Unmanned aerial mini sar real-time imaging and emergency communication system for disaster relief

Abstract

The application discloses a UAV-borne MiniSAR real-time imaging and emergency communication system for disaster relief, relates to the technical field of emergency communication and radar remote sensing, and comprises a radar path inversion modeling module, which performs multi-time flight over a target area, collects radar echo signals at multiple times, simultaneously acquires flight attitude data and imaging azimuth information of a flight platform, inverts the propagation path of the radar echo signals based on the flight attitude data and the imaging azimuth information, extracts the incident angle and the reflection path length, and generates a reflection path distribution diagram.The application can accurately identify and eliminate false target points under multi-path interference through multi-time radar path modeling and image stability evaluation, realize image structure repair by combining a dynamic extension algorithm, construct a dynamic confidence distribution diagram, realize high-confidence target positioning and rescue area optimization, and enhance the anti-interference and response accuracy of the system.

Description

Technical Field

[0001] This invention relates to the fields of emergency communication and radar remote sensing technology, specifically to an unmanned aerial vehicle (UAV)-borne MiniSAR real-time imaging and emergency communication system for disaster relief. Background Technology

[0002] The UAV-borne MiniSAR real-time imaging and emergency communication system for disaster relief is an intelligent UAV mission platform integrating a small synthetic aperture radar (MiniSAR) and an emergency communication module. It is specifically designed for rapid perception and communication support at the scene of sudden disasters such as earthquakes, floods, and fires. Equipped with X, Ku, or dual-band MiniSAR radar, the system continuously acquires high-resolution radar images in complex weather, low-light, and even complete darkness environments, enabling key functions such as building collapse detection, surface deformation monitoring, water inundation analysis, and fire boundary tracking. Simultaneously, the system possesses real-time image processing and transmission capabilities, allowing for the temporary establishment of an aerial communication network via its built-in satellite or microwave communication relay module in the event of frontline communication disruptions, ensuring information interconnection between rescue forces and the command center. Its lightweight and modular platform is compatible with various small and lightweight UAVs, possessing core advantages such as rapid deployment, all-weather operation, precise positioning, and collaborative command, significantly improving the timeliness and scientific rigor of disaster emergency response.

[0003] Existing technologies have the following shortcomings: In extreme disaster scenarios such as urban earthquakes, target areas typically contain numerous collapsed building structures, metal rubble, and steel remnants—highly reflective objects—creating a non-ideal imaging region with an extremely complex electromagnetic environment. When a MiniSAR radar system performs high-resolution imaging in such areas, its emitted radar signals are prone to multiple reflections and scatterings between various highly reflective surfaces, forming typical multipath echo paths. These non-direct echo signals, when superimposed on normal reflected signals at the radar receiver, may produce strong echo virtual images in the radar image that are inconsistent with the actual target location, i.e., "false target points." Especially when the image resolution is high or the target background structure is complex, the radar characteristics of false target points (such as reflection intensity and contour structure) are easily confused with the echo signals of real personnel, causing the system to misjudge inanimate objects such as rubble and metal remnants as trapped personnel. This seriously interferes with the results of personnel search and rescue assessments, misleads emergency resource dispatch routes, delays the rescue time for truly trapped targets, and may even trigger secondary safety risks.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide an unmanned aerial vehicle (UAV) MiniSAR real-time imaging and emergency communication system for disaster relief. By using multi-moment radar path modeling and image stability assessment, it accurately identifies and eliminates false target points under multi-path interference. Combined with a dynamic extension algorithm, it achieves image structure repair, constructs a dynamic confidence distribution map, realizes high-confidence target positioning and rescue area optimization, and enhances the system's anti-interference and response accuracy, thereby solving the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a UAV-borne MiniSAR real-time imaging and emergency communication system for disaster relief, comprising a radar path inversion modeling module, a multipath interference identification module, a target stability assessment module, a false target removal and image purification module, an image structure repair module, and a target update and response linkage module:

[0007] The radar path inversion modeling module performs multi-time flight over the target area, collects radar echo signals at multiple times, and simultaneously acquires flight attitude data and imaging azimuth information of the flight platform. Based on the flight attitude data and imaging azimuth information, it inverts the propagation path of the radar echo signal, extracts the incident angle and reflection path length, and generates a reflection path distribution map.

[0008] The multipath interference identification module calculates the arrival time difference and phase offset of radar echo signals based on the reflection path distribution map, extracts the path overlap and signal divergence of image pixels, and identifies the interference-sensitive areas where multipath echoes converge.

[0009] The target stability assessment module analyzes the reflection intensity changes, phase fluctuations, and spatial position consistency of image pixels in the interference-sensitive area in the radar echo sequence at multiple times, calculates the target stability score, and generates a preliminary authenticity judgment map.

[0010] The false target removal and image purification module initially determines the target stability score of each image point in the image based on its authenticity, filters out image pixels with scores below the threshold that are located in interference-sensitive areas, determines them as false target image points, performs image removal operation, and outputs the image mask after structural purification.

[0011] The image structure restoration module extracts the edge structure and texture distribution trend of the surrounding area based on the image mask, and uses a dynamic extension algorithm to perform structural completion and image detail reconstruction on the removed area to generate the image structure restoration result.

[0012] The target update and response linkage module performs pixel-level fusion comparison between the image structure repair results and the original image to construct a dynamic confidence distribution map, update the location information of the trapped target and the priority rescue area, and form a closed-loop linkage mechanism for identification and response.

[0013] Preferably, the radar path inversion modeling steps are as follows:

[0014] Unmanned aerial vehicles equipped with microwave synthetic aperture radar will be deployed over the target area to perform multi-time coverage flight missions.

[0015] At each moment of flight, radar echo signals are collected, and flight attitude data and imaging azimuth information are recorded simultaneously.

[0016] Based on the flight attitude data and imaging azimuth information at each moment, the outgoing direction and propagation path of the radar wave are inverted, and the corresponding incident angle and reflection path length are calculated.

[0017] By integrating path parameters for all pixels within the image region, a reflection path distribution map is constructed, which includes incident angle, path length, and path dispersion indices.

[0018] Preferably, the multipath interference identification steps are as follows:

[0019] For each radar echo path, the theoretical arrival time is calculated based on its reflection path length and compared with the actual reception time to obtain the time difference of arrival.

[0020] Extract the complex components of the radar echo signal for each path, calculate their phase offset, and form a phase perturbation map;

[0021] Based on the time difference spectrum and phase perturbation map, the path overlap and signal divergence of each image pixel are calculated.

[0022] By jointly analyzing path overlap and signal divergence, a multipath interference sensitivity heatmap is generated, and interference-sensitive areas are delineated.

[0023] Preferably, the target stability assessment steps are as follows:

[0024] Extract the radar echo intensity sequence of each image point at multiple time points and calculate the amplitude of reflection intensity change;

[0025] Extract the radar echo phase value of each image point at multiple time points and calculate the degree of phase fluctuation.

[0026] Analyze the spatial projection position of each image point at multiple time points and calculate the spatial position consistency.

[0027] A target stability scoring function is constructed based on the magnitude of reflection intensity change, the degree of phase fluctuation, and the consistency of spatial location, generating a preliminary determination map of the authenticity of image points.

[0028] Preferably, the specific steps for false target removal and image purification are as follows:

[0029] Statistical analysis was performed on the target stability score of each image point, and a stability score threshold was set.

[0030] Image points with scores below a threshold and spatial locations within interference-sensitive areas are selected and a pseudo-target point mask is generated.

[0031] Perform edge consistency screening and hole repair operations on the pseudo-target point mask image to construct a structurally optimized image purification mask;

[0032] The image is cleaned by masking corresponding image points in the original image and outputting the image mask after structural cleansing.

[0033] Preferably, the image structure restoration steps are as follows:

[0034] By scanning the image mask using edge detection operators, pixel boundary lines from the real area to the removed area are extracted, and an edge distribution point set is constructed.

[0035] Based on the edge distribution point set, the texture gradient direction in the boundary neighborhood is analyzed, and the texture direction consistency coefficient and edge direction continuity index are calculated.

[0036] Based on the obtained indicators, a dynamic extension algorithm is used to perform pixel-by-pixel completion along the main texture direction to complete the structural filling of the hole areas in the image;

[0037] The edges, grayscale, and textures between the repaired area and the real area are smoothed to output the image structure repair result.

[0038] Preferably, the target update and response linkage steps are as follows:

[0039] The image structure restoration result is pixel-level registered and weighted fusion with the original image to generate a fused image;

[0040] Based on the fused image, the confidence value is calculated pixel by pixel by combining stability score, repair degree, image intensity contrast, edge continuity and texture consistency to construct a dynamic confidence distribution map;

[0041] Based on the high-confidence areas in the dynamic confidence distribution map, the trapped target is relocated and its contour is extracted, and the location information and priority rescue ranking results are output.

[0042] The target location information and priority rescue ranking results are uploaded to the mission dispatch command terminal, linking the trajectory planning, communication deployment and path simulation processes to form a closed-loop mechanism for identification and response.

[0043] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0044] This invention constructs a reflection path distribution map based on multi-time radar data and combines it with multi-dimensional indicators such as path overlap, signal divergence, and image point stability scores to accurately identify and eliminate "false target points" caused by multi-path interference. This solves the problem of recognition failure caused by misjudging trapped personnel due to highly reflective objects in existing technologies. Furthermore, a dynamic extension algorithm is used to repair the image structure boundaries and complete the texture, ensuring image continuity and structural consistency. Then, by fusing the original and repaired images at the pixel level, a dynamic confidence distribution map is constructed, enabling high-confidence positioning of trapped targets and intelligent updating of priority rescue areas. This method not only establishes a closed-loop data linkage mechanism between radar image recognition and rescue mission execution but also significantly enhances the system's anti-interference capability, interpretation reliability, and resource scheduling accuracy in complex environments. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0046] Figure 1 This is a schematic diagram of the modules of the UAV-borne MiniSAR real-time imaging and emergency communication system for disaster relief according to the present invention. Detailed Implementation

[0047] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0048] This invention provides, for example Figure 1 The UAV-borne MiniSAR real-time imaging and emergency communication system for disaster relief, shown, includes a radar path inversion modeling module, a multipath interference identification module, a target stability assessment module, a false target removal and image purification module, an image structure restoration module, and a target update and response linkage module.

[0049] The radar path inversion modeling module performs continuous multi-time flight over the target area, collects radar echo signals at multiple times, and simultaneously acquires flight attitude data and imaging azimuth information of the flight platform. Based on the flight attitude data and imaging azimuth information, it inverts the propagation path of each radar echo signal, extracts the corresponding incident angle and reflection path length, and generates a complete reflection path distribution map.

[0050] To effectively identify false target points under multi-path interference conditions in extreme disaster scenarios such as urban earthquakes, it is first necessary to acquire radar echo signals of the target area at multiple consecutive time points. Then, combining this with the dynamic attitude information and imaging azimuth data of the flight platform, the propagation path of the radar echo signal at each time point is accurately inverted to construct a complete reflection path distribution map. This includes the following operational steps:

[0051] Unmanned aerial vehicles (UAVs) equipped with microwave synthetic aperture radar (SAR) will be deployed over the target disaster area to perform multi-time coverage flight missions. This mission employs a pre-defined flight trajectory planning strategy, enabling the UAV to illuminate and image the same ground target area from different spatial angles at different flight times. During flight, radar echo signals are continuously collected at different time points, and the corresponding flight attitude information, including six degrees of freedom parameters such as pitch angle, yaw angle, roll angle, geographical location, flight altitude, and heading angle, is recorded in real time through onboard navigation equipment and an inertial measurement unit. Simultaneously, imaging azimuth information, such as the angle between the camera imaging axis and the radar main beam direction, must also be collected. The radar illumination process and its reception response at each moment are closely related to the UAV's spatial attitude at that moment; therefore, high-precision recording with spatiotemporal synchronization must be ensured during data acquisition.

[0052] Based on the collected radar echo data and the corresponding attitude and imaging azimuth information, path reconstruction analysis is performed on each radar echo signal using a ground station or remote computing center. Specifically, for each signal transmission point, the outgoing direction of the radar beam is calculated based on the flight attitude and heading data at that time; then, the first reflection path that the radar wave may experience after propagating to the ground is calculated through inversion, considering the possible multiple reflection paths formed by high reflectivity structures on the ground (such as metal rubble, steel rebar ruins, etc.), and based on the known ground digital elevation model and building debris distribution information, the path structure that the radar wave may pass through from the transmission point to each echo receiving point is simulated to form a multipath propagation estimation map.

[0053] For each radar signal propagation path, the theoretical arrival time and corresponding propagation delay are calculated based on its geometric path length and the phase velocity of the radar wave in different media. Simultaneously, considering the phase change of the electromagnetic wave during reflection, the phase offset introduced by each path is derived. Through statistical analysis of the propagation time and phase difference of all paths, the structural characteristics of the reflection path in the current scene are further derived. Each ground image unit corresponds to a radar image pixel. The parameters of multiple echo paths received by this pixel can be centrally calculated to extract the path incident angle distribution and reflection path length set corresponding to that pixel. This set of parameters is stored in a path distribution description table as the basic dataset for subsequent interference identification.

[0054] After analyzing the path parameters of all pixels, a complete reflection path distribution map is generated based on the integrated path information of the entire image region. This map is presented as a two-dimensional image, with each pixel labeled with key parameters such as the main incident angle, average path length, path dispersion index, and path quantity statistics. Color or grayscale encoding is also used to visualize the path structure complexity. This reflection path distribution map comprehensively reflects the distribution of multipath signals and potential interference risk zones within the entire imaging region, providing a physical modeling basis and spatial constraint information for subsequent false target point identification and stability scoring analysis. This step ensures a high degree of consistency and physical authenticity in the mapping process from imaging data to path geometry, thereby improving the recognition accuracy of the image processing results.

[0055] The core function of this step is to provide a physically and geometrically meaningful radar signal propagation model to support the subsequent identification and elimination of false target points caused by multipath reflection in disaster scenarios. Specifically, in complex disaster environments such as urban earthquakes and fires, building collapses create a large amount of highly reflective rubble, metal structures, and irregular boundaries. These structures cause radar waves to undergo multiple reflections and scatterings in the target area, resulting in non-ideal multipath propagation phenomena that severely interfere with radar imaging accuracy and the reliability of image interpretation. To accurately identify which image points may be affected by multipath interference, it is necessary to precisely invert the radar echo signal propagation path at each moment and, combined with the flight platform's attitude data and imaging azimuth information, reconstruct the radar beam's propagation trajectory in three-dimensional space. By extracting the incident angle and reflection path length of each path, a correspondence between image pixels and radar propagation paths can be established, thereby identifying areas where the signal may experience path overlap, multiple reflections, or angular deflections during propagation. This path inversion based on physical propagation mechanisms not only reflects the spatial distribution trend of signal disturbance but also provides structural parameters for subsequent calculations of indicators such as path overlap, signal divergence, and image point stability. Therefore, this step is not only fundamental to radar image interference modeling but also a prerequisite for constructing a complete interference identification and image purification algorithm system, playing a crucial role in achieving accurate image restoration and real target extraction.

[0056] The multipath interference identification module, based on the reflection path distribution map, calculates the arrival time difference and phase offset of each radar echo signal, extracts the path overlap and signal divergence parameters of each pixel in the image, and identifies interference-sensitive areas with multipath echo aggregation characteristics.

[0057] To effectively identify radar image false target interference areas caused by multipath effects in urban disaster scenarios, it is necessary to conduct more in-depth time and phase domain analysis of radar echo signals based on the constructed reflection path distribution map, thereby quantifying the multipath interference characteristics carried by each image pixel. This implementation includes the following steps:

[0058] For each radar echo path, the theoretical arrival time is calculated based on its reflection path length and the constant propagation speed of radar waves in free space. This time is then time-registered with the actual received radar echo signal to determine the arrival time difference between each path. This time difference not only reflects whether the signal experienced propagation delay on different paths but also reveals the concentration of arrival times of multiple echoes at a given pixel. For direct paths generated by real targets, the arrival times of the echo signals are highly concentrated, while echo signals generated by non-direct paths exhibit discrete time delays. By comparing the arrival times of all paths, a time difference spectrum for each pixel can be established, providing fundamental data for subsequent interference clustering assessment.

[0059] After calculating the time difference, the phase information of the radar wave signal carried by each reflection path is further analyzed. Since electromagnetic waves propagate along different paths, causing different phase changes, especially after passing through multiple refraction and reflection surfaces, phase disturbances are often more severe. By extracting the complex signal components of the radar echo and calculating their phase offset relative to the main path signal, a phase disturbance map is constructed. This phase disturbance map can intuitively reflect the complexity of the path the signal experiences during propagation, especially at path intersections and multi-refraction path points, where significant phase abrupt changes are likely to occur. These areas usually coincide highly with areas prone to multipath interference.

[0060] By combining the aforementioned time difference spectrum with the phase perturbation map, for each radar image pixel, the path concentration and phase consistency indices of all associated reflection paths are statistically analyzed to calculate the path overlap and signal divergence of that pixel. Path overlap refers to a weighted comprehensive score of the directional angle, path length difference, and spatial projection overlap between the received echo paths at that point, reflecting whether the signal propagation path has significant multipath overlap characteristics. Signal divergence measures the standard deviation and fluctuation frequency of the radar signal phase offset carried by different paths at that point, used to judge the stability and coherence of the signal. Real target areas typically exhibit high path overlap but low signal divergence, while false target cluster areas exhibit a combination of high path overlap and high signal divergence.

[0061] Based on the above calculation results, the path overlap and signal divergence of all pixels are jointly visualized and analyzed in the image domain to construct a multipath interference sensitivity heatmap. In this heatmap, regions that simultaneously meet the conditions of path overlap exceeding a preset upper limit and signal divergence greater than a specific threshold are marked, and high-sensitivity regions for multipath interference are defined according to their heat level. These regions typically appear near dense building ruins, piles of metal rubble, or high-rise structural remains, and are the spatial locations where false target signals are most likely to accumulate.

[0062] The purpose of this step is to accurately identify high-risk areas with multipath echo interference in radar images under complex electromagnetic environments such as urban disasters, providing a basis for subsequent false target point identification and image purification. In disaster scenarios such as earthquakes, target areas often have large amounts of highly reflective objects such as metal rubble and steel rebar debris. These structures easily cause radar waves to undergo multiple reflections and scattering during propagation, forming complex multipath echo paths. When these multipath signals are superimposed at the receiver, they may produce strong echoes in the radar image that are inconsistent with the real target, leading to false image information. To effectively identify these interference areas, this step first quantifies the signal propagation delay characteristics by calculating the arrival time difference corresponding to each reflection path; then, it further analyzes the phase shift between each path to reveal the consistency of signal propagation. Combining the time difference and phase perturbation, an index of path overlap and signal divergence can be constructed for each image pixel. Areas with higher path overlap and more severe phase fluctuations are more likely to be multipath interference clusters. By performing this type of index analysis on the entire imaging area, spatial positioning of interference-sensitive areas can be achieved. This operation not only improves the accuracy of radar image interference modeling, but also establishes the first screening mechanism for target authenticity analysis, which is a key intermediate judgment link in the entire image purification and target recognition process.

[0063] The target stability assessment module analyzes the reflection intensity variation, phase fluctuation degree, and spatial position consistency of each image point in the interference-sensitive area in the radar echo sequence at multiple times, calculates the corresponding target stability score, and generates a preliminary determination map of the authenticity of the image point based on the score.

[0064] To more accurately determine whether each pixel in the image contains false target features caused by multipath interference, it is necessary to conduct in-depth analysis of the signal response characteristics of each image point in the radar imaging sequence at multiple time points within the identified interference-sensitive area. This step aims to construct a stability scoring model by extracting time-series features and spatial consistency indicators, ultimately generating a preliminary determination map of the authenticity of image points. This implementation method includes the following steps:

[0065] Target image points located within interference-sensitive areas are selected, and their corresponding radar echo intensity value sequences at multiple different times are extracted. Since the MiniSAR radar system images the same ground area multiple times during flight from different perspectives and at different times, each image point will have corresponding echo reflection intensities at multiple time points. By calculating the amplitude of the reflection intensity variation of this image point in the time dimension, the consistency of its signal response can be determined. Typically, real ground targets (such as human bodies or stable structures) exhibit small fluctuations in echo intensity across multiple times, demonstrating stable reflection characteristics. However, pseudo-target points caused by multipath interference show large fluctuations in their intensity sequences due to path changes and phase interference, exhibiting high instability. Therefore, this step uses the amplitude of intensity variation as the first stability evaluation index.

[0066] For the radar echo signals of the aforementioned image points at various time points, their phase information is further extracted, and the degree of phase variation in the time series is analyzed. By calculating the phase value of the complex radar echo signal at each time point, a time series phase variation curve is obtained. A phase fluctuation index is constructed by statistically analyzing the range of variation and the frequency of phase jumps. At the location of the true target point, the echo phase changes slowly due to the stable path, resulting in a smooth curve; while at the false target point, the unstable path causes frequent irregular phase jumps in the echo. By setting a phase fluctuation amplitude threshold, image points with significant phase disturbances can be identified, providing a second dimension parameter for subsequent scoring.

[0067] To comprehensively consider spatial geometric location factors, it is necessary to analyze whether each image point consistently projects to the same physical spatial location across multiple time points. Due to the viewing angle differences in radar imaging, if an image point projects to a relatively consistent ground position across multiple viewing angles, it indicates that the target truly exists and its structure is continuous. Conversely, if the position of the same image point shifts significantly in multiple imaging iterations, it may be due to multipath interference or path distortion leading to false target imaging. By analyzing the spatial overlap rate of the projection results of image points across multiple time points, a spatial consistency index is constructed. Points with high spatial consistency have stronger credibility, while points with large spatial drift have lower credibility. This index serves as the third component parameter of the stability score, further enhancing the overall accuracy of the judgment model.

[0068] By combining the three indicators mentioned above—the amplitude of reflection intensity variation, the degree of phase fluctuation, and spatial consistency—a target stability scoring function is constructed using a weighted function to score each image point within the interference-sensitive area. Image points with high scores are judged as target points with high realism, while those with low scores are judged as possible false target points. Based on this scoring result, a preliminary image point authenticity assessment map is generated. This map visualizes the distribution of target credibility within the image area in terms of heat or grayscale, serving as the basis for subsequent image purification and false target removal.

[0069] The main function of this step is to assess the credibility of image points located within interference-sensitive areas by analyzing the temporal characteristics and spatial consistency of radar imaging data at multiple time points, thereby determining whether they possess genuine target attributes. This processing is based on a spatiotemporal joint judgment mechanism constructed from radar imaging principles and electromagnetic wave propagation characteristics, aiming to address multipath interference problems caused by complex structures at disaster sites. In practical application scenarios, such as urban earthquake or building collapse areas, highly reflective objects such as steel rebar debris and metal rubble are common in the environment. These structures easily trigger multiple reflections of radar waves, thus forming false high-intensity echo points in the image. Traditional SAR image processing methods often cannot distinguish the radar echo characteristics of these false target points from those of real trapped personnel, easily leading to misidentification and interfering with rescue route planning. This step comprehensively constructs a "target stability score" index by analyzing the amplitude of radar echo intensity changes, the continuity and frequency of phase change curves, and the spatial consistency of imaging positions at different viewpoints for image points at multiple time points. A higher score indicates that the image point has a stable temporal response, continuous phase, and a high degree of spatial overlap, thus increasing its reliability. Conversely, a low score suggests that the point is more likely to be a false target caused by multipath interference. By establishing such a quantification mechanism, not only can the scientific accuracy of false target identification be improved, but a mathematical basis and judgment boundary can also be provided for subsequent image cleanup processing, significantly enhancing the practicality and robustness of radar image analysis in disaster emergency response.

[0070] The false target removal and image purification module initially determines the target stability score of each image point in the image based on its authenticity, sets a score threshold, filters out image points with stability scores below the threshold and located in interference-sensitive areas, determines them as false target image points, performs image removal operation, and outputs the image mask after structural purification.

[0071] An image mask is a binary or multi-valued image used to mark specific regions in an image. Essentially, it's an auxiliary layer of the same size as the original image, where each pixel value represents the usage status or semantic attribute of the corresponding location in the original image. In radar image false target removal and image inpainting applications, the main function of an image mask is to distinguish which pixels belong to credible real target areas and which pixels have been identified as interference components (such as false target points) and need to be masked. Typically, a pixel value of "1" in the mask represents the area to be retained (i.e., valid image content), and a value of "0" represents the area to be removed or repaired (i.e., invalid or interference parts). By applying masks, selective processing of the original image content can be achieved, such as avoiding false targets in target recognition calculations, accurately locating areas to be repaired in image inpainting, or compressing redundant information in data transmission. As a bridge between structural processing and semantic separation, image masks can significantly improve the accuracy and efficiency of subsequent image analysis, reconstruction, and interpretation operations, making them a fundamental control tool in robust image processing workflows.

[0072] To further eliminate false target interference in radar images caused by multipath interference, it is necessary to reasonably classify the stability scores of image points based on the generated preliminary authenticity assessment map, and, based on certain judgment rules, identify and remove image points with low confidence, thereby obtaining a structurally cleaned image mask. This includes the following steps:

[0073] Statistical analysis was performed on the target stability score of each image point in the preliminary authenticity assessment image, calculating statistical characteristic parameters such as the mean, median, and standard deviation of the score across the entire image. Based on these statistical characteristics, and considering the fault tolerance requirements of the target detection task and the complexity of interference in the field environment, a threshold for the stability score was set. This threshold should not be a fixed constant but should be dynamically adjusted according to different application scenarios. For example, in areas with dense buildings or many residual metal structures, a stricter threshold control strategy is recommended to improve the robustness of false target identification. The scoring threshold can be set using quantile methods (lower quartile minus one standard deviation) or K-means clustering boundary points to ensure that the screening logic has data-driven characteristics and environmental adaptability.

[0074] Based on a set scoring threshold, all image points in the initial authenticity assessment image are traversed and filtered to identify those that meet two conditions: first, their target stability score is lower than the set threshold; and second, their spatial location falls within the aforementioned interference-sensitive area. Only image points that simultaneously meet both conditions are judged as highly suspicious pseudo-target image points. This dual-condition judgment strategy effectively avoids the erroneous rejection of edge-credible target points, improving the processing accuracy and confidence. During the filtering process, each candidate point is marked, and a binary result image is formed in pixel space, where the area judged as a pseudo-target point is set to "0", and the reserved area is set to "1", thus initially constructing a pseudo-target point mask image.

[0075] Based on the generated pseudo-target point mask image, edge consistency screening and hole repair operations are performed on the mask image to ensure structural integrity in subsequent image purification processing. Specifically, connected component analysis is used to determine whether there are isolated pseudo-target misclassification points caused by local scoring biases. Combined with regional morphological features such as area, perimeter, and compactness, points that clearly do not conform to spatial distribution patterns are screened out. In addition, for single credible points interspersed in continuous pseudo-target point regions (which may be caused by noise or algorithm drift), structural consistency correction should also be performed through a local voting mechanism to enhance the structural coherence and semantic rationality of the mask, ultimately forming an image purification mask with optimized structure.

[0076] Based on structural purification masks, pixel-level masking operations are performed on the original radar images to shield all image points marked as false targets, thus eliminating their participation in subsequent image recognition, target localization, and command decision-making. The elimination operation employs a non-destructive processing method, preserving the original image data while establishing an independent mask control layer to support subsequent data fusion, image restoration, and manual verification. This process not only removes false information caused by multipath interference but also effectively compresses the transmission bandwidth of redundant interference data, improving overall recognition accuracy and decision-making response efficiency.

[0077] The purpose of this step is to further identify and remove potentially false target interference areas in the image based on the previously constructed preliminary authenticity assessment map, thereby generating an image mask with structural purification features. This provides high-confidence data support for subsequent target reconstruction and rescue decisions. Specifically, in complex disaster environments such as urban earthquakes and fires, highly reflective objects such as collapsed buildings and metal debris can easily cause multiple scattering and multipath propagation of radar signals, resulting in "false target points" in radar images that have high reflection intensity and clear shapes but are not actually present. Although the authenticity of image points can be assessed through a stability scoring mechanism, the interference intensity varies significantly in different regions. Without regional screening, this could lead to judgment bias. Therefore, this step combines two dimensions: scoring threshold and spatial location, to perform dual screening. Only image points in interference-sensitive areas with significantly low stability scores are retained as false target points, greatly improving the accuracy and robustness of the removal operation. Based on this, these false target areas are removed from the original radar image using an image mask, generating a purified layer with good structural continuity and natural edge transitions. This layer can be used for subsequent image restoration and target reconstruction, and can also serve as auxiliary data input to the command system for personnel search and rescue path optimization and task priority adjustment. In summary, this step plays a crucial role in filtering and structural restoration within the entire false target suppression mechanism. It is a core link in realizing the transformation from multipath contaminated images to high-reliability imaging results, possessing strong practical value and technological innovation.

[0078] The image structure restoration module extracts the edge structure and texture distribution trend of the surrounding area based on the image mask, and uses a dynamic extension algorithm to complete the boundary structure and reconstruct the image details of the removed image area, and outputs the image structure restoration result.

[0079] To improve the overall readability and structural continuity of radar images after false target removal, effective boundary completion and detail reconstruction of the removed regions are necessary to ensure the final image possesses complete visual coherence and spatial semantic consistency. This step, based on the generated image mask, uses edge structure extraction and texture extension algorithms to reconstruct believable image content within the image's hole regions. The entire process includes the following steps:

[0080] Based on the image mask results, edge localization is performed on the regions marked as pseudo-target points, and the image regions along their outer boundaries are extracted as edge support regions for structural completion. Specifically, an edge detection operator is used to scan the image mask, identifying all pixel boundary lines transitioning from the "real region" to the "removed region," resulting in a set of well-defined and geometrically coherent edge distribution points. Edge structure extraction not only requires identifying first-order grayscale variation information but also combining it with the texture gradient distribution trend in the image to extract local spatial gradient directions and grayscale variation patterns, providing directional constraints and texture priors for subsequent image detail completion.

[0081] To scan an image mask using edge detection operators and accurately identify all pixel boundaries transitioning from the "real area" to the "cancelled area," the following steps are required: First, normalize the image mask data to a single-channel binary image, setting pixel values ​​in the real area to 1 and pixel values ​​in the cancelled area to 0. Second, perform a convolution scan on the binary image using an edge detection operator (such as Sobel, Prewitt, Roberts, or the more advanced Canny operator) to extract boundary regions where pixel values ​​undergo significant jumps; the Sobel operator calculates gradients in both the horizontal and vertical directions to obtain edge direction information. Third, in the detected edge response map, filter out blurry or discontinuous edge points based on the gradient direction information, retaining edge pixels with high gradient magnitudes, clear directions, and good continuity, forming a candidate set of edge points. Finally, apply a pixel connectivity analysis algorithm (such as eight-neighborhood connectivity) to aggregate these edge points, forming a set of geometrically coherent edge segments or edge point sets, marked as transition boundaries from the real area to the cancelled area. These edge distribution points not only contain location information but can also be appended with geometric attributes such as gradient direction and local curvature to guide the texture extension and structural completion processes in subsequent image reconstruction. Through this step, the interface between the real region and the hole in the pseudo-target in the image mask can be accurately depicted, forming the key input condition for the image structure restoration algorithm.

[0082] Based on the extracted edge structure features, texture modeling analysis is performed on the boundary's adjacent region. Multiple concentric annular texture distribution windows are established within the neighborhood of the edge region, and parameters such as grayscale distribution, spatial frequency, and directional gradient are statistically analyzed to generate a multi-scale texture description vector. By calculating the texture direction consistency coefficient and edge direction continuity index, the potential image extension trend of the removed region is determined. The core of this step lies in transforming the task of filling image holes from pure pixel interpolation into texture direction propagation guided by the boundary structure, thereby ensuring that the filled region possesses natural extension at the image visual level.

[0083] The purpose of calculating the texture direction consistency coefficient and edge direction continuity index is to accurately determine the structural extension trend of the removed areas in the image, ensuring a natural transition between texture filling and edge structure during image restoration, without breaks or discontinuities. The specific steps are as follows:

[0084] Multiple local windows (such as 5×5 or 7×7) are selected within the edge neighborhood of the removed region. The direction of the pixel gradient in each window is statistically analyzed. The Sobel or Scharr operators can be used to calculate the gradient components in the horizontal and vertical directions respectively, thereby obtaining the principal gradient direction angle (θ) of each window, which is the principal direction of the local texture.

[0085] Cosine vector mapping is performed on the main direction values ​​of all windows, and then the mean magnitude of their vectors is calculated to obtain the texture direction consistency coefficient. The closer the value is to 1, the more consistent the texture direction of each region is; the lower the value, the more discrete the direction is and the texture is discontinuous.

[0086] The contour lines composed of edge pixels are extracted, and the edge direction is modeled using polynomial fitting or least squares method. Parameters such as the rate of change of direction, curvature continuity and offset stability of the fitted curve are calculated to obtain the edge direction continuity index.

[0087] Combining these two indicators helps determine the image extension trend within the removed area: a high texture direction consistency coefficient and good edge continuity indicate a clear and coherent extension direction, allowing for guided dynamic extension and filling along the main direction; conversely, texture fusion or multi-directional sampling strategies are needed to prevent structural abrupt changes. These steps not only quantify the continuity of the image structure but also provide directional support for subsequent image inpainting, achieving more natural and accurate structural restoration.

[0088] A dynamic expansion algorithm is employed to perform pixel-by-pixel completion of the removed image regions. Starting from image edge points, the algorithm expands pixel-level according to their local gradient directions and the principal axis of the texture. During each expansion step, target pixel values ​​are generated based on the texture features of the completed region using methods such as weighted averaging, pattern matching, and directional interpolation. Spatial consistency and structural compatibility constraints are applied to each filling point to ensure that the image structure does not experience abrupt changes or artifacts during the filling process. Dynamic expansion can not only repair small-scale structural holes but is also suitable for coherent completion needs with irregular edge shapes, making it particularly suitable for processing complex radar image structures at disaster sites.

[0089] In the image restoration process described in this invention, the "dynamic extension algorithm" refers to a pixel-by-pixel guided filling method based on boundary structure and texture direction. Its core idea is to simulate the "natural growth" of image content from known to unknown areas, thereby achieving structural reconstruction and texture detail restoration of the removed image regions. The algorithm's role here is to fill the image voids caused by false target removal in a coherent, reasonable, and high-fidelity manner, ensuring that the restored image does not exhibit structural breaks, textural distortion, or abrupt boundary transitions.

[0090] The specific completion operation includes the following steps:

[0091] The boundary of the removed region is initialized as the starting line for repair. Several points with the most stable local features are selected from the set of boundary points as "priority filling seed points", and their spatial coordinates, gradient direction and main texture direction are recorded.

[0092] A local window (e.g., a 5×5 or 7×7 neighborhood) is constructed centered on each seed point. Based on its texture gradient and edge direction, the most similar local image patch in the known region is searched as a reference template.

[0093] The pixel values ​​in the reference template are mapped to the current position to be filled in the culled area. The filling value is calculated by directional interpolation, weighted averaging or texture synthesis. At the same time, the gradient direction and texture label at this position are updated to maintain the consistency of the extension direction and the continuity of the texture mode.

[0094] The filling area is advanced along the normal direction of the repair boundary, and the boundary point set is dynamically updated. This process is repeated until the entire removed area is completely filled with pixels. The entire extension process adopts a stepwise growth strategy, which makes the image content continue naturally in spatial structure and is continuous and coordinated in texture representation. This ensures that the repaired image not only conforms to the realism of visual perception but also has structural coherence. It is a key step in achieving semantic integrity and geometric consistency of the image.

[0095] After the image completion process is complete, a smoothing process is performed on the entire repaired area and its surrounding real image areas to output the image structure restoration result. This smoothing operation includes three aspects: edge blending, grayscale coordination, and texture transition. First, an edge filter eliminates abrupt gradient changes at the boundary between the old and new areas. Second, local histogram matching is performed on the grayscale mean of the filled area to ensure its overall brightness level is consistent with the surrounding areas. Third, a texture similarity mapping is constructed to correct the differences in frequency domain between the detailed textures and the original image, thus ensuring a natural transition at the pixel level and maintaining structural continuity and uniformity in the restored image. The final output image structure restoration result possesses completeness, realism, and recognizability, effectively enhancing the interpretive value of the image and providing high-quality image input for subsequent image-based target localization, dynamic analysis, and rescue response.

[0096] This step aims to address the "holes" or structural breaks left in the removed areas of the image after false target point removal. Through edge structure analysis and texture trend modeling, a dynamic extension method is used to continuously repair the image content, ensuring the overall structural integrity and visual coherence of the image. Since false target points typically exhibit strong reflectivity and high contrast during radar image processing, direct removal can easily cause contour interruptions, edge distortion, or texture loss, affecting the accuracy of subsequent image recognition, target localization, and decision analysis. Based on an image mask, this step first extracts the edge direction and texture orientation of the real area surrounding the removed region. By modeling the grayscale gradient, texture frequency, and principal direction angle of these local areas, the natural spatial extension trend of the image content is analyzed. On this basis, the dynamic extension algorithm guides the image content "inward" pixel by pixel, guiding edge curves and smooth textures from the real area into the removed area, achieving coordinated structural and textural completion. This process not only preserves the original geometric shape of the image but also effectively reduces the "information vacuum" effect caused by removal, enhancing the semantic integrity and interpretability of the image. The final output image structure restoration result can achieve high-quality reconstruction of image content without introducing artifacts or noise, providing a reliable visual data foundation for accurate interpretation of radar images and rescue missions at disaster sites. Therefore, this step is a key technical link in transitioning from "false target removal" to "high-quality reconstruction," possessing high practicality and engineering value.

[0097] The target update and response linkage module performs pixel-level fusion comparison between the image structure repair result and the original image, constructs a dynamic confidence distribution map based on the fusion result, updates the location information of the trapped target and the priority rescue area, and forms a closed-loop linkage mechanism between the target recognition result based on radar image recognition and the rescue mission response strategy.

[0098] To further realize the intelligent recognition decision-making closed loop after radar image false target removal and structural repair processing, the image structural repair results need to be fused and compared with the original image at the pixel level to construct a dynamic confidence distribution map. Based on this map, the location information of the trapped target and the priority rescue area are updated, ultimately forming a closed-loop linkage mechanism between target recognition results based on radar image recognition and rescue mission response strategies. This step includes the following steps:

[0099] The structurally repaired image is precisely registered and fused with the original radar image at the pixel level. During image alignment, a combination of feature point registration and geometric projection mapping is used to ensure complete consistency of the image structure in spatial coordinates. Subsequently, the two images are superimposed using a weighted fusion method, where the weights are adaptively adjusted based on the stability score of the image region, the repair intensity, and the original signal strength. For regions with high stability and good original image quality, the original image information is retained first; for regions where false targets have been removed and their neighborhoods, more information from the repaired image is preserved. The fused image not only integrates the true reflection features of the original radar imaging but also encompasses the correction results of false target removal and structural completion, constructing comprehensive image data with greater physical realism and structural integrity.

[0100] Based on the fused image, the confidence value of each point in the image is calculated pixel-by-pixel to construct a dynamic confidence distribution map. The confidence calculation incorporates multiple indicators, including pixel stability score, degree of participation in restoration, image intensity contrast, edge continuity, and texture consistency. Specifically, pixels with high restoration levels, natural boundary transitions, and smooth texture extension in the removed areas are assigned higher confidence values; while areas with obvious restoration traces, structural abrupt changes, or low texture similarity are assigned lower confidence values. This confidence distribution map is presented in a two-dimensional matrix format and visualized using color heatmaps or grayscale images, enabling commanders to clearly understand the credibility and reliability of information in each region of the image, providing quantitative support for subsequent intelligent analysis and manual intervention.

[0101] Based on the high-confidence regions in the confidence distribution map, a joint target recognition algorithm is used to relocate trapped individuals in the image. The same detection process as the previous target recognition algorithm can be used here, but the confidence map is introduced as a filtering weight factor. Target recognition and contour analysis are only performed in areas where the confidence value exceeds a set threshold, thus avoiding false target areas misleading the recognition model and improving recognition accuracy. Based on the relocation, geographic coordinate transformation and scene semantic analysis are further combined to extract the precise spatial location of each trapped target. A priority ranking list is generated based on factors such as confidence level, entrapment depth (which can be calculated from a self-collapse model), and density of surrounding obstacles, serving as the basis for rescue dispatch decisions.

[0102] The updated target location information and priority rescue ranking results are uploaded to the mission scheduling command terminal, linking emergency resource management processes such as UAV trajectory planning, communication relay deployment, and ground rescue route simulation, forming a complete response closed loop from radar image processing → target location and identification → confidence assessment → mission command push. This closed-loop mechanism can not only dynamically optimize target identification results and improve the reliability and effectiveness of radar image identification in complex disaster scenarios, but also realize data-driven collaborative command and dispatch, ensuring that rescue resources are efficiently allocated to the most needed locations, and maximizing the response speed and execution efficiency of post-disaster rescue.

[0103] This step involves precisely fusing and comparing the image after false target removal and image structure restoration with the original radar image. It extracts confidence features of changed and stable regions, further constructing a dynamic confidence distribution map. This map is used to reassess and optimize the location of trapped targets and corresponding priority rescue areas, ultimately achieving a data-driven closed-loop linkage mechanism between image recognition and emergency response. In real-world disaster scenarios, especially in complex and information-chaotic environments such as earthquakes, fires, or building collapses, traditional radar images are often subject to strong multipath reflections and false target interference, leading to significant errors in identifying personnel locations. This step allows the system to compare pixel differences, stability scores, texture consistency, and edge continuity between the restored image and the original image, quantitatively modeling the confidence of each region in the image and constructing a spatially continuous image confidence map with a clear confidence distribution. This map can not only eliminate potential recognition errors in low-confidence regions but also guide the target re-identification process, limiting the execution scope of the recognition algorithm to high-confidence regions, thereby significantly improving the accuracy and reliability of trapped target identification. Furthermore, the spatial information carried by the confidence map can be directly mapped to task priority reference coordinates, providing intuitive and quantifiable regional guidance for the rescue dispatch system and realizing closed-loop control from image perception to decision execution. This mechanism effectively opens up the information channel between the image processing end and the task response end, not only improving the disaster emergency system's real-time judgment capability of the target status, but also enhancing the scientific nature of resource allocation and the efficiency of rescue operations, possessing high application value and strategic significance.

[0104] This invention constructs a reflection path distribution map based on multi-time radar data and combines it with multi-dimensional indicators such as path overlap, signal divergence, and image point stability scores to accurately identify and eliminate "false target points" caused by multi-path interference. This solves the problem of recognition failure caused by misjudging trapped personnel due to highly reflective objects in existing technologies. Furthermore, a dynamic extension algorithm is used to repair the image structure boundaries and complete the texture, ensuring image continuity and structural consistency. Then, by fusing the original and repaired images at the pixel level, a dynamic confidence distribution map is constructed, enabling high-confidence positioning of trapped targets and intelligent updating of priority rescue areas. This method not only establishes a closed-loop data linkage mechanism between radar image recognition and rescue mission execution but also significantly enhances the system's anti-interference capability, interpretation reliability, and resource scheduling accuracy in complex environments.

[0105] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0106] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0107] It should be noted that, in this document, the use of relational terms such as "first" and "second" is merely for distinguishing one entity or operation from another, and does not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0108] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0109] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0110] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0111] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0112] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0113] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0114] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A UAV-borne MiniSAR real-time imaging and emergency communication system for disaster relief, characterized in that, It includes a radar path inversion modeling module, a multipath interference identification module, a target stability assessment module, a false target removal and image purification module, an image structure restoration module, and a target update and response linkage module. The radar path inversion modeling module performs multi-time flight over the target area, collects radar echo signals at multiple times, and simultaneously acquires flight attitude data and imaging azimuth information of the flight platform. Based on the flight attitude data and imaging azimuth information, it inverts the propagation path of the radar echo signal, extracts the incident angle and reflection path length, and generates a reflection path distribution map. The multipath interference identification module calculates the arrival time difference and phase offset of radar echo signals based on the reflection path distribution map, extracts the path overlap and signal divergence of image pixels, and identifies the interference-sensitive areas where multipath echoes converge. The steps for multipath interference identification are as follows: For each radar echo path, the theoretical arrival time is calculated based on its reflection path length and compared with the actual reception time to obtain the time difference of arrival. Extract the complex components of the radar echo signal for each path, calculate their phase offset, and form a phase perturbation map; Based on the time difference spectrum and phase perturbation map, the path overlap and signal divergence of each image pixel are calculated. By jointly analyzing path overlap and signal divergence, a multipath interference sensitivity heatmap is generated, and interference-sensitive areas are delineated. The target stability assessment module analyzes the reflection intensity changes, phase fluctuations, and spatial position consistency of image pixels in the interference-sensitive area in the radar echo sequence at multiple times, calculates the target stability score, and generates a preliminary authenticity judgment map. The false target removal and image purification module initially determines the target stability score of each image point in the image based on its authenticity, filters out image pixels with scores below the threshold that are located in interference-sensitive areas, determines them as false target image points, performs image removal operation, and outputs the image mask after structural purification. The image structure restoration module extracts the edge structure and texture distribution trend of the surrounding area based on the image mask, and uses a dynamic extension algorithm to perform structural completion and image detail reconstruction on the removed area to generate the image structure restoration result. The image structure restoration steps are as follows: By scanning the image mask using edge detection operators, pixel boundary lines from the real area to the removed area are extracted, and an edge distribution point set is constructed. Based on the edge distribution point set, the texture gradient direction in the boundary neighborhood is analyzed, and the texture direction consistency coefficient and edge direction continuity index are calculated. Based on the obtained indicators, a dynamic extension algorithm is used to perform pixel-by-pixel completion along the main texture direction to complete the structural filling of the hole areas in the image; The edges, grayscale, and texture between the repaired area and the real area are smoothed to output the image structure repair result. The target update and response linkage module performs pixel-level fusion comparison between the image structure repair results and the original image to construct a dynamic confidence distribution map, update the location information of the trapped target and the priority rescue area, and form a closed-loop linkage mechanism for identification and response.

2. The UAV-borne MiniSAR real-time imaging and emergency communication system for disaster relief as described in claim 1, characterized in that, The radar path inversion modeling steps are as follows: Unmanned aerial vehicles equipped with microwave synthetic aperture radar will be deployed over the target area to perform multi-time coverage flight missions. At each moment of flight, radar echo signals are collected, and flight attitude data and imaging azimuth information are recorded simultaneously. Based on the flight attitude data and imaging azimuth information at each moment, the outgoing direction and propagation path of the radar wave are inverted, and the corresponding incident angle and reflection path length are calculated. By integrating path parameters for all pixels within the image region, a reflection path distribution map is constructed, which includes incident angle, path length, and path dispersion indices.

3. The UAV-borne MiniSAR real-time imaging and emergency communication system for disaster relief as described in claim 1, characterized in that, The steps for assessing the stability of the target are as follows: Extract the radar echo intensity sequence of each image point at multiple time points and calculate the amplitude of reflection intensity change; Extract the radar echo phase value of each image point at multiple time points and calculate the degree of phase fluctuation. Analyze the spatial projection position of each image point at multiple time points and calculate the spatial position consistency. A target stability scoring function is constructed based on the magnitude of reflection intensity change, the degree of phase fluctuation, and the consistency of spatial location, generating a preliminary determination map of the authenticity of image points.

4. The UAV-borne MiniSAR real-time imaging and emergency communication system for disaster relief as described in claim 1, characterized in that, The specific steps for false target removal and image purification are as follows: Statistical analysis was performed on the target stability score of each image point, and a stability score threshold was set. Image points with scores below a threshold and spatial locations within interference-sensitive areas are selected and a pseudo-target point mask is generated. Perform edge consistency screening and hole repair operations on the pseudo-target point mask image to construct a structurally optimized image purification mask; The image is cleaned by masking corresponding image points in the original image and outputting the image mask after structural cleansing.

5. The UAV-borne MiniSAR real-time imaging and emergency communication system for disaster relief according to claim 1, characterized in that, The steps for target update and response linkage are as follows: The image structure restoration result is pixel-level registered and weighted fusion with the original image to generate a fused image; Based on the fused image, the confidence value is calculated pixel by pixel by combining stability score, repair degree, image intensity contrast, edge continuity and texture consistency to construct a dynamic confidence distribution map; Based on the high-confidence areas in the dynamic confidence distribution map, the trapped target is relocated and its contour is extracted, and the location information and priority rescue ranking results are output. The target location information and priority rescue ranking results are uploaded to the mission dispatch command terminal, linking the trajectory planning, communication deployment and path simulation processes to form a closed-loop mechanism for identification and response.

Images