all reliable pixels in the set of reliable regions whose spatial distance from the geometric center is less than a preset distance threshold; Calculate the median of the velocity values ​​of all the found reliable pixels, and determine the median as the anchor velocity.

8. The adaptive defuzzification method for isolated echo radial velocity according to claim 1, characterized in that, Based on the sign relationship between the overall velocity statistics of the region and the anchoring velocity, a sign consistency correction is performed on the velocity values ​​within the region, including: Calculate the mean of the original velocities of all pixels within the second type of region; Compare the sign of the mean with that of the anchoring speed; If the mean value has the opposite sign to the anchor velocity, then a sign flip correction is applied to all pixels in the second type region whose velocity signs are opposite to the anchor velocity; wherein, the sign flip correction is: adding or subtracting a preset correction amount to the original velocity value of the pixel, so that the sign of the corrected velocity of the pixel becomes consistent with the sign of the anchor velocity.

9. An electronic device comprising a memory, a processor, and a computer program or instructions stored in the memory, characterized in that, The processor executes the computer program or instructions to implement the adaptive deblurring method for isolated echo radial velocity as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, When the computer program or instructions are executed by the processor, they implement the adaptive deblurring method for isolated echo radial velocity as described in any one of claims 1 to 8.

Images