A forest scatterer type identification and positioning system based on radar polarization decomposition

Abstract

The application relates to the technical field of radar target detection, and discloses a forest scatterer type identification and positioning system based on radar polarization decomposition, which comprises a polarization feature preprocessor, a main controller and a threshold calibration unit; the main controller performs asymmetric scheduling and review; the threshold calibration unit uses the clutter echo obtained through the review to calculate the background clutter feature baseline in real time, and dynamically updates the classification threshold used by the polarization feature preprocessor; the application uses the review information flow to realize closed-loop self-calibration of the classification threshold, makes the system free from the dependence on the priori of the static environment, can automatically track the feature drift of the background clutter caused by meteorological changes, and avoids the risk that the scheduling logic is disabled due to the saturation of the classifier.

Description

Technical Field

[0001] This invention relates to a forest scatterer type identification and localization system based on radar polarization decomposition, belonging to the field of radar target detection technology. Background Technology

[0002] Currently, its core task is to achieve reliable detection and accurate positioning of targets in various environments. The mainstream technical systems in this field, such as pulse Doppler radar or continuous wave radar, rely heavily on a basic premise in their design philosophy and working mechanism: the energy of the target's echo signal at the receiver must be higher than or different from the energy of the background clutter. Therefore, statistical detectors based on energy or amplitude, such as constant false alarm rate detectors, constitute the standard tools for target determination in such systems. However, when the above systems are applied to emergency search and rescue in strong scattering environments such as forests, such as searching for crashed aircraft, vehicles, or missing persons in dense forests, the energy detection premise on which they rely no longer holds. Forest canopies, tree trunks, and vegetation are extremely strong radar scatterers. For a target located under the canopy, its radar cross-section may be much smaller than the total radar cross-section of the massive vegetation within the same radar resolution unit. This results in the energy of the target echo being completely submerged by the energy of the massive forest clutter, and the signal-to-clutter ratio is far below the detection threshold.

[0003] To address this, some researchers in this field have attempted to separate clutter from targets using feature-based filtering logic. For example, Chinese invention patent CN118097873A discloses a forest fire early warning method and system based on weather radar. This scheme attempts to separate suspected fire smoke echoes through a series of filtering steps based on specific thresholds, such as filtering clear-sky echoes, precipitation echoes, and ground object echoes, and finally confirms them with satellite hotspots. However, this technical approach is essentially an open-loop filtering logic that relies on static prior knowledge. The various filtering thresholds it depends on, such as the velocity threshold for ground object echoes or the intensity threshold for clear-sky echoes, once set, cannot respond to dynamic changes in the environment. For example, when the scattering characteristics of background clutter, such as vegetation, change due to meteorological changes such as rainfall or dew, the clutter will become more sensitive to environmental changes. Over time, these fixed static thresholds will face the risk of failure, causing the system to misclassify a large number of clutter as targets or vice versa, making it difficult to guarantee the reliability of detection under complex conditions. At this time, the system faces a dilemma: on the one hand, insisting on using conventional energy detectors will fail due to the inability to distinguish between targets and strong clutter, resulting in an extremely high false negative rate; on the other hand, emergency search and rescue missions impose strict constraints on the real-time performance of the system, making it impossible to use technologies such as synthetic aperture radar to perform complex polarimetric imaging analysis that requires several hours of offline processing. Although the latter can theoretically distinguish ground features on images through complex decomposition algorithms, its time cost is unacceptable in emergency scenarios. This makes it difficult for the system to balance real-time performance and detection reliability in strong clutter environments.

[0004] Therefore, the technical problem to be solved by this invention is how to devise a completely new system working method that, under the premise of strictly meeting the constraints of real-time detection and positioning, gets rid of the single dependence on energy domain detection and utilizes other physical dimension characteristics of the echo to transform an undetectable target that is submerged in energy by clutter into a target that can be reliably detected and located by the system with extremely low computational cost. Summary of the Invention

[0005] This invention provides a forest scatterer type identification and positioning system based on radar polarization decomposition. Its main purpose is to solve the problem that traditional energy detection fails due to target being submerged by clutter in strong scattering environments such as forests, while complex imaging technology cannot meet real-time constraints, resulting in the system being unable to reliably locate targets that are not detectable in the energy domain in real time.

[0006] To achieve the above objectives, the present invention provides a forest scatterer type identification and localization system based on radar polarization decomposition, comprising:

[0007] A polarization feature preprocessor is used to calculate the polarization entropy of each resolution unit in the wide-area search region in real time, and divide the resolution unit into high-entropy and low-entropy regions according to a dynamic classification threshold to generate region of interest masks.

[0008] The main controller is used to receive the region of interest mask and perform asymmetric resource scheduling, which includes scheduling detection resources to focus on performing fine detection in low-entropy regions and skipping fine detection in high-entropy regions; the main controller is also used to perform asymmetric verification, which includes forcibly allocating a small amount of detection resources to verification units in high-entropy regions to obtain their original polarization feature data.

[0009] The threshold calibration unit is specifically used to receive the original polarization feature data of the verification unit identified as the high-entropy region when performing asymmetric verification. Based on the original polarization feature data, a background clutter feature baseline characterizing the clutter background of the current high-entropy region is calculated in real time, and the dynamic classification threshold is dynamically calculated and updated based on the background clutter feature baseline.

[0010] The polarization feature preprocessor is also used to employ a dynamic classification threshold, which is updated in real time by the threshold calibration unit, as the current basis for dividing the high-entropy region into the low-entropy region.

[0011] Preferably, the system further includes: a Doppler feature analysis unit, used to calculate the Doppler spectral width of the echo signal of the resolution unit in parallel; the main controller is also used to, when performing asymmetric resource scheduling, only identify the resolution units that simultaneously satisfy the polarization entropy below the dynamic classification threshold and the Doppler spectral width below the preset spectral width threshold as low-entropy regions, and identify high-entropy regions and low-entropy but high-spectral-width regions as non-interest regions.

[0012] Preferably, the polarization feature preprocessor is also used to: calculate a feature gradient map representing the spatial rate of change of polarization entropy in real time based on the original data of polarization entropy, and the logic of the main controller for performing the asymmetric verification is further limited to allocating a small amount of detection resources to regions with higher feature gradient values ​​in high-entropy regions according to the feature gradient map.

[0013] Preferably, the logic of the main controller for performing asymmetric verification is further defined as a transmitter of the instruction system, which transmits a sequence of active probe pulses with multiple different preset polarization states to the verification unit; the polarization feature preprocessor is also used to receive and analyze the echo response sequence excited by the active probe pulse sequence, and to identify the camouflaged target by identifying whether the echo response sequence exhibits a jump that does not conform to clutter statistical characteristics as the transmission polarization state switches; the main controller is also used to schedule detection resources to focus on the camouflaged target.

[0014] Preferably, the system further includes: a logic integrity monitoring unit, used to monitor in real time the global proportion of regions identified as low-entropy regions in the region of interest mask; the main controller is also used to automatically bypass asymmetric resource scheduling and fall back to the global detection mode that performs standard detection on the entire range of the wide-area search region when the logic integrity monitoring unit detects that the global proportion exceeds the preset system load threshold.

[0015] Preferably, the logic of the threshold calibration unit for dynamically calculating and updating the dynamic classification threshold includes: performing statistical analysis on the received raw polarization feature data to calculate the background clutter feature baseline. And based on the background clutter characteristic baseline The dynamic classification threshold is calculated using the following rules. : ,in This is a preset entropy offset used to ensure the separation of the target from clutter.

[0016] Preferably, the threshold calibration unit is used to calculate the background clutter characteristic baseline. The statistical analysis includes: constructing a statistical histogram from the received raw polarization feature data, and calculating the preset quantiles of the statistical histogram to use as the background clutter feature baseline. .

[0017] Preferably, the logic of the polarization feature preprocessor for calculating the feature gradient map includes: applying a preset convolution kernel to the original data of polarization entropy to calculate the local rate of change of polarization entropy in space.

[0018] Preferably, the logic of the Doppler feature analysis unit for calculating the Doppler spectral width includes: performing Doppler processing on the echo signal to obtain the Doppler spectrum, and measuring the spectral peak width of the Doppler spectrum to use it as the Doppler spectral width.

[0019] Preferably, the logic of the polarization feature preprocessor for identifying echo response sequences includes: establishing clutter statistical characteristics, which characterize a preset smooth response mode in which the clutter echo amplitude changes with the switching of the transmit polarization state; identification includes: when the echo amplitude change of adjacent pulses in the echo response sequence exceeds the preset range defined by the smooth response mode, the verification unit is identified as a camouflaged target.

[0020] Compared with the prior art, the beneficial effects of the present invention are:

[0021] 1. This invention provides a system operation mode for multi-dimensional feature cascade filtering. When using radio waves for positioning or presence detection, it utilizes the polarization scattering characteristics of the echo to identify and skip high-entropy volume scattering clutter regions before they enter the core detection process. For the remaining low-entropy regions, the system does not immediately treat them all as targets. Instead, it further utilizes the system's inherent Doppler processing capability and introduces the Doppler spectral width criterion, which is orthogonal in physical mechanism, to effectively separate rigid targets with zero spectral width characteristics from benign interference sources such as water bodies with non-zero spectral width characteristics. This progressive information purification from scattering mechanism to motion characteristics allows the system's detection resources to focus only on high-value rigid targets, solving the dual masking problem of strong clutter and benign interference while maintaining real-time detection.

[0022] 2. By reusing the internal information flow of the system, the asymmetric verification mechanism that ensures the robustness of the system is coupled in a closed loop with the effectiveness of its classification logic. When the system performs a small number of verification probes to prevent missed detections, it does not discard the acquired clutter echo data, but uses it as a real-time sample of the clutter characteristics of the current environmental background. These real-time samples are used to continuously calculate and update a dynamic classification threshold, which is then fed back to the polarization feature preprocessor. This design transforms the system from an open-loop scheduling method that relies on static prior knowledge into a closed-loop adaptive system that can automatically track external environmental changes such as weather changes. Without increasing additional detection overhead, it ensures the continuous effectiveness of its core scheduling basis under complex operating conditions.

[0023] 3. By monitoring the statistical saturation of scheduling instructions, a logical integrity guarantee mechanism is constructed for the system. When the system encounters environmental changes caused by non-target factors, such as severe weather, which leads to global contamination of the core classification criteria and causes the global proportion of the low-entropy region to exceed the preset system load threshold, the system will automatically bypass the asymmetric scheduling logic based on polarization characteristics and force a rollback to the original baseline working mode that performs standard testing on the entire range. This design transforms an uncontrollable logical failure that could paralyze system resources into a controllable, pre-designed graceful degradation to baseline performance, ensuring the system's most basic functions remain online and available under any extreme environment. Attached Figure Description

[0024] Figure 1 This is a block diagram of the multi-unit collaborative and dual-closed-loop control logic of the system of the present invention;

[0025] Figure 2 This is a schematic diagram illustrating the adaptive tracking effect of dynamic classification threshold under environmental drift conditions according to the present invention.

[0026] Figure 3 This is the timing diagram for the dynamic threshold closed-loop calibration based on asymmetric verification in this invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the scope of protection of the invention.

[0028] This invention provides a forest scatterer type identification and localization system based on radar polarization decomposition. Its core components include a polarization feature preprocessor, a main controller, and a threshold calibration unit. The system as a whole uses the polarization feature preprocessor to perform real-time polarization feature analysis on echoes from a wide-area search region, generating a basis for dividing high- and low-entropy regions. The main controller performs asymmetric detection resource scheduling and verification based on this division. The threshold calibration unit uses the data obtained from the verification to reverse-calibrate the classification threshold of the polarization feature preprocessor, forming a closed-loop adaptive detection and localization system. In a specific system deployment, this system can be integrated into a radar platform with full-polarization or dual-polarization transceiver capabilities. This platform targets a wide-area search region, taking a densely forested area as an example, to perform presence detection and localization tasks. The primary challenge for the polarization feature preprocessor in this context is to initially screen out potential target cues from the massive energy-dominated forest clutter. To address this, the polarization feature preprocessor is used to calculate the polarization entropy of each resolution cell in the wide-area search region in real time. This calculation process follows a defined signal processing procedure: the input is the raw polarization echo data sequence acquired by the radar receiver in that resolution cell; the processing logic includes constructing the polarization covariance matrix or coherence matrix of that resolution cell; then, multi-look smoothing is performed to suppress speckle noise; and eigenvalues ​​of the matrix are obtained through eigenvalue decomposition. And ultimately based on the formula The polarization entropy of the unit was calculated. Value, of which These are normalized feature values; the preprocessor uses a dynamic classification threshold to... Cells with values ​​higher than this threshold are identified as high-entropy regions, which typically correspond to forest clutter with volume scattering characteristics, and will... Cells with values ​​below this threshold are identified as low-entropy regions, which may correspond to man-made targets or specific ground features, thus generating a global region of interest mask.

[0029] The main controller receives the region of interest mask generated by the polarization feature preprocessor. Its core task is to change the system's uniform resource allocation mode to perform asymmetric resource scheduling. Given that the system's detection resources, such as dwell time and transmit power, are limited, the main controller must prioritize them for the highest-value regions. Specifically, this asymmetric resource scheduling manifests as follows: the main controller controls the radar's beam pointing and dwell time manager to skip fine-grained detection of high-entropy regions, i.e., avoiding long pulse accumulation or high duty cycle detection in these areas, thereby diverting the vast majority of detection resources away from ineffective clutter. Released from the background wave; simultaneously, detection resources are scheduled to focus on low-entropy regions for fine-grained detection. For example, the instruction system increases the dwell time of these low-entropy region units, uses high-resolution waveforms, or initiates subsequent fine-grained identification processes to confirm the existence of the target and perform precise localization. However, the effectiveness of the above scheduling logic depends entirely on the accuracy of classification. If a target, for example a camouflaged target, is misclassified as a high-entropy region, the system will permanently skip that target, resulting in missed detection. To avoid this risk at the system level, the main controller is also used to perform asymmetric verification. This verification logic is designed such that the main controller strongly... A small portion of the detection resources, such as 1% to 5% of the total temporal detection resources, is selectively allocated to verification units within regions already identified as high-entropy areas. These verification units can be selected through uniform random sampling within the high-entropy regions or based on a risk-weighted strategy. The main controller instructs the system to perform standard probes on these verification units to obtain their raw polarization characteristic data. This data stream serves two purposes: preventing missed detections and providing crucial data input for system closed-loop calibration. The threshold calibration unit addresses the background drift problem faced by the system in real-world environments, specifically the polarization characteristics of forest clutter. The polarization characteristics can change with weather conditions, such as rainfall or dew, causing the static threshold to fail. Therefore, this unit is specifically designed to receive the raw polarization characteristic data of the verification units, identified as high-entropy regions, acquired during asymmetric verification. This data represents real-time sampling of the current real clutter background. Based on the raw polarization characteristic data, the threshold calibration unit calculates a background clutter characteristic baseline characterizing the current high-entropy clutter background in real time. This calculation process can follow a defined statistical procedure. For example, in one implementation, this unit collects the polarization entropy of all verification samples within a 5-minute sliding time window. The values ​​were calculated, and a statistical histogram was constructed for these samples. The 50th percentile of this statistical histogram, i.e., the median, was calculated and used as the baseline for background clutter characteristics. ; obtain Subsequently, the threshold calibration unit dynamically calculates and updates the dynamic classification threshold based on the background clutter characteristic baseline; this calculation logic can follow a rule to ensure separation, such as... ,in It is a preset entropy offset used to ensure the separation of the target from clutter, which can be calibrated to 0.08; finally, the polarization feature preprocessor uses a dynamic classification threshold updated in real time by the threshold calibration unit as the current basis for dividing the high-entropy region and the low-entropy region.

[0030] In some application scenarios, especially forest search and rescue, relying solely on polarization entropy is insufficient to distinguish high-value targets (such as metal debris) from benign interference sources (such as still water surfaces), both of which may exhibit low-entropy characteristics, causing the system to waste a significant amount of resources focusing on water bodies. To address this issue, the system of this invention may further include: a Doppler feature analysis unit, which utilizes the system's inherent Doppler processing capabilities to calculate the Doppler spectral width of the echo signal from the resolving unit in parallel. Its calculation logic includes: performing Doppler processing on the echo signal within a coherent processing interval, such as a fast Fourier transform, to obtain the Doppler spectrum, and measuring the spectral peak width of the Doppler spectrum, which can be a -3dB bandwidth or a root mean square bandwidth, as the Doppler spectral width. Correspondingly, the main controller is also used to introduce serial decision logic when performing asymmetric resource scheduling: based on physical priors, Rigid targets exhibit high value with low entropy and zero spectral width, while water interference sources exhibit low entropy but high spectral width. The main controller only identifies resolvable units that simultaneously satisfy a polarization entropy below the dynamic classification threshold and a Doppler spectral width below a preset spectral width threshold of 0.05 m / s as low-entropy regions. High-entropy regions (forest clutter) and low-entropy but high-spectral-width regions (water interference) are identified as non-interest regions. This progressive purification from scattering mechanism to motion characteristics ensures that detection resources are focused solely on high-value rigid targets. To improve the efficiency of the asymmetric kernel, the kernel logic of the main controller can be optimized. The polarization feature preprocessor is also used to: calculate, in real time, a feature gradient map representing the spatial change rate of polarization entropy based on the original polarization entropy data obtained during polarization entropy calculation. This calculation logic may include: applying a preset convolution kernel, such as a... The Sobel operator is used to calculate the local rate of change of polarization entropy in space; correspondingly, the logic of the main controller for performing the asymmetric verification is further defined as follows: according to the feature gradient map, a small amount of detection resources are preferentially allocated to regions with higher feature gradient values ​​in the high-entropy region, because these regions are the boundary between high-entropy and low-entropy, and are the regions where misjudged or disguised targets are most likely to hide.

[0031] To counter a baseline camouflaged target whose passive polarization characteristics match clutter, the system can introduce an active discrimination mechanism. The logic of the main controller performing asymmetric verification is further defined as follows: the transmitter of the command system transmits a sequence of active probe pulses with multiple preset polarization states to the verification unit. This sequence can sequentially include horizontally polarized, vertically polarized, left-hand circularly polarized, and right-hand circularly polarized pulses. The polarization characteristic preprocessor is also used to receive and analyze the echo response sequence excited by the active probe pulse sequence. Its discrimination logic is as follows: the system pre-establishes clutter statistical characteristics, which characterize a preset smooth response pattern where the echo amplitude of clutter (such as a forest) changes with the switching of the transmitted polarization state. Discrimination includes: when the echo amplitude change of adjacent pulses in the echo response sequence, for example, the amplitude ratio of the H-pulse echo to the V-pulse echo, exceeds a preset range defined by the smooth response pattern, i.e., it exhibits a non-consistent clutter statistical characteristic. When there is a drastic change in characteristics, the verification unit is identified as a camouflaged target, and the main controller then schedules detection resources to focus on the camouflaged target. Finally, to ensure that the system's detection function remains available when encountering unexpected extreme conditions that cause global contamination of polarization features, the system also includes: a logic integrity monitoring unit; this unit is used to monitor in real time the global proportion of regions identified as low-entropy areas in the region of interest mask; the main controller is also used to: have a built-in preset system load threshold, such as 40% of the global proportion. When the logic integrity monitoring unit detects that the global proportion exceeds this threshold, it means that the classification criteria on which the scheduling logic depends may have globally failed; at this time, the main controller automatically bypasses asymmetric resource scheduling and forces a fallback to a global detection mode that performs standard detection on the entire range of the wide-area search region. This fallback to baseline performance ensures the basic availability and logical robustness of the system under any circumstances.

[0032] Example 1: The system of the present invention operates in an emergency search and rescue scenario, which involves locating and searching for a crashed small aircraft in a dense, humid forest area following a sudden heavy rainfall. This condition presents the system with two challenges: First, the radar cross-section of the target, i.e., the aircraft wreckage, is much smaller than the massive amount of humid vegetation within its resolution unit, causing it to be completely submerged in the energy domain, rendering conventional CFAR detectors ineffective. Second, the recent rainfall has caused an unknown drift in the overall polarization characteristics of the forest clutter, while large areas of still water have formed in low-lying areas within the forest. The system initiates a wide-area search, and the polarization feature preprocessor begins calculating the global polarization entropy. Simultaneously, the main controller activates the asymmetric verification logic, which forcibly allocates a small amount of detection resources, such as 1%, randomly to the preliminarily identified high-entropy forest for detection. The threshold calibration unit specifically receives the raw polarization characteristic data from these verification units in the high-entropy region and calculates a dynamic background clutter characteristic baseline in real time by calculating the median through statistical analysis. ,Should This reflects the actual polarization characteristics of the current humid forest; the threshold calibration unit then uses a base-based... Dynamically updated classification thresholds Feedback is given to the polarization feature preprocessor; this closed-loop mechanism enables the system's classification criteria. Capable of dynamically tracking the background clutter characteristic baseline caused by environmental changes and rainfall. The drift.

[0033] The polarization feature preprocessor uses this dynamic classification threshold to generate a region of interest mask. This mask identifies the vast majority of regions, such as 95%, as high-entropy, humid forest areas. Upon receiving this mask, the main controller immediately executes asymmetric resource scheduling, instructing the system to skip fine-grained detection of these 95% of regions and focus all detection resources, such as long-staying, high-power pulses, on the remaining 5% low-entropy region. At this point, the system faces a low-entropy degeneracy problem; these 5% low-entropy regions contain both high-value rigid aircraft wreckage and large areas of still water formed by benign interference sources like rainfall. At this critical juncture, the cascade filtering mechanism of the Doppler feature analysis unit is activated. This unit calculates the Doppler spectral width of all resolving units within these 5% low-entropy regions in parallel. The main controller then executes the final resource allocation... During scheduling decisions, a series of decision-making logics were applied: for cells with low entropy but high spectral width exceeding a preset spectral width threshold, they were identified as water bodies and classified as non-interest regions; while only those cells with low entropy and zero spectral width below the preset spectral width threshold were ultimately identified as high-priority interest points. After this series of series filtering operations based on polarization and motion characteristics, the aircraft debris, which was previously submerged in strong clutter and benign interference, has low entropy and zero spectral width, and is identified as a high-priority target to be detected in the system's main controller logic. The system's core detection resources, such as high signal-to-noise ratio long pulse accumulation, are fully focused on this target, and the standard energy detector is activated at this point, enabling reliable detection and localization of the target in a context where clutter and interference have been filtered out by previous processes.

[0034] Example 2: This example objectively verifies the improvement in target detection performance of the system of the present invention compared with conventional technology under strong clutter and complex interference environments through a system simulation test. The test platform is constructed as a numerical simulation environment to simulate and generate radar echo data containing targets, ground clutter, and interference. The platform can accurately control the radar cross section (RCS) and polarization entropy of the echoes of each scatterer. ) and Doppler spectral width ( The experiment sets up a search and rescue scenario where the target (T) is a small metal debris with an RCS of -25 dBsm and low entropy physical characteristics. ) and zero spectral width ( The target exists within a radar resolution cell, which is simultaneously covered by strong forest clutter (C1). The RCS of the forest clutter is set to -10 dBsm, i.e., the signal-to-clutter ratio (S / C) is -15 dB, and the target energy is lower than the clutter energy. Four system configuration samples were set up for comparison: Control group A: Simulates a conventional system, using only an energy-based constant false alarm rate (CFAR) detector without any polarization processing; Control group B: Employs simplified polarization filtering, using a static classification threshold based on prior knowledge of dry forests (…). The control group C: uses the polarization filtering and closed-loop calibration of the present invention, which includes a polarization feature preprocessor, a main controller and a threshold calibration unit, but does not include a Doppler feature analysis unit; the sample group of the present invention: uses the complete system of the present invention, which includes all core units, namely the polarization feature preprocessor, the main controller, the threshold calibration unit and the Doppler feature analysis unit.

[0035] The experiment was conducted in three different scenarios, with each scenario being run repeatedly. The statistical detection probability ( ) and false alarm probability ( S / C remains -15dB in all scenarios: Scenario 1 (baseline): Dry forest environment, polarization entropy of forest clutter C1 The statistical mean is 0.92; Scenario 2 (Environmental Drift): Simulating a humid forest environment after rainfall, the polarization entropy of forest clutter C1. The statistical mean drifted to 0.76; Scenario 3 (compound interference): dry forest environment ( The mean value is 0.92), but a large-area low-entropy interference source, namely water (C2), is introduced into the scene. Its characteristic is low entropy ( ) and non-zero spectral width ( (m / s), the experimental data are summarized in Table 1.

[0036] Table 1: Comparison of Detection Performance of Different System Configurations in Search and Rescue Scenarios

[0037]

[0038] Referring to Table 1, in all scenarios with an S / C of -15dB, the detection probability of control group A (energy detection only) is lower because the target is overwhelmed by clutter energy. All values ​​were below 1%, confirming the failure of conventional systems under this condition; in scenario 1 benchmark, control groups B and C, as well as the sample group of this invention, could effectively filter out high-entropy forests. ), and focus on low-entropy objectives ( All of them received more than 98% Compared to less than 0.5% In scenario 2, during environmental drift, the static threshold of control group B ( ) fails because the entropy of the humid forest ( If the value falls below the static threshold, the system will misidentify large-area clutter as points of interest, increasing the false alarm probability. The threshold calibrator rose to 88.4%, causing the system to malfunction. In contrast, the control group C and the sample group of this invention obtained their threshold calibration units through asymmetric verification. The clutter samples are used to calculate new background clutter characteristic baselines in real time. And dynamically update the classification thresholds. This dynamically adjusts the classification threshold to a lower value (e.g., 0.68), adapting to environmental drift and maintaining high accuracy. With low In scenario 3, with compound interference, although control group C can correctly handle forest clutter through dynamic thresholding, it cannot distinguish low-entropy targets. ) and low-entropy water bodies ( This caused it to classify all bodies of water as points of interest. The figure rose to 14.8%; after performing polarization decision, the sample group of this invention further activated the Doppler feature analysis unit to perform spectral width decision on all low-entropy units, which will target units with zero spectral width ( ) are identified as high-priority points of interest, while water bodies with non-zero spectral width are also identified. ) is marked as a non-interest area and skipped. It eventually remained at 0.5%.

[0039] Example 3: This example combines Figures 1 to 3 This document describes a forest scatterer type identification and localization system based on radar polarization decomposition, as follows: Figure 1As shown, the system uses the raw polarization echo of the wide-area search region as the raw input data. This raw echo is sent to the polarization feature preprocessor and the Doppler feature analysis unit, respectively. The polarization feature preprocessor is used to calculate the polarization entropy and generate the region of interest mask based on the dynamic threshold. The Doppler feature analysis unit is used to calculate the Doppler spectral width of the echo signal in parallel. The region of interest mask and Doppler spectral width data generated by both are submitted to the main controller. The main controller performs asymmetric resource scheduling and asymmetric verification. On the one hand, it schedules detection resources to focus on the low-entropy region to achieve the system output of detection resource focusing and target localization. On the other hand, it sends the high-entropy region data obtained from the verification to the threshold calibration unit. The threshold calibration unit uses the verification data to calculate the clutter baseline, dynamically updates the classification threshold, and feeds back the dynamic classification threshold to the polarization feature preprocessor, forming the main closed loop. At the same time, the region of interest mask is also sent to the logic integrity monitoring unit to monitor the global proportion of the low-entropy region to prevent system overload. This unit then sends a global proportion / back-off command to the main controller, forming a logic integrity monitoring loop.

[0040] like Figure 2 As shown in the figure, the horizontal axis represents time (in minutes), and the vertical axis on the left represents polarization entropy. The value, with the right vertical axis representing rainfall intensity (%), is shown in the figure. The figure contains three curves, representing the characteristic baseline of background clutter. Dynamic classification threshold And the rainfall intensity simulation, as shown in the figure, shows that as the rainfall intensity simulation curve rises from time 0 and reaches its peak at 40 minutes, the background clutter characteristic baseline... The curve characterizing environmental clutter properties decreased accordingly from 0.92 to 0.76, while the dynamic classification threshold... Curve, which follows, for example The rules are always automatically tracked. The drift decreased synchronously from 0.84 to 0.68; such as Figure 3As shown in the figure, this diagram depicts the interaction between the radar receiver, polarization feature preprocessor, main controller, and threshold calibration unit. The process begins with the system initiating a wide-area search. The radar receiver sends raw polarization echo data to the polarization feature preprocessor. The polarization feature preprocessor performs a series of internal processes, including constructing the polarization covariance matrix, performing multi-look smoothing, calculating the polarization entropy value, and dividing the region into high-entropy and low-entropy regions based on a dynamic threshold to generate a region of interest mask. This mask is then sent to the main controller. After analyzing the region of interest mask, the main controller makes an asymmetric resource scheduling decision. On the one hand, it instructs the radar receiver to schedule detection resources to focus on the low-entropy region; on the other hand, it instructs the radar receiver to forcibly allocate a small amount of resources to the high-entropy region verification unit. The radar receiver sends the acquired high-entropy region verification unit data to the threshold calibration unit. The threshold calibration unit collects the polarization entropy values ​​of the verification samples, constructs a statistical histogram, calculates the background clutter feature baseline, and finally dynamically updates the classification threshold. It then feeds back the updated dynamic threshold to the polarization feature preprocessor, which then uses the new threshold to continue the next round of classification.

[0041] Example 4: This example illustrates a standardized engineering calibration procedure for determining key preset parameters of the system of the present invention. This procedure is executed before system deployment and aims to provide a parameter basis for the reliable operation of subsequent dynamic logic. The core objective of this procedure is to determine the preset spectral width threshold for the Doppler feature analysis unit and the entropy offset for the threshold calibration unit. The calibration was conducted at a radar test range with controllable target and environmental settings. The first step involved calibrating a preset spectral width threshold to distinguish between rigid and non-rigid interference. The initial targets included a metallic corner reflector as a rigid target with a theoretically zero Doppler spectral width, and a large water tank as a non-rigid benign interference source, whose microscopic surface ripples resulted in a non-zero spectral width. The enabling environment for calibration consisted of a radar with sufficient specifications to resolve velocity differences on the order of 0.01 m / s. The calibration procedure was as follows: the system was aligned with the metallic corner reflector, and data was collected under stationary conditions. The echo of each coherent processing interval, among which The value was set to 1000, and the Doppler spectral width for each interval was calculated. The mean of the statistical distribution of the spectral width was measured to be 0.005 m / s. The maximum value was calculated by adding three times the standard deviation to the mean. m / s; then, the system was aimed at the water tank and collected data under light wind conditions. One echo, of which With the value set to 1000, the mean of its spectral width statistical distribution was measured to be 0.15 m / s. The minimum value was calculated by subtracting three times the standard deviation from the mean. m / s; to maintain a decision margin between the rigid body target and water disturbance, the preset spectral width threshold is set at and Between them, a definite calculation rule is to take the arithmetic mean of the two, that is... m / s, this value is fixed and written into the Doppler feature analysis unit.

[0042] The second step is to calibrate the entropy offset used by the threshold calibration unit to calculate the dynamic classification threshold. This parameter It is a fixed offset, which means that it defines... This judgment boundary is relative to the current clutter mean. The statistical separation degree was determined. The initial object used for calibration was a high-density dry vegetation area, which was used as the reference for high-entropy clutter, and the aforementioned metal corner reflector was used as the reference for low-entropy targets. The calibration procedure is as follows: System data acquisition... Polarized echo samples from arid vegetation areas, among which Set the value to 5000 and calculate its polarization entropy. The statistical distribution of the values ​​is used to obtain their mean. Standard deviation Simultaneous collection One corner reflector sample, among which The polarization entropy was measured by setting it to 1000. The maximum value, which is the mean plus three standard deviations. The decision-making process involves selecting one. Value, making It can effectively filter out the vast majority of clutter while ensuring Not filtered out by errors; a defined rule is to... Set as clutter standard deviation A preset multiple , here Choose 2, this setting is... This is designed to filter out approximately 97.7% of clutter; based on this, the entropy offset is determined to be... This calculation yields This value is higher than This satisfies the constraint that the target will not be filtered out; this value The threshold calibration unit is embedded in the data. Through the above procedure, the two key preset parameters required for system operation are the preset spectral width threshold and the entropy offset. All of these were determined through an objective calibration process.

[0043] Example 5: This example illustrates an offline calibration procedure for establishing clutter statistical characteristics for a polarization feature preprocessor. This procedure is a prerequisite for the system to execute the active excitation and dynamic response discrimination mechanism. The calibration is performed in a pre-surveyed benchmark test area that is confirmed to contain no man-made targets and consists only of high-density forest clutter. The initial object used for calibration is the mass resolution unit in the benchmark test area, and the enabling environment used is the system of this invention, which is placed in a dedicated offline data acquisition mode. In this mode, the main controller is used to cyclically and forcibly instruct the system's transmitter to repeatedly transmit active probe pulse sequences with multiple different preset polarization states to different verification units, i.e., benchmark clutter units, in the benchmark test area. For example, a four-pulse sequence containing horizontal H, vertical V, left-handed LHC, and right-handed RHC is used. The polarization feature preprocessor is used to receive and store echo response sequences known to be pure clutter from the pulse sequence to construct a benchmark clutter response database.

[0044] The next step in the calibration procedure is to establish a smooth response mode, a polarization characteristic preprocessor, or an offline processing unit to perform statistical analysis on all echo response sequences in the reference clutter response database. This analysis quantifies the response changes between adjacent pulses within the sequence, expressed as the ratio of the H-pulse echo amplitude to the V-pulse echo amplitude. For example, the processor calculates all the data in the database. group samples Statistical mean of values and standard deviation The preset range defined by the smooth response model is objectively determined as a high-confidence interval of the statistical distribution, and this range is set as follows: The system repeats this statistical process for each pair of adjacent polarization state switching in the active probe pulse sequence, such as V-LHC or LHC-RHC, thereby establishing a quantified smooth response range for each switching. These determined ranges together constitute the benchmark model of clutter statistical characteristics, which is stored in the polarization feature preprocessor as an objective basis for identifying jumps that do not conform to clutter statistical characteristics in real-time detection.

[0045] Example 6: This example illustrates the standardized engineering calibration procedure required before deployment of the system of the present invention to determine key control logic thresholds. Its purpose is to determine a preset system load threshold for the logic integrity monitoring unit and to determine the resource allocation benchmark and prioritization method for the main controller when performing asymmetric verification. The first step is to calibrate the preset system load threshold, which is set based on the maximum throughput capacity of the system's backend fine-tuning processor. The initial object used for calibration is the system hardware platform, whose key functional specifications are determined as follows: its signal processor, when performing a full-process fine-tuning detection of a resolution unit, including long-term pulse accumulation and high-precision CFAR decision, has a maximum processing rate of [missing information]. One resolution unit per second; the calibration procedure is as follows: determine the wide-area search period of the system. This refers to the time required to complete one global scan; based on this, the maximum processing capacity of the system within one scan cycle can be calculated. Then determine the total number of resolution units of the system in wide-area scanning mode. The preset system load threshold Calculated as Taking a specific deployment as an example, if 50,000 resolution units per second, If it is 4 seconds, then For a resolution of 200,000 units, if the wide-area search region contains With a total of 1,000,000 resolution units, this threshold... identified as That is, 20%, and this value is fixed and written into the logic integrity monitoring unit.

[0046] The second step is to calibrate the resource allocation benchmark for asymmetric verification. The setting of this benchmark needs to balance two constraints: first, to provide updates for the threshold calibration unit. Minimum required clutter sample size Secondly, a robust detection lower limit is set to prevent missed detections. The calibration procedure is as follows: Through offline simulation, determine the optimal calibration method. Once the calculated variance stabilizes at an acceptable value, such as below 0.005, the threshold calibration unit will be in its update cycle. Minimum sample size required For 200 samples; calculations are performed on... Within a period of 60 seconds, for example, the number of wide-area scans performed by the system. ,Right now This determines the minimum number of calibration samples required for each scan. ,Right now 1; at the same time, the system sets a fixed robust detection ratio. Taking 1% as an example; when the main controller is running, it obtains the total number of units judged as high-entropy regions within the current scan cycle. And calculate the number of units required for robust detection. The total number of units corresponding to the resources ultimately used by the main controller for asymmetric verification. It was identified as The larger of the two values ​​is taken, ensuring that the verification mechanism always meets the dual requirements of statistical calibration and robust detection. The third step clarifies the specific implementation method of the main controller performing priority allocation based on the feature gradient map. This implementation method is a defined algorithmic sorting process: when the main controller determines that the total number of units to be verified in the current scan cycle is... ,by For example, its execution logic is as follows: The input is the polarization feature preprocessor that generates all The feature gradient maps of each high-entropy region unit and its corresponding gradient value; the processing logic is that the main controller... The units are sorted in descending order based on their gradient values; the output is that the main controller selects the top elements from the sorted list. This unit The high-entropy region unit with the highest gradient value is selected as the object of the asymmetric verification kernel and allocated a small amount of detection resources.

[0047] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0048] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A forest scatterer type identification and localization system based on radar polarization decomposition, characterized in that, include: A polarization feature preprocessor is used to calculate the polarization entropy of each resolution unit in the wide-area search region in real time, and divide the resolution unit into high-entropy and low-entropy regions according to a dynamic classification threshold to generate region of interest masks. The main controller is used to receive the region of interest mask and perform asymmetric resource scheduling, which includes scheduling detection resources to focus on performing fine detection in low-entropy regions and skipping fine detection in high-entropy regions; the main controller is also used to perform asymmetric verification, which includes forcibly allocating a small amount of detection resources to verification units in high-entropy regions to obtain their original polarization feature data. The threshold calibration unit is specifically used to receive the original polarization feature data of the verification unit identified as the high-entropy region when performing asymmetric verification. Based on the original polarization feature data, a background clutter feature baseline characterizing the clutter background of the current high-entropy region is calculated in real time, and the dynamic classification threshold is dynamically calculated and updated based on the background clutter feature baseline. The polarization feature preprocessor is also used to employ a dynamic classification threshold, which is updated in real time by the threshold calibration unit, as the current basis for dividing the high-entropy region into the low-entropy region.

2. The forest scatterer type identification and localization system based on radar polarization decomposition according to claim 1, characterized in that, The system also includes: a Doppler feature analysis unit, used to calculate the Doppler spectral width of the echo signal of the resolution unit in parallel; the main controller is also used to identify only the resolution units that simultaneously satisfy the polarization entropy below the dynamic classification threshold and the Doppler spectral width below the preset spectral width threshold as low-entropy regions when it performs asymmetric resource scheduling, and to identify high-entropy regions and low-entropy but high-spectral-width regions as non-interest regions.

3. The forest scatterer type identification and localization system based on radar polarization decomposition according to claim 1, characterized in that, The polarization feature preprocessor is also used to: calculate a feature gradient map representing the spatial rate of change of polarization entropy in real time based on the raw data of polarization entropy. The logic of the main controller for performing asymmetric verification is further limited to allocating a small amount of detection resources to regions with higher feature gradient values ​​in high-entropy regions according to the feature gradient map.

4. A forest scatterer type identification and localization system based on radar polarization decomposition according to claim 1, characterized in that, The logic of the main controller for performing asymmetric verification is further defined as the transmitter of the instruction system, which transmits a series of active probe pulses with multiple different preset polarization states to the verification unit. The polarization feature preprocessor is also used to receive and analyze the echo response sequence excited by the active probe pulse sequence. By identifying whether the echo response sequence exhibits a jump that does not conform to clutter statistics as the transmit polarization state switches, it can identify camouflaged targets. The main controller is also used to schedule detection resources to focus on camouflaged targets.

5. A forest scatterer type identification and localization system based on radar polarization decomposition according to claim 1, characterized in that, The system also includes: a logic integrity monitoring unit, used to monitor in real time the global proportion of regions identified as low-entropy areas in the region of interest mask; the main controller is also used to automatically bypass asymmetric resource scheduling and fall back to the global detection mode that performs standard detection on the entire range of the wide-area search region when the logic integrity monitoring unit detects that the global proportion exceeds the preset system load threshold.

6. A forest scatterer type identification and localization system based on radar polarization decomposition according to claim 1, characterized in that, The threshold calibration unit's logic for dynamically calculating and updating the dynamic classification threshold includes: performing statistical analysis on the received raw polarization feature data to calculate the background clutter feature baseline. And based on the background clutter characteristic baseline The dynamic classification threshold is calculated using the following rules. : ,in This is a preset entropy offset used to ensure the separation of the target from clutter.

7. A forest scatterer type identification and localization system based on radar polarization decomposition according to claim 6, characterized in that, The threshold calibration unit is used to calculate the characteristic baseline of background clutter. The statistical analysis includes: constructing a statistical histogram from the received raw polarization feature data, and calculating the preset quantiles of the statistical histogram to use as the background clutter feature baseline. .

8. A forest scatterer type identification and localization system based on radar polarization decomposition according to claim 3, characterized in that, The logic of the polarization feature preprocessor for calculating the feature gradient map includes: applying a pre-defined convolution kernel to the raw data of polarization entropy to calculate the local rate of change of polarization entropy in space.

9. A forest scatterer type identification and localization system based on radar polarization decomposition according to claim 2, characterized in that, The logic of the Doppler feature analysis unit for calculating the Doppler spectral width includes: performing Doppler processing on the echo signal to obtain the Doppler spectrum, and measuring the spectral peak width of the Doppler spectrum to use it as the Doppler spectral width.

10. A forest scatterer type identification and localization system based on radar polarization decomposition according to claim 4, characterized in that, The logic of the polarization feature preprocessor for identifying echo response sequences includes: establishing clutter statistical characteristics, which characterize the change of clutter echo amplitude with the switching of transmit polarization state in a preset smooth response mode; identification includes: when the echo amplitude change of adjacent pulses in the echo response sequence exceeds the preset range defined by the smooth response mode, the verification unit is identified as a camouflaged target.

Images