Method and product for searching for people in the field based on a drone sar image

Abstract

The application discloses a kind of based on unmanned aerial vehicle SAR image field personnel search method and product. Including: based on the background scattering characteristic model of target search area being constructed to multi-temporal SAR image sequence;The SAR image of current time sequence node and reference time sequence node in multi-temporal SAR image sequence is carried out interference coherence variation detection, and the suspected change target area that changes relative to background in target search area is obtained in combination with background scattering characteristic model;Motion state feature and spatial scale feature of suspected change target area are extracted from multi-temporal SAR image sequence, and motion state detection and spatial scale detection are carried out to suspected change target area in turn;In the case where detection passes, suspected change target area is determined as suspected personnel target;Based on the position change information of suspected personnel target in multi-temporal SAR image sequence, positioning analysis is carried out, and search positioning result is output.The application can improve the accuracy of positioning result.

Description

Technical Field

[0001] This application relates to the field of environmental change monitoring technology, and more particularly to the field of fire risk identification technology, specifically to a method for searching for personnel in the field based on UAV SAR images, a training method for forest fire identification models, and related devices, equipment, media, and products. Background Technology

[0002] In complex natural environments such as forests and mountains, incidents of people going missing occur frequently. Due to the undulating terrain, dense vegetation cover, and complex environmental conditions, the search for missing persons faces significant challenges. Especially at night, in inclement weather, or when visibility is limited, commonly used field search techniques mainly include manual ground search, optical imaging, and infrared imaging.

[0003] However, manual ground searches rely on rescue personnel venturing into complex terrain, limiting the search area and efficiency, which is constrained by personnel physical strength and environmental conditions. Furthermore, conducting ground searches in mountainous and forested areas can pose a threat to the safety of rescue personnel. Optical imaging methods have some applications in personnel searches, but these methods are highly dependent on lighting conditions, with significantly reduced imaging effectiveness at night, in fog, smoke, or under dense vegetation cover. While infrared imaging can overcome lighting limitations to some extent, in dense vegetation or environments with small temperature differences, human thermal characteristics are easily weakened, leading to unstable identification results.

[0004] In recent years, drone technology has been applied to some extent in the field of emergency search and rescue. By carrying optical or infrared sensors, it can conduct aerial patrols of large areas. However, existing drone search methods mostly rely on single imaging results, making it difficult to take full advantage of the fact that missing persons may change their location. Their ability to locate continuously moving missing persons is limited, resulting in inaccurate location output. Summary of the Invention

[0005] In view of this, embodiments of this application provide a method, device, system, medium, and product for searching for personnel in the field based on UAV SAR images, which can improve the accuracy of positioning results.

[0006] In a first aspect, embodiments of this application provide a method for searching for personnel in the field based on UAV SAR images. The method includes: acquiring synthetic aperture radar (SAR) images of a target search area at multiple time-series nodes to obtain a multi-temporal SAR image sequence; constructing a background scattering characteristic model of the target search area based on the multi-temporal SAR image sequence; performing interferometric coherence change detection on the SAR images of the current time-series node and a reference time-series node in the multi-temporal SAR image sequence, and combining this with the background scattering characteristic model to obtain suspected changing target areas within the target search area that have changed relative to the background; and extracting the motion state of the suspected changing target areas from the multi-temporal SAR image sequence. Based on motion state characteristics and spatial scale characteristics, motion state detection and spatial scale detection are sequentially performed on suspected change target areas. Motion state detection is used to evaluate whether the motion of the suspected change target area conforms to the preset human motion characteristic model, and spatial scale detection is used to evaluate whether the suspected change target area matches the size range of the scattering characteristics of a human body in the SAR image. If both motion continuity detection and spatial scale detection pass, the suspected change target area is identified as a suspected human target. Based on the position change information of the suspected human target in the multi-temporal SAR image sequence, localization analysis is performed, and the search and localization results are output.

[0007] Secondly, embodiments of this application provide a field personnel search device based on UAV SAR images. The device includes: an acquisition module for acquiring synthetic aperture radar (SAR) images of a target search area at multiple time-series nodes to obtain a multi-temporal SAR image sequence; a construction module for constructing a background scattering characteristic model of the target search area based on the multi-temporal SAR image sequence; a detection module for detecting interferometric coherence changes in the SAR images of the current time-series node and a reference time-series node within the multi-temporal SAR image sequence, and, combined with the background scattering characteristic model, obtaining suspected changed target areas within the target search area that have changed relative to the background; the detection module is further used to extract suspected changed target areas from the multi-temporal SAR image sequence. The system identifies the motion state and spatial scale characteristics of the target area, and then sequentially performs motion state detection and spatial scale detection on suspected change target areas based on these characteristics. Motion state detection assesses whether the motion of the suspected change target area conforms to a pre-defined human motion characteristic model, while spatial scale detection assesses whether the suspected change target area matches the size range of the scattering characteristics of a human body in a SAR image. A determination module identifies the suspected change target area as a suspected human target if both motion continuity and spatial scale detection are passed. A localization module performs localization analysis based on the positional change information of the suspected human target in a multi-temporal SAR image sequence and outputs the search and localization results.

[0008] Thirdly, embodiments of this application provide an electronic device, including: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, it implements the steps of the field personnel search method based on UAV SAR images as described in the first aspect.

[0009] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer program instructions, which, when executed by a processor, implement the steps of the field personnel search method based on UAV SAR images as described in the first aspect.

[0010] Fifthly, embodiments of this application provide a computer program product stored in a non-volatile storage medium, which, when executed by a processor, implements the steps of the field personnel search method based on UAV SAR imagery as described in the first aspect.

[0011] In a sixth aspect, embodiments of this application provide a chip including a processor and a communication interface, the communication interface and the processor being coupled together, the processor being used to run programs or instructions to implement the steps of the field personnel search method based on UAV SAR images as described in the first aspect.

[0012] This application provides a method, device, system, medium, and product for searching for personnel in the field based on UAV SAR imagery. Utilizing synthetic aperture radar (SAR) for imaging, its active microwave imaging principle eliminates reliance on natural lighting conditions and can penetrate vegetation cover to a certain extent, ensuring stable acquisition of image data of the target area even at night, in inclement weather, and under vegetation cover, overcoming the inherent limitations of optical and infrared imaging methods. Secondly, by acquiring multi-temporal SAR image sequences of the same target area and shifting the detection focus from static feature identification of a single image to analysis of changes within the sequence, potential movement of missing persons no longer constitutes interference but becomes a key detectable signal. Specifically, by detecting interferometric coherence changes and combining them with a background scattering characteristic model, interference from complex static backgrounds such as vegetation and terrain can be effectively suppressed, initially screening out potential target areas changed due to movement. Furthermore, by extracting motion state features and spatial scale features and performing joint detection, based on a preset personnel movement model and human body scattering scale, potential changing targets can be accurately screened, effectively distinguishing between personnel activity and non-target changes such as animal interference and environmental noise, reducing the false alarm rate. Thus, by analyzing the location changes of suspected individuals across multiple time phases through screening, more reliable search and location results can be output. Compared to search methods that rely solely on single-shot imaging, this improves the success rate of location. Consequently, in complex natural environments, it enables effective, stable, and accurate long-distance detection and location of missing persons who may be in motion. Attached Figure Description

[0013] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings of the embodiments of this application will be briefly described below.

[0014] Figure 1 This is a flowchart illustrating a method for searching for personnel in the field based on UAV SAR images provided in an embodiment of this application; Figure 2 This is a flowchart illustrating a method for searching for personnel in the field based on UAV SAR images, provided in another embodiment of this application. Figure 3 This is an exemplary schematic diagram of a method for searching for personnel in the field based on UAV SAR images provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of a field personnel search device based on UAV SAR images provided in an embodiment of this application; Figure 5 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0015] The principles and spirit of this application will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided to make the principles and spirit of this application clearer and more thorough, enabling those skilled in the art to better understand and implement the principles and spirit of this application. The exemplary embodiments provided herein are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments described herein without inventive effort are within the scope of protection of this application.

[0016] In this document, terms such as first, second, and third are used only to distinguish one entity (or operation) from another entity (or operation), and are not intended to require or imply any order or relationship between these entities (or operations).

[0017] The following description, in conjunction with the accompanying drawings, details the field personnel search method based on unmanned aerial vehicle synthetic aperture radar (SAR) images provided in this application, through specific embodiments and application scenarios.

[0018] Figure 1 This is a schematic flowchart of a field personnel search method based on UAV SAR images provided in an embodiment of this application. The execution subject of this field personnel search method based on UAV SAR images can be a field personnel search device based on UAV SAR images.

[0019] The following example illustrates the field personnel search method based on UAV SAR imagery, using a field personnel search device based on UAV SAR imagery as the executing entity. It should be noted that the aforementioned executing entity and application scenario do not constitute a limitation on this application.

[0020] like Figure 1 As shown, the field personnel search method based on UAV SAR images provided in this application embodiment may include steps 110-150.

[0021] Step 110: Obtain synthetic aperture radar (SAR) images of the target search area at multiple time nodes to obtain a multi-temporal SAR image sequence; Step 120: Based on the multi-temporal SAR image sequence, construct a background scattering characteristic model of the target search area; Step 130: Detect the interferometric coherence change of the SAR images of the current time node and the reference time node in the multi-temporal SAR image sequence, and combine it with the background scattering characteristic model to obtain the suspected changing target area that changes relative to the background within the target search area. Step 140: Extract motion state features and spatial scale features of suspected changed target areas from multi-temporal SAR image sequences, and perform motion state detection and spatial scale detection on suspected changed target areas based on motion state features and spatial scale features. Step 150: If both motion continuity detection and spatial scale detection pass, the suspected change target area is identified as a suspected human target. Step 160: Based on the location change information of the suspected personnel target in the multi-temporal SAR image sequence, perform location analysis and output the search and location results.

[0022] The field personnel search method based on UAV SAR imagery provided in this application utilizes synthetic aperture radar for imaging. Its active microwave imaging principle does not rely on natural lighting conditions and can penetrate vegetation cover to a certain extent, thus ensuring stable acquisition of image data of the target area even at night, in inclement weather, and under vegetation obstruction, overcoming the inherent limitations of optical and infrared imaging methods. Secondly, by acquiring multi-temporal SAR image sequences of the same target area and shifting the detection focus from static feature identification of a single image to the analysis of changes in the sequence, the potential movement of missing persons no longer constitutes interference but becomes a key detectable signal. Specifically, by detecting interferometric coherence changes and combining them with a background scattering characteristic model, interference from complex static backgrounds such as vegetation and terrain can be effectively suppressed, initially screening out potential target areas caused by movement. Furthermore, by extracting motion state features and spatial scale features and performing joint detection, based on a preset personnel movement model and human body scattering scale, potential changing targets can be accurately screened, effectively distinguishing between personnel activity and non-target changes such as animal interference and environmental noise, reducing the false alarm rate. Thus, by performing location analysis based on the location change information of suspected personnel targets in multiple time phases through screening, more reliable search and location results can be output, thereby enabling effective, stable, and accurate long-distance detection and location of missing persons who may be in motion in complex natural environments.

[0023] The specific implementation of the above steps will be described in detail below with reference to specific embodiments.

[0024] Step 110 involves acquiring synthetic aperture radar (SAR) images of the target search area at multiple time nodes to obtain a multi-temporal SAR image sequence.

[0025] In step 110, a time-series node refers to the time when data acquisition is completed after a flight mission to image the target search area and a complete SAR image of that area is acquired. A multi-temporal SAR image sequence refers to a data set composed of multiple SAR images of the same target search area acquired by multiple independent flight imaging missions completed at different times, arranged in chronological order. This data serves as the basis for subsequent temporal change analysis.

[0026] In practice, an unmanned aerial vehicle (UAV) platform equipped with synthetic aperture radar (SAR) can be used to repeatedly scan and image the target search area—the pre-defined geographical area where personnel need to be searched—following a pre-planned flight path. Multiple flight scans of the target search area can be performed at different times, such as 30 minutes or 1 hour intervals. Each mission acquires one frame of SAR image of the area, thus obtaining SAR images of the area at multiple consecutive or discrete moments, forming a multi-temporal SAR image sequence. The time interval between two adjacent flight imaging missions can be set according to the UAV system performance, endurance, and specific search mission requirements.

[0027] For example, in a search mission, the UAV performed three independent flight imaging missions, and the completion times of each mission were t1, t2, and t3 respectively. The corresponding three time nodes are t1, t2, and t3, thus obtaining a data sequence containing three-phase SAR images. The time interval between adjacent phases is (t2-t1) and (t3-t2).

[0028] According to an embodiment of this application, optionally, after obtaining the multi-temporal SAR image sequence, the process may further include: preprocessing the multi-temporal SAR images, the preprocessing including: filtering the multi-temporal SAR image sequence to suppress image noise; performing geometric correction on the filtered multi-temporal SAR image sequence to match the SAR images with geographic coordinates; and performing registration processing on the geometrically corrected multi-temporal SAR images to align the SAR images of different time nodes in space.

[0029] Specifically, SAR images often contain noise, which can be removed using median filtering. By replacing each pixel with the median of its neighborhood window, random noise is effectively suppressed while preserving edge structures and improving image clarity. The directly observed backscattering coefficients are extracted to provide input for subsequent vegetation cover extraction. Combining the UAV's GPS position, IMU attitude parameters, and imaging parameters, geometric corrections are performed on the image, including slant range correction, azimuth correction, and map projection transformation. This achieves accurate matching between the image and map locations, providing fundamental support for fire area localization and multi-temporal image alignment.

[0030] For multi-temporal SAR images, cross-correlation can be used to calculate the correlation between corresponding regions in the images, obtain the correlation matrix between the two images, and adjust the transformation parameters of the images to maximize the correlation coefficient between the two images, so as to ensure the spatial consistency of the image data.

[0031] For example, correlation can be calculated based on the pixel values ​​of the images, for two SAR images. and The correlation can be calculated using the cross-correlation function of the following formula (1). : (1) Where μ1 and μ2 are respectively and The average pixel value, These are pixel coordinates.

[0032] Step 120 involves constructing a background scattering characteristic model of the target search area based on multi-temporal SAR image sequences.

[0033] In step 120, based on the acquired multi-temporal SAR image sequence, the scattering characteristics of the stable and unchanging natural background within the target search area can be learned or statistically analyzed to construct a model that reflects the typical scattering state of the background, thereby providing a quantitative reference for the "unchanged" or "normal" state for subsequent steps.

[0034] Therefore, the background scattering characteristic model can be used to characterize a set of reference data or mathematical models of the inherent radar scattering characteristics formed by natural environmental elements such as terrain and vegetation within the target search area, without interference from human activities. Based on this model, a discrimination benchmark can be provided for subsequent identification of changing areas that differ from the background scattering characteristics.

[0035] According to an embodiment of this application, optionally, the specific construction process of step 120 above may include: selecting a frame of SAR image from a multi-temporal SAR image sequence as an initial background reference image; selecting a region in the initial background reference image that satisfies the scattering characteristics of natural vegetation and has a continuous spatial distribution as a background candidate region; extracting radar scattering characteristic parameters of the background candidate region, the radar scattering characteristic parameters including echo amplitude statistical characteristics, interferometric coherence characteristics and spatial distribution continuity characteristics; and constructing a parameterized background scattering characteristic model based on the radar scattering characteristic parameters.

[0036] The initial background reference image can be a specific frame selected from a multi-temporal SAR image sequence to represent the initial state of the background; typically, the earliest frame acquired or confirmed to be free of human activity is selected. Background candidate regions can be areas in the initial background reference image that are automatically identified based on prior knowledge or image features, conforming to the scattering characteristics of a stable natural environment, such as homogeneous woodland or grassland, and are spatially continuous. Radar scattering characteristic parameters can be numerical indicators used to quantitatively describe the scattering characteristics of the region, calculated directly or indirectly from the SAR image.

[0037] Specifically, this application can select a frame from the sequence as an initial background reference image, and then use image analysis algorithms, such as amplitude-based thresholding or texture analysis, to select contiguous regions in the image that conform to the statistical characteristics of natural vegetation scattering as background candidate regions. Next, the radar scattering characteristic parameters of the pixels within the background candidate regions are calculated, such as the mean and standard deviation of the backscattering intensity (amplitude) of all pixels, or the mean of the interferometric coherence of the local region. Finally, these parameters, such as the mean or variance, are combined to construct a parameterized mathematical model, which is the background scattering characteristic model. For example, it can be expressed as the background region amplitude values ​​following a normal distribution with a mean of μ and a standard deviation of σ.

[0038] In this embodiment, a stable statistical model of the background scattering characteristics of the target area can be systematically quantified and established from multi-temporal SAR image sequences. This provides an objective reference for subsequent steps, enabling accurate differentiation between abnormal scattering changes caused by real human activities and inherent background fluctuations caused by slow natural processes such as vegetation growth and gradual changes in soil moisture during change detection. By setting statistical parameter benchmarks, interference from background noise and slow environmental changes can be effectively suppressed, thus providing a reliable analytical starting point for the entire method, improving the quality of input data for subsequent change detection steps, and ultimately enhancing the accuracy of the positioning results.

[0039] In step 130, the SAR images of the current time node and the reference time node in the multi-temporal SAR image sequence are subjected to interferometric coherence change detection, and combined with the background scattering characteristic model, the suspected change target area that changes relative to the background in the target search area is obtained.

[0040] In step 130, an earlier time-point image representing the background state can be selected as a reference time-series node, while the current time-series node is a later time-point image to be analyzed. Suspected change target areas can be image areas initially identified in this step that may have undergone significant changes in radar scattering characteristics due to factors such as human activity.

[0041] Specifically, this application first processes two SAR images from the current time-series node and a reference time-series node to generate a coherence coefficient map reflecting the degree of surface change. By setting a threshold, regions that may have changed can be initially identified from the map. Then, in conjunction with the background scattering characteristic model established in step 120, these initially identified change regions are verified to filter out the first region, which, although having low coherence, still has scattering characteristics consistent with the background model, such as those caused by slight fluctuations in the background itself or system noise. Finally, the second region, which differs significantly from the background model, is retained as a suspected change target region.

[0042] According to an embodiment of this application, optionally, step 130 may specifically include: calculating the interferometric coherence coefficient map between two SAR images of the current time-series node and the reference time-series node; identifying pixel regions with coherence lower than a preset change threshold from the interferometric coherence coefficient map to obtain an initial change region; comparing the radar scattering characteristic parameters of the initial change region with the parameters at the corresponding positions in the background scattering characteristic model; deleting regions in the initial change region whose radar scattering characteristic parameters differ from the parameters of the background scattering characteristic model less than a preset background difference threshold, and determining the remaining regions as suspected change target regions.

[0043] Among them, the interferometric coherence coefficient map refers to the grayscale image generated by interferometric processing of two SAR images, calculating the coherence coefficient value corresponding to each resolution unit; the initial change region refers to the set of all connected pixels with coherence lower than the threshold after binarizing the coherence coefficient map by setting a threshold; the preset background difference threshold is a pre-set value used to measure whether the difference between the scattering features of the current detection area and the background model features is significant enough.

[0044] Specifically, firstly, the two SAR images of the current time-series node and the reference time-series node are registered and interferometrically processed, and their coherence is calculated pixel by pixel to obtain an interferometric coherence coefficient map. Then, a coherence threshold is set to measure the coherence... Figure 2 Value-enhanced analysis extracts all pixels with coherence below this threshold, yielding the initial change region.

[0045] For each initial change region, radar scattering characteristic parameters such as the average backscattering intensity in the current temporal image are calculated. This calculated value is compared with reference parameters for the corresponding geographical location in the background scattering characteristic model, such as the average background amplitude, and the difference is calculated, such as the absolute difference or ratio. If the difference is less than a preset background difference threshold, the change is considered insignificant and may originate from background noise or minor disturbances; therefore, it (the first region) is removed from the initial change regions. Otherwise, it is retained. Finally, all retained regions (the second region) are identified as suspected change target regions.

[0046] In this embodiment, interferometric coherence detection is highly sensitive to subtle movements or structural changes in ground features, and can initially identify all possible areas of change. However, its results are easily affected by coherent noise and minor environmental disturbances. By introducing background model comparison, the initially identified areas are further screened, effectively eliminating false changes that, although coherence is reduced, still exhibit overall scattering characteristics consistent with background statistical features, such as uniform vegetation swaying in a light breeze. Thus, this two-stage processing strategy enhances the ability to separate genuine areas of change related to human activity and exhibiting significant differences in scattering characteristics from complex and dynamic natural backgrounds, reducing false alarms and providing a cleaner and more reliable set of candidate targets for subsequent target screening steps.

[0047] Step 140 involves extracting motion state features and spatial scale features of suspected changed target areas from multi-temporal SAR image sequences, and performing motion state detection and spatial scale detection on the suspected changed target areas based on the motion state features and spatial scale features.

[0048] In step 140, the motion state features can be quantitative information derived from the image sequence to describe the motion behavior of the suspected changing target area over time; the spatial scale features can be quantitative information measured from a single image to describe the physical size of the area.

[0049] Specifically, for each suspected changed target region obtained in step 130, its motion state features can be extracted by analyzing its location and time information in multi-temporal image sequences. Simultaneously, its spatial scale features can be extracted based on the region's morphology in a single frame image. Motion state detection is performed on the region based on the extracted motion state features to assess whether its motion pattern matches the expected human activity characteristics; spatial scale detection is performed on the region based on the extracted spatial scale features to assess whether its physical size matches the representation of human targets in SAR images.

[0050] According to an embodiment of this application, optionally, the motion state features include displacement velocity features, displacement trajectory features, and motion pattern features. The motion state detection of the suspected changed target area based on the motion state features in step 140 above may include at least the following two: displacement amplitude detection of the suspected changed target area based on displacement velocity features; displacement continuity detection of the suspected changed target area based on displacement trajectory features; and motion pattern rationality detection of the suspected changed target area based on motion pattern features.

[0051] Among them, displacement velocity characteristics refer to the instantaneous or average velocity calculated from the displacement and time interval of the target between adjacent time phases; displacement trajectory characteristics refer to the characteristics reflecting the changes in displacement amount calculated from the position sequence of the target in multiple consecutive time phases, such as the sequence of displacement distance; motion pattern characteristics refer to the high-level characteristics extracted from longer time series or complete trajectories to describe the macroscopic laws of motion, such as the regularity of motion direction or acceleration characteristics.

[0052] In practice, for the same suspected changing target area, the system will start tracking from the first time phase in which it appears, and determine its accurate location in each time phase image through image registration and target association technology.

[0053] Based on this location information, three types of features can be extracted: First, displacement velocity features, calculated based on the position and time differences between two adjacent time phases; second, displacement trajectory features, connecting the position points of multiple consecutive time phases to form a displacement vector sequence, and calculating the magnitude, i.e., the distance, between adjacent displacement vectors in the sequence; third, motion pattern features, analyzing the overall trend of the position sequence, such as calculating the curvature of the trajectory, the number of abrupt changes in direction, or the standard deviation of the velocity. Motion state detection can then select at least two of the above detection methods, for example, simultaneously performing displacement amplitude detection based on displacement velocity features and displacement continuity detection based on displacement trajectory features.

[0054] In this embodiment, a single motion feature, such as instantaneous speed, may be insufficient to accurately distinguish between people and certain animals or environmental disturbances. Multi-feature fusion analysis, however, can more comprehensively reflect the target's motion patterns. For example, simultaneously considering the reasonableness of speed (displacement amplitude detection) and the smoothness of trajectory (displacement continuity detection) can more effectively distinguish between purposeful, rhythmic walking by humans and random, discontinuous movement by animals, or the swaying of vegetation caused by wind. Thus, this comprehensive motion analysis enhances the accuracy and robustness of identifying human target motion patterns, effectively verifying from a kinematic perspective whether a suspected target is a human body, and reducing the risk of positioning errors due to misjudgment of motion patterns.

[0055] According to an embodiment of this application, optionally, the above-mentioned displacement amplitude detection of a suspected changing target area based on displacement velocity characteristics may include: obtaining the position of the suspected changing target area in the images of two adjacent time nodes in a multi-temporal SAR image sequence, and calculating the displacement distance between the suspected changing target area and the two adjacent time nodes based on the position; obtaining the imaging time interval between the two adjacent time nodes, and calculating the displacement velocity characteristics based on the displacement distance and the imaging time interval; if the displacement velocity characteristics are within the range of personnel movement speed, then the displacement amplitude detection is determined to pass.

[0056] Specifically, the time interval between adjacent imaging sequences is For a suspected changed target area, the first step is to locate it at times t and t+ in a multi-temporal SAR image sequence. The specific pixel coordinates of the two frames of images are obtained, and these pixel coordinates are converted into actual geographic coordinates using geocoding parameters. The Euclidean distance between the two geographic locations is then calculated, which is the displacement distance d.

[0057] Then, the displacement amplitude can be detected using the following formula (2): (2) in, and These represent the minimum and maximum movement speed thresholds that a person may reach in a field environment. When the equivalent velocity of a changing target in an adjacent imaging phase is within the range of the velocity thresholds, the changing target is determined to meet the displacement amplitude constraint.

[0058] For example, the system presets a speed range that a person can walk at, such as 0.5 meters per second to 2 meters per second. If the calculated average speed v falls within this preset range, the target is determined to have passed the displacement amplitude detection.

[0059] In this embodiment, in field SAR images, the instantaneous velocity of many moving targets may fall within the range of human speed, but their motion patterns may not conform to human characteristics. Based on this, this application performs rapid screening at the most basic level of physical motion probability, efficiently eliminating non-human targets whose speeds far exceed human physiological limits (flying birds) or are far below normal human activity levels (slowly moving shadows). This enables rapid initial screening throughout the motion state detection process, narrowing the target range required for subsequent, more complex motion analysis, and improving overall processing efficiency and the initial reliability of the positioning results.

[0060] According to an embodiment of this application, optionally, the above-mentioned displacement continuity detection of a suspected changing target area based on displacement trajectory features may include: obtaining the position of the suspected changing target area in the images of at least three consecutive time nodes in a multi-temporal SAR image sequence; calculating the displacement distance between each pair of adjacent time nodes based on the position to obtain a displacement distance sequence as a displacement trajectory feature; calculating the difference between each pair of adjacent displacement distances based on the displacement distance sequence; and determining that the displacement continuity detection passes if each difference is less than or equal to a preset maximum displacement change threshold.

[0061] Among them, the displacement distance sequence refers to the sequence formed by arranging the displacement distances between each pair of adjacent nodes in time order across multiple consecutive time nodes.

[0062] Specifically, for a continuously tracked suspected change target region, its position in at least three consecutive time-series node images is obtained, and the displacement distance d1 in the first time interval and the displacement distance d2 in the second time interval are calculated to form a displacement distance sequence. Calculate the absolute value of the difference between every two displacement distances. The system can preset a maximum displacement change threshold. This threshold is used to characterize the maximum displacement change that a person can achieve within adjacent time periods. If the absolute value of the calculated difference is less than or equal to this threshold, If the displacement continuity test is passed, then the sequence is considered to have passed. For longer sequences, this condition must be met if the difference between the distances of all adjacent displacements in the sequence satisfies the requirement.

[0063] In this embodiment, specific requirements for the continuity of motion are proposed by analyzing whether the change in displacement distance of a target over a continuous time period is smooth. When humans walk on the ground, their stride length and speed usually change continuously over a short period of time, without drastic or irregular jumps. However, many natural disturbances, such as tree canopies blown by gusts of wind, small-scale thermal convection disturbances on the ground surface, or sensor noise, may cause random jumps in the detected target position over time. Based on this, by setting a maximum displacement change threshold, targets whose displacement trajectories exhibit high-frequency jitter or abrupt changes can be identified and filtered out, thereby ensuring that the motion trajectory of the detected target has the continuity and stability characteristic of human walking, further enhancing the accuracy of personnel identification based on motion characteristics.

[0064] According to an embodiment of this application, optionally, the above-mentioned motion pattern rationality detection of a suspected changed target area based on motion pattern features may include: acquiring position change information of the suspected changed target area in multiple time-series images in a multi-temporal SAR image sequence; extracting motion pattern features based on the position change information, wherein the motion pattern features are used to characterize the comprehensive motion trajectory morphology of the target in multiple temporal phases; determining whether the motion trajectory of the suspected changed target area conforms to a preset motion behavior pattern based on the motion pattern features, wherein the preset motion behavior pattern includes a motion speed exceeding a preset speed change threshold and a motion direction change exceeding a preset angle threshold between consecutive time-series nodes; if the motion trajectory does not conform to the preset motion behavior pattern, then the motion pattern rationality detection is passed.

[0065] Among them, position change information is a general term for the position sequence of the target in multiple time phases; comprehensive motion trajectory morphology is used to characterize the overall characteristics of the entire position sequence, such as path, direction change, and velocity distribution; preset motion behavior mode refers to a set of regularized conditions used to judge whether the motion is abnormal, such as prohibiting physically impossible speed changes or instantaneous changes in direction.

[0066] Specifically, the system acquires a series of position coordinates of a suspected change target area over a relatively long period of time, across multiple time-series nodes. Based on these coordinates, its motion trajectory can be reconstructed, and higher-level motion pattern features can be extracted, such as calculating the instantaneous velocity sequence and direction angle sequence of each point on the trajectory, or analyzing the change pattern of the velocity-time curve. Preset abnormal behavior patterns may include, but are not limited to: velocity abrupt changes, i.e., the velocity change exceeding a preset velocity abrupt change threshold within an extremely short time interval, such as acceleration from rest to more than 5 meters per second; and direction abrupt changes, i.e., the angle of change of motion direction between consecutive position points exceeding a preset angle threshold, such as an acute angle turn exceeding 120 degrees. The system analyzes the extracted motion pattern features to check for the presence of any of the above abnormal behavior patterns. If the motion trajectory does not contain these abnormal patterns that do not conform to human movement biomechanics and common behavioral logic, it is determined that it has passed the motion pattern rationality detection.

[0067] In this embodiment, based on the basic judgment of instantaneous velocity and displacement continuity, further analysis of the comprehensive motion trajectory pattern can identify and exclude targets that have passed the aforementioned basic detection but whose overall motion pattern clearly does not conform to human behavioral characteristics. For example, some wild animals or special natural phenomena may produce seemingly continuous and reasonably fast movements, but their motion trajectories may frequently exhibit abnormal patterns such as sharp turns at large angles that violate human kinematics, prolonged high-speed sprints in a straight line followed by sudden stops. Based on this, by pre-setting rules such as velocity change thresholds and direction change angle thresholds, such abnormal behaviors can be captured, thereby enhancing the system's ability to distinguish in complex scenarios. In complex and ever-changing outdoor scenarios, this greatly reduces misjudgments caused by strange target behavior patterns, ensuring a reliable final positioning result.

[0068] According to an embodiment of this application, optionally, spatial scale detection of a suspected changed target region based on spatial scale characteristics may include: obtaining the area and maximum external dimensions of the suspected changed target region to obtain spatial scale characteristics; determining whether the area is within the interval formed by a first preset area threshold and a second preset area threshold, and whether the maximum external dimensions are within the interval formed by a first preset size threshold and a second preset size threshold; if the area and the maximum external dimensions are within the corresponding preset intervals, then the suspected changed target region is determined to have passed the spatial scale detection; wherein, the first preset area threshold, the second preset area threshold, the first preset size threshold, and the second preset size threshold are preset according to the size range of the scattering characteristics of the human body in the SAR image.

[0069] The area can be the actual projected area of ​​the suspected changed target region on the geo-corrected SAR image; the maximum external size can be the length of the longest side of the smallest bounding rectangle of the region, or the equivalent diameter, or other dimensions that characterize its maximum extension. The first preset area threshold, the second preset area threshold, the first preset size threshold, and the second preset size threshold together define the reasonable upper and lower limits of the area and size that a human target may present after imaging in a SAR image.

[0070] Specifically, once a suspected variable target region is identified in a single-frame SAR image, the number of pixels occupied by the region is first converted into the actual ground projection area S based on the geometric correction parameters of the image. Simultaneously, the minimum bounding rectangle of the region is calculated, and its longest side length L is obtained as the maximum external dimension. The system pre-stores area threshold ranges derived from typical human body dimensions (e.g., shoulder width or height) and SAR imaging geometry (e.g., resolution or viewing angle). and size threshold range When making a judgment, if the spatial scale parameters of a suspected changing target simultaneously meet the above-mentioned area threshold range and scale threshold range, the changing target is determined to meet the spatial scale constraint; otherwise, it is determined to not meet the spatial scale constraint and is marked as a candidate for a non-personnel changing target.

[0071] In this embodiment, the backscattered energy distribution and imaging size of a human body in a SAR image have a predictable range at a certain resolution and viewing angle. By calculating the projected area or maximum circumscribed size of the detected change region and comparing it with a preset threshold range for human target size, non-human targets that are significantly too large (e.g., large animal groups, vehicles) or too small (e.g., small animals, isolated noise points) can be quickly and directly eliminated. This effectively filters out a large number of interferences that are significantly inconsistent with human size in the early stages of detection, reducing the number of candidate targets flowing into subsequent complex motion analysis stages. This not only improves overall processing efficiency but also avoids the waste of subsequent analysis resources and potential misjudgments that may result from targets with fundamentally inconsistent sizes, thus improving the accuracy of the final positioning result.

[0072] In step 150, if both motion continuity detection and spatial scale detection pass, the suspected change target area is identified as a suspected human target.

[0073] In step 150, a suspected person target refers to a region of suspected change in target area detected by motion state detection and spatial scale detection in step 140. Based on this, the region not only matches the characteristics of a person in terms of motion behavior, but its physical size also matches that of a human target, thus being determined as a high-probability target corresponding to the missing person.

[0074] Specifically, for a suspected area of ​​change, the system will determine whether it simultaneously meets two conditions: first, it is determined to conform to preset human movement characteristics in motion state detection; second, it is determined to be within the preset human target size range in spatial scale detection. Only when both conditions are met will the area be ultimately marked as a suspected human target.

[0075] According to an embodiment of this application, optionally, after step 140 performs motion state detection and spatial scale detection on the suspected changed target area based on motion state features and spatial scale features, it may further include: obtaining the number of times the suspected changed target area continuously appears in a multi-temporal SAR image sequence of multiple consecutive time nodes; when the number of continuous appearances is greater than or equal to a preset minimum duration threshold, the time persistence detection is determined to be passed.

[0076] Specifically, the number of times (N_p) the target is successfully associated and detected in a temporally consecutive series of image frames is counted. A minimum duration threshold (N_min) is set, for example, N_min equals 3. If the counted number of occurrences (N_p) is greater than or equal to N_min, the target is considered to have passed the temporal persistence detection.

[0077] If both motion continuity detection and spatial scale detection pass, step 150 above will identify the suspected changed target area as a suspected person target. This may include identifying the suspected changed target area as a suspected person target if at least two of motion state detection, spatial scale detection, and time continuity detection pass.

[0078] For example, such as Figure 2 As shown, the system collects all constraint results obtained from the analysis of each suspected changing target. These results are based on the comparison between the target features extracted from multi-temporal SAR images and the preset prior model of human behavior. Specifically, they include: spatial scale constraint results, used to determine whether the target size matches the human body; temporal scale constraint results, used to determine whether the target continues to appear; motion continuity analysis results and motion pattern analysis results, used to evaluate the trajectory smoothness of the target's motion and the overall rationality of its behavior, respectively.

[0079] The system counts the number of constraints that have passed, N. It then determines if N >= 2, requiring the target to pass at least two of the aforementioned constraints.

[0080] Based on the judgment results: if N is greater than or equal to 2, it is judged as "comprehensive constraint passed", and the system marks the target as a human target and includes it in the personnel target set as the basis for subsequent location analysis; if N is less than 2, it is judged as "comprehensive constraint failed", and the system marks the target as a non-personnel target and includes it in the exclusion target set, thereby filtering it out.

[0081] In this application embodiment, a continuity requirement in the time dimension is introduced, extending the instantaneous detection of a single time phase to continuous observation and verification across multiple time phases. In field SAR monitoring, many spurious changes caused by transient interference, such as flicker noise caused by a single flock of birds flying over or brief changes in illumination, only appear in a single or very few time phases. Based on this, by setting a minimum threshold for the number of times a suspected target appears continuously, it is required that the suspected target must appear stably in multiple consecutive observation time phases, which can effectively filter out such short-lived, transient interference signals. In this way, it can be ensured that the object ultimately identified as a suspected human target is a time-continuous entity, rather than accidental noise or transient phenomena, which can improve the stability and reliability of the detection results in the time domain. This allows subsequent positioning analysis to be based on a real target trajectory with time continuity, thereby enhancing the credibility of the final positioning results.

[0082] Step 160 involves performing a location analysis based on the location change information of the suspected human target in the multi-temporal SAR image sequence, and outputting the search and location results.

[0083] In step 160, the location change information may refer to the series of geographic location coordinates corresponding to the suspected person target in each frame of the image sequence; the search and location result may refer to the geospatial information generated by the spatial analysis algorithm based on the above spatiotemporal location information, which is used to indicate the possible location of the missing person.

[0084] Specifically, this application can statistically analyze the location coordinates of suspected individuals across all relevant time-series images. Based on the spatiotemporal sequence formed by these coordinates, further trajectory analysis, location prediction, or probability statistics can be performed, thereby generating more accurate and intuitive location information. The final search and location results are intended to provide clear, geographic coordinate-based spatial guidance for ground search and rescue operations.

[0085] According to an embodiment of this application, optionally, step 160 above, based on the location change information of the suspected personnel target in the multi-temporal SAR image sequence, performs location analysis and outputs search and location results, which may include: combining the location change information of the suspected personnel target, the geographical information of the target search area, and the prior activity information of the missing person to generate at least one suspected activity area; associating the suspected activity area with the latest location of the suspected personnel target, and outputting search and location results including location confidence.

[0086] The prior activity information of the missing person may include known information such as the last known location of the missing person, planned route or possible destination; the suspected activity area is the spatial range in which the missing person may hide or pass through, which is derived from the analysis of trajectory and prior information; the location confidence is a quantitative assessment of the reliability of the location result.

[0087] Specifically, after obtaining the spatiotemporal location sequence of suspected individuals, the system combines geographic information such as digital elevation models and map vectors to analyze whether their movement trajectory is constrained by terrain, such as following ridgelines or bypassing steep slopes, and predicts their possible activity boundaries. Simultaneously, it correlates and extrapolates the starting point or trend of the trajectory with the last known location of the missing person. Through spatial analysis algorithms such as buffer analysis, path planning inversion, or probability diffusion models, it generates one or more suspected activity areas in geographic space. The system also assigns a location confidence score to the area based on the continuity of the trajectory, the stability of the detection, and the degree of matching with prior information, expressed as a percentage.

[0088] In this embodiment, by integrating geographical constraints and prior knowledge, the positioning results are made more consistent with the logic of the real world, making the predicted area more accurate and instructive. The output confidence score helps search and rescue commanders distinguish the priority levels of different clues, rationally allocate search and rescue resources, and enhance decision support value in actual search and rescue operations.

[0089] According to an embodiment of this application, optionally, after outputting the search and location results in step 160 above, the method may further include: controlling a UAV equipped with synthetic aperture radar to perform supplementary flight scans on the target search area or the key area indicated by the search and location results to obtain SAR images of newly added time-series nodes; adding the SAR images of newly added time-series nodes to a multi-temporal SAR image sequence, and re-executing the steps of constructing a background scattering characteristic model, obtaining suspected changing target areas, performing detection, determining suspected personnel targets, and outputting search and location results to update the search and location results.

[0090] The key area refers to the geographical range that needs to be given priority attention, as determined by the search and positioning results output in step 160. The SAR image of the newly added time-series node refers to the latest SAR image data obtained by controlling the UAV to fly again after the closed-loop process is started.

[0091] Specifically, after outputting the initial search and location results, the system can replan the UAV's flight path based on these results, instructing the UAV to perform supplementary scans of the most probable key areas at higher frequencies or higher resolutions to acquire new temporal SAR images. The new images are added to the existing multi-temporal SAR image sequence, and then the system uses the updated image sequence as input to re-execute the complete process from steps 120 to 160. Because the new data contains the latest state information of the target area, the recalculated background model, detected changing targets, and the final location results will all be updated, thus reflecting the possible latest location of the missing person.

[0092] In this embodiment, after obtaining the initial positioning result, the system does not terminate but uses this result as feedback to intelligently guide subsequent UAV data collection and focus observation on high-probability areas. Newly acquired data is incorporated into the processing flow in real time, updating the background model, re-detecting changes, and performing filtering and positioning analysis. This closed-loop mechanism enables the system to continuously track the target's dynamic movement, constantly correcting and refining the positioning estimate, thus effectively addressing the reality that missing persons may still be moving. This not only improves the speed of confirming stationary targets but also enhances the continuous tracking and prediction capabilities for moving targets, ensuring that the search and positioning results continuously converge, update, and approximate the true situation over time and with data accumulation, thereby maintaining high reliability and timeliness in dynamically changing search tasks.

[0093] As a specific example, taking the search for a missing person in a forest environment as a scenario, the specific steps include: Step 1: Control the UAV equipped with X-band synthetic aperture radar to perform three flight scan missions over the preset target forest area at different times in the morning, such as T1=08:00, T2=09:30, and T3=11:00, so as to obtain a multi-temporal SAR image sequence of the area.

[0094] Step 2: Based on the multi-temporal SAR image sequence, perform coherence change detection. Specifically, the image at time T2 is interferometrically processed with the image at time T1 (as a reference background) to generate an interferometric coherence map. Suspected change regions are extracted from the coherence map by setting a coherence threshold. In this example, three connected low coherence regions, A, B, and C, were detected, which may be caused by human movement, animal activity, or environmental disturbances.

[0095] Step 3: The system tracks each extracted suspected change region (targets A, B, and C) in images from three different time phases and constructs a target feature database. This database stores refined features extracted from the following three dimensions for each target: Spatial feature extraction: The morphology of each target in a single-phase image is measured. For example, the area and size of target A in the image at time T2 are calculated to obtain its spatial scale features.

[0096] Temporal feature extraction: Analyze the occurrence pattern of the target throughout the entire observation period. For example, record the occurrence time sequence of target A (detected in both T2 and T3) and its duration (from T2 to T3, lasting approximately 1.5 hours).

[0097] Motion feature extraction: Analyzing the dynamic information of the target across time phases. By associating the positions of target A in time phases T2 and T3, displacement calculation and velocity analysis are performed. Furthermore, combining its positions in the three time phases (if associative), trajectory analysis can be performed to obtain its motion path.

[0098] Step 4: The system compares the extracted features with preset prior constraints that characterize the activity characteristics of personnel. Spatial Constraints: Determine whether the area and size of the target fall within the human body target area, for example, an area of ​​0.5-3 square meters and a length of 1-2.5 meters. Target A's area and size both fall within this range, so the constraint result is "passed".

[0099] Time scale constraint: Determines whether the target appears in a sufficient number of consecutive time phases, for example, requiring it to appear in at least two time phases. If target A appears continuously in time phases T2 and T3, the constraint result is passed.

[0100] Motion characteristic constraints: Determine whether the target's speed is reasonable, for example, 0.1-2 m / s, and whether the trajectory is continuous and smooth. Target A's speed is approximately 0.22 m / s, which is consistent with walking speed, and its displacement direction is stable; therefore, the constraint result is "pass".

[0101] Step 5: The system makes a comprehensive judgment on the above constraints of targets A, B, and C.

[0102] For example: Target A passes all three constraints; Target B is a deer, which is too fast and fails the motion characteristic constraint; Target C could be a bush that is continuously blown by the wind, which is too large and fails the spatial constraint, and its motion trajectory is a high-frequency swaying motion in place, thus failing the motion characteristic constraint.

[0103] Therefore, the system determines that target A is a human target, while targets B and C are non-human targets. For target A, which is determined to be a human target, the system outputs its spatial location information (the latest geographic coordinates at time T3) and movement trajectory information (the movement path from T2 to T3), forming a high-confidence search and location result that can be directly provided to the ground search and rescue team to guide them to the vicinity of the designated coordinates for precise verification and rescue.

[0104] Corresponding to the method embodiments of this application, this application also provides a field personnel search device based on UAV SAR images.

[0105] Figure 4 This is a schematic diagram of a field personnel search device based on UAV SAR imagery provided in an embodiment of this application. Figure 4 As shown, the field personnel search device 400 based on UAV SAR images may include: an acquisition module 410, a construction module 420, a detection module 430, a determination module 440, and a positioning module 450.

[0106] The acquisition module 410 is used to acquire synthetic aperture radar (SAR) images of the target search area at multiple time-series nodes to obtain a multi-temporal SAR image sequence; the construction module 420 is used to construct a background scattering characteristic model of the target search area based on the multi-temporal SAR image sequence; the detection module 430 is used to perform interferometric coherence change detection on the SAR images of the current time-series node and the reference time-series node in the multi-temporal SAR image sequence, and, combined with the background scattering characteristic model, to obtain suspected changed target areas within the target search area that have changed relative to the background; the detection module 430 is also used to extract the motion state characteristics and spatial scale characteristics of the suspected changed target areas from the multi-temporal SAR image sequence. The system identifies suspected changing target areas by sequentially performing motion state detection and spatial scale detection based on motion state characteristics and spatial scale characteristics. Motion state detection is used to assess whether the motion of the suspected changing target area conforms to a preset human motion characteristic model, and spatial scale detection is used to assess whether the suspected changing target area matches the size range of the scattering characteristics of a human body in the SAR image. The determination module 440 is used to identify the suspected changing target area as a suspected human target if both motion continuity detection and spatial scale detection are passed. The positioning module 450 is used to perform positioning analysis based on the position change information of the suspected human target in the multi-temporal SAR image sequence and output the search and positioning results.

[0107] The field personnel search device based on UAV SAR imagery provided in this application utilizes synthetic aperture radar for imaging. Its active microwave imaging principle does not rely on natural lighting conditions and can penetrate vegetation cover to a certain extent, thus ensuring stable acquisition of image data of the target area even at night, in inclement weather, and under vegetation obstruction, overcoming the inherent limitations of optical and infrared imaging methods. Secondly, by acquiring multi-temporal SAR image sequences of the same target area and shifting the detection focus from static feature identification of a single image to analysis of changes in the sequence, the potential movement of missing persons no longer constitutes interference but becomes a key detectable signal. Specifically, by detecting interferometric coherence changes and combining them with a background scattering characteristic model, interference from complex static backgrounds such as vegetation and terrain can be effectively suppressed, initially screening out suspected change areas caused by movement. Furthermore, by extracting motion state features and spatial scale features and performing joint detection, suspected change targets can be accurately screened based on a preset personnel movement model and human body scattering scale, effectively distinguishing between personnel activity and non-target changes such as animal interference and environmental noise, reducing the false alarm rate. Thus, by performing location analysis based on the location change information of suspected personnel targets in multiple time phases through screening, more reliable search and location results can be output, thereby enabling effective, stable, and accurate long-distance detection and location of missing persons who may be in motion in complex natural environments.

[0108] In some embodiments, the construction module is specifically used to: select a frame of SAR image from a multi-temporal SAR image sequence as an initial background reference image; select a region in the initial background reference image that satisfies the scattering characteristics of natural vegetation and has a continuous spatial distribution as a background candidate region; extract radar scattering characteristic parameters of the background candidate region, the radar scattering characteristic parameters including echo amplitude statistical characteristics, interferometric coherence characteristics and spatial distribution continuity characteristics; and construct a parameterized background scattering characteristic model based on the radar scattering characteristic parameters.

[0109] In some embodiments, the detection module is specifically used to: calculate the interferometric coherence coefficient map between two SAR images of the current time-series node and the reference time-series node; identify pixel regions with coherence lower than a preset change threshold from the interferometric coherence coefficient map to obtain the initial change region; compare the radar scattering characteristic parameters of the initial change region with the parameters at the corresponding positions in the background scattering characteristic model; delete regions in the initial change region whose radar scattering characteristic parameters differ from the parameters of the background scattering characteristic model less than a preset background difference threshold, and determine the remaining regions as suspected change target regions.

[0110] In some embodiments, the motion state features include displacement velocity features, displacement trajectory features, and motion pattern features. The detection module includes at least two of the following: a first detection unit, used to detect the displacement amplitude of the suspected changing target area based on the displacement velocity features; a second detection unit, used to detect the displacement continuity of the suspected changing target area based on the displacement trajectory features; and a third detection unit, used to detect the rationality of the motion pattern of the suspected changing target area based on the motion pattern features.

[0111] In some embodiments, the first detection unit is specifically used to: obtain the position of the suspected changed target area in the images of two adjacent time nodes in a multi-temporal SAR image sequence, and calculate the displacement distance between the suspected changed target area and the two adjacent time nodes based on the position; obtain the imaging time interval between the two adjacent time nodes, and calculate the displacement velocity feature based on the displacement distance and the imaging time interval; if the displacement velocity feature is within the range of the personnel movement speed, then it is determined that the displacement amplitude detection has passed.

[0112] In some embodiments, the second detection unit is specifically used to: acquire the position of the suspected changed target region in the images of at least three consecutive time nodes in a multi-temporal SAR image sequence; calculate the displacement distance between each pair of adjacent time nodes based on the position to obtain a displacement distance sequence as a displacement trajectory feature; calculate the difference between each pair of adjacent displacement distances based on the displacement distance sequence; and determine that the displacement continuity detection has passed if each difference is less than or equal to a preset maximum displacement change threshold.

[0113] In some embodiments, the third detection unit is specifically used to: acquire position change information of the suspected changed target area in multiple time-series images in a multi-temporal SAR image sequence; extract motion pattern features based on the position change information, the motion pattern features being used to characterize the comprehensive motion trajectory morphology of the target in multiple temporal phases; determine whether the motion trajectory of the suspected changed target area conforms to a preset motion behavior pattern based on the motion pattern features, wherein the preset motion behavior pattern includes a motion speed exceeding a preset speed change threshold and a motion direction change exceeding a preset angle threshold between consecutive time-series nodes; if the motion trajectory does not conform to the preset motion behavior pattern, then the motion pattern rationality detection is passed.

[0114] In some embodiments, the detection module is further configured to: obtain the area and maximum external dimensions of the suspected changed target region to obtain spatial scale features; determine whether the area is within the interval formed by a first preset area threshold and a second preset area threshold, and whether the maximum external dimensions are within the interval formed by a first preset size threshold and a second preset size threshold; if the area and the maximum external dimensions are within the corresponding preset intervals, then the suspected changed target region is determined to have passed the spatial scale detection; wherein, the first preset area threshold, the second preset area threshold, the first preset size threshold, and the second preset size threshold are preset according to the scattering feature size range of the human body in the SAR image.

[0115] In some embodiments, the detection module is further configured to, after performing motion state detection and spatial scale detection on the suspected changed target area based on motion state features and spatial scale features in sequence, obtain the number of times the suspected changed target area continuously appears in a multi-temporal SAR image sequence of multiple consecutive time nodes; and determine that the time persistence detection is passed when the number of continuous appearances is greater than or equal to a preset minimum persistence threshold.

[0116] In some embodiments, the positioning module is specifically used to: combine the location change information of the suspected person target, the geographical information of the target search area, and the prior activity information of the missing person to generate at least one suspected activity area; associate the suspected activity area with the latest location of the suspected person target, and output the search and positioning result including the location confidence.

[0117] In some embodiments, the acquisition module is further configured to, after outputting the search and location results, control the UAV equipped with synthetic aperture radar to perform supplementary flight scans on the target search area or the key area indicated by the search and location results to acquire SAR images of newly added time-series nodes; the update module is configured to add the SAR images of newly added time-series nodes to the multi-temporal SAR image sequence, and re-execute the steps of constructing the background scattering characteristic model, obtaining the suspected changing target area, performing detection, determining the suspected personnel target, and outputting the search and location results to update the search and location results.

[0118] The field personnel search device based on UAV SAR imagery provided in this application embodiment can achieve... Figure 1-3 The various processes implemented in the method embodiments can achieve the same technical effect, and will not be described again here to avoid repetition.

[0119] Figure 5 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application.

[0120] like Figure 5As shown, the electronic device 500 includes a memory 501, a processor 502, and a computer program stored in the memory 501 and executable on the processor 502.

[0121] In one example, the processor 502 described above may include a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0122] Memory 501 may include read-only memory (ROM), random access memory (RAM), disk storage media devices, optical storage media devices, flash memory devices, electrical, optical, or other physical / tangible memory storage devices. Therefore, typically, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the field personnel search method based on UAV SAR imagery according to the embodiments of the first aspect of this application.

[0123] The processor 502 runs a computer program corresponding to the executable program code stored in the memory 501 to implement the field personnel search method based on UAV SAR images in the first aspect embodiment described above.

[0124] In some examples, the electronic device 500 may also include a communication interface 503 and a bus 510. For example, Figure 5 As shown, the memory 501, processor 502, and communication interface 503 are connected through bus 510 and complete communication with each other.

[0125] The communication interface 503 is mainly used to realize communication between various modules, systems, units and / or devices in the embodiments of this application. Input devices and / or output devices can also be connected through the communication interface 503.

[0126] Bus 510 includes hardware, software, or both, that couples components of electronic device 500 together. For example, and not limitingly, bus 510 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-E) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 510 may include one or more buses. Although specific buses are described and illustrated in the embodiments of this application, this application considers any suitable bus or interconnection.

[0127] The electronic device provided in this application embodiment is capable of achieving Figure 1-3 The various processes implemented in the method embodiments can achieve the same technical effect, and will not be described again here to avoid repetition.

[0128] In conjunction with the field personnel search method based on UAV SAR imagery described in the above embodiments, this application embodiment can provide a computer storage medium for implementation. This computer storage medium stores computer program instructions; when executed by a processor, these computer program instructions implement any step in the above method embodiments.

[0129] In conjunction with the field personnel search method based on UAV SAR imagery in the above embodiments, this application embodiment can provide a computer program product to implement it. This (computer) program product is stored in a non-volatile storage medium, and when executed by at least one processor, it implements any step in the above method embodiments.

[0130] This application also provides a chip, which includes a processor and a communication interface. The communication interface and the processor are coupled. The processor is used to run programs or instructions to implement the various processes of the above method embodiments and achieve the same technical effect. To avoid repetition, it will not be described again here.

[0131] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0132] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0133] The functional blocks shown in the above-described block diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0134] It should also be noted that the exemplary embodiments mentioned in this application describe some methods or systems based on a series of steps or systems. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0135] The aspects of this disclosure have been described above with reference to flowchart illustrations and / or block diagrams of methods, systems (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing system to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing system, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by special-purpose hardware performing the specified functions or actions, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0136] The above description is merely a specific implementation of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.

Claims

1. A method for searching for personnel in the field based on UAV SAR imagery, characterized in that, include: Synthetic aperture radar (SAR) images of the target search area at multiple time points are acquired to obtain a multi-temporal SAR image sequence; Based on the multi-temporal SAR image sequence, a background scattering characteristic model of the target search area is constructed; Interferometric coherence change detection is performed on the SAR images of the current time node and the reference time node in the multi-temporal SAR image sequence, and combined with the background scattering characteristic model, the suspected changing target area that changes relative to the background within the target search area is obtained; The motion state features and spatial scale features of the suspected changed target area are extracted from the multi-temporal SAR image sequence, and the motion state detection and spatial scale detection are performed on the suspected changed target area in sequence based on the motion state features and spatial scale features. The motion state detection is used to evaluate whether the motion of the suspected changed target area conforms to the preset human motion feature model, and the spatial scale detection is used to evaluate whether the suspected changed target area matches the size range of the scattering features of the human body in the SAR image. If both motion continuity detection and spatial scale detection pass, the suspected change target area is identified as a suspected human target; Based on the location change information of the suspected human target in the multi-temporal SAR image sequence, a location analysis is performed, and the search and location results are output.

2. The method according to claim 1, characterized in that, The construction of the background scattering characteristic model of the target search area includes: One frame of SAR image is selected from the multi-temporal SAR image sequence as the initial background reference image; In the initial background reference image, regions that satisfy the scattering characteristics of natural vegetation and have a continuous spatial distribution are selected as background candidate regions; The radar scattering characteristic parameters of the background candidate region are extracted. These radar scattering characteristic parameters include echo amplitude statistical characteristics, interferometric coherence characteristics, and spatial distribution continuity characteristics. Based on the radar scattering characteristic parameters, a parameterized background scattering characteristic model is constructed.

3. The method according to claim 1, characterized in that, The step involves detecting interferometric coherence changes in the SAR images of the current time node and the reference time node within the multi-temporal SAR image sequence, and combining this with the background scattering characteristic model to obtain suspected changing target regions within the target search area that exhibit changes relative to the background, including: Calculate the interferometric coherence coefficient map between two SAR images of the current time node and the reference time node; Pixel regions with coherence below a preset change threshold are identified from the interference coherence coefficient map to obtain the initial change region; The radar scattering characteristic parameters of the initial change region are compared with the parameters at the corresponding positions in the background scattering characteristic model; Delete the regions in the initial change region whose radar scattering characteristic parameters differ from the background scattering characteristic model parameters by less than a preset background difference threshold, and determine the remaining regions as the suspected change target regions.

4. The method according to claim 1, characterized in that, The motion state features include displacement velocity features, displacement trajectory features, and motion pattern features. Motion state detection of the suspected changed target region based on these motion state features includes at least two of the following: Based on the displacement velocity characteristics, the displacement amplitude of the suspected change target area is detected to verify whether its instantaneous movement speed is within the speed range that people can walk on. Based on the displacement trajectory features, displacement continuity detection is performed on the suspected change target area to verify whether the changes between adjacent displacements are smooth and continuous. Based on the motion pattern characteristics, the motion pattern rationality of the suspected change target area is detected to verify whether its comprehensive motion trajectory conforms to human motion behavior characteristics.

5. The method according to claim 4, characterized in that, Based on the displacement velocity characteristics, displacement amplitude detection is performed on the suspected change target region, including: The position of the suspected changed target region in the images of two adjacent time nodes in the multi-temporal SAR image sequence is obtained, and the displacement distance of the suspected changed target region between the two adjacent time nodes is calculated based on the position; The imaging time interval between two adjacent time nodes is obtained, and the displacement velocity feature is calculated based on the displacement distance and the imaging time interval. If the displacement velocity characteristic is within the range of the person's movement speed, then the displacement amplitude detection is deemed successful.

6. The method according to claim 4, characterized in that, Based on the displacement trajectory features, displacement continuity detection is performed on the suspected changed target area, including: The location of the suspected changed target region in the images of at least three consecutive time nodes in the multi-temporal SAR image sequence is obtained; Based on the location, the displacement distance between each pair of adjacent time nodes of the suspected change target area is calculated to obtain a displacement distance sequence, which is used as the displacement trajectory feature; Based on the displacement distance sequence, calculate the difference between every two adjacent displacement distances; If each of the aforementioned differences is less than or equal to a preset maximum displacement change threshold, then the displacement continuity detection is deemed successful.

7. The method according to claim 4, characterized in that, Based on the motion pattern features, the reasonableness of the motion pattern in the suspected changed target area is detected, including: Obtain the positional change information of the suspected changed target area in the SAR images of multiple time nodes in the multi-temporal SAR image sequence; Based on the position change information, the motion pattern features are extracted, and the motion pattern features are used to characterize the comprehensive motion trajectory morphology of the target in multiple time phases; Based on the motion pattern characteristics, determine whether the motion trajectory of the suspected change target area conforms to a preset motion behavior pattern, wherein the preset motion behavior pattern includes the motion speed exceeding a preset speed change threshold and the motion direction undergoing a change exceeding a preset angle threshold between consecutive time nodes; If the movement trajectory does not conform to the preset movement behavior pattern, then the movement pattern rationality test is passed.

8. The method according to claim 1, characterized in that, Spatial scale detection of the suspected changed target region based on spatial scale characteristics includes: The area and maximum external dimensions of the suspected change target region are obtained to determine its spatial scale characteristics. Determine whether the area is within the range formed by the first preset area threshold and the second preset area threshold, and whether the maximum external dimension is within the range formed by the first preset size threshold and the second preset size threshold; If the area and the maximum external dimension are within the corresponding preset range, then the suspected change target area is determined to have passed the spatial scale detection. The first preset area threshold, the second preset area threshold, the first preset size threshold, and the second preset size threshold are preset according to the size range of the scattering characteristics of the human body in the SAR image.

9. The method according to claim 1, characterized in that, After performing motion state detection and spatial scale detection on the suspected changed target region based on motion state features and spatial scale features, the process further includes: The number of times the suspected changed target region appears continuously in a multi-temporal SAR image sequence with multiple consecutive time nodes is obtained; When the number of occurrences is greater than or equal to a preset minimum duration threshold, the time duration detection is deemed to have passed.

10. The method according to claim 1, characterized in that, Based on the location change information of the suspected human target in the multi-temporal SAR image sequence, a location analysis is performed, and the search and location results are output, including: By combining the location change information of the suspected person, the geographical information of the target search area, and the prior activity information of the missing person, at least one suspected activity area is generated; The suspected activity area is associated with the latest location of the suspected person, and the search and location results, including location confidence, are output.

11. The method according to claim 1, characterized in that, After outputting the search and location results, it also includes: Control the UAV equipped with synthetic aperture radar to perform supplementary flight scans of the target search area or the key area indicated by the search and positioning results, and acquire SAR images of newly added time-series nodes; The SAR image of the newly added time-series node is added to the multi-temporal SAR image sequence, and the steps of constructing the background scattering characteristic model, obtaining the suspected changing target area, performing detection, determining the suspected human target, and outputting the search and location results are re-executed to update the search and location results.

12. A computer program product, characterized in that, The computer program product is stored in a non-volatile storage medium, and when executed by a processor, the computer program product implements the method as described in any one of claims 1-11.

Images