CN-121978647-A - Phased array radar signal self-adaptive optimization algorithm based on deep learning

CN121978647ACN 121978647 ACN121978647 ACN 121978647ACN-121978647-A

Abstract

The invention discloses a phased array radar signal self-adaptive optimization algorithm based on deep learning, which relates to the technical field of radar signal processing and comprises the following steps: S1, acquiring an electromagnetic propagation environment disturbance intensity curve in a detection task, and overlapping the disturbance intensity curve point by point into an original radar echo to generate a radar echo set with disturbance marks. According to the invention, by introducing a polar phase spiral traction structure based on disturbance response characteristics, self-adaptive return and stable correction of a radar main lobe in a complex propagation environment are realized, the distance error caused by path jump is remarkably reduced, meanwhile, a beam regulation and control framework with immunity and continuity is constructed by combining a dynamic rhythm constraint and a traction path guiding mechanism, stable control from disturbance perception to direction recovery is realized, and the target recognition precision and the system reliability are improved.

Inventors

MENG ZHENG
QIN XIAONING
LI DANNI
ZHOU LINGMIN
PENG YONG
YANG FEI
CUI HONGJIE

Assignees

广州中科莱斯科技发展有限公司

Dates

Publication Date: 20260505
Application Date: 20251229

Claims (10)

1. The phased array radar signal self-adaptive optimization algorithm based on deep learning is characterized by comprising the following steps of: s1, acquiring an electromagnetic propagation environment disturbance intensity curve in a detection task, and overlapping the disturbance intensity curve point by point into an original radar echo to generate a radar echo set with a disturbance mark; s2, reading disturbance intensity changes from a radar echo set with disturbance marks, segmenting continuous radar echo fragments according to fluctuation characteristics of the disturbance intensity in a time dimension, identifying propagation path jump abrupt sections in the fragments, and constructing jump risk index bands; S3, inputting the jump risk index band into a deep learning processing link, locking a characteristic channel corresponding to the middle layer of the deep learning network through a marked abrupt section in the index band, and generating a remote characteristic staggered layer list according to the locked characteristic channel; s4, rearranging the beam pointing update sequence according to the remote characteristic staggered layer list, and limiting the adjustment rhythm of the near pointing in the update process to adjust the beam pointing to form an anti-jump beam adjustment baseline; S5, arranging a polar phase spiral traction ring row along the anti-bending wave beam adjustment base line, and when the electromagnetic disturbance intensity is rapidly increased, carrying out segmented traction through the polar phase spiral traction ring row, so that the main lobe is directed to be gradually migrated back to a real target reflection position along a spiral traction path.
2. The deep learning based phased array radar signal adaptive optimization algorithm of claim 1, wherein step S1 comprises: Deploying environment-aware receiving equipment with broadband receiving capability to acquire disturbance characteristics in an electromagnetic propagation medium; performing interpolation smoothing and normalization processing on the electromagnetic propagation environment disturbance intensity curve and mapping the electromagnetic propagation environment disturbance intensity curve into a radar working coordinate system; carrying out time alignment fusion on the disturbance intensity curve point by point and the original radar echo data to generate a radar echo sequence with a disturbance mark; and re-ordering echo samples according to the disturbance intensity change and adding disturbance gradient weight values to construct a high-robustness input echo set for identifying propagation path anomalies.
3. The deep learning based phased array radar signal adaptive optimization algorithm of claim 2, wherein step S2 comprises: Based on a radar echo set with a disturbance intensity mark, constructing a disturbance intensity time sequence for a continuous echo sample according to an acquisition time sequence, identifying disturbance inflection points according to a disturbance intensity change rate, and dividing the echo sequence into a plurality of time segments with independent dynamic characteristics; Carrying out propagation path stability evaluation on each time segment, and identifying a mutation section with propagation path jump signs by analyzing the offset distance and the offset slope of the main lobe energy distribution; Combining the disturbance intensity change characteristic and the spatial offset characteristic, performing time stamp calibration on the initial frame and the final frame of the abrupt section, and extracting a disturbance peak value, duration time, spatial offset amplitude and energy morphological variation degree to construct a jump risk unit; And aggregating the risk units according to the time continuity and the spatial overlapping characteristics of the risk units to form a risk cluster, and adding the risk cluster to an original radar echo sequence to generate a jump risk index band with space-time directional characteristics.
4. A deep learning based phased array radar signal adaptive optimization algorithm according to claim 3, wherein propagation path stability assessment comprises constructing a main lobe energy profile within each time segment, determining a transition region by calculating the energy offset distance and offset slope of azimuth and pitch directions, determining the segment as a propagation path transition abrupt segment when the energy profile of successive echo frames deviates beyond a preset threshold for several frames, and for generating a transition risk unit.
5. A deep learning based phased array radar signal adaptive optimization algorithm according to claim 3, wherein step S3 comprises: Selecting a radar echo sample consistent with the time stamp of the jump risk index band as an input sequence, inputting the echo sample with the jump risk mark into a deep learning processing link, and synchronously loading a mutation section in the jump risk index band to guide feature response positioning; Monitoring the characteristic channel response of each layer in the network in the deep learning process, locking an abnormal response channel by analyzing the matching relation between the time response gradient and the disturbance characteristic of the transition section, and determining a characteristic channel group which is related to the abnormal height of the propagation path in the middle layer; extracting space-time distribution information of feature responses according to the locked feature channels, constructing a channel feature cluster spectrum, identifying a remote staggered layer channel through a multi-channel cross-validation mechanism, and generating a remote feature staggered layer list; mapping and correlating the remote characteristic staggered layer list with the jump risk index band, establishing a space-time corresponding relation between the propagation abnormality and the characteristic response, and outputting parameter input for correcting the beam adjustment strategy.
6. The adaptive optimization algorithm for deep learning-based phased array radar signals according to claim 5, wherein the deep learning process performs space-time characteristic peak analysis on the locked characteristic channels, determines a distant staggered layer position through characteristic peak frame numbers and spatial positioning parameters, identifies a near-distance high-energy false reflection structure in a channel characteristic distribution diagram, and uses a channel group which keeps consistency of a spatial offset direction and response amplitude as a distant staggered layer channel set for improving accuracy and stability of beam adjustment.
7. The deep learning based phased array radar signal adaptive optimization algorithm of claim 6, wherein step S4 comprises: Reading a characteristic channel number, a spatial offset position and a response frame number in a remote characteristic staggered layer list, and constructing a staggered layer intensity distribution map by combining the response intensity of each channel, wherein the staggered layer intensity distribution map is used for identifying a direction section and a target distance section which are affected by disturbance and protrude in a radar scanning range; Rearranging beam pointing update sequences according to high-risk areas in the staggered intensity distribution diagram, and setting time buffer windows in low-risk near-distance directions to delay update rhythms so as to enable the far-distance directions to form an update mechanism with priority repair and inertia response in the near-distance directions; Introducing an anti-jump adjustment weight distribution table on the basis of updating sequence adjustment, determining a stability rating of beam pointing according to the response density and the spatial energy stability of a staggered channel, and setting an energy transition gap to prevent signal mutation caused by the pointing jump; after the beam updating sequence and rhythm constraint design is completed, the dynamic evolution track of the main lobe pointing in the three-dimensional space is extracted and fitted into an anti-jump beam adjustment base line to be used as a dynamic adjustment reference and deviation limiting basis for real-time beam adjustment.
8. The adaptive optimization algorithm of the phased array radar signal based on deep learning according to claim 7 is characterized in that when a staggered intensity distribution map is constructed, normalization processing is carried out according to dual parameters of a beam coverage angle and a target distance, staggered risk grades are divided by energy density gradients, when an anti-refraction adjustment weight distribution table is introduced, a multi-time confirmation mechanism is set for beam directions with lowest stability grades, and energy buffer gaps with fixed duration are inserted in a direction switching process, so that smooth transition of a main lobe is ensured in an updating process.
9. The deep learning based phased array radar signal adaptive optimization algorithm of claim 7, wherein step S5 comprises: Constructing a polar spiral traction path according to track parameters of an anti-jump beam adjustment baseline in a three-dimensional space, and forming a polar spiral traction structure consisting of a plurality of adjacent spiral rings, wherein the polar spiral traction structure is used for guiding a main lobe to point to move back to a real target reflection position step by step along a preset path; A direction traction instruction point with phase control capability is distributed on a spiral ring key node of a spiral traction path to form a polar phase spiral traction ring row, and fine adjustment control of the main lobe direction is realized by controlling phase difference and interval between traction points; When the electromagnetic disturbance intensity is rapidly increased, the segmented traction is implemented according to the angle difference value between the current position of the main lobe and the target spiral node, so that the main lobe is propelled step by step along the spiral traction path and forms a smooth continuous returning track; And extracting track data of the traction node in a three-dimensional space after the spiral traction path is completed, and determining the lowest offset point of the main lobe track to construct a main lobe stable target position which is used as an initial reference and a dynamic regulation reference for the adjustment of the next periodic wave beam.
10. The adaptive optimization algorithm for the phased array radar signals based on deep learning according to claim 9 is characterized in that traction actions of the polar phase spiral traction ring columns dynamically adjust traction angles and steps according to the change rate of disturbance intensity, when the increase rate of disturbance intensity is increased, rotation density of a spiral path is synchronously increased to enhance the track constraint capacity of a main lobe, smooth transition of the main lobe direction is achieved through phase continuous compensation between adjacent traction ring columns in the traction process, and angle response stability and energy distribution balance of the main lobe in segmented transition are kept, so that the pointing accuracy and the transition consistency of the main lobe in a strong disturbance environment are further improved.

Description

Phased array radar signal self-adaptive optimization algorithm based on deep learning Technical Field The invention relates to the technical field of radar signal processing, in particular to a phased array radar signal self-adaptive optimization algorithm based on deep learning. Background The phased array radar signal self-adaptive optimization algorithm based on deep learning refers to an intelligent technical path which tightly combines artificial intelligence with phased array radar signal processing. The method has the core ideas that the original echo received by the radar is analyzed in real time by using a deep learning model, noise suppression, clutter rejection and multi-target separation are automatically completed under complex environments such as noise interference, electromagnetic straying and strong echo shielding, and the beam direction, weighting coefficient and array configuration of the phased array radar are dynamically adjusted under the condition of environmental parameter changes such as weather, sea state, atmospheric refraction and electromagnetic interference by combining with the space-time reference provided by Beidou positioning, so that the radar always maintains the optimal detection state in complex application scenes. By the method, the radar can realize higher target detection precision and tracking stability in dynamic environments such as maritime monitoring, civil aviation guiding, special detection and the like, reduces manual parameter adjustment, improves the intelligent and self-adaptive capacity of the system, and reflects novel technical advantages of automatic signal optimization and parameter adjustment by utilizing AI driving. The prior art has the following defects: Under a complex electromagnetic environment, strong space-time disturbance can be generated in an atmosphere medium, so that the electromagnetic wave propagation path jumps in a very short time. When the probe wave of the phased array radar passes through such an unstable medium, an abnormal turn-back occurs in an originally continuous stable propagation path, resulting in a disturbed spatial hierarchy of echo signals. When the deep learning model analyzes echo data affected by disturbance, the reflection characteristics of a long-distance target are easily misjudged as short-distance target characteristics, so that the expression of the characteristics of the internal space of the model is disordered. Therefore, the radar system can generate significant deviation in target identification and distance judgment, and can incorrectly identify a non-threat target as a close-range dangerous body, thereby causing incorrect early warning and control instructions. The above information disclosed in the background section is only for enhancement of understanding of the background of the disclosure and therefore it may include information that does not form the prior art that is already known to a person of ordinary skill in the art. Disclosure of Invention The invention aims to provide a phased array radar signal self-adaptive optimization algorithm based on deep learning so as to solve the problems in the background technology. In order to achieve the purpose, the invention provides the following technical scheme that the phased array radar signal self-adaptive optimization algorithm based on deep learning comprises the following steps: s1, acquiring an electromagnetic propagation environment disturbance intensity curve in a detection task, and overlapping the disturbance intensity curve point by point into an original radar echo to generate a radar echo set with disturbance marks for providing input data for subsequent propagation path anomaly analysis; S2, reading disturbance intensity changes from a radar echo set with disturbance marks, segmenting continuous radar echo fragments according to fluctuation characteristics of the disturbance intensity in a time dimension, identifying propagation path jump mutation sections in the fragments, and constructing jump risk index bands for marking source areas of space propagation distortion; S3, inputting the jump risk index band into a deep learning processing link, locking a characteristic channel corresponding to the middle layer of the deep learning network through a marked abrupt section in the index band, and generating a remote characteristic staggered layer list according to the locked characteristic channel for providing a correction basis of a beam adjustment strategy; S4, rearranging the beam pointing update sequence according to the remote characteristic staggered layer list, and limiting the adjustment rhythm of the near pointing in the update process, so that the beam pointing is adjusted to form an anti-jump beam adjustment baseline for establishing stable dynamic adjustment reference; S5, arranging a polar phase spiral traction ring row along the anti-bending wave-jump beam adjustment base line, and when the electromagnetic disturbance intensity