Intelligent sorting method and system for multi-dimensional features of radar signals

Abstract

The application discloses a radar signal multi-dimensional feature intelligent sorting method and system, relates to the radar signal sorting technical field and comprises the following steps: collecting the amplitude, the repetition period and the time of arrival features from the long-range search to the low-power stage, extracting the associated features before and after the amplitude sudden drop under the unified time scale to form a mode conversion precursor set; according to the precursor set, extending tracking and re-merging the sudden drop interval pulses to construct a low-power maintenance judgment interval; combining the environment energy separation with the overall energy lifting to obtain the relative strength profile of the low-power stage; and according to the same, completing trajectory splicing and repositioning the mode conversion node. The application constructs the mode conversion precursor set and combines the extension tracking and continuous splicing to avoid the low-power stage pulse false rejection, thereby maintaining the time continuity of the radiation source trajectory. By introducing the environment energy separation and repositioning the mode conversion node, the full-time domain identity continuous expression is realized, and the sorting and situation analysis reliability is improved.

Description

Technical Field

[0001] This invention relates to the field of radar signal sorting technology, specifically to a method and system for intelligent sorting of multi-dimensional features of radar signals. Background Technology

[0002] Intelligent sorting of radar signals based on multi-dimensional features refers to the process of jointly characterizing and comprehensively judging received radar radiation source signals under complex electromagnetic environments using multiple parameters. Specifically, it first extracts multi-dimensional features such as frequency, pulse width, repetition period, arrival time, amplitude, phase, modulation pattern, and time-spectral structure from the original pulse sequence to construct a unified feature representation space. Then, within this feature space, it comprehensively analyzes the correlation, clustering trends, and change trajectories between different signals, automatically classifying signals from different radiation sources or operating modes, and accurately aggregating pulse streams from the same radiation source. This process differs from traditional sorting methods based on a single parameter threshold; instead, it uses multi-dimensional feature coupling relationships and dynamic change patterns for joint judgment, thus enabling refined identification and grouping of radar signal sources, types, and operating states even under conditions of high density, overlap, complex modulation, and parameter jitter.

[0003] The existing technology has the following shortcomings:

[0004] In complex electromagnetic environments, when a radar switches from a long-range search phase to a low-power silent scanning phase, its transmitted signal amplitude experiences a significant and sudden drop. If this sudden drop coincides with another radar entering an enhanced scanning phase and increasing its transmission power, the overall signal strength distribution at the receiver will be dynamically reconstructed. During this process, sorting processes can easily misjudge the low-power phase signal as a signal disappearance or equipment failure, thus removing this portion of the pulse stream from the original group set and ignoring its true mode transition node. This misjudgment can lead to a break in the trajectory of the same radiation source on the time axis, thereby affecting subsequent parameter evolution analysis and situation prediction results. In severe cases, it can cause misidentification of the radiation source and deviations in threat level assessment.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for intelligent sorting of multi-dimensional features of radar signals to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a multi-dimensional feature intelligent sorting method for radar signals, comprising the following steps:

[0008] Step 1: Collect amplitude variation curves, repetition period variation curves, and arrival time distribution trajectories of the target radiation source from the long-range search phase to the low-power silent scanning phase. Compare and analyze the amplitude variation curves, repetition period variation curves, and arrival time distribution trajectories under a unified time scale, extract the correlation features of continuous pulses before and after the amplitude drop, and form a set of mode conversion precursors.

[0009] Step 2: Based on the mode transition precursor set, extend the arrival time distribution trajectory within the amplitude drop interval, and re-merge the pulses within the amplitude drop interval according to the rhythm characteristics in the mode transition precursor set to construct the low power hold-off determination interval.

[0010] Step 3: Around the low power hold judgment interval, the synchronous environment energy fluctuation curve is introduced for comparative analysis, and combined with the amplitude change curve within the low power hold judgment interval, the overall energy rise caused by the enhanced scanning stage is separated to obtain the relative intensity profile of the low power stage.

[0011] Step 4: Based on the relative intensity profile of the low-power phase, the radiation source trajectory on the original time axis is continuously spliced, and the rhythmic characteristics in the mode transition precursor set are incorporated into the continuous splicing process to achieve continuous connection of pulses before and after the amplitude drop.

[0012] Step 5: After completing the continuous splicing of the radiation source trajectory, the overall evolution trend after splicing is repositioned by combining the mode transition precursor set and the relative intensity profile of the low-power stage, the mode transition node position is determined, and the radiation source identity is maintained in a continuous expression across the entire time domain.

[0013] Preferably, the correlation features of consecutive pulses before and after the sudden drop in amplitude are extracted to form a set of precursors to mode transition, including the following steps:

[0014] Within the observation interval covering the long-range search phase to the low-power silent scan phase, all pulses are recorded pulse by pulse, and the arrival times are numbered and sorted according to a unified time scale. The amplitude values ​​and the time interval between adjacent pulses are recorded synchronously to form amplitude variation curves, repetition period variation curves, and arrival time distribution trajectories, which are then unfolded in parallel under a unified time scale.

[0015] Scan the amplitude change curve along the time axis, mark the boundary of the amplitude drop interval, construct a continuous pulse sequence around the boundary of the amplitude drop interval, and combine the repeating period change curve and the arrival time distribution trajectory to select continuous pulse association segments that meet the requirements of continuous amplitude transition, rhythm structure continuation and numbering sequence connection.

[0016] Around the continuous pulse correlation segment, the amplitude change curve, the repetition period change curve, and the arrival time distribution trajectory are compared point by point to form correlation features including amplitude change trend segment, repetition period rhythm segment, and arrival time sequence segment, which are arranged in chronological order as feature combination sequence;

[0017] The feature combination sequences are aggregated as a whole, integrating the continuous amplitude transition segments, the repeating periodic rhythm continuation segments, and the arrival time sequence connection segments to form a set of pattern transition precursors covering the time range before and after the sudden drop in amplitude.

[0018] Preferably, the amplitude change trend segments, repetitive periodic rhythm segments, and arrival time sequence segments in the mode transition precursor set are arranged point by point on a unified time scale, and the continuous pulse associated segments span the time range on both sides of the amplitude drop interval, so as to ensure that the mode transition precursor set fully covers the transition interval from the long-distance search stage to the low-power silent scan stage.

[0019] Preferably, constructing the low-power hold-up determination interval includes the following steps:

[0020] Read the start and end times, arrival time number sequence, and repeat cycle value sequence from the premonitory set of mode conversion, determine the starting point number of extended tracking, sort the arrival time numbers in the amplitude drop interval one by one along a unified time scale, and match the time interval values ​​between adjacent arrival time numbers with the repeat cycle value sequence item by item, and filter the rhythm continuation pulse to form an extended tracking sequence.

[0021] Arrival time numbers are continuously arranged around the extended tracking sequence, and continuous time periods are divided according to the time interval values. The time interval values ​​within the continuous time periods are matched with the repeating period value sequence to construct a pulse group with a unified rhythm identifier, thereby realizing the re-merging of pulses within the amplitude drop interval.

[0022] The constructed pulse group is expanded under a unified time scale. The boundary number of the continuous time segment is determined based on the earliest arrival time number and the latest arrival time number. The segments whose time interval values ​​meet the condition of continuous arrangement are merged to form a continuous time segment.

[0023] Within a continuous time interval, the arrival time number, time interval value, and amplitude value are recorded, and a correspondence is maintained with the repetitive period value sequence in the mode transition precursor set to construct a low-power hold-up determination interval covering the amplitude drop interval.

[0024] Preferably, the starting point number of the extended tracking is the last arrival time number within the coverage area of ​​the mode transition precursor set. The construction of the pulse group is based on the correspondence between the time interval value and the repetition period value sequence. The continuous time segments maintain the sequential connection of arrival time numbers under a unified time scale and continue the rhythmic characteristics in the mode transition precursor set in terms of rhythmic structure.

[0025] Preferably, obtaining the relative intensity profile in the low-power stage includes the following steps:

[0026] Within the low-power hold-up determination interval, the amplitude values ​​corresponding to the arrival time number are read according to a unified time scale to form an amplitude change curve within the low-power hold-up determination interval. At the same time, the total energy data is received within the same time range to form a synchronous environmental energy fluctuation curve. The environmental energy values ​​and the synchronous environmental energy fluctuation curve are aligned point by point under a unified time scale.

[0027] Based on the point-by-point alignment results, the synchronous environmental energy fluctuation curve is divided into time segments. The environmental energy value sequence and amplitude value sequence within each time segment are read and compared in parallel to mark the segments that affect the overall energy rise.

[0028] For each segment affected by the overall energy rise, the environmental energy value and amplitude value are read point by point, and the amplitude component that is synchronized with the environmental energy change is separated to form a new amplitude change sequence;

[0029] The new amplitude change sequences are arranged sequentially according to their arrival time numbers and correspond to the rhythmic features in the mode transition precursor set to construct a relative intensity profile of the low-power phase.

[0030] Preferably, when separating the amplitude component that is synchronized with the change in environmental energy, the amplitude value in the intensity maintenance segment of the target radiation source itself is used as a reference segment. The amplitude change part corresponding to the fluctuation curve of the synchronized environmental energy is stripped point by point under a unified time scale, and the new amplitude change sequence is kept in one-to-one correspondence with the arrival time number.

[0031] Preferably, based on the relative intensity profile of the low-power phase, the radiation source trajectory on the original time axis is continuously spliced, including the following steps:

[0032] Unfold the radiation source trajectory on the original time axis under a unified time scale, identify the end time number of the trajectory segment before the amplitude drop and the start time number of the trajectory segment after the amplitude drop, insert the arrival time number interval covered by the relative intensity profile of the low power phase into the corresponding position on the original time axis in chronological order, and match the time interval between the inserted numbers with the rhythm features in the mode transition precursor set to construct a continuous splicing sequence.

[0033] The relative intensity values ​​are read point by point around the continuous splicing sequence and connected with the amplitude values ​​of the trajectory segment before the amplitude drops. At the same time, the time intervals between adjacent arrival time numbers are marked accordingly, so that the relative intensity profile of the low power stage is continuously expressed in terms of time axis and rhythm structure.

[0034] The spliced ​​single continuous sequence is expanded as a whole, and the rhythmic feature segments in the mode transition precursor set are mapped to the single continuous sequence, so that the segments before amplitude drop, the low power stage segments and the segments after amplitude drop are connected in terms of time interval arrangement and relative intensity expression, thus realizing the continuous connection of pulses before and after amplitude drop.

[0035] Preferably, the overall evolution trend after splicing is repositioned, including the following steps:

[0036] Under a unified time scale, the trajectory of the continuously spliced ​​radiation source is unfolded, and the arrival time number, relative intensity value and time interval value are read point by point. The amplitude change trend segment and the repetitive periodic rhythm characteristic segment in the mode switching precursor set are mapped to the corresponding time segment to form a reference segment.

[0037] The relative intensity profile values ​​of the low-power phase are read along the time axis around the reference section. The relative intensity profile values ​​and amplitude change trend segments are continuously arranged, and the time interval values ​​are matched with the repetition cycle rhythm characteristics one by one. The transition section of mode conversion is marked.

[0038] For each point in the mode transition transition zone, the relative intensity value and time interval value are analyzed and aligned with the amplitude change trend segment and the repetition period rhythm characteristics to determine the arrival time number from the continuous decrease in amplitude to the stable arrangement, thus forming the mode transition node position.

[0039] Using the mode transition node position as the boundary, the complete time series is segmented and the arrival time number, relative intensity value and rhythm characteristics are continuously arranged in the entire time domain to realize the continuous expression of the radiation source identity.

[0040] The radar signal multi-dimensional feature intelligent sorting system includes a mode precursor extraction module, a low-power preservation construction module, an energy separation processing module, a trajectory continuous stitching module, and an evolution trend positioning module.

[0041] The mode precursor extraction module collects the amplitude change curve, repetition period change curve, and arrival time distribution trajectory of the target radiation source from the long-range search phase to the low-power silent scan phase. It compares and analyzes the amplitude change curve, repetition period change curve, and arrival time distribution trajectory under a unified time scale, extracts the correlation features of continuous pulses before and after the amplitude drop, and forms a set of mode conversion precursors.

[0042] The low-power hold-up construction module extends the tracking of arrival time distribution trajectories within the amplitude drop interval based on the mode transition precursor set, and re-merges the pulses within the amplitude drop interval according to the rhythm characteristics in the mode transition precursor set to construct the low-power hold-up determination interval.

[0043] The energy separation and processing module, focusing on the low-power hold-in determination interval, introduces the synchronous environment energy fluctuation curve for comparative analysis, and combines it with the amplitude change curve within the low-power hold-in determination interval to separate the overall energy rise caused by the enhanced scanning stage, thereby obtaining the relative intensity profile of the low-power stage.

[0044] The trajectory continuous stitching module continuously stitches the radiation source trajectories on the original time axis based on the relative intensity profile of the low-power phase, and incorporates the rhythmic features in the mode transition precursor set into the continuous stitching process.

[0045] The evolution trend localization module, after completing the continuous splicing of radiation source trajectories, combines the pre-mode transition precursor set with the relative intensity profile of the low-power stage to reposition the overall evolution trend after splicing and determine the location of the mode transition node.

[0046] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0047] This invention constructs a set of precursors to mode transitions and integrates amplitude variation curves, repetition period variation curves, and arrival time distribution trajectories on a unified time scale. This establishes a criterion for determining rhythm continuity before a sudden drop in amplitude occurs, preventing pulses in the low-power phase from being classified solely based on amplitude strength, thus avoiding false rejection of pulse streams due to power decline. Through extended tracking and continuous stitching, the radiation source trajectory is fully represented on the time axis, ensuring continuous connection of the mode transition process in the time dimension and improving the stability of the radiation source trajectory and the consistency of the sorting results.

[0048] This invention introduces a synchronous environmental energy fluctuation curve and separates the overall energy rise caused by the enhanced scanning phase to obtain the relative intensity profile of the low-power phase, enabling the independent expression of the true intensity changes of the target radiation source in a complex electromagnetic environment. By combining this with a set of mode-switching precursors to reposition the overall evolution trend, the location of mode-switching nodes can be accurately determined, maintaining the continuous expression of the radiation source's identity across the entire time domain, thereby improving the reliability of parameter evolution analysis and situation assessment results. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0050] Figure 1 This is a flowchart of the method of the present invention;

[0051] Figure 2 Here is a flowchart of the method for constructing the low-power hold-up determination interval according to the present invention;

[0052] Figure 3 This is a flowchart of the method for obtaining the relative intensity profile in the low-power stage according to the present invention.

[0053] Figure 4 This is a schematic diagram of the modules of the present invention. Detailed Implementation

[0054] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0055] This invention provides, for example Figures 1 to 3 The radar signal multi-dimensional feature intelligent sorting method shown includes the following steps:

[0056] Step 1: Collect amplitude variation curves, repetition period variation curves, and arrival time distribution trajectories of the target radiation source from the long-range search phase to the low-power silent scanning phase. Compare and analyze the amplitude variation curves, repetition period variation curves, and arrival time distribution trajectories under a unified time scale, extract the correlation features of continuous pulses before and after the amplitude drop, and form a set of mode conversion precursors.

[0057] The correlation features of consecutive pulses before and after the sudden drop in amplitude are extracted to form a set of precursors to mode transition. The specific steps are as follows:

[0058] Within the observation interval covering the entire continuous time period from the long-range search phase to the low-power silent scanning phase, all pulses emitted by the target radiation source are recorded pulse by pulse. The arrival time of each pulse is numbered and sorted according to a unified time scale, and the corresponding amplitude value and the time interval between adjacent pulses are recorded simultaneously. An amplitude variation curve is formed by arranging the amplitude values ​​of each pulse in chronological order, a repetitive periodic variation curve is formed by arranging the time intervals between adjacent pulses in chronological order, and an arrival time distribution trajectory is formed based on the distribution position of all pulses on the unified time scale. Under this unified time scale, the amplitude variation curve, the repetitive periodic variation curve, and the arrival time distribution trajectory are unfolded in parallel, so that the three curves maintain a one-to-one correspondence on the same time axis. That is, any time scale point can simultaneously correspond to the amplitude value, the repetitive periodic value, and the arrival time position, thus forming a continuous three-dimensional time evolution record covering the two working phases and the transition interval.

[0059] In three curves unfolded on a unified time scale, the amplitude change curve is scanned point by point along the time axis. When the amplitude value is detected to transition from a stable segment to a decreasing segment between multiple consecutive pulses, the corresponding start and end time points of the transition are marked as the boundary of the amplitude drop interval. A fixed number of pulse samples are extended forward and backward around the boundary of the amplitude drop interval to form a continuous pulse sequence containing the segment before and after the amplitude drop. For the continuous pulse sequence, the corresponding time interval changes are read point by point on the repetition period change curve to observe whether the time interval before and after the amplitude drop maintains a rhythmic continuity. At the same time, the pulse time arrangement is checked on the arrival time distribution trajectory to see if the pulse time arrangement maintains continuous numbering and there are no time gaps. Pulse segments that simultaneously satisfy the requirements of continuous amplitude drop, continuous rhythmic structure of repetition period value, and continuous arrival time numbering are extracted to form continuous pulse association segments corresponding to the amplitude drop process. The continuous pulse association segments are stored in chronological order to provide a candidate data set for subsequent feature combination.

[0060] After obtaining the continuous pulse correlation segment, the three curves are re-examined and compared in detail around the continuous pulse correlation segment. The three curves are the amplitude change curve, the repetition period change curve, and the arrival time distribution trajectory. The amplitude values ​​in the continuous pulse correlation segment are arranged point by point to observe whether there is a continuous transition relationship between the amplitude values ​​before and after the sudden drop. At the same time, the corresponding repetition period values ​​are read point by point to confirm that there is no abrupt change in the repetition period in the entire continuous pulse correlation segment. The arrival time number of the corresponding pulse in the arrival time distribution trajectory is compared one by one with the arrival time numbers of the preceding and following pulses to confirm that the arrival time numbers are sequentially connected. Under the premise that the above three curves simultaneously satisfy the continuous amplitude transition, the rhythm structure continuation, and the sequential connection of the numbering, the amplitude change trend segment, the repetition period rhythm segment, and the arrival time sequence segment in the corresponding continuous pulse correlation segment are combined into a correlation feature. Multiple correlation features are arranged in chronological order to form a feature combination sequence containing multiple continuous transition segments, so that the amplitude change curve, the repetition period change curve, and the arrival time distribution trajectory complete cross-correspondence and joint expression at the same time scale.

[0061] Finally, the feature combination sequences arranged in chronological order are aggregated as a whole. The amplitude transition segments, repetitive rhythmic segments, and arrival time sequence connection segments recorded in each associated feature are integrated to form a set of pattern transition precursors covering the time range before and after the amplitude drop. This set of pattern transition precursors contains specific amplitude change value sequences, corresponding repetitive cycle value sequences, and arrival time number sequences. These three types of sequences maintain a point-to-point correspondence under a unified time scale and completely cover the transition interval from the long-range search phase to the low-power silent scan phase. In this way, the original single phenomenon of amplitude decrease is transformed into a multi-dimensional joint feature expression containing amplitude change trends, repetitive rhythmic structures, and arrival time sequence structures. This forms a set of pattern transition precursors on the timeline that can describe the initiation and transition processes of pattern transition, providing continuous feature evidence for subsequent extended tracking and pulse re-categorization around the amplitude drop interval.

[0062] Step 2: Based on the mode transition precursor set, extend the arrival time distribution trajectory within the amplitude drop interval, and re-merge the pulses within the amplitude drop interval according to the rhythm characteristics in the mode transition precursor set to construct the low power hold-off determination interval.

[0063] The specific steps for constructing the low-power hold-off decision interval are as follows:

[0064] The system reads the start and end times, corresponding arrival time sequence, and repetition period sequence of each record from the mode transition precursor set. The last arrival time number covered by the mode transition precursor set is determined as the extended tracking start point number, and its position on a unified time scale is used as the starting time point for extended tracking. After the starting time point, the arrival time numbers of all pulses within the amplitude drop interval are read sequentially. Each arrival time number is then arranged sequentially with the extended tracking start point number, and the time interval between two adjacent arrival time numbers is calculated. Simultaneously, the corresponding time interval value is compared with the repetition period sequence recorded in the mode transition precursor set. The repeated periodic value sequence is compared item by item. When the time interval value falls within the range covered by the repeated periodic value sequence in the mode transition precursor set, the corresponding pulse is identified as a rhythm continuation pulse and added to the extended tracking sequence. If the time interval value of a certain pulse deviates from the repeated periodic value sequence in the mode transition precursor set, the arrival time number of the next pulse is read and the time interval is calculated with the arrival time number of the previously identified rhythm continuation pulse. In this way, the pulses that meet the rhythm continuation condition within the amplitude drop interval form a continuous number sequence on the time axis, thereby achieving extended tracking across the amplitude drop interval on the arrival time distribution trajectory.

[0065] After obtaining the complete extended tracking sequence, all pulses in the extended tracking sequence are sorted in ascending order of arrival time number. The sorted pulses are then divided into several consecutive time segments, and the corresponding time interval values ​​are read point by point within each consecutive time segment. These time interval values ​​are then matched segment by segment with the repetition periodic value sequence recorded in the mode transition precursor set. When multiple consecutive time interval values ​​fall within the same rhythm interval of the repetition periodic value sequence in the mode transition precursor set, the corresponding multiple consecutive pulses are grouped into the same pulse group, and a unified rhythm identifier is assigned to the pulse group. Pulses that cannot be directly grouped into any pulse group are retained. The arrival time number and time interval value are matched, and the corresponding pulse is inserted between adjacent pulse groups. By comparing the relationship between the time interval value of the corresponding pulse and the rhythm indicators of the preceding and following pulse groups, the direction of the corresponding pulse is determined. When the time interval value of the corresponding pulse forms a continuous arrangement with the time interval value within the adjacent pulse group after insertion, the corresponding pulse is added to the corresponding pulse group. Through pulse sorting, segmentation, and insertion and merging, the pulses in the amplitude drop interval are re-formed on the arrival time distribution trajectory to form a continuous rhythm structure consistent with the rhythm characteristics in the mode transition precursor set, thereby completing the re-merging of pulses in the amplitude drop interval.

[0066] After regrouping into multiple continuous pulse groups, the earliest and latest arrival times of each pulse group are used as the boundary numbers of the continuous time segments. All pulse groups are expanded as a whole under a unified time scale, and segments with time intervals between pulse groups less than a preset time span are merged into a single continuous time segment. The corresponding continuous time segment is defined as the low-power hold-up determination interval. Within the low-power hold-up determination interval, the arrival time number, corresponding time interval value, and corresponding amplitude value of each pulse are recorded point by point. These data are matched item by item with the repetition period value sequence recorded in the mode transition precursor set, so that the arrival time distribution trajectory within the low-power hold-up determination interval maintains a connection with the mode transition precursor set in terms of numbering order, while continuing the rhythmic characteristics of the mode transition precursor set in terms of rhythmic structure. By continuously unfolding the process of extending the tracking sequence, re-merging pulse groups, and merging time segments, a low-power hold-off determination interval covering the range of sudden amplitude drop is constructed under a unified time scale. This ensures that the low-power hold-off determination interval is continuously expressed on the arrival time distribution trajectory and maintains a consistent rhythm in the value of the repetition period. It also provides complete time segment data for subsequent comparative analysis of environmental energy fluctuation curves around the low-power hold-off determination interval.

[0067] It should be noted that the preset time span refers to the maximum allowable time interval threshold used to determine whether adjacent pulse groups belong to the same continuous time segment under a unified time scale. This time span is not arbitrarily set, but is determined based on the repeating period value sequence recorded in the mode transition precursor set. Usually, the upper limit range of the repeating period value sequence or its stable fluctuation range is taken as a reference, and the time interval distribution formed by extended tracking within the amplitude drop interval is combined for setting.

[0068] When the time interval between adjacent pulse groups does not exceed the preset time span, the rhythmic structure is considered to remain continuous and can be merged into the same continuous time segment; when the time interval exceeds the preset time span, the rhythmic structure is considered to be broken and will not be merged.

[0069] By using this time span determined with rhythmic characteristics as a reference, a reasonable low-power holding judgment interval can be constructed while maintaining rhythmic continuity.

[0070] Step 3: Around the low power hold judgment interval, the synchronous environment energy fluctuation curve is introduced for comparative analysis, and combined with the amplitude change curve within the low power hold judgment interval, the overall energy rise caused by the enhanced scanning stage is separated to obtain the relative intensity profile of the low power stage.

[0071] The overall energy rise caused by the enhanced scanning phase is separated to obtain the relative intensity profile of the low-power phase. The specific steps are as follows:

[0072] Within the defined low-power hold-off interval, the amplitude value corresponding to each arrival time number is read point by point according to a unified time scale. All amplitude values ​​are then arranged sequentially in chronological order to form an amplitude variation curve within the low-power hold-off interval. Simultaneously with the formation of this curve, real-time total energy data across the entire receiving frequency domain is collected between the same starting and ending time points. The total received energy value corresponding to each time scale point is recorded point by point and arranged in chronological order to form a synchronous environmental energy fluctuation curve. Subsequently, the amplitude variation curve within the low-power hold-off interval and the synchronous environmental energy fluctuation curve are aligned point by point on a unified time scale, ensuring that each arrival time number corresponds to both an amplitude value and an environmental energy value. A hyperbolic parallel recording table is established using the arrival time number as a unique index, thus creating a synchronous comparison structure on the time axis between local amplitude variations within the low-power hold-off interval and overall environmental energy variations.

[0073] It should be noted that the total received energy value can be obtained by integrating and accumulating the total signal power within the received frequency domain at the corresponding time scale point. Specifically, within the sampling window corresponding to each time scale point, the instantaneous signal power detected by the receiving device across the entire operating frequency band is sampled. The power values ​​of each frequency unit within that time window are accumulated or integrated to obtain the overall energy value at that time scale point. Subsequently, the overall energy values ​​calculated for each time scale point are arranged in chronological order, forming a sequence of total energy values ​​reflecting the overall electromagnetic environment intensity changes, which is used to construct the synchronous environment energy fluctuation curve.

[0074] After point-by-point alignment, the synchronous environmental energy fluctuation curve is continuously scanned along the time axis, dividing the entire time range covered by the low-power maintenance judgment interval into continuous time segments. Each time segment contains a fixed number of time scale points, and the corresponding environmental energy value sequence is read within each time segment. Simultaneously, the amplitude value sequence corresponding to the amplitude change curve within the time segment is read. The environmental energy value sequence and the amplitude value sequence are arranged in parallel, and the direction and amplitude of change of the two sequences within the same time segment are compared. When the environmental energy value continuously increases within a certain time segment, and the amplitude value increases synchronously within the same time segment, the corresponding time segment is marked as the overall energy rise impact segment, and the start and end time numbers of the corresponding time segment are recorded. When the environmental energy value changes while the amplitude value remains fluctuating around a predetermined rhythm range, the corresponding time segment is marked as the target radiation source's own intensity maintenance segment. Through segment-by-segment scanning and marking, the low-power maintenance judgment interval is decomposed into multiple time segments with clear attribute identifiers, and it is determined which time segments are affected by the overall energy rise during the enhanced scanning phase.

[0075] After marking the time segments, separation processing is performed on the time ranges marked as affected by the overall energy rise. Specifically, within each affected segment, the environmental energy value and amplitude value are read point by point. The changes in the environmental energy value within the corresponding time segment are listed in chronological order, and the changes in the environmental energy value are compared with the average amplitude level of the target radiation source's own intensity in the segment. Under a unified time scale, for each time scale point, the amplitude component synchronized with the environmental energy change is subtracted from the original amplitude value, retaining the amplitude part that is consistent with the pre-mode transition precursor set in terms of rhythm characteristics, thus forming a new amplitude value at each time scale point. The new amplitude values ​​of all time scale points are arranged in chronological order to form an amplitude change sequence after the overall energy rise separation processing, and recorded in parallel with the original amplitude change curve, so as to compare and observe the differences before and after separation on the time axis.

[0076] After separating all segments affected by the overall energy rise, the separated amplitude change sequences are aggregated within the low-power hold-up determination interval and arranged sequentially according to arrival time numbers from smallest to largest, forming a relative intensity profile of the low-power phase covering the entire low-power hold-up determination interval. The corresponding relative intensity profiles are mapped point-by-point to the arrival time number sequence on a unified time scale, and the rhythmic structure continues the rhythmic characteristics recorded in the mode transition precursor set. This ensures that the relative intensity profile of the low-power phase eliminates the influence of the overall energy rise during the enhanced scanning phase while maintaining the true intensity change trend of the target radiation source within the low-power phase. Through this continuous processing of point-by-point alignment, segmentation marking, point-by-point separation, and overall aggregation, a relative intensity profile that accurately reflects the intensity change state of the low-power phase is constructed on the time axis, providing stable and coherent intensity representation data for subsequent continuous stitching around the radiation source trajectory and mode transition node localization.

[0077] Step 4: Based on the relative intensity profile of the low-power phase, the radiation source trajectory on the original time axis is continuously spliced, and the rhythmic characteristics in the mode transition precursor set are incorporated into the continuous splicing process to achieve continuous connection of pulses before and after the amplitude drop.

[0078] The radiation source trajectories on the original timeline are continuously spliced ​​together. The specific steps are as follows:

[0079] Under a unified time scale, the radiation source trajectory on the original time axis is fully unfolded, clearly identifying the end time number of the trajectory segment before the amplitude drop and the start time number of the trajectory segment after the amplitude drop. Simultaneously, all arrival time number intervals covered by the relative intensity profile of the low-power phase are read. Each arrival time number in the relative intensity profile of the low-power phase is arranged chronologically and inserted one by one into the corresponding time position on the original time axis, ensuring that the end time number of the trajectory segment before the amplitude drop and the start time number in the relative intensity profile of the low-power phase are adjacent on the time axis. During insertion, the time interval between each newly inserted arrival time number and the previous time number is recorded point by point, and the corresponding time interval is mapped to the rhythmic feature interval in the mode transition precursor set. When the time interval falls within the repetition period range recorded in the mode transition precursor set, the corresponding time interval is marked as a rhythmic continuation interval, and the corresponding pulse is included in the continuous splicing sequence, thereby establishing a continuous structure spanning the amplitude drop interval at the time number level.

[0080] After the time numbers are arranged consecutively, the relative intensity value corresponding to each arrival time number within the inserted segment is read point by point. The corresponding relative intensity values ​​are then connected in chronological order to the amplitude values ​​at the end of the trajectory segment before the amplitude drop, ensuring that the relative intensity profile of the low-power phase is numerically adjacent to the original trajectory. During this connection process, the time interval between two adjacent arrival time numbers is checked point by point again to ensure that the corresponding time interval is consistent with the rhythmic characteristics in the mode transition precursor set. The time intervals within the relative intensity profile segment of the low-power phase are then identified as rhythmic segments corresponding to the rhythmic characteristics of the mode transition precursor set. Through this synchronous arrangement of time numbers and relative intensity values, the relative intensity profile of the low-power phase forms a continuous segment on the time axis connected to the trajectory segment before the amplitude drop, while maintaining a consistent arrangement order with the mode transition precursor set in terms of rhythmic structure.

[0081] After connecting the time numbers and relative intensity values, the entire spliced ​​radiation source trajectory is unfolded as a whole. The trajectory segment before the amplitude drop, the relative intensity contour segment of the low-power phase, and the trajectory segment after the amplitude drop are arranged in chronological order into a single continuous sequence. In the single continuous sequence, each arrival time number and corresponding relative intensity value are read point by point, and the rhythmic feature segments in the mode transition precursor set are mapped to the single continuous sequence, so that the rhythmic features are continuously expressed between the segment before the amplitude drop, the segment of the low-power phase, and the segment after the amplitude drop. Subsequently, the time intervals in the entire continuous sequence are arranged point by point to ensure that there are no gaps in the time intervals or number jumps, so that the radiation source trajectory on the original time axis forms a continuous expression state after the intervention of the relative intensity contour of the low-power phase. Thus, the continuous splicing of the radiation source trajectory on the original time axis based on the relative intensity contour of the low-power phase is completed, and the rhythmic features in the mode transition precursor set are integrated in the continuous splicing process, realizing the continuous connection of the pulses before and after the amplitude drop.

[0082] Step 5: After completing the continuous splicing of the radiation source trajectory, the overall evolution trend after splicing is repositioned by combining the mode transition precursor set and the relative intensity profile of the low power stage, the mode transition node position is determined, and the radiation source identity is continuously expressed in the entire time domain.

[0083] The overall evolution trend after splicing is redefined, and the specific steps are as follows:

[0084] Under a unified time scale, the continuously spliced ​​radiation source trajectories are unfolded sequentially according to their arrival time numbers from smallest to largest, forming a complete time series covering the long-range search phase, the low-power phase, and subsequent phases. Within the complete time series, the relative intensity value corresponding to each arrival time number and the time interval between adjacent pulses are read point by point. The amplitude change trend segments and repetitive periodic rhythmic feature segments in the mode transition precursor set are mapped to the corresponding time segments of the complete time series, making the mode transition precursor set a clearly marked reference segment on the time axis. Subsequently, the continuous transition positions of amplitude changes, the order of time intervals, and the rhythmic segment divisions are recorded point by point within the reference segment. These records are used as the starting reference for repositioning the overall evolutionary trend, so that the complete time series simultaneously possesses the continuous expression after splicing and the rhythmic identification of the mode transition precursor set on the time axis.

[0085] Subsequently, starting from the position corresponding to the end time number of the mode transition precursor set, the relative intensity profile values ​​of the low-power phase are read point by point along the time axis. The relative intensity profile values ​​are then continuously arranged according to the time sequence and the amplitude change trend at the end of the mode transition precursor set, so that the two amplitude values ​​form a continuous connecting curve on the time axis. During this process, the time interval value corresponding to each arrival time number is recorded point by point, and the time interval value is compared with the repetitive periodic rhythm characteristics in the mode transition precursor set item by item, so that the relative intensity profile of the low-power phase continues the rhythmic segment of the mode transition precursor set in terms of rhythm arrangement. When the relative intensity profile of the low-power phase shows a change pattern from the amplitude decrease segment to the stable segment in a certain continuous time segment, and the corresponding time interval value maintains the same arrangement order as the rhythmic characteristics of the mode transition precursor set in the corresponding segment, the corresponding continuous time segment is marked as the mode transition transition segment, and the start arrival time number and end arrival time number of the corresponding segment are recorded, so that the candidate range of the mode transition node is clearly defined on the time axis.

[0086] It should be noted that the relative intensity profile value refers to the actual intensity value of the target radiation source corresponding to each arrival time number within the low-power maintenance judgment interval, after overall energy enhancement separation processing. This value is not the original received amplitude, but rather, after removing the influence of synchronous environment energy fluctuations, retaining only the intensity component related to the target radiation source, thus reflecting the true intensity change state of the target radiation source itself during the low-power phase. Arranging these separated intensity values ​​in order of arrival time number constitutes the relative intensity profile during the low-power phase, and its point-by-point value is the relative intensity profile value.

[0087] After marking the transitional zones for mode transition, each arrival time number within the corresponding zone is expanded point by point. The relative intensity values ​​are arranged in chronological order and aligned point by point with the amplitude change trend segments in the pre-mode transition precursor set, forming a continuous change curve on the time axis. Simultaneously, the time interval values ​​within the corresponding zone are arranged in chronological order and correspond point by point with the repetitive periodic rhythm characteristics in the pre-mode transition precursor set, forming a cross-segment continuity structure on the time axis. After alignment, the specific time number where the relative intensity values ​​transition from a continuous decrease to a stable arrangement is found, and it is confirmed that there is no rhythmic break in the time interval values ​​at the corresponding time number. This time number is then determined as the mode transition node position. Subsequently, this mode transition node position is uniquely identified in the complete time series, making this node a clear boundary between the long-range search phase and the low-power phase.

[0088] After determining the location of the mode transition node, the complete time series is segmented according to this node location. The arrival time number segment before the node is marked as the long-range search stage evolution segment, and the arrival time number segment after the node until the end of the low-power stage is marked as the low-power silent scan stage evolution segment. Throughout the time series, the arrival time numbers, relative intensity values, and rhythmic features are maintained in a continuous sequence. This repositioning method, using the set of precursors to mode transition as a reference and the relative intensity profile of the low-power stage as a basis, allows the spliced ​​overall evolutionary trend to form a single, continuous expression structure on the time axis, maintaining a continuous expression of the radiation source identity across the entire time domain. This completes the repositioning of the spliced ​​overall evolutionary trend and the determination of the mode transition node location.

[0089] This invention provides, for example Figure 4 The radar signal multi-dimensional feature intelligent sorting system shown includes a pattern precursor extraction module, a low power preservation construction module, an energy separation processing module, a trajectory continuous stitching module, and an evolution trend positioning module.

[0090] The mode transition precursor extraction module collects the amplitude change curves, repetition period change curves, and arrival time distribution trajectories of the target radiation source from the long-range search phase to the low-power silent scan phase. It compares and analyzes the amplitude change curves, repetition period change curves, and arrival time distribution trajectories under a unified time scale, extracts the correlation features of continuous pulses before and after the amplitude drop, and forms a set of mode transition precursors.

[0091] The low-power hold-up construction module extends and tracks the arrival time distribution trajectory within the amplitude drop interval based on the mode transition precursor set. It then re-merges the pulses within the amplitude drop interval according to the rhythm characteristics in the mode transition precursor set to construct the low-power hold-up determination interval.

[0092] The energy separation processing module, focusing on the low-power hold judgment interval, introduces the synchronous environment energy fluctuation curve for comparative analysis, and combines it with the amplitude change curve within the low-power hold judgment interval to separate the overall energy rise caused by the enhanced scanning stage, thereby obtaining the relative intensity profile of the low-power stage.

[0093] The trajectory continuous stitching module continuously stitches together the radiation source trajectories on the original time axis based on the relative intensity profile of the low-power phase, and incorporates the rhythmic features in the mode transition precursor set into the continuous stitching process.

[0094] The evolution trend localization module, after completing the continuous splicing of radiation source trajectories, combines the pre-mode transition precursor set with the relative intensity profile of the low-power stage to reposition the overall evolution trend after splicing and determine the location of the mode transition node.

[0095] The intelligent sorting method for multi-dimensional features of radar signals provided in this embodiment of the invention is implemented through the aforementioned intelligent sorting system for multi-dimensional features of radar signals. For details of the specific methods and processes of the intelligent sorting system for multi-dimensional features of radar signals, please refer to the embodiments of the intelligent sorting method for multi-dimensional features of radar signals described above, which will not be repeated here.

[0096] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for intelligent sorting of radar signals based on multi-dimensional features, characterized in that, Includes the following steps: The amplitude variation curves, repetition period variation curves, and arrival time distribution trajectories of the target radiation source during the long-range search phase to the low-power silent scanning phase are collected and compared and analyzed under a unified time scale. The correlation characteristics of continuous pulses before and after the amplitude drop are extracted to form a set of mode conversion precursors. Based on the mode transition precursor set, the arrival time distribution trajectory within the amplitude drop interval is extended and tracked. According to the rhythm characteristics in the mode transition precursor set, the pulses within the amplitude drop interval are re-merged to construct the low power hold-off determination interval. Around the low-power hold-off determination interval, the synchronous environment energy fluctuation curve is introduced for comparative analysis. Combined with the amplitude change curve within the low-power hold-off determination interval, the overall energy rise caused by the enhanced scanning stage is separated to obtain the relative intensity profile of the low-power stage. Based on the relative intensity profile of the low-power phase, the radiation source trajectories on the original time axis are continuously spliced ​​together, and the rhythmic features in the mode transition precursor set are incorporated into the continuous splicing process. After completing the continuous splicing of radiation source trajectories, the overall evolution trend after splicing is repositioned by combining the pre-mode transition set with the relative intensity profile of the low-power phase, and the location of the mode transition node is determined.

2. The intelligent sorting method for multi-dimensional features of radar signals according to claim 1, characterized in that, Extracting the correlation features of consecutive pulses before and after a sudden drop in amplitude to form a set of mode transition precursors includes the following steps: Within the observation interval covering the long-range search phase to the low-power silent scan phase, all pulses are recorded pulse by pulse, and the arrival times are numbered and sorted according to a unified time scale. The amplitude values ​​and the time interval between adjacent pulses are recorded synchronously to form amplitude variation curves, repetition period variation curves, and arrival time distribution trajectories, which are then unfolded in parallel under a unified time scale. Scan the amplitude change curve along the time axis, mark the boundary of the amplitude drop interval, construct a continuous pulse sequence around the boundary of the amplitude drop interval, and combine the repeating period change curve and the arrival time distribution trajectory to select continuous pulse association segments that meet the requirements of continuous amplitude transition, rhythm structure continuation and numbering sequence connection. Around the continuous pulse correlation segment, the amplitude change curve, the repetition period change curve, and the arrival time distribution trajectory are compared point by point to form correlation features including amplitude change trend segment, repetition period rhythm segment, and arrival time sequence segment, which are arranged in chronological order as feature combination sequence; The feature combination sequences are aggregated as a whole, integrating the continuous amplitude transition segments, the repeating periodic rhythm continuation segments, and the arrival time sequence connection segments to form a set of pattern transition precursors covering the time range before and after the sudden drop in amplitude.

3. The intelligent sorting method for multi-dimensional features of radar signals according to claim 2, characterized in that, The amplitude change trend segments, repetitive periodic rhythm segments, and arrival time sequence segments in the premonition set of mode transitions are arranged point by point on a unified time scale, and the continuous pulse-related segments span the time range on both sides of the boundary of the amplitude drop interval.

4. The intelligent sorting method for multi-dimensional features of radar signals according to claim 2, characterized in that, Constructing the low-power hold decision interval includes the following steps: Arrival time numbers within the amplitude drop interval are sorted one by one along a unified time scale, and the time interval values ​​between adjacent arrival time numbers are matched with the repeating period value sequence one by one. The rhythm continuation pulse is then selected to form an extended tracking sequence. Arrival time numbers are continuously arranged around the extended tracking sequence, and continuous time periods are divided according to the time interval values. The time interval values ​​within the continuous time periods are matched with the repeating period value sequence to construct a pulse group with a unified rhythm identifier. The constructed pulse group is expanded under a unified time scale. The boundary number of the continuous time segment is determined based on the earliest arrival time number and the latest arrival time number. The segments whose time interval values ​​meet the condition of continuous arrangement are merged to form a continuous time segment. Within a continuous time interval, the arrival time number, time interval value, and amplitude value are recorded, and a correspondence is maintained with the repetitive period value sequence in the mode transition precursor set to construct a low-power hold-up determination interval covering the amplitude drop interval.

5. The intelligent sorting method for multi-dimensional features of radar signals according to claim 4, characterized in that, The extended tracking starting point number is the last arrival time number within the coverage area of ​​the mode transition precursor set. The construction of pulse groups is based on the correspondence between the time interval value and the repetition period value sequence. Continuous time segments maintain the sequential connection of arrival time numbers under a unified time scale, and continue the rhythmic characteristics in the mode transition precursor set in terms of rhythmic structure.

6. The intelligent sorting method for multi-dimensional features of radar signals according to claim 4, characterized in that, Obtaining the relative intensity profile in the low-power phase involves the following steps: The environmental energy values ​​and synchronous environmental energy fluctuation curves are aligned point by point under a unified time scale. Based on the point-by-point alignment results, the synchronous environmental energy fluctuation curve is divided into time segments. The environmental energy value sequence and amplitude value sequence within each time segment are read and compared in parallel to mark the segments that affect the overall energy rise. For each segment affected by the overall energy rise, the environmental energy value and amplitude value are read point by point, and the amplitude component that is synchronized with the environmental energy change is separated to form a new amplitude change sequence; The new amplitude change sequences are arranged sequentially according to their arrival time numbers and correspond to the rhythmic features in the mode transition precursor set to construct a relative intensity profile of the low-power phase.

7. The intelligent sorting method for multi-dimensional features of radar signals according to claim 6, characterized in that, When separating the amplitude component that is synchronized with the changes in environmental energy, the amplitude value in the segment where the intensity of the target radiation source itself is maintained is used as a reference segment. The amplitude change part corresponding to the fluctuation curve of the synchronized environmental energy is stripped point by point under a unified time scale, and the new amplitude change sequence is kept in one-to-one correspondence with the arrival time number.

8. The intelligent sorting method for multi-dimensional features of radar signals according to claim 6, characterized in that, Based on the relative intensity profile of the low-power phase, the radiation source trajectories on the original time axis are continuously stitched together, including the following steps: Unfold the radiation source trajectory on the original time axis under a unified time scale, identify the end time number of the trajectory segment before the amplitude drop and the start time number of the trajectory segment after the amplitude drop, insert the arrival time number interval covered by the relative intensity profile of the low power phase into the corresponding position on the original time axis in chronological order, and match the time interval between the inserted numbers with the rhythm features in the mode transition precursor set to construct a continuous splicing sequence. The relative intensity values ​​are read point by point around the continuous splicing sequence, and then connected and arranged with the amplitude values ​​of the trajectory segments before the sudden drop in amplitude. At the same time, the time intervals between adjacent arrival time numbers are marked accordingly. The spliced ​​single continuous sequence is expanded as a whole, and the rhythmic feature segments in the mode transition precursor set are mapped to the single continuous sequence, so that the segments before the amplitude drop, the low power stage segments, and the segments after the amplitude drop remain consistent in terms of time interval arrangement and relative intensity expression.

9. The intelligent sorting method for multi-dimensional features of radar signals according to claim 8, characterized in that, The overall evolution trend after splicing is redefined, including the following steps: Under a unified time scale, the trajectory of the continuously spliced ​​radiation source is unfolded, and the arrival time number, relative intensity value and time interval value are read point by point. The amplitude change trend segment and the repetitive periodic rhythm characteristic segment in the mode switching precursor set are mapped to the corresponding time segment to form a reference segment. The relative intensity profile values ​​of the low-power phase are read along the time axis around the reference section. The relative intensity profile values ​​and amplitude change trend segments are continuously arranged, and the time interval values ​​are matched with the repetition cycle rhythm characteristics one by one. The transition section of mode conversion is marked. For each point in the mode transition transition zone, the relative intensity value and time interval value are analyzed and aligned with the amplitude change trend segment and the repetition period rhythm characteristics to determine the arrival time number from the continuous decrease in amplitude to the stable arrangement, thus forming the mode transition node position. Using the mode transition node position as the boundary, the complete time series is segmented, and the arrival time number, relative intensity value and rhythm characteristics are arranged continuously in the entire time domain.

10. A radar signal multi-dimensional feature intelligent sorting system, used to implement the radar signal multi-dimensional feature intelligent sorting method according to any one of claims 1-9, characterized in that, It includes a pattern precursor extraction module, a low-power preservation construction module, an energy separation and processing module, a trajectory continuous stitching module, and an evolution trend localization module; The mode precursor extraction module collects the amplitude change curve, repetition period change curve, and arrival time distribution trajectory of the target radiation source from the long-range search phase to the low-power silent scan phase. It compares and analyzes the amplitude change curve, repetition period change curve, and arrival time distribution trajectory under a unified time scale, extracts the correlation features of continuous pulses before and after the amplitude drop, and forms a set of mode conversion precursors. The low-power hold-up construction module extends the tracking of arrival time distribution trajectories within the amplitude drop interval based on the mode transition precursor set, and re-merges the pulses within the amplitude drop interval according to the rhythm characteristics in the mode transition precursor set to construct the low-power hold-up determination interval. The energy separation and processing module, focusing on the low-power hold-in determination interval, introduces the synchronous environment energy fluctuation curve for comparative analysis, and combines it with the amplitude change curve within the low-power hold-in determination interval to separate the overall energy rise caused by the enhanced scanning stage, thereby obtaining the relative intensity profile of the low-power stage. The trajectory continuous stitching module continuously stitches the radiation source trajectories on the original time axis based on the relative intensity profile of the low-power phase, and incorporates the rhythmic features in the mode transition precursor set into the continuous stitching process. The evolution trend localization module, after completing the continuous splicing of radiation source trajectories, combines the pre-mode transition precursor set with the relative intensity profile of the low-power stage to reposition the overall evolution trend after splicing and determine the location of the mode transition node.

Images