Target library based radar dynamic multiple hypothesis track-before-detect processing method and system

Abstract

The application provides a target library-based radar dynamic multi-hypothesis detection pre-tracking processing method and system, which comprises the following steps: step 1, a single frame processing procedure comprises single frame data processing, tracking, parameter solving and data correction; step 2, a target state hypothesis obtained through parameter solving of a previous frame is used for tracking processing of a next frame, and the process is repeated; step 3, the multi-frame processing results are used for multi-frame coherent accumulation and detection to obtain a current scanning result; step 4, a target library is established for the multi-scanning results, and the target library is classified and managed; step 5, backward filtering processing is carried out between the multi-scanning, and information of detected but not tracked targets and detected and tracked targets in the current scanning is fed back to the next scanning; and step 6, forward filtering processing is carried out between the multi-scanning, and information of newly detected but not tracked targets and newly detected and tracked targets in the current scanning is fed back to the previous scanning.

Description

Technical Field

[0001] This invention relates to the field of radar signal processing technology, and more specifically, to a radar dynamic multi-hypothesis detection pre-tracking processing method and system based on a target database. Background Technology

[0002] As an active detection method, radar can detect moving targets in all weather conditions, at all times, and at all altitudes, featuring long operating range, wide coverage, ability to track multiple targets, and high positioning accuracy. The radar processing method for target detection and tracking directly determines its detection performance. Weak targets emit weak echo signals, making them difficult to detect effectively. Traditional detection-then-tracking processing is based on point data that has undergone threshold processing by the detector. While this method can reduce the amount of data and allows for separate design of the detector and tracker, it can lead to false alarms and missed detections when the signal-to-noise ratio (SNR) is low. Therefore, it is necessary to process the data over a longer timescale through tracking.

[0003] Therefore, the "track before detection" approach was proposed. Radar dynamic multi-hypothesis detection-before-tracking is a typical strongly nonlinear problem, which can be accumulated to achieve better results. This method treats each possible target trajectory as a hypothesis, proposing a multi-hypothesis approach. This hypothesis-oriented approach considers not only target correlation but also the possibility of false alarms and new target generation. The current-moment hypothesis is generated by interconnecting data with the previous-moment hypothesis, realizing dynamic multi-hypothesis detection-before-tracking. Through interconnection, iterative processing is performed to determine the probability of the target's actual existence. Dynamic comparison also avoids false alarms from single scans. Similarly, this scheme is also applicable to the detection of weak targets with low signal-to-noise ratios, avoiding missed detections of weak targets in single processing. During the dynamic updating of the target database, target information is transferred between multiple scans, mainly transmitting the target speed, angle, and distance information detected in the current processing. This corrects prior information from other processing, avoiding false alarms and missed detections caused by inaccurate prior information leading to signal processing errors.

[0004] The doctoral dissertation, "Research on Multi-Frame Pre-Detection Tracking Methods for Airborne Radar" (Li Wujun, University of Electronic Science and Technology of China), proposed adaptive multi-frame pre-detection tracking algorithms for airborne radar, multi-frequency multi-frame pre-detection tracking algorithms, and adaptive estimation multi-frame pre-detection tracking algorithms based on historical learning of maneuver characteristics. Compared with airborne radar, space-based radar faces more complex clutter environments and relative motion relationships, and there are differences in detection modes and capabilities, as well as processing procedures.

[0005] Patent document CN201610506878.1, titled "A Dynamic Programming Pre-Detection Tracking Method Based on Multiple Hypothesis Testing," discloses a method for target detection and trajectory estimation using data within a sliding window, which can improve the detection probability of targets appearing later than the start time of the sliding window. However, this method primarily involves sliding window processing and does not consider feeding the current processing result back to the previous processing step.

[0006] The paper "Research on Pre-Detection Tracking Algorithm for Multi-Frame Detection in Networked Radar" (Wang Jinghe, Yi Wei, Kong Lingjiang. Journal of Radar, 2019, 8(4):490-500.) proposes a pre-detection tracking algorithm for multi-frame detection in networked radar systems based on point sequence fusion, and implements it using a point sequence fusion algorithm based on particle filtering. This method mainly fuses the point sequences obtained by each radar in the networked radar system, which can simultaneously improve the target detection probability and tracking performance.

[0007] The paper "A Tracking Algorithm for Weak Targets Before Detection Based on Measurement Data Accumulation" (Yu Ruofeng, Yang Wei, Fu Yaowen. Shanghai Aerospace, 2020, 37(5):56-66.) accumulates measurement data of the resolution cells occupied by the target's potential trajectory across multiple frames, and estimates the average strength of the target based on the expected value of the accumulated data. This method can accurately estimate the target strength parameters while achieving joint detection and tracking of weak targets. This approach focuses more on improving the target detection capability in the current processing and only provides useful target strength information for subsequent target classification or recognition.

[0008] The paper "A Pre-Detection Tracking Algorithm for Weak Target Detection by Radar" (Lü Xiaojun, Zhang Jiaqi. Electronic Technology and Software Engineering, 2022, 10: 134-138.) proposes a pre-detection tracking algorithm that first uses Hough transform to obtain prior information about the target for track initiation, and then uses dynamic programming for track accumulation. This algorithm addresses the problem of undetectable targets when radar detects weak targets. The method focuses on how to improve target detection capabilities by obtaining prior information. Summary of the Invention

[0009] To address the shortcomings of existing technologies, the purpose of this invention is to provide a radar dynamic multi-hypothesis detection pre-tracking processing method and system based on a target library.

[0010] The radar dynamic multi-hypothesis detection pre-tracking processing method based on a target library provided by the present invention includes:

[0011] Step S1: Perform single-frame processing on each frame of radar echo data. The single-frame processing includes tracking processing, parameter calculation and data correction, and outputs the target state assumption for this frame. Step S2: Use the target state assumptions obtained from the parameter calculation of the previous frame as the initial conditions for the tracking processing of the current frame, and establish the inter-frame assumption transfer relationship through multi-assumption interconnection; Step S3: Perform multi-frame coherent accumulation and target detection on the multi-frame processing results to obtain the detection results of the current scan. The multi-frame coherent accumulation is energy accumulation along the potential target track, and no single-frame detection threshold is set. Step S4: Establish a target library based on the detection results of multiple scans, and classify and manage the targets. The target library includes four categories of targets: detected but not tracked, detected and tracked, newly detected but not tracked, and newly detected and tracked. Step S5: Perform back-filtering between multiple scans to feed back the information of untracked targets and tracked targets detected in the current scan to the clutter suppression and detection stage of the next scan. Step S6: Perform forward filtering between multiple scans, and feed back the information of newly detected untracked targets and newly detected tracked targets in the current scan to the signal processing stage of the previous scan for secondary detection and tracking processing.

[0012] Preferably, in step S1: The tracking process includes: performing data association and filtering on the raw echo data based on prior knowledge of the target's kinematics and the similarity of the energy received by the sensor, and generating multiple candidate target motion trajectory hypotheses; The parameter calculation includes: calculating the target's state parameters from the tracking results, the state parameters including velocity, azimuth angle, and distance; The data correction includes: performing geometric correction and error compensation on the calculated target state, eliminating measurement deviations caused by platform motion or sensor attitude changes, and outputting the corrected target state hypothesis. The inter-frame hypothesis transmission in step S2 includes: constructing a multi-hypothesis tree structure, interconnecting multiple target state hypotheses from the previous frame with the observation data of the current frame to form multiple candidate hypotheses at the current moment; scoring each candidate hypothesis with confidence, the scoring being based on the degree of matching between the hypothesis and the observation data and the motion continuity of the hypothesis; retaining hypotheses with scores higher than a preset threshold for subsequent frame processing, and eliminating hypotheses with low confidence.

[0013] Preferably, the multi-frame coherent accumulation in step S3 adopts the following energy accumulation formula. :

[0014] Where k is the current frame number being processed, and N is the number of pulses contained in each frame. The echo complex signal of the nth pulse in the i-th frame. Let T be the Doppler frequency of the target in the i-th frame, T be the pulse repetition period, and j be the imaginary unit. When energy accumulates If the detection threshold is exceeded, the target is determined to exist, and the corresponding trajectory information of the target is output.

[0015] Preferably, the establishment and classification management of the target library in step S4 includes: Based on the detection results of each scan, the targets are divided into four categories: detected but not tracked, detected and tracked, newly detected but not tracked, and newly detected and tracked. The target database is dynamically updated: if a target is not detected in multiple consecutive scans, it is deleted from the target database; if a target is detected for the first time in the current scan, it is added to the target database as a new target. A confidence score is accumulated for each target. The confidence score is calculated based on the detection consistency and track continuity of the target in multi-frame scans. When the confidence score is lower than a threshold, the target is identified as a false alarm and removed from the target database.

[0016] Preferably, the backward filtering process in step S5 includes: Extract information on untracked targets and tracked targets detected in this scan from the target database. The information includes speed, azimuth angle, and distance. The above information is fed back to the clutter suppression stage of the next scan to construct the target steering vector, which is constructed as follows:

[0017] in, For spatial guidance vectors, For velocity compensation vector, For distance compensation vector, For Hadamard product; At the same time, the above information is fed back to the inter-frame coherent fusion stage of the next scan to correct inter-frame movement.

[0018] Preferably, the forward filtering process in step S6 includes: Identify newly detected targets in this scan and determine whether the target was not detected in the previous scan due to potential energy loss; The information of newly detected targets is fed back to the signal processing module of the previous scan, and clutter suppression, detection and tracking are performed again. If the target is detected again in the previous scan, the confidence level of the target in the target database is increased and its track history is updated; The secondary detection uses a multi-frame energy accumulation formula. :

[0019] Where K is the total number of frames. This is the echo complex signal of the nth pulse in the kth frame. Let the Doppler frequency of the target be assumed for the k-th frame.

[0020] Preferably, step S4 further includes target feature recording and analysis: Record the historical trajectory, motion characteristics, detection frequency, and energy change trend of each target; Based on the above characteristics, behavioral pattern analysis is performed on the targets to identify maneuvering or constant-speed targets, and the classification labels of various targets in the target database are dynamically adjusted. The analysis results are used as prior information for hypothesis generation and confidence assessment in subsequent scans.

[0021] Preferably, the spatial guidance vector The velocity compensation vector is determined based on the antenna array manifold. and distance compensation vector Calculate using the following formulas respectively:

[0022]

[0023] in, For Doppler frequency, This represents the frequency shift caused by distance.

[0024] Preferably, the multi-frame energy accumulation formula in the secondary detection... Further considering inter-frame energy weighting, the following form is adopted:

[0025] in, The weighting coefficient for the k-th frame is dynamically adjusted based on the target motion characteristics or historical signal-to-noise ratio values.

[0026] The radar dynamic multi-hypothesis detection pre-tracking processing system based on a target library provided by the present invention includes: Module M1: Performs single-frame processing on each frame of radar echo data. The single-frame processing includes tracking processing, parameter calculation, and data correction, and outputs the target state assumption for this frame. Module M2: It uses the target state assumptions obtained from the parameter calculation of the previous frame as the initial conditions for the tracking processing of the current frame, and establishes the inter-frame hypothesis transfer relationship through multi-hypothesis interconnection. Module M3: Performs multi-frame coherent accumulation and target detection on the multi-frame processing results to obtain the detection results of the current scan. The multi-frame coherent accumulation is energy accumulation along the potential target track, and no single-frame detection threshold is set. Module M4: Establishes a target library based on the detection results of multiple scans and classifies and manages the targets. The target library includes four categories of targets: detected but not tracked, detected and tracked, newly detected but not tracked, and newly detected and tracked. Module M5: Performs back-filtering between multiple scans, feeding back information on untracked targets and tracked targets detected in the current scan to the clutter suppression and detection stage of the next scan; Module M6: Performs forward filtering between multiple scans, feeding back information on newly detected untracked targets and newly detected tracked targets in the current scan to the signal processing stage of the previous scan for secondary detection and tracking.

[0027] Compared with the prior art, the present invention has the following beneficial effects: This invention utilizes target motion correlation to accumulate energy between frames, which can effectively improve the decision signal-to-noise ratio and achieve effective detection of weak targets. By sharing information between multiple scans, it can assist in single scan signal processing, dynamically adjust prior information, and improve target detection performance. At the same time, it establishes a target library and uses the processing results of each scan to implement dynamic updates, which is beneficial for information management of multiple target types. Attached Figure Description

[0028] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a schematic diagram of the inter-frame detection and tracking processing flow; Figure 2 This is a schematic diagram illustrating the interaction of target information between the target library and multiple scans. Detailed Implementation

[0029] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0030] Example 1 To address the optimization design requirements of radar signal processing and target detection frameworks, the technical problem this invention aims to solve is to provide a radar dynamic multi-hypothesis detection pre-tracking processing method based on a target database, comprising the following steps: (1) The single-frame processing procedure consists of single-frame data tracking, parameter calculation, and data correction; (2) The target state obtained by solving the parameters of the previous frame is assumed to be used for tracking processing in the next frame, and so on; (3) Use the multi-frame processing results for multi-frame coherent accumulation and detection to obtain the current scan result; (4) Establish a target library based on the results of multiple scans and manage it by category; (5) Perform backward filtering between multiple scans to feed back the information of undetected and tracked targets detected in the current scan to the next scan; (6) Perform forward filtering between multiple scans to feed back information on newly detected untracked targets and newly detected tracked targets in the current scan to the previous scan.

[0031] In step (1), the original sensor data is directly processed, and the information obtained from the processing is accumulated over time. The decision on the target is located in the last step of the entire processing chain.

[0032] In step (2), data association and filtering are performed using prior knowledge of target kinematics and similarity of energy received by the sensor, and inter-frame energy is accumulated using target motion correlation to improve the decision signal-to-noise ratio.

[0033] In step (3), the target is first tracked, and energy is further accumulated along the trajectory of the potential target until the detection performance can be well guaranteed. No detection threshold is set for single frame data, which preserves the original data observed by the sensor to the greatest extent and avoids the problem of missing weak targets.

[0034] In step (4), the target library contains four types of targets: detected but not tracked, detected and tracked, newly detected but not tracked, and newly detected and tracked.

[0035] In step (5), information on untracked and tracked targets detected in the current scan, including velocity, azimuth angle, and distance, is fed back to the next scan. This is used to optimize prior information used in clutter suppression, detection, and other processes, and then signal-level and information-level processing is performed again. In clutter suppression processing, a more accurate target guidance vector is constructed to avoid energy loss caused by beam center assumption. In inter-frame coherent fusion processing, the fed-back target information is used to correct inter-frame movement and improve multi-frame coherent accumulation gain.

[0036] In step (6), the target database is updated based on the scan results. If the target still does not form a trace, it may be a false alarm target. If the target forms a trace after the scan, it may indicate that the previous signal processing may have missed an alarm. For newly detected targets obtained in this scan, it is considered whether the target was not detected due to severe energy loss in the previous scan. This type of target information is fed back to the previous processing stage for signal-level and information-level processing again to see if the target can be detected and tracked, and the target database is updated. If the target can be detected, the confidence in the actual existence of the target will also increase. Finally, the confidence in the target is used to make a judgment on each target in the target database.

[0037] like Figure 1 As shown, the single-frame processing process consists of single-frame data, tracking, parameter calculation, and data correction; the target state assumption obtained from the parameter calculation of the previous frame is used for the tracking processing of the next frame, and so on; the multi-frame processing results are used for multi-frame coherent accumulation and detection to obtain the current scan result.

[0038] like Figure 2 As shown, the target database contains four types of targets: detected but not tracked, detected and tracked, newly detected but not tracked, and newly detected and tracked. Back-up filtering is performed between scans to feed back information on both detected and tracked targets from the current scan to the next scan; forward filtering is performed between scans to feed back information on newly detected and tracked targets from the current scan to the previous scan.

[0039] Example 2 Example 2 is a preferred example of Example 1.

[0040] This invention provides a radar dynamic multi-hypothesis detection pre-tracking processing method based on a target database, including: Step S1: Perform single-frame processing on each frame of radar echo data. The single-frame processing includes tracking processing, parameter calculation and data correction, and outputs the target state assumption for this frame. Step S2: Use the target state assumptions obtained from the parameter calculation of the previous frame as the initial conditions for the tracking processing of the current frame, and establish the inter-frame assumption transfer relationship through multi-assumption interconnection; Step S3: Perform multi-frame coherent accumulation and target detection on the multi-frame processing results to obtain the detection results of the current scan. The multi-frame coherent accumulation is energy accumulation along the potential target track, and no single-frame detection threshold is set. Step S4: Establish a target library based on the detection results of multiple scans, and classify and manage the targets. The target library includes four categories of targets: detected but not tracked, detected and tracked, newly detected but not tracked, and newly detected and tracked. Step S5: Perform back-filtering between multiple scans to feed back the information of untracked targets and tracked targets detected in the current scan to the clutter suppression and detection stage of the next scan. Step S6: Perform forward filtering between multiple scans, and feed back the information of newly detected untracked targets and newly detected tracked targets in the current scan to the signal processing stage of the previous scan for secondary detection and tracking processing.

[0041] Preferably, in step S1: The tracking process includes: performing data association and filtering on the raw echo data based on prior knowledge of the target's kinematics and the similarity of the energy received by the sensor, and generating multiple candidate target motion trajectory hypotheses; The parameter calculation includes: calculating the target's state parameters from the tracking results, the state parameters including velocity, azimuth angle, and distance; The data correction includes: performing geometric correction and error compensation on the calculated target state, eliminating measurement deviations caused by platform motion or sensor attitude changes, and outputting the corrected target state hypothesis. The inter-frame hypothesis transmission in step S2 includes: constructing a multi-hypothesis tree structure, interconnecting multiple target state hypotheses from the previous frame with the observation data of the current frame to form multiple candidate hypotheses at the current moment; scoring each candidate hypothesis with confidence, the scoring being based on the degree of matching between the hypothesis and the observation data and the motion continuity of the hypothesis; retaining hypotheses with scores higher than a preset threshold for subsequent frame processing, and eliminating hypotheses with low confidence.

[0042] Preferably, the multi-frame coherent accumulation in step S3 adopts the following energy accumulation formula. :

[0043] Where k is the current frame number being processed, and N is the number of pulses contained in each frame. The echo complex signal of the nth pulse in the i-th frame. Let T be the Doppler frequency of the target in the i-th frame, T be the pulse repetition period, and j be the imaginary unit. When energy accumulates If the detection threshold is exceeded, the target is determined to exist, and the corresponding trajectory information of the target is output.

[0044] Preferably, the establishment and classification management of the target library in step S4 includes: Based on the detection results of each scan, the targets are divided into four categories: detected but not tracked, detected and tracked, newly detected but not tracked, and newly detected and tracked. The target database is dynamically updated: if a target is not detected in multiple consecutive scans, it is deleted from the target database; if a target is detected for the first time in the current scan, it is added to the target database as a new target. A confidence score is accumulated for each target. The confidence score is calculated based on the detection consistency and track continuity of the target in multi-frame scans. When the confidence score is lower than a threshold, the target is identified as a false alarm and removed from the target database.

[0045] Preferably, the backward filtering process in step S5 includes: Extract information on untracked targets and tracked targets detected in this scan from the target database. The information includes speed, azimuth angle, and distance. The above information is fed back to the clutter suppression stage of the next scan to construct the target steering vector, which is constructed as follows:

[0046] in, For spatial guidance vectors, For velocity compensation vector, For distance compensation vector, For Hadamard product; At the same time, the above information is fed back to the inter-frame coherent fusion stage of the next scan to correct inter-frame movement.

[0047] Preferably, the forward filtering process in step S6 includes: Identify newly detected targets in this scan and determine whether the target was not detected in the previous scan due to potential energy loss; The information of newly detected targets is fed back to the signal processing module of the previous scan, and clutter suppression, detection and tracking are performed again. If the target is detected again in the previous scan, the confidence level of the target in the target database is increased and its track history is updated; The secondary detection uses a multi-frame energy accumulation formula. :

[0048] Where K is the total number of frames. This is the echo complex signal of the nth pulse in the kth frame. Let the Doppler frequency of the target be assumed for the k-th frame.

[0049] Preferably, step S4 further includes target feature recording and analysis: Record the historical trajectory, motion characteristics, detection frequency, and energy change trend of each target; Based on the above characteristics, behavioral pattern analysis is performed on the targets to identify maneuvering or constant-speed targets, and the classification labels of various targets in the target database are dynamically adjusted. The analysis results are used as prior information for hypothesis generation and confidence assessment in subsequent scans.

[0050] Preferably, the spatial guidance vector The velocity compensation vector is determined based on the antenna array manifold. and distance compensation vector Calculate using the following formulas respectively:

[0051]

[0052] in, For Doppler frequency, This represents the frequency shift caused by distance.

[0053] Preferably, the multi-frame energy accumulation formula in the secondary detection... Further considering inter-frame energy weighting, the following form is adopted:

[0054] in, The weighting coefficient for the k-th frame is dynamically adjusted based on the target motion characteristics or historical signal-to-noise ratio values.

[0055] This invention also provides a radar dynamic multi-hypothesis detection pre-tracking processing system based on a target database, comprising: Module M1: Performs single-frame processing on each frame of radar echo data. The single-frame processing includes tracking processing, parameter calculation, and data correction, and outputs the target state assumption for this frame. Module M2: It uses the target state assumptions obtained from the parameter calculation of the previous frame as the initial conditions for the tracking processing of the current frame, and establishes the inter-frame hypothesis transfer relationship through multi-hypothesis interconnection. Module M3: Performs multi-frame coherent accumulation and target detection on the multi-frame processing results to obtain the detection results of the current scan. The multi-frame coherent accumulation is energy accumulation along the potential target track, and no single-frame detection threshold is set. Module M4: Establishes a target library based on the detection results of multiple scans and classifies and manages the targets. The target library includes four categories of targets: detected but not tracked, detected and tracked, newly detected but not tracked, and newly detected and tracked. Module M5: Performs back-filtering between multiple scans, feeding back information on untracked targets and tracked targets detected in the current scan to the clutter suppression and detection stage of the next scan; Module M6: Performs forward filtering between multiple scans, feeding back information on newly detected untracked targets and newly detected tracked targets in the current scan to the signal processing stage of the previous scan for secondary detection and tracking.

[0056] Those skilled in the art will understand that, in addition to implementing the system, apparatus, and their modules provided by this invention in purely computer-readable program code, the same program can be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system, apparatus, and their modules provided by this invention can be considered a hardware component, and the modules included therein for implementing various programs can also be considered structures within the hardware component; alternatively, modules for implementing various functions can be considered both software programs implementing the method and structures within the hardware component.

[0057] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A radar dynamic multi-hypothesis detection pre-tracking processing method based on a target database, characterized in that, include: Step S1: Perform single-frame processing on each frame of radar echo data. The single-frame processing includes tracking processing, parameter calculation and data correction, and outputs the target state assumption for this frame. Step S2: Use the target state assumptions obtained from the parameter calculation of the previous frame as the initial conditions for the tracking processing of the current frame, and establish the inter-frame assumption transfer relationship through multi-assumption interconnection; Step S3: Perform multi-frame coherent accumulation and target detection on the multi-frame processing results to obtain the detection results of the current scan. The multi-frame coherent accumulation is energy accumulation along the potential target track, and no single-frame detection threshold is set. Step S4: Establish a target library based on the detection results of multiple scans, and classify and manage the targets. The target library includes four categories of targets: detected but not tracked, detected and tracked, newly detected but not tracked, and newly detected and tracked. Step S5: Perform back-filtering between multiple scans to feed back the information of untracked targets and tracked targets detected in the current scan to the clutter suppression and detection stage of the next scan. Step S6: Perform forward filtering between multiple scans, and feed back the information of newly detected untracked targets and newly detected tracked targets in the current scan to the signal processing stage of the previous scan for secondary detection and tracking processing.

2. The radar dynamic multi-hypothesis detection pre-tracking processing method based on a target database according to claim 1, characterized in that, In step S1: The tracking process includes: performing data association and filtering on the raw echo data based on prior knowledge of the target's kinematics and the similarity of the energy received by the sensor, and generating multiple candidate target motion trajectory hypotheses; The parameter calculation includes: calculating the target's state parameters from the tracking results, the state parameters including velocity, azimuth angle, and distance; The data correction includes: performing geometric correction and error compensation on the calculated target state, eliminating measurement deviations caused by platform motion or sensor attitude changes, and outputting the corrected target state hypothesis. The inter-frame hypothesis transmission in step S2 includes: constructing a multi-hypothesis tree structure, interconnecting multiple target state hypotheses from the previous frame with the observation data of the current frame to form multiple candidate hypotheses at the current moment; scoring each candidate hypothesis with confidence, the scoring being based on the degree of matching between the hypothesis and the observation data and the motion continuity of the hypothesis; retaining hypotheses with scores higher than a preset threshold for subsequent frame processing, and eliminating hypotheses with low confidence.

3. The radar dynamic multi-hypothesis detection pre-tracking processing method based on a target database according to claim 2, characterized in that, The multi-frame coherent accumulation in step S3 adopts the following energy accumulation formula. : Where k is the current frame number being processed, and N is the number of pulses contained in each frame. The echo complex signal of the nth pulse in the i-th frame. Let T be the Doppler frequency of the target in the i-th frame, T be the pulse repetition period, and j be the imaginary unit. When energy accumulates If the detection threshold is exceeded, the target is determined to exist, and the corresponding trajectory information of the target is output.

4. The radar dynamic multi-hypothesis detection pre-tracking processing method based on a target database according to claim 3, characterized in that, The establishment and classification management of the target library in step S4 includes: Based on the detection results of each scan, the targets are divided into four categories: detected but not tracked, detected and tracked, newly detected but not tracked, and newly detected and tracked. The target database is dynamically updated: if a target is not detected in multiple consecutive scans, it is deleted from the target database; if a target is detected for the first time in the current scan, it is added to the target database as a new target. A confidence score is accumulated for each target. The confidence score is calculated based on the detection consistency and track continuity of the target in multi-frame scans. When the confidence score is lower than a threshold, the target is identified as a false alarm and removed from the target database.

5. The radar dynamic multi-hypothesis detection pre-tracking processing method based on a target database according to claim 4, characterized in that, The backward filtering process in step S5 includes: Extract information on untracked targets and tracked targets detected in this scan from the target database. The information includes speed, azimuth angle, and distance. The above information is fed back to the clutter suppression stage of the next scan to construct the target steering vector, which is constructed as follows: in, For spatial guidance vectors, For velocity compensation vector, For distance compensation vector, For Hadamard product; At the same time, the above information is fed back to the inter-frame coherent fusion stage of the next scan to correct inter-frame movement.

6. The radar dynamic multi-hypothesis detection pre-tracking processing method based on a target database according to claim 5, characterized in that, The forward filtering process in step S6 includes: Identify newly detected targets in this scan and determine whether the target was not detected in the previous scan due to potential energy loss; The information of newly detected targets is fed back to the signal processing module of the previous scan, and clutter suppression, detection and tracking are performed again. If the target is detected again in the previous scan, the confidence level of the target in the target database is increased and its track history is updated; The secondary detection uses a multi-frame energy accumulation formula. : Where K is the total number of frames. This is the echo complex signal of the nth pulse in the kth frame. Let the Doppler frequency of the target be assumed for the k-th frame.

7. The radar dynamic multi-hypothesis detection pre-tracking processing method based on a target database according to claim 6, characterized in that, Step S4 also includes target feature recording and analysis: Record the historical trajectory, motion characteristics, detection frequency, and energy change trend of each target; Based on the above characteristics, behavioral pattern analysis is performed on the targets to identify maneuvering or constant-speed targets, and the classification labels of various targets in the target database are dynamically adjusted. The analysis results are used as prior information for hypothesis generation and confidence assessment in subsequent scans.

8. The radar dynamic multi-hypothesis detection pre-tracking processing method based on a target database according to claim 7, characterized in that, The spatial guidance vector The velocity compensation vector is determined based on the antenna array manifold. and distance compensation vector Calculate using the following formulas respectively: in, For Doppler frequency, This represents the frequency shift caused by distance.

9. The radar dynamic multi-hypothesis detection pre-tracking processing method based on a target database according to claim 8, characterized in that, The multi-frame energy accumulation formula in the secondary detection Further considering inter-frame energy weighting, the following form is adopted: in, The weighting coefficient for the k-th frame is dynamically adjusted based on the target motion characteristics or historical signal-to-noise ratio values.

10. A radar dynamic multi-hypothesis detection pre-tracking processing system based on a target database, characterized in that, The radar dynamic multi-hypothesis detection pre-tracking processing method based on a target library, as described in any one of claims 1 to 9, includes: Module M1: Performs single-frame processing on each frame of radar echo data. The single-frame processing includes tracking processing, parameter calculation, and data correction, and outputs the target state assumption for this frame. Module M2: It uses the target state assumptions obtained from the parameter calculation of the previous frame as the initial conditions for the tracking processing of the current frame, and establishes the inter-frame hypothesis transfer relationship through multi-hypothesis interconnection. Module M3: Performs multi-frame coherent accumulation and target detection on the multi-frame processing results to obtain the detection results of the current scan. The multi-frame coherent accumulation is energy accumulation along the potential target track, and no single-frame detection threshold is set. Module M4: Establishes a target library based on the detection results of multiple scans and classifies and manages the targets. The target library includes four categories of targets: detected but not tracked, detected and tracked, newly detected but not tracked, and newly detected and tracked. Module M5: Performs back-filtering between multiple scans, feeding back information on untracked targets and tracked targets detected in the current scan to the clutter suppression and detection stage of the next scan; Module M6: Performs forward filtering between multiple scans, feeding back information on newly detected untracked targets and newly detected tracked targets in the current scan to the signal processing stage of the previous scan for secondary detection and tracking.

Images